diff --git a/nvfp/Meta-Llama-3-8B/0.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/0.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..7f426e31c040dbf466db8a325b9b36b56ee9c37b
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/0.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dec5250399545659e8e880a97d3a6c6c960f6f50a16ad7c8e91942374819048e
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/0.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/0.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..59a89504b457d277304396edf951764f10cd8715
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/0.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f2f03d2cd2fdb83c1b58b6854fc629ccb1afa0015a18daaa49a01fedf0104d8
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/0.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/0.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..16b14dbdc2ab20f4ce868e54feaaf5fb2e5b471f
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/0.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:384d43bd25e1a2af6478c6687686ea663a47a2625d300395d3622eeb41891ed3
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/0.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/0.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d2c0195d4493802bcec80c495e2e4b48652e79c6
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/0.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:023a540d812ddf2f91f81d910ded8df0a923cddcb1c798aee312a19bac47c987
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/0.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/0.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e34226a1a44f6197295de67caa6e4016a5ab7c22
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/0.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f949075e0d9d8946f483f67ac93ed711677dd0956b721d6dae18afc5933aeb8d
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/0.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/0.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..7efc2bf8eee8f620628b6a0a875ac68bf054d7dc
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/0.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:507f0ed697c6ba7490393b8307466296018acc58f0437b3cff02f69c8750f795
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/0.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/0.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ebe4aaf53d580e2524c7c12527e81a79d6a4a837
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/0.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3b18cc5705bf9f970fc3a0a74c1e80d16dd80ec2af38913787df781871bd82c
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/1.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/1.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c1f314fb3ef1c244e6de99bc19fb31a4c7336a41
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/1.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1deadf9d1c480af5232322fedee6c5a105505fb3e895686ab471929b6032cfaf
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/1.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/1.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1115611fd6b4cb629a6a763ce6454d46e6653ff9
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/1.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ef3a8d5f31f0a1c747245ed989939df85aac7e64dee9e2064752c74ff9b0fe7
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/1.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/1.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..07d27e045ab0e0ebc118bc34c7e536c500f5e24f
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/1.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94c4c0bc5cb6042c38d17479fbd25c889e2ea4d6a7ef7b295bd69413fd885631
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/1.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/1.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c3da9537121fb69f6217fa0fcb75922ef37f11d6
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/1.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df998ce02ab3a85e1265bfc6a38382d3e44f21971b2b03ecd5dd25986ab3bafc
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/1.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/1.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3c913d64e2f70fa891d93c0bafaa2be7e38d7497
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/1.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e80ce4146c2b654a45fbb33a6fedb455071a9c806fcb2c369416af1056450d89
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/1.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/1.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a7dda5feaa38264d4d7d2245bb12846b1061cc3f
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/1.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4a76381282f356556ac990ad314f2fac47129c4c8a88f9e24605b4ca371dff1
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/1.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/1.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8ec8c9e98fc8702b217abc67ed201d634488bcc6
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/1.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64ffc6443ba2f726fc4a6878ea6cc6b5da494b4c3f01271ff9e2f175d70c7d0a
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/10.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/10.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..50093d262d6c90d296fed3fabcf05aa745709a96
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/10.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4feaf72103059e0356f6d9684e2d673791146174a45c68f027e6386d2841215e
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/10.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/10.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0a4ea3c985cdf1560f4b824d002f8a40008b2d6c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/10.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30658d161ac52f3bbcaeb0df70895ab561bbe72ebf0468fc8610d10873135ef0
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/10.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/10.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..325df94787a55a54d2b0f7ed3bab6ceccfcedca0
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/10.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cf6c9eea6e126000eb47992560c8fabad2afe0537582fd6e40b36e619843142
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/10.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/10.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..47c4d17d54ce302e49662af1c987e525f39b9b71
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/10.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:771fd8af4ef30b990810e869804a0fc78749664a26168ceb82e1de6cd097e7b7
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/10.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/10.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1cb8b70cf1a402c1bd11249f16482b376620f3bb
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/10.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1e73326c3d1f5925d907d7cc6880495ebc7b252410dc12ae006fc45f3a81531
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/10.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/10.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..324fec0e41c9989630a01df4d9c5858b683dbb28
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/10.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6231275e14bc5fdd831c83c8b9fd1c6f216fa4ba78396f6ff53bd2192103ad45
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/10.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/10.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..7e0a047e877c8aec0471984c717e429728b86b93
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/10.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b19234fa3e65d2b78b1e4905704e6fc9e3834e64763bacd23edb2c23618a821
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/11.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/11.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f3e449c4f77780d14c9fd6f0869f73fee323da55
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/11.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94cfce45b4d6c18e212549f2cf3a5d7d69ab5200feb946cd0da80c64ff837628
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/11.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/11.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ae111f04399f7d4820832f124474b30341c34962
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/11.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ce356b72a26d875f95733a9e95f567731287f4c974aa606536674283937e590
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/11.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/11.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f740c2da2fb51bf917d5bfe5855071855e1adc23
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/11.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7b4c330f75d75e0282c2ef8d9f9b37a6e57091e5d7dd058351f202ad7776f5e
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/11.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/11.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d9f3fb3b50ab25255aaed69934ebf57a8cbb26d2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/11.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e319dd00d4f1cdc987baa3edb4bce4cfe122278d0a7a97f750d335fef728c02a
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/11.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/11.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0076c4e3d4949c366d8acd86ff74cc5bcc1856a3
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/11.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3302d7a0da7a4ea0bcadb5375dedc7285e541e27c6334e5ccc5aeb1d398a5f18
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/11.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/11.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2f5934fe7a045baf03d94a539f79ea3dcaa51ef9
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/11.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f22bda2f2e16212e1161267079f1987aa3e499999094eb6d083a7e56a2fd5e5
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/11.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/11.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f7b84fa39b647d0131fec5393ee1250a58f82143
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/11.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18c7545f271a687793aac92df6790991495473f93b2731f1e54533282a82317f
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/12.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/12.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..13508d8c88d42225945fd48641723d4c56c1fc42
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/12.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6d8a3c2e362bb38edc6baebc7e44d22c7aca9224546ab92df8133b8b8cec6bd
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/12.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/12.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..17510f6c2a6e62713f43b32d2afe8c0f073473dd
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/12.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e3d15479d7e7c45407213f95ab40d4fbab7bf40b8a4dd48d00a34d8d663cf75
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/12.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/12.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0b253a40be13cd37a1507c38592a1c47d27dea50
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/12.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:399ef6f9ae555aa704a00018ce97990de2dd833958bc55b71c0959e56ff9ec6d
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/12.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/12.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2cef5bfda9336780577d536083ec5460cc71e5e5
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/12.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc70b7e95c58399135a10c10259daec27c6faefa3826fe3d0937c7c527c30aa7
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/12.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/12.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c6faec9402a0b133fb0bef0a64bd96ace07c6904
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/12.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb271b8a2c6f5089457b31b4be117ea6c150d5737a3e811206a45373822461e7
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/12.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/12.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d5496302e9aa44132f877b69c7d730b709a051b5
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/12.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c714e93c6d0d326431178c4f6608c1dbf2625ce43ec45303882aa6b72676473
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/12.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/12.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..9495b4198cf4fae8af4333d2c32e39bb901b2483
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/12.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1946f9138dacbab3f0e6de98bc54a01ee640220f910ee1f708d47c00636975b
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/13.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/13.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d64086202f7186ee097104ae77bf7e193e95e060
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/13.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e0af089765b85fb9c0a49caa679f06dcb396beccf0becd288dcddb80bceae3b
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/13.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/13.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e2209bc1d33fd6190b1b98c945fd35dd0590d704
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/13.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e61f247e6db3c5f6ddafe8af02b9a6b6d890cc2c9dec6ef488890c4c827e2fe1
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/13.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/13.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..50948916a8efbdd81452d08bf2a94e440e606326
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/13.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ca6a4290495679666ebd7a94360038c8c717b30dd8e8e6be60d84fa4779e1b3
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/13.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/13.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..66f9a6859e96c44c874b32f6d6d270d520d4bb9e
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/13.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:287866aff43321ffcaeb27c6227d01f00e52a42f84b7b01d72e938ce0cd752d7
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/13.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/13.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c9d4cb4bd568166a2f283e33abc0663eaae022a1
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/13.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85c8a98368b7a2e4e617aabdb97ecca2e1f3273df2a85ed17adeb48ba3b6626a
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/13.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/13.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..84e335eae63403ab0185d5669fd149639d4cc3e5
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/13.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:657bcb9fb2ab168053c59834930eaf15d07857ca59268dea2802972af309aa35
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/13.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/13.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3d09cd55eb755b1c80098ab634f18e3bfd7360f9
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/13.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b67c587466c783f95ed5177273f3d1ee51a7c891ca5346925a151b4a15c2725
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/14.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/14.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..30a367b0ce73d33cdf182513dca16f33ea28bbd9
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/14.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d07ddd3c8c07b6252513158ac274fb470b5fab6f35dc7f39e44678423b3fadb
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/14.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/14.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1e71014184193f49b0ec901bff3b25eb366e5d6e
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/14.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cf08d0267c2395318ea1e38d2e92d8b2c4534477b872895cb0cd84404ed4039
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/14.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/14.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..722170262281232b32190b2f058187055298208c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/14.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec538d2e50e1a46a17c2d8e0bbbf765428c8f7e4e671547c7ddcf0249d9821ca
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/14.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/14.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0529633755abe47a437915261d95ea4dfd4b0671
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/14.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05ed8907312ac3778617058e912405aabc27faed854105a649bb968a3016e16b
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/14.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/14.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b67d0c8b79b37ed77d05ccff9c27740851fbfcdd
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/14.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d07aabf937a6c140f0dec03b8db120cfb41fe72d70a594cf0ae304d85fadc725
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/14.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/14.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..70889f968bcb730408ebd7956a17bb1ddc1e11f1
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/14.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e69676eee8492a7e4e49ce5b17c53536373c5e6882bf35f5134d4bd734730c0
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/14.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/14.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1f521a10743c75146a30cb7f9a23e24d60379ec8
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/14.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:257cf56cb192b77394606019abff2aaa75cc6c83566d22b0725e616187394f1d
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/15.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/15.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..aed548bb98830061aba518d20f51a2a6c40f1a91
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/15.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:932fb542b0fa5dbe333ae536bba85eeaed8a23e12d8df6452ab342051c31cc25
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/15.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/15.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..bc38e8d259eda170582b0fb1a50aaa7b89016c8a
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/15.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c295efdff49a36237f9364116a84862043c62b5a9f2578817a15c0de5f1eae0
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/15.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/15.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..df8457315ae4ae3b8ebc7858670d41a213ff5162
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/15.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cac6e9e812e5e37d7b4794df49eca2bb9d950009d4fc834739b502f4f85b8ff2
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/15.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/15.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c9a77d65ff3d7e412a57402e74490e6df3605dbb
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/15.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:caf4c8ee61558b0905a1fe96629daa67c29789d12db6237dec3754d39bb9a5cb
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/15.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/15.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ad09108430c23c13aebd335900dc7b311bd38528
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/15.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ac583e06bdfd2eea1cc3c22a9c42608a8cdf539fd1172774e2d1841f61dbad4
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/15.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/15.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a6d7cac6c183cef580d6225d30820a3654436de1
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/15.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cc7056d90c2528bcb04ad25d31cc8d63c03fd63e023ff10ee51fc01151155da
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/15.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/15.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..30ad0a12f7b6522fdf3badb7cffdb94719b5146f
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/15.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b957f2290519e265243bc07fa38ca0673417c926641df81e4863c50c0245432
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/16.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/16.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1bf74d9e51b08ce788708a0e22a5dbb3cf19e1ee
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/16.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c13f9805ccdf3d7ecd456b2222ed5d5342943e17cbd52578ffb681f6de2b56a
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/16.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/16.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0932b9aebf3e5cc23cb794004392277719910f43
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/16.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4663ff686671881626d7d43bfe52d4fb5d33fa898d68c1af52bb1dc5459eaf4e
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/16.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/16.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e1de4298c8ab98676ab8e8f289d040ad4159934c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/16.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02a02a315a85a93d071811587c372e4978aab2038c960a552ab3f58338cf092e
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/16.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/16.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a8c7aed466a5294987d99b42e7edd283e8e7cd45
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/16.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b57de3f081d069803a7a87695f55067f62f3b0ce6138e541b7e97785092591a
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/16.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/16.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..69f0276ab51b15a2016b94e95190cd06190163fe
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/16.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab18577f7ace48c3ac557ee261e15f4e3cc2f27310a602a61cb28970acbbdb6d
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/16.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/16.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..efe8ccf014457c5fbf6ac24126230c28c9558127
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/16.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1dc391654bcd6b83d8fc3adfee78a6edffa55fddd922b34016085f4ad319cb3
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/16.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/16.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..477110f0e4857b69ef1f02dd8e25957b6816e99d
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/16.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afd740935134b2b4a71ca96704bb7368bf794becc6bdd3ac8f69f8abba783059
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/17.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/17.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..36c72aa2a272eb9f956974dce2b0d559e4193a29
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/17.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:feda936734831c11f63312aa4bde16c71efec9b50c1470fe5df17b38defe5714
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/17.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/17.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c910347eeffc555bd0f4134e7b0cca7837d32ed2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/17.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6cd9cf8f15e0bf29689d9329c6062e7362e24be9d993ca03d2bcbad201a08f1
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/17.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/17.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..713d3913cff2ca851cd672d297185d18854c7a40
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/17.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9aff9e1bfe0eedcbf0c939eb922986fb72c8ab25a658204efba299e56bb1393e
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/17.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/17.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f97547201e5bfbb5416cf24ba37bf5fb630fa0dd
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/17.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bfbe6e32970ec2aa004698be17fc423867b62b675cb5942d38ad943817f7162
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/17.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/17.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d1e0eb16236c92b63e64a43d750d6b8f0690a4ee
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/17.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fae0d9e04cb1ef9662c63383fd3ebe5ba86e41d2fe7379a641e0dae445f2ab5f
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/17.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/17.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..402625d6a8809194233a1f0652a61777b26632ea
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/17.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3c96dc5ebc55d0f6889dddb4d42e3b456bb910f9f16e02e67933ecd1f08ce98
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/17.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/17.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..512fa49bcc8418fac7f4358444981420dfe31657
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/17.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd7663c0204ba9b7b0b66d795151791bb482830ccc48e449a108fb2dca53e25d
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/18.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/18.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..69dc0712417d7909802d76eaa6337130066d9e14
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/18.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bbda30ec00c93316f984a482191a8f6b9a507c2a5fd8c12f84db88227ed92d0
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/18.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/18.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a959678c5871b5cab5047e6b3ff0041e213b01bb
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/18.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96183ece8b5dc900d3d0cb0381e7d2912a1c48534a1889c48e074dcc93fa9aa4
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/18.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/18.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..022a45577bbd51dffcaeff98252dacccdbad5a0c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/18.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d9154669b020463042a91994501c10ea79342d0b507322c5d3206b5252d87df
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/18.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/18.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8f20b74c355fccfa44f987ac832b11d02b4e3cd6
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/18.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11cf3724bc00cf9ad386c3318e7e3e875d20c3bd541594839ecb151d4c385812
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/18.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/18.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e8502668c56362eef5e00f45c3eeaba5a960b96b
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/18.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65b64e3ff56d68f03a3d67e4101e414f7b7cedfb1eac4bef3fa3e1722ef500cf
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/18.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/18.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ebdeefe0db901369411c83eebb4e64350adaf686
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/18.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6658d91bf02e17882c32dbbe65dfecaeab09ad11baa6f7422a3ab5f5d1128c93
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/18.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/18.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..eb2ea3764188f6a7c26ec1f18241c5652d356152
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/18.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f25bd218f660b1eacabb02521319a308fac2dce05dad3c75fed6f27f4f7160e
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/19.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/19.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1ddd1daa6f4d9e9520e823529b24bdac7123bdfc
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/19.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9054b99916387c94f77fd045f94a540cf04f96426d44591c6d7c674b46d0594
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/19.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/19.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c32c54b91acdf02168c7210f5998d92c9253246f
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/19.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:baa03a99f38a7ea50c5670ccd8838b762872e5819dab34a13cd30804da4e04a1
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/19.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/19.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c4102d2292be3ee7e39ff23614f4211f48b17db8
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/19.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94c2447b7092f2be4bb3ebd8cd906bebbf8e5f390afea2a68589d6f1352b56e3
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/19.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/19.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4a3df0648e4aa3e5dd2a6dd75ad3ee386ef751df
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/19.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48d5f5acc658e0ce61e51f73a99d8146004e07f85b9b68a09eccd4b5f32ffb9a
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/19.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/19.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..60790d7b1e05972544985ee26207cffd170089c8
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/19.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d2e4edc9c1e74da523d84f7136c6760918ecfdb0074a7d1831ecbc799c91241
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/19.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/19.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4235d67fa006a215f2d1dcc469269fa0728a4a74
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/19.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0deae2bf75507ba809703eb829876bd913ce1240509974622a30f5eed0b8d1c7
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/19.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/19.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..104ffdb80516d2de1f830bb9bc84b10ea419415f
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/19.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:531e1c2ce7a18e7cc919ec10124da490135cd547ba0fe1b0dea99ddb5bc8e427
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/2.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/2.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b1935d1ebe708eb212c6c7136fd9c442e32dd51a
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/2.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa1f126f9fa21d29cd2637163d72099f6d9827c84217eb753f20fbafbb0fa6e9
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/2.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/2.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..add9cd51553d4a804e937781ae1c2b63be8046f7
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/2.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e036a19e51b9c93de0bd43402b1bdbe7d328a2936391a3b1120f48caae43362
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/2.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/2.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8d6f7f24e899d41ec3bd8f39650e9b181eab862b
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/2.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7fa1665de547dcbcf34e579b7017fbeb49b27b807abca6d10b6ff0703865fcc
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/2.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/2.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f03037d9a11c0702fb469890b78bef2e80864f20
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/2.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef12bc0189355e156b7c3ac67e3234e6c93cd8b73bcb079ba151dea577a00ee4
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/2.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/2.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..71df9147f8bfa9cd87d292e70cd77ac990df11e2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/2.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:014b773aa0773dc3d1b002b5b92c7373e57ee159ce6723f2e55230146fea263b
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/2.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/2.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..cd2675aa1c964bb4a64b2bce27cdf2bebb8de7e2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/2.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6c9dfeb0809c1f25acd6614e895232e11b2ef9ebeca3957cf2ca4c8584777db
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/2.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/2.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..50a55cec4bc0af8568d56003ab924151cc151359
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/2.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc6fee5fe8ecb1fd4093f989101783e3e7ceb5c6454ac14522f42def03811548
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/20.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/20.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1cc2a41ca42cabb5229d40a88e038a6f9756b0f2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/20.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74842c592f187c802aadb174d5608b3aaa8058aea12b648ab7e752f79f090379
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/20.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/20.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f2b97269fd74b1b472d02d5f119089da167bb728
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/20.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01ab21b6089be83db262924eb7a88ece7284f27f9604946d1ee0f2ad6c087fe2
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/20.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/20.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f41f00d248fe79428ef8ffda9adce8e145f4dbfc
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/20.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a91445ee998d88189289415e7160f7f1a6929d1540c295a65902cdd02ef9710
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/20.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/20.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a5038dfece9cc87056381b65edc77e1b7fd8cddb
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/20.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acd5b79843be39a64b0062d94334062969fe30e5cc0cc71d9584fc24f1813c8e
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/20.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/20.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..50aeb81f19ea0edb525bd29823e57d32d135d0a2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/20.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f3b0b5e061379c887f8886edea11af90f4fbcd9b39bafa0b3af596dbe5a30d0
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/20.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/20.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d622345b533512de60da3e0e8c8530b65ceeba30
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/20.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d663d563c933b7b3183e27046134f8585f42f0ba8ad3ea7b2dd549ca78f4629
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/20.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/20.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..592d3a8f7ae62487e432f3c68f622981d4ba7772
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/20.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71fd884bb14ee4a16955f8b9a561ac355a5652a4c343f7098854d5f591f3f922
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/21.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/21.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c43116e263ee5b2250144166ec9d31ccf70878ff
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/21.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0478475700d320304226b21e23632b2ee098c48bec03f9a4525785c7bf2d2e2
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/21.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/21.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..56515e5dd4a358cbbabcf38a51bd5268f0b689bb
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/21.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76c66daaaff722ec6a951d07134b5c836f7ee4be1e6149defb90ec4f67112d0f
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/21.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/21.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a31c7c8464786043c580394d35df8a6c693e0ce0
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/21.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a9bd8f4fe3c04adf54845d8769088c3b7be54800dcaa726d30c7c6978209905
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/21.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/21.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..92a49a2b6107b9c6b95387e27c7e65d5ff128668
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/21.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b17241cd46b2487c5c8a8ab4a5814175753d816d06e3795471e96423813907a
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/21.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/21.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2207c6e96ce5f503b2399f79722b2932dff0061a
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/21.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1da4f417991b4de6d70bb874f5fad0e3fcf2ff3270778d060fad1a747433a1f
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/21.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/21.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8973852f4b3cacd73283fb65e2e09b4c9c2fb5e9
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/21.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b598f5f4eaa42721e047ae2c04423b2ddb7e027b0ad66f92c9a94cc2af1e6d7
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/21.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/21.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2c724ebc69be70d6b538ea013f4c92c07983f32d
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/21.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:792eb15fb49c187cf2fb452872140e02efd585b83fc13efbcb41d0f3a239d2a8
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/22.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/22.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a6bc68ba41a7c11a081a117d50ddde1f90e51a08
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/22.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fe4596ec55fb54d967e282f95e1e8a33c27665e4e7829f981cd8d01837b1357
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/22.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/22.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d17a2cab44a460b9d06c6f570a2bd36bceb1d735
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/22.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db3dd4e3de1670bc8f34318b0f1b47c6aa0ff632eac0d3652a0a4691f0fc484c
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/22.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/22.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..cb27e84e76d67a19e39049c9093c329c531e6638
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/22.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a4912cbb864a602e6b27a662035c437b5e22d08056ea1f048766347019cca81
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/22.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/22.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8576879211f1de9bb0d149c2efcb6517e46cb016
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/22.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:706e3f6a9fc0f087bb386036c2e94d41d931247dc277b6ef059d6bcc41862d67
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/22.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/22.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..68b3a6fff4b60324426e3237265d5c9dbf3b792d
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/22.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aef69248fc1b4015877f5d52063ecbe1be79e27f9e01360f6791dc41bbe88c5f
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/22.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/22.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ccbeedacb13b1df0c2654a5ba083d35137c2f4da
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/22.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37805830a3390d280e7d85c11a387b3405e4ad071febcfb5e9c8457c0c7ee1f2
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/22.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/22.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..9862d0ef27044d5ce7e3478e2ff2771e879e9dd4
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/22.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9da1b90c228c03c998a5668024596e1766fe48751b4169bf24c5f9067f71fc17
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/23.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/23.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..34c72e66ed8c90054e0814e2fa9db235ea0cfbae
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/23.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f64ec9d84dda4638be1516fbf6e93c7d60c8819fe0cfc15a304612eca58163a8
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/23.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/23.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..eefcd1f2d4c155044964300e12752b2e75b2699c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/23.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7411a16a69eaa61b5c53817a8f8aff4db14b903bf6c2180aff14eb583ab7b75f
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/23.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/23.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..7de17c427e27b396cf81db542c3d47b6850504d3
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/23.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:519c2547c52781e6619b2c28b096e660b0e37c8186d544f444d4cffbd18662c6
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/23.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/23.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..048258cf4f4f4ef2a3e27943809be18cf13d036a
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/23.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a2147e56199fe5a931165a2c773c1c8aeac46065a1cb4c1e5ed90bdb88ffdda
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/23.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/23.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4949c4936fd4886c02bba3746810483fe8683504
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/23.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50361f68f637aa523b57e6c0ee7a18a8a9253d882e0d45cb444cd3761da7d24a
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/23.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/23.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..811d615386eab3eaa6391a93a8d2d91f82d214c8
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/23.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf71545383616be8069d62b7173f14d749126737be90dc8e935847a61a743c92
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/23.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/23.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0db5b7d00f417174a51669ffa60dc6a224a81dda
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/23.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae5cf861e5a1be534e8d528ec4ee84e83682246eadafa0774903d9bc438ebc3e
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/24.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/24.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d4b37bf9ae81fadffbd5ab5a26c230d2eb6f0458
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/24.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2325d57ccf766fdd094db25f7e2f8dec9ac29ba215761dd1c5dcdd449f72eed9
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/24.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/24.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8988b313627b83e49261062e3bcdc330e9b49beb
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/24.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3ac455e943820afb8a630ac0cec5d2b0780377be9fc27b889b10586f0fb4a2e
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/24.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/24.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b27b2736fa2818e516a74f915d4f8e50b66693b2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/24.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a005220f1b8367b926851f079b9e0f5eb842c122fafe3e57649d0dc04186a8cd
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/24.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/24.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ab12f36fabd6b22b4c0f6b7002fa6437ad83c386
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/24.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:883402a7e87af4bc4183265229f8c6f15d95662cbf6561751b810ad8b2886042
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/24.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/24.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..13e102292772e932263955a093e9b8bbf3ce0eea
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/24.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb9d877d97e2ac756483d3d71d4ec050238618f2ff129978b496466273273832
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/24.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/24.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a1e337614308e8a2e54e6928ef68852223d1b596
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/24.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f72c338225c081498d4a2819cfbeefda04138d1b98647fc2dc713626c4de834
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/24.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/24.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f6f9f443c8da08ed3fa64de0d33abace3cf96b6d
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/24.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bf2fa75fe610e1e5048ff83ac4de4f3f6f24f7c49e93bf942371aadace8b631
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/25.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/25.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..7f0ea5e2889a0e86a469c7c93c53d65719a63889
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/25.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f24e0454b054a3c86788cd8fcfafa1b3d9422207872221d2b52d66ff615ece2
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/25.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/25.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..c775722a1254acb83ee350590d007dc3aaac55e2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/25.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4066b39c93ab96ca02f6ddc228251ea2f2f9535194e1e83351a3abee77a30c5
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/25.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/25.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..65ed71d2f8fda763b1bccb0cc2e023bae0f120fd
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/25.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39bfbc7bf0e1cd6fb9316bfaca35756b44a000df27163852dd26b99b6ebb2164
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/25.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/25.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4d8a7643f5be6d27d54739a32e18a686586eb207
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/25.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46fa19d2c7a9cd28d0a14732bfdcde944d0ec645a82762dbd13283c2d04967b8
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/25.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/25.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..edd93d8b88b5d465405c0b4d9e11b82526bbd7c8
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/25.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:506c2f096ff579e226e4bbd790abfc0fca73fd752e3b02c202807cb6920904cb
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/25.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/25.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e9d310f335b9e4f0b223a6fb3a243c3552bdec36
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/25.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75ff97ac74e6f008c5c518800355d493123babd4addb9722387b80185aa5670e
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/25.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/25.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..42a8008e119962a87242b98eaf51f822051b70dc
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/25.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e71ee03daed612af633cba9324f4237c1bda69111f5c03792f69dfeda595ca23
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/26.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/26.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a50a199b77115fe27cafb3dc006aeaff594b52d1
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/26.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd0ae913bd2b016fc1b4f577a77dbcf5d44acc3aae1106017fec8c4162525761
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/26.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/26.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..99f830ac34d0a1260a92cf219715fee87b859fc4
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/26.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39da3314481caa81eba64a65e6c01375339128b2160760e7b99ca2832042c96f
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/26.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/26.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3d627584070d9b3297352cb9ed69ae4818ab10f9
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/26.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2008487a55f07706833588cf8d61f6c2bdf3dedbeec7f26e84341c3714bbb3d
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/26.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/26.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..698b7219b941811871459eab951bfddf849a3273
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/26.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9afad83d20550784b9d1ed66316b93c4d97d54961c49ad0aaf424960034f7b8b
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/26.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/26.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..12074e15ba45fbfd31a6a0ac936c5390e90425bf
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/26.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5755b91115aa6ff081d08e4bcc1d897de5108580414d101d2609076172ec5b8c
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/26.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/26.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..bf2e4fabb8800159dd77580e22d55bc5150f5b0e
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/26.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40f433c92b3a9ae0ede3ec8465a00402bd700c453c472d92fdf4cc05aa1f083c
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/26.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/26.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..bb3f18896c5d7dc98465f3551e852694f02ba619
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/26.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7a3c4c396ad99be9c04554fb4cb9c1901a155325e34bc37409b3b43ee3860b8
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/27.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/27.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..52742456f4f033651dd01783f54c04b904379351
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/27.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b53cda162d90d1a45db1d45e217eea8cbe2b545a34dc9045e8af62544b3b791d
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/27.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/27.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a8a9dcc9f18e91e23ac0085eef17fb00aeb4b25f
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/27.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:649683e37f7c81afb288f933bbf082ba0ab323ed3af5e4aa102a5d698458ef6d
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/27.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/27.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..5026724e1e7f3eaaed607308f040e906f59ebf86
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/27.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ebb91a2204ffd2af24532ef1da89416777b586481f4fce00d9506c7facc17d9
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/27.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/27.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a9482fdd41843daeb7f8252969399d8e8cc44b80
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/27.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37144cc29af1ebfceed6b45310ae72960f8bfe4a39ea3cbc596e4e300be318f2
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/27.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/27.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d009ef011b1da0f3b7d5e4dc9197298fc3a0337d
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/27.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:636216a7cbfd814b6d386e85846b3489b07cb28f28a55228c2121ae6a30c95fc
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/27.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/27.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ff915977803d05b9255333e6feb1a0ea1a982bc8
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/27.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dd5df5f13901299dde0fb076c00d24b0eb9398b7e0157e1f04c98131433d8b7
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/27.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/27.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1a91e81412a9938414ecab3f570e98372684b8c9
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/27.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35a1558c1859806e64630f52eaa2dd98542d0f7f8bc4a0c482ba06999c031fea
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/28.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/28.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..94789d24b85addbd61656db6e6c89145d9b478af
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/28.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:746618512b034b7718a4bdc0d7090e34df5f92baaf72331cdbddd6df980d797c
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/28.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/28.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0b5449a7c3c3fe22a7ba9c8cb2949eb6c161d7c6
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/28.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15fc9ca7fcf57cc1ed35898eafa105cd91d45bb899649a2b2b7b666b9e487bf5
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/28.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/28.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3011c5ac752adb0fb61d78293ed588caf6bbc950
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/28.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63ad8ec20cc4c6cf5ac8ca1f7fac65a97cef115e34124a7cdd0df5e5a1ad0654
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/28.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/28.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6dd55226cfe45c709db3ad5427b1adacaa9a294a
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/28.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5923fa1f840858deb776eb98280dfd5c9a43d37bcb2e2a8ee75b043821181f88
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/28.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/28.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..be2c2d121851e9f32630661863e31fd9df441dd2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/28.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f9085d0c5de10533bb4d543f8ad733d9517bee825ab2cfd01db66f387990f0f
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/28.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/28.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6e2c38a6dc6972647944f5fe61b221b484bd5c06
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/28.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c957db3ff7c7281b83d9ac0d4af229d5fa25d7a4f4a2cf92f213ff02c5c8097
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/28.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/28.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..64f8dea21a1dd3f0e257ada6e84a5de441b19828
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/28.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54b8b2c2bd1b08a0ba1ffdfeb2bb766e1afbf352a468de8a56df26b83d1f2ea3
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/29.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/29.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..64cccc659d95e45560879c8ab671c04c9ec4a31c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/29.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:109410df73f9dbc233815cd2d4425cdf0543ef075a4065811fa3f9390b3c5488
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/29.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/29.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6aa7b7e2d49cfa30df153fef976b0e55cedc75cb
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/29.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea986a41cd9ab804a86cef1ed8187c85ffacebd2ccf2076522c4a66cc2eed79a
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/29.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/29.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ef282b5254649f0fd8bcf9299408543c350fe1a0
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/29.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f5b6c603336b5879c9585698906bcef3cdbd2d6df69d239d5d9f8010ecf20c9
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/29.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/29.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..eeb191ca23937bb257735a6caf636212742c7d7c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/29.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08a80e156e3d1cf2b95dc470251d670b9e51c49f2d97080da54a86945fdcad26
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/29.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/29.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6be288c245f7a67d322886c509cf6aa2aa3158fb
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/29.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fdc00e3eae5b2d09de3ff2b771a67d13d3f72f1ec781c169d48756a9fdf5a05
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/29.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/29.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..9d141c3690b41aaee116a313908a24259b88ec2f
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/29.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70ce87064601aea4c8697a52be2beef78de5fcbea0b2c1dfc356cbb52f4ae615
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/29.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/29.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6a9b40d9f433163f834e266ca5e60608013d77b6
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/29.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85a37b492615367b36d587fef25d7f52cc6c464ebcd442a7e763cb71aa8c8f10
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/3.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/3.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a4d7c1ad283a68b257233b65da97a486f970a91e
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/3.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:249ca96ce8e75e74d92b11f12d35aa9c32a92acc4465d1c046ddd2fe9e0b7882
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/3.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/3.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..46a4abfb1867646ba365c2066834c9225a8ac163
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/3.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c12e2141ad174e3e21acbcfb6eec0603acbf11510ae13765ac2b409d0eb1e0bd
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/3.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/3.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..99ce1d2cafdb1b9b0f76a5057f7b6e72f07c9d32
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/3.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a1a4de51476370203765609b04d5db47a83ac9836afdfa93f0357dfe3597997
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/3.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/3.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b0abea404cb603740dd4e2ee610ed71beb4b5428
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/3.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfb6cafb6a3f967e80d1d583bbfd7647ddf5760ac3f5a2447a5345f91a7c74be
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/3.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/3.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..cfbc0d74789a55859793dc0d8596c566557e9f37
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/3.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:329766979ff2670eb203d9f0d8a95c7d886d13062957b28e81073ea40b960a5f
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/3.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/3.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4730d28012799fa4c8f16ec377360fc144605ac8
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/3.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6de13aa4d74b356b6e052d5becde166104fca0895e55d31cb5c6d60731a3563f
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/3.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/3.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2a0be3c64f85701aa801ed24d35a80e772de94ac
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/3.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ac3ac7db00b0a59c1f251aa04de28d1f19d099ddd68dd262067c8010e8f4570
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/30.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/30.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a39b0fa3cb80cfa701b3dcbb16213e11f3d63f4a
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/30.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06015ef7bdd7ede53d46cf6787fb738d6be4c3a53e3d5763537693d2100c18fc
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/30.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/30.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2db68c9c69f9794e3b735bad8fc71f571c6d6200
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/30.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbf42df4a558d0dab68c15c2fb977bd5afc6c9de24606414397cea2fe5068fce
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/30.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/30.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e415955d0ac40f02503073bbd4aca4cce5219426
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/30.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:909e0b8f45c28f33fa79362a1399f2ad11611e35f446b2976e6a658721726c37
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/30.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/30.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..de3c46c5e6344645e0541e98d18773d61227b3f6
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/30.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50cadcbb7a5a5e1a64c34ad944d85598ee13d4a21d43b62c1e0f2004accf6aa1
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/30.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/30.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..5e7406b3cfa9b6946bef75e4b2545fb928485b01
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/30.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a82725cb6a08f99c8933367ed7b10b0bcd523cbde71d697b16ff681064f1c3b5
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/30.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/30.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0f9e67a40cb56eabdb2e35f29f66aa67e133a167
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/30.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:461418d320cbf4a36446c5f22c08f5cfbf755b004ebb4d7a0bc4a51a2b6d4eb3
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/30.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/30.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..bd93b508348d0fbcb5f9150317f665440c9b0f9e
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/30.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:227d19ce8edf3de58fa98f9f5896137ac3838300c3e8cc1a345f52b943e68874
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/31.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/31.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..aec48921ab97041cab95da23a3e10200ae2e5250
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/31.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d07e0d8447e0c300c401cffb022f4e87dbfa5c0228e7d3cced0f69098a2eb9c3
+size 233493
diff --git a/nvfp/Meta-Llama-3-8B/31.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/31.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..cf77e08dfba24c2160e252531553dddd4e1d6a80
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/31.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d263ae8492e097f45944ce88eb0f8a9e188ff67ccf6ee5d34c673e48e2dda8ff
+size 69653
diff --git a/nvfp/Meta-Llama-3-8B/31.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/31.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..53053061eef026880552d3a71e144c6706ca2a8d
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/31.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be99f7305d23ed45289402547643970f8d9d3e0276867805aea97b888a1d6953
+size 69621
diff --git a/nvfp/Meta-Llama-3-8B/31.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/31.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1933e1e1ec15caf134691ed3b4930bb7f14b809c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/31.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3268c99f3022180654a6d21253557041a2020a52937463d11c256f0934da3fb2
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/31.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/31.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..80d974ac938e61712ae98125962fa955abc9fc2c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/31.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fff8c887ca916d49b180c1e8428401fc7768ddbd83a91bb5fd7540efadad3c1
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/31.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/31.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ad30aa167e29ad918efbf87a982bf936f87c13da
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/31.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5aba208761397a811892f87dfcd59b987f157cb28cbdc97202f4d677745869e6
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/31.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/31.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..d9f3bf5f4b8faebbbc70cac5e9a1bf6959edc061
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/31.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3fc1fe0913d697ab89bea8144c5545352ff6ddb9f81d9c95c30245a1d197c37
+size 69701
diff --git a/nvfp/Meta-Llama-3-8B/4.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/4.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..55c4f6b8e547758bb43f8dcac9c4ad3789bf8336
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/4.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a05475700ae680b8c9fc3c531062eb424d1d2733d9a9c12310398da13073b5a
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/4.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/4.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..72e61a364d9354ade84422f74ac2adbbd6b1e0c4
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/4.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8d71264666db1dcd787ea8ece52201d06574db4f5e9ae97cb956063e80854a0
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/4.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/4.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0a8dcc26815bcb881efe3dafbeb1adf2d6856251
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/4.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8df8958337943bcfc99d7863695f106f36c439de27436c522bd357671ce2245
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/4.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/4.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a5824b8f3954407cdafc042e6c6c5e76f8bae17e
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/4.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e112c9cc1ac9e48f7fa76e93c2f5fa733bdcffb86e4ef3e07ae5ef3866246904
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/4.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/4.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f0ad853cc371efe40daad82f5fd796d3d28d39da
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/4.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0af0b7465e82ea74681c28e9f0107cb80e2279ddb6ee75c2be51e2737136f62b
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/4.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/4.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ea086a5d840da0a8c958562fe3ae77b8b08e58e2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/4.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70ff5f6bdef4ae648b9c234bee4b051156f8c0ff6120d07d948cf3308129dd82
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/4.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/4.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f2a49b86e745f0135ae31757aa944588bca79302
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/4.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4fecd5d6f14106c6f4f73dd631ddd408126c2de2815fcd610e2fd597e163795
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/5.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/5.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..9c735dd4bb319a1a65157144ae0cb49ecb8ea832
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/5.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9a45de11ab7197b17e9700bc6279993b2445104355f38f061fb792fdaf83307
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/5.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/5.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1cd96ba755125cf00ab2243da7ad07cf37182c7d
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/5.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44aeebc6fdcaa46d9bba6e4f0bb838355acea22fbaba1f409f1faad17acc61eb
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/5.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/5.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e9b7d35f21a4750ec0950c1e4088b9b1bd412cb3
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/5.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3ca8933f718c94e7fd410d9b86fe852a5accafbe351b805fb26476a4647de02
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/5.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/5.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..30a21c56a1e334ef79f6e1b5b144dae552096e19
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/5.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95433e888dfbc9f2995eb83e0fc71d4ebe0ea66da0709b264f49757ddf3df3aa
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/5.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/5.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4b4db14333218da3f9c3f23569110298b806fb9e
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/5.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28f5ea97df436329e568f5603f4a661d0a76768a9ee36d17f822179da22cfea3
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/5.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/5.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..dcf2b58734b606990d44f129fc80344d67063945
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/5.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:550c1adeddf1b4cca3964fd05f84277532f02edaff5bf6227887d3cc374f086f
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/5.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/5.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..0f9aa53e5a969988c33cf5d60b7a7db092d6025c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/5.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e3ce440a3411499e8a4f029ce5343da99ce1858c5ecaff958d785bf2f8abef5
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/6.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/6.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4f676421d3dc29c33fd8048b8cb870ddea2eece1
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/6.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f45a101748b23ed395a07aa364137e19acb1f6d637db4818bb7cdc4e14aa6e12
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/6.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/6.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2bc1717b620bd3ef93fda8f032ec970748ff5439
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/6.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8227a36a58222e9a981154fe1410f58568cbd3e74fb96049e5f1f0c5fd404789
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/6.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/6.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b1094ba2770a2944c3119c8078cec36f3e68a960
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/6.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9201097ca9e95c2ce34d64ae0a060026e1ba5d85978983af0f1c75574246f8e
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/6.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/6.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6635955bf8a65db5015abbd63645b7a7bddb7ded
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/6.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5200b30d8a03684a28aee45d5482e2e3d6888446bd89b7b8e4e0720fa486a47a
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/6.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/6.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f90a3418caf6df51648f4f4f75d6407a50fd5795
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/6.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6c847fc744475fc4080f71ee4c9911c4e45c58f86137b99f9521f6b6b5c04dc
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/6.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/6.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6c1d6cc965c07277c8fabe0870a93c0b6516f59b
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/6.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82fb71e0523566eed191ece2877cc7d0a08f9dcea6714aa524740b5a7b272482
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/6.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/6.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3ae13ddd426ae9ae003322019c4d3288b94ea430
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/6.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:209db3fafa818ba60aa109b31502f9e41dfd5f3f00649fe3f8d936cbff34dca0
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/7.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/7.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6dd56c24b87d03a0e3e98ede02652ab958f811e0
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/7.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03a39c7619aba37ef78b82e7bf28269893ac425cef3e415ea8d1c483ec160010
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/7.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/7.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ae94ea61b186ef8ca4c104ff0638b5dc07892052
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/7.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c52afee6816fa0650f668f0969b1b30be2fb3b0666ce99032a01612b6505e886
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/7.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/7.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8b9354c04c0b6fdef0c78ea5117f416383ac3ef8
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/7.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2ac52ae86a294d7157ecb76d9a1a604f237f50cd084b4dbe6d011c69600c4f1
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/7.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/7.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..26f8ea4bc8d05995b154b124d81ad2700da6efc2
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/7.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65c0736e1c098397034032039fd405af23305791e735a42a7430af094d45d3ce
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/7.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/7.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..65addea8e911e2d887640a346f6601c4c4ed6e4a
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/7.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed312787b2856d3acadf2a83c3bd0032fa53784c0c0854e8569df4613ae25caf
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/7.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/7.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..65eae76de25562fed9b7455987d39f04adebb6e0
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/7.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d37c2c45fc705cb06a6ab49327aeaa2d968c4e27dfc302b62c94de0990da40b
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/7.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/7.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..5eb52047c5ef2c60b44d464f3af0ee5fda6269a7
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/7.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffc8235f29990fe29783caf1d69619f692ef723627a44c0763d19074a1b5f207
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/8.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/8.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..16734e14bf67cfa72fbf478cec18e1ab4fe34adf
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/8.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51ea304df8648aadcb0a178bb480260262d7656d084488c03fa240d31afd0189
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/8.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/8.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b5d6914ee676013aa60a102105ffa35c8e2455a3
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/8.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1a97ad2c61709b4c0350b908423471c786de7f985c27284c311024a795980a7
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/8.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/8.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8ff13a44af4d33b28df3af9f81125d9381a8fc1e
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/8.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b58b98015bca8316db4ce4c2dae2febcc6c09a0b07a90fce640a92a1693ce4ee
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/8.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/8.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..5d2fefc8af785dd08db366f0afaa104cc7e75df5
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/8.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2aa64306d3c313a939a14dd2f1786ad882cb890b5d4a9f3b5296bdea1ad612fc
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/8.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/8.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..1f3ca688af02be115454fdc54ea4cf6c758b3cfa
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/8.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc4c4715df3aad3271eb5e5dc932eaaf41d3b8c86206c175fcfc1a269f88bedd
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/8.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/8.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e9eabee5d74461c3f934b971efe24db7561487ef
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/8.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c7728f8926bb9cb5d9b2f7da369e931c4d406aa5574c875136c5d7311b4a910
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/8.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/8.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..116d117dc18e9989eff2b271a6591ce60dc7e76e
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/8.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cbeb71679dd2e5841c781109c3452bf7ea653d012269b905261f017c76e7cdd
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/9.mlp.down_proj.pt b/nvfp/Meta-Llama-3-8B/9.mlp.down_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2bcbfcae8a30304ae56a7bfc27c67b288b9999dc
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/9.mlp.down_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d28786512cf6473fdb0c9f84d08252744854877d2ba4a335e71bdf4979f629c5
+size 233477
diff --git a/nvfp/Meta-Llama-3-8B/9.mlp.gate_proj.pt b/nvfp/Meta-Llama-3-8B/9.mlp.gate_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..7c3abd737c76160ef8d871eef4586c94cda93afd
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/9.mlp.gate_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f35f616415c45aa7702be005e449654fb53b5fb63ddba0800ad530122b33c1b
+size 69637
diff --git a/nvfp/Meta-Llama-3-8B/9.mlp.up_proj.pt b/nvfp/Meta-Llama-3-8B/9.mlp.up_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..12c3b1834510548fda01fccf9bed3811ca25f7b7
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/9.mlp.up_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a56f3e7abb014ab94dd76b7c5f782c9995395001238398f547eb89639b67b0f0
+size 69605
diff --git a/nvfp/Meta-Llama-3-8B/9.self_attn.k_proj.pt b/nvfp/Meta-Llama-3-8B/9.self_attn.k_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..acee5089685259e422164ba2f09671308e2be1f5
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/9.self_attn.k_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:452b9a243ec3fa12ed9110ffc34ed57d00500be3b1f4fca5e9fe45dbda5b8ca6
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/9.self_attn.o_proj.pt b/nvfp/Meta-Llama-3-8B/9.self_attn.o_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..88129236e0f89395eafccc954968a79ebb52192b
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/9.self_attn.o_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bbdd6630f50216e340cc531e6be62802137b03c40c5f35cd5ae204489de78c3
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/9.self_attn.q_proj.pt b/nvfp/Meta-Llama-3-8B/9.self_attn.q_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ba06952f6678fbf6a091430b5d2c65f5437944de
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/9.self_attn.q_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:976f47b368f29d07cfdbaf036d7770abc3ddb86e707f86aacc0da4be364d5241
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/9.self_attn.v_proj.pt b/nvfp/Meta-Llama-3-8B/9.self_attn.v_proj.pt
new file mode 100644
index 0000000000000000000000000000000000000000..620befdd998b80ace1eea5a4e289f338f4f17149
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/9.self_attn.v_proj.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64f1af1ca1eeb76b7644e4cdd5a6e8ff2e805314e9502c99b904c27d3ce391f5
+size 69685
diff --git a/nvfp/Meta-Llama-3-8B/activation_mse_comparison.png b/nvfp/Meta-Llama-3-8B/activation_mse_comparison.png
new file mode 100644
index 0000000000000000000000000000000000000000..2d2ae0341e38aac18627ced805887502d85fce19
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/activation_mse_comparison.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7857ac70a6546678c0bc9de9bc96bb0d2a7b96c49ae94dc31f8e21765f8414e
+size 698244
diff --git a/nvfp/Meta-Llama-3-8B/activation_mse_comparison_rotation_only.png b/nvfp/Meta-Llama-3-8B/activation_mse_comparison_rotation_only.png
new file mode 100644
index 0000000000000000000000000000000000000000..3401febd32ceb00c78cac477af86ac708019d579
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/activation_mse_comparison_rotation_only.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:220051b68ab9e679b4129e9fd11e1f19a52fe80c091d6aacda0065bc61e04975
+size 719849
diff --git a/nvfp/Meta-Llama-3-8B/activation_mse_comparison_scale_only.png b/nvfp/Meta-Llama-3-8B/activation_mse_comparison_scale_only.png
new file mode 100644
index 0000000000000000000000000000000000000000..1c85d60b378e6bb17afd88c72092129dd9b83bed
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/activation_mse_comparison_scale_only.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70d2a2332cf4e06b21ddbad76967d4782a77c81048f39f8670e27f5fb7f89893
+size 750416
diff --git a/nvfp/Meta-Llama-3-8B/args.json b/nvfp/Meta-Llama-3-8B/args.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b44500a915478afbc7933ed9f7b152c8c16f166
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/args.json
@@ -0,0 +1,35 @@
+{
+  "model": "meta-llama/Meta-Llama-3-8B",
+  "params": [
+    "channel_scales:0.05,angles:0.04"
+  ],
+  "epochs": [
+    20
+  ],
+  "weight_decay": 0.01,
+  "betas": [
+    0.9,
+    0.95
+  ],
+  "eps": 1e-10,
+  "loss": "smooth_l1",
+  "group_size": 16,
+  "n_bit": 4,
+  "quant_mode": 3,
+  "num_rotations": 4,
+  "skipped_modules": [],
+  "datasets": [
+    "redpajama"
+  ],
+  "val_dataset": "pileval",
+  "train_size": 512,
+  "validation_size": 64,
+  "batch_size": 16,
+  "val_batch_size": null,
+  "seqlen": 2048,
+  "cache_shards": 1,
+  "output_dir": "./nvfp3/",
+  "resume": true,
+  "checkpointing": false,
+  "seed": 0
+}
\ No newline at end of file
diff --git a/nvfp/Meta-Llama-3-8B/run6.log b/nvfp/Meta-Llama-3-8B/run6.log
new file mode 100644
index 0000000000000000000000000000000000000000..53eb5ae855b93bd55dd7d0956376768e6e742460
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/run6.log
@@ -0,0 +1,9234 @@
+nohup: ignoring input
+INFO 12-22 14:29:48 [__init__.py:241] Automatically detected platform cuda.
+`torch_dtype` is deprecated! Use `dtype` instead!
+Config(model='meta-llama/Meta-Llama-3-8B', params=['channel_scales:0.05,angles:0.04'], epochs=[20], weight_decay=0.01, betas=(0.9, 0.95), eps=1e-10, loss='smooth_l1', group_size=16, n_bit=4, quant_mode=3, num_rotations=4, skipped_modules=[], datasets=['redpajama'], val_dataset='pileval', train_size=512, validation_size=64, batch_size=16, val_batch_size=None, seqlen=2048, cache_shards=1, output_dir='./nvfp3/', resume=True, checkpointing=False, seed=0)
+[INFO] Loading model...
+
Loading checkpoint shards:   0%|          | 0/4 [00:00<?, ?it/s]
Loading checkpoint shards:  25%|██▌       | 1/4 [00:01<00:04,  1.63s/it]
Loading checkpoint shards:  50%|█████     | 2/4 [00:03<00:03,  1.74s/it]
Loading checkpoint shards:  75%|███████▌  | 3/4 [00:10<00:04,  4.24s/it]
Loading checkpoint shards: 100%|██████████| 4/4 [00:11<00:00,  2.75s/it]
Loading checkpoint shards: 100%|██████████| 4/4 [00:11<00:00,  2.78s/it]
+Using the latest cached version of the dataset since SubSir/RedPajama-Data-1T-Sample-backup couldn't be found on the Hugging Face Hub
+WARNING:datasets.load:Using the latest cached version of the dataset since SubSir/RedPajama-Data-1T-Sample-backup couldn't be found on the Hugging Face Hub
+Found the latest cached dataset configuration 'default' at /raid/home/zhijian/.cache/huggingface/datasets/SubSir___red_pajama-data-1_t-sample-backup/default/0.0.0/bb07e0ea4d3f6760b67a2991247e2266706e3250 (last modified on Mon Nov 17 02:36:23 2025).
+WARNING:datasets.packaged_modules.cache.cache:Found the latest cached dataset configuration 'default' at /raid/home/zhijian/.cache/huggingface/datasets/SubSir___red_pajama-data-1_t-sample-backup/default/0.0.0/bb07e0ea4d3f6760b67a2991247e2266706e3250 (last modified on Mon Nov 17 02:36:23 2025).
+Repo card metadata block was not found. Setting CardData to empty.
+WARNING:huggingface_hub.repocard:Repo card metadata block was not found. Setting CardData to empty.
+[INFO] Preparing datasets...
+[INFO] Capturing first layer input...
+
Processing Blocks:   0%|          | 0/32 [00:00<?, ?it/s]
                                                         

Processing Blocks:   0%|          | 0/32 [00:06<?, ?it/s]
Processing Blocks:   3%|▎         | 1/32 [00:32<16:58, 32.86s/it]
                                                                 

Processing Blocks:   3%|▎         | 1/32 [00:33<16:58, 32.86s/it]
Processing Blocks:   6%|▋         | 2/32 [01:03<15:48, 31.61s/it]
                                                                 

Processing Blocks:   6%|▋         | 2/32 [01:04<15:48, 31.61s/it]
Processing Blocks:   9%|▉         | 3/32 [01:33<14:59, 31.00s/it]
                                                                 

Processing Blocks:   9%|▉         | 3/32 [01:35<14:59, 31.00s/it]
Processing Blocks:  12%|█▎        | 4/32 [02:03<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [02:04<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [02:22<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [02:22<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [04:44<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [04:44<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [05:48<14:15, 30.56s/it][INFO] Layer 0 already done. Loading results...
+[INFO] Layer 1 already done. Loading results...
+[INFO] Layer 2 already done. Loading results...
+[INFO] Layer 3 already done. Loading results...
+[INFO] Initializing quantization parameters for Layer 4...
+[INFO] Starting Sequential Training for Layer 4
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.00117, val_og_loss=0.00141]
+
  5%|▌         | 1/20 [00:16<05:13, 16.48s/iter, val_loss=0.00117, val_og_loss=0.00141]
+
  5%|▌         | 1/20 [00:18<05:13, 16.48s/iter, val_loss=0.00118, val_og_loss=0.00141]
+
 10%|█         | 2/20 [00:18<02:20,  7.78s/iter, val_loss=0.00118, val_og_loss=0.00141]
+
 10%|█         | 2/20 [00:19<02:20,  7.78s/iter, val_loss=0.0012, val_og_loss=0.00141] 
+
 15%|█▌        | 3/20 [00:19<01:25,  5.00s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 15%|█▌        | 3/20 [00:21<01:25,  5.00s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 20%|██        | 4/20 [00:21<00:59,  3.71s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 20%|██        | 4/20 [00:23<00:59,  3.71s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 25%|██▌       | 5/20 [00:23<00:44,  2.98s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 25%|██▌       | 5/20 [00:24<00:44,  2.98s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 30%|███       | 6/20 [00:24<00:35,  2.54s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 30%|███       | 6/20 [00:26<00:35,  2.54s/iter, val_loss=0.00121, val_og_loss=0.00141]
+
 35%|███▌      | 7/20 [00:26<00:29,  2.26s/iter, val_loss=0.00121, val_og_loss=0.00141]
+
 35%|███▌      | 7/20 [00:28<00:29,  2.26s/iter, val_loss=0.00122, val_og_loss=0.00141]
+
 40%|████      | 8/20 [00:28<00:24,  2.08s/iter, val_loss=0.00122, val_og_loss=0.00141]
+
 40%|████      | 8/20 [00:30<00:24,  2.08s/iter, val_loss=0.00121, val_og_loss=0.00141]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.96s/iter, val_loss=0.00121, val_og_loss=0.00141]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.96s/iter, val_loss=0.0012, val_og_loss=0.00141] 
+
 50%|█████     | 10/20 [00:31<00:18,  1.87s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 50%|█████     | 10/20 [00:33<00:18,  1.87s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.81s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.81s/iter, val_loss=0.00121, val_og_loss=0.00141]
+
 60%|██████    | 12/20 [00:35<00:14,  1.77s/iter, val_loss=0.00121, val_og_loss=0.00141]
+
 60%|██████    | 12/20 [00:36<00:14,  1.77s/iter, val_loss=0.0012, val_og_loss=0.00141] 
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.76s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.76s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 70%|███████   | 14/20 [00:38<00:10,  1.75s/iter, val_loss=0.0012, val_og_loss=0.00141]
+
 70%|███████   | 14/20 [00:40<00:10,  1.75s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.73s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.73s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
 80%|████████  | 16/20 [00:41<00:06,  1.73s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
 80%|████████  | 16/20 [00:43<00:06,  1.73s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.73s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.73s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.73s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.73s/iter, val_loss=0.00118, val_og_loss=0.00141]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.72s/iter, val_loss=0.00118, val_og_loss=0.00141]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.72s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
100%|██████████| 20/20 [00:48<00:00,  1.71s/iter, val_loss=0.00119, val_og_loss=0.00141]
100%|██████████| 20/20 [00:48<00:00,  2.44s/iter, val_loss=0.00119, val_og_loss=0.00141]
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [06:37<14:15, 30.56s/it]INFO:ParoQuant:Best val loss: 0.0011654181871563196, Original val loss: 0.0014149447670206428
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [06:41<14:15, 30.56s/it]2025-12-22 14:36:55,483 - INFO - Best val loss: 0.0011654181871563196, Original val loss: 0.0014149447670206428
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=0.00288, val_og_loss=0.00347]
+
  5%|▌         | 1/20 [00:13<04:25, 13.99s/iter, val_loss=0.00288, val_og_loss=0.00347]
+
  5%|▌         | 1/20 [00:15<04:25, 13.99s/iter, val_loss=0.00289, val_og_loss=0.00347]
+
 10%|█         | 2/20 [00:15<01:58,  6.60s/iter, val_loss=0.00289, val_og_loss=0.00347]
+
 10%|█         | 2/20 [00:16<01:58,  6.60s/iter, val_loss=0.00293, val_og_loss=0.00347]
+
 15%|█▌        | 3/20 [00:16<01:12,  4.25s/iter, val_loss=0.00293, val_og_loss=0.00347]
+
 15%|█▌        | 3/20 [00:18<01:12,  4.25s/iter, val_loss=0.00297, val_og_loss=0.00347]
+
 20%|██        | 4/20 [00:18<00:50,  3.13s/iter, val_loss=0.00297, val_og_loss=0.00347]
+
 20%|██        | 4/20 [00:19<00:50,  3.13s/iter, val_loss=0.00297, val_og_loss=0.00347]
+
 25%|██▌       | 5/20 [00:19<00:37,  2.51s/iter, val_loss=0.00297, val_og_loss=0.00347]
+
 25%|██▌       | 5/20 [00:21<00:37,  2.51s/iter, val_loss=0.00293, val_og_loss=0.00347]
+
 30%|███       | 6/20 [00:21<00:29,  2.13s/iter, val_loss=0.00293, val_og_loss=0.00347]
+
 30%|███       | 6/20 [00:22<00:29,  2.13s/iter, val_loss=0.00296, val_og_loss=0.00347]
+
 35%|███▌      | 7/20 [00:22<00:24,  1.89s/iter, val_loss=0.00296, val_og_loss=0.00347]
+
 35%|███▌      | 7/20 [00:23<00:24,  1.89s/iter, val_loss=0.00295, val_og_loss=0.00347]
+
 40%|████      | 8/20 [00:23<00:20,  1.73s/iter, val_loss=0.00295, val_og_loss=0.00347]
+
 40%|████      | 8/20 [00:25<00:20,  1.73s/iter, val_loss=0.00292, val_og_loss=0.00347]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.63s/iter, val_loss=0.00292, val_og_loss=0.00347]
+
 45%|████▌     | 9/20 [00:26<00:17,  1.63s/iter, val_loss=0.00292, val_og_loss=0.00347]
+
 50%|█████     | 10/20 [00:26<00:15,  1.56s/iter, val_loss=0.00292, val_og_loss=0.00347]
+
 50%|█████     | 10/20 [00:28<00:15,  1.56s/iter, val_loss=0.00292, val_og_loss=0.00347]
+
 55%|█████▌    | 11/20 [00:28<00:13,  1.51s/iter, val_loss=0.00292, val_og_loss=0.00347]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.51s/iter, val_loss=0.00293, val_og_loss=0.00347]
+
 60%|██████    | 12/20 [00:29<00:11,  1.48s/iter, val_loss=0.00293, val_og_loss=0.00347]
+
 60%|██████    | 12/20 [00:30<00:11,  1.48s/iter, val_loss=0.00293, val_og_loss=0.00347]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.45s/iter, val_loss=0.00293, val_og_loss=0.00347]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.45s/iter, val_loss=0.0029, val_og_loss=0.00347] 
+
 70%|███████   | 14/20 [00:32<00:08,  1.44s/iter, val_loss=0.0029, val_og_loss=0.00347]
+
 70%|███████   | 14/20 [00:33<00:08,  1.44s/iter, val_loss=0.00291, val_og_loss=0.00347]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.42s/iter, val_loss=0.00291, val_og_loss=0.00347]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.42s/iter, val_loss=0.0029, val_og_loss=0.00347] 
+
 80%|████████  | 16/20 [00:35<00:05,  1.42s/iter, val_loss=0.0029, val_og_loss=0.00347]
+
 80%|████████  | 16/20 [00:36<00:05,  1.42s/iter, val_loss=0.00291, val_og_loss=0.00347]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.41s/iter, val_loss=0.00291, val_og_loss=0.00347]
+
 85%|████████▌ | 17/20 [00:37<00:04,  1.41s/iter, val_loss=0.00291, val_og_loss=0.00347]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.41s/iter, val_loss=0.00291, val_og_loss=0.00347]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.41s/iter, val_loss=0.0029, val_og_loss=0.00347] 
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.40s/iter, val_loss=0.0029, val_og_loss=0.00347]
+
 95%|█████████▌| 19/20 [00:40<00:01,  1.40s/iter, val_loss=0.00289, val_og_loss=0.00347]
+
100%|██████████| 20/20 [00:40<00:00,  1.40s/iter, val_loss=0.00289, val_og_loss=0.00347]
100%|██████████| 20/20 [00:40<00:00,  2.03s/iter, val_loss=0.00289, val_og_loss=0.00347]
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [07:21<14:15, 30.56s/it]INFO:ParoQuant:Best val loss: 0.0028841234743595123, Original val loss: 0.0034707761369645596
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [07:23<14:15, 30.56s/it]2025-12-22 14:37:39,754 - INFO - Best val loss: 0.0028841234743595123, Original val loss: 0.0034707761369645596
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=0.000215, val_og_loss=0.000278]
+
  5%|▌         | 1/20 [00:13<04:11, 13.26s/iter, val_loss=0.000215, val_og_loss=0.000278]
+
  5%|▌         | 1/20 [00:14<04:11, 13.26s/iter, val_loss=0.000214, val_og_loss=0.000278]
+
 10%|█         | 2/20 [00:14<01:52,  6.28s/iter, val_loss=0.000214, val_og_loss=0.000278]
+
 10%|█         | 2/20 [00:16<01:52,  6.28s/iter, val_loss=0.000216, val_og_loss=0.000278]
+
 15%|█▌        | 3/20 [00:16<01:08,  4.05s/iter, val_loss=0.000216, val_og_loss=0.000278]
+
 15%|█▌        | 3/20 [00:17<01:08,  4.05s/iter, val_loss=0.000216, val_og_loss=0.000278]
+
 20%|██        | 4/20 [00:17<00:48,  3.00s/iter, val_loss=0.000216, val_og_loss=0.000278]
+
 20%|██        | 4/20 [00:18<00:48,  3.00s/iter, val_loss=0.000216, val_og_loss=0.000278]
+
 25%|██▌       | 5/20 [00:18<00:36,  2.43s/iter, val_loss=0.000216, val_og_loss=0.000278]
+
 25%|██▌       | 5/20 [00:20<00:36,  2.43s/iter, val_loss=0.000217, val_og_loss=0.000278]
+
 30%|███       | 6/20 [00:20<00:29,  2.08s/iter, val_loss=0.000217, val_og_loss=0.000278]
+
 30%|███       | 6/20 [00:21<00:29,  2.08s/iter, val_loss=0.000217, val_og_loss=0.000278]
+
 35%|███▌      | 7/20 [00:21<00:24,  1.86s/iter, val_loss=0.000217, val_og_loss=0.000278]
+
 35%|███▌      | 7/20 [00:23<00:24,  1.86s/iter, val_loss=0.000216, val_og_loss=0.000278]
+
 40%|████      | 8/20 [00:23<00:20,  1.71s/iter, val_loss=0.000216, val_og_loss=0.000278]
+
 40%|████      | 8/20 [00:24<00:20,  1.71s/iter, val_loss=0.000215, val_og_loss=0.000278]
+
 45%|████▌     | 9/20 [00:24<00:17,  1.61s/iter, val_loss=0.000215, val_og_loss=0.000278]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.61s/iter, val_loss=0.000215, val_og_loss=0.000278]
+
 50%|█████     | 10/20 [00:25<00:15,  1.55s/iter, val_loss=0.000215, val_og_loss=0.000278]
+
 50%|█████     | 10/20 [00:27<00:15,  1.55s/iter, val_loss=0.000215, val_og_loss=0.000278]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.50s/iter, val_loss=0.000215, val_og_loss=0.000278]
+
 55%|█████▌    | 11/20 [00:28<00:13,  1.50s/iter, val_loss=0.000215, val_og_loss=0.000278]
+
 60%|██████    | 12/20 [00:28<00:11,  1.47s/iter, val_loss=0.000215, val_og_loss=0.000278]
+
 60%|██████    | 12/20 [00:30<00:11,  1.47s/iter, val_loss=0.000214, val_og_loss=0.000278]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.45s/iter, val_loss=0.000214, val_og_loss=0.000278]
+
 65%|██████▌   | 13/20 [00:31<00:10,  1.45s/iter, val_loss=0.000214, val_og_loss=0.000278]
+
 70%|███████   | 14/20 [00:31<00:08,  1.43s/iter, val_loss=0.000214, val_og_loss=0.000278]
+
 70%|███████   | 14/20 [00:32<00:08,  1.43s/iter, val_loss=0.000213, val_og_loss=0.000278]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.43s/iter, val_loss=0.000213, val_og_loss=0.000278]
+
 75%|███████▌  | 15/20 [00:34<00:07,  1.43s/iter, val_loss=0.000213, val_og_loss=0.000278]
+
 80%|████████  | 16/20 [00:34<00:05,  1.41s/iter, val_loss=0.000213, val_og_loss=0.000278]
+
 80%|████████  | 16/20 [00:35<00:05,  1.41s/iter, val_loss=0.000213, val_og_loss=0.000278]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.40s/iter, val_loss=0.000213, val_og_loss=0.000278]
+
 85%|████████▌ | 17/20 [00:37<00:04,  1.40s/iter, val_loss=0.000212, val_og_loss=0.000278]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.40s/iter, val_loss=0.000212, val_og_loss=0.000278]
+
 90%|█████████ | 18/20 [00:38<00:02,  1.40s/iter, val_loss=0.000212, val_og_loss=0.000278]
+
 95%|█████████▌| 19/20 [00:38<00:01,  1.40s/iter, val_loss=0.000212, val_og_loss=0.000278]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.40s/iter, val_loss=0.000212, val_og_loss=0.000278]
+
100%|██████████| 20/20 [00:39<00:00,  1.40s/iter, val_loss=0.000212, val_og_loss=0.000278]
100%|██████████| 20/20 [00:39<00:00,  1.99s/iter, val_loss=0.000212, val_og_loss=0.000278]
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [08:03<14:15, 30.56s/it]INFO:ParoQuant:Best val loss: 0.00021175446454435587, Original val loss: 0.00027780188247561455
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [08:05<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [08:05<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [08:38<14:15, 30.56s/it]2025-12-22 14:38:21,268 - INFO - Best val loss: 0.00021175446454435587, Original val loss: 0.00027780188247561455
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=4.3e-6, val_og_loss=4.56e-6]
+
  5%|▌         | 1/20 [00:16<05:14, 16.54s/iter, val_loss=4.3e-6, val_og_loss=4.56e-6]
+
  5%|▌         | 1/20 [00:18<05:14, 16.54s/iter, val_loss=4.3e-6, val_og_loss=4.56e-6]
+
 10%|█         | 2/20 [00:18<02:20,  7.81s/iter, val_loss=4.3e-6, val_og_loss=4.56e-6]
+
 10%|█         | 2/20 [00:19<02:20,  7.81s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 15%|█▌        | 3/20 [00:19<01:25,  5.02s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 15%|█▌        | 3/20 [00:21<01:25,  5.02s/iter, val_loss=4.31e-6, val_og_loss=4.56e-6]
+
 20%|██        | 4/20 [00:21<00:59,  3.70s/iter, val_loss=4.31e-6, val_og_loss=4.56e-6]
+
 20%|██        | 4/20 [00:23<00:59,  3.70s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 25%|██▌       | 5/20 [00:23<00:44,  2.97s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 25%|██▌       | 5/20 [00:24<00:44,  2.97s/iter, val_loss=4.31e-6, val_og_loss=4.56e-6]
+
 30%|███       | 6/20 [00:25<00:35,  2.54s/iter, val_loss=4.31e-6, val_og_loss=4.56e-6]
+
 30%|███       | 6/20 [00:26<00:35,  2.54s/iter, val_loss=4.3e-6, val_og_loss=4.56e-6] 
+
 35%|███▌      | 7/20 [00:26<00:29,  2.26s/iter, val_loss=4.3e-6, val_og_loss=4.56e-6]
+
 35%|███▌      | 7/20 [00:28<00:29,  2.26s/iter, val_loss=4.31e-6, val_og_loss=4.56e-6]
+
 40%|████      | 8/20 [00:28<00:24,  2.08s/iter, val_loss=4.31e-6, val_og_loss=4.56e-6]
+
 40%|████      | 8/20 [00:30<00:24,  2.08s/iter, val_loss=4.34e-6, val_og_loss=4.56e-6]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.97s/iter, val_loss=4.34e-6, val_og_loss=4.56e-6]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.97s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 50%|█████     | 10/20 [00:31<00:18,  1.90s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 50%|█████     | 10/20 [00:33<00:18,  1.90s/iter, val_loss=4.29e-6, val_og_loss=4.56e-6]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.83s/iter, val_loss=4.29e-6, val_og_loss=4.56e-6]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.83s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 60%|██████    | 12/20 [00:35<00:14,  1.79s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 60%|██████    | 12/20 [00:36<00:14,  1.79s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.75s/iter, val_loss=4.32e-6, val_og_loss=4.56e-6]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.75s/iter, val_loss=4.31e-6, val_og_loss=4.56e-6]
+
 70%|███████   | 14/20 [00:38<00:10,  1.73s/iter, val_loss=4.31e-6, val_og_loss=4.56e-6]
+
 70%|███████   | 14/20 [00:40<00:10,  1.73s/iter, val_loss=4.28e-6, val_og_loss=4.56e-6]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.72s/iter, val_loss=4.28e-6, val_og_loss=4.56e-6]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.72s/iter, val_loss=4.27e-6, val_og_loss=4.56e-6]
+
 80%|████████  | 16/20 [00:41<00:06,  1.71s/iter, val_loss=4.27e-6, val_og_loss=4.56e-6]
+
 80%|████████  | 16/20 [00:43<00:06,  1.71s/iter, val_loss=4.27e-6, val_og_loss=4.56e-6]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.70s/iter, val_loss=4.27e-6, val_og_loss=4.56e-6]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.70s/iter, val_loss=4.27e-6, val_og_loss=4.56e-6]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.70s/iter, val_loss=4.27e-6, val_og_loss=4.56e-6]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.70s/iter, val_loss=4.25e-6, val_og_loss=4.56e-6]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.70s/iter, val_loss=4.25e-6, val_og_loss=4.56e-6]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.70s/iter, val_loss=4.25e-6, val_og_loss=4.56e-6]
+
100%|██████████| 20/20 [00:48<00:00,  1.71s/iter, val_loss=4.25e-6, val_og_loss=4.56e-6]
100%|██████████| 20/20 [00:48<00:00,  2.44s/iter, val_loss=4.25e-6, val_og_loss=4.56e-6]
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [09:27<14:15, 30.56s/it]INFO:ParoQuant:Best val loss: 4.248792720318306e-06, Original val loss: 4.564087248581927e-06
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [09:27<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [09:27<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [10:24<14:15, 30.56s/it]2025-12-22 14:39:45,237 - INFO - Best val loss: 4.248792720318306e-06, Original val loss: 4.564087248581927e-06
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:38<?, ?iter/s, val_loss=0.00052, val_og_loss=0.000525]
+
  5%|▌         | 1/20 [00:38<12:04, 38.16s/iter, val_loss=0.00052, val_og_loss=0.000525]
+
  5%|▌         | 1/20 [00:40<12:04, 38.16s/iter, val_loss=0.000523, val_og_loss=0.000525]
+
 10%|█         | 2/20 [00:40<05:11, 17.31s/iter, val_loss=0.000523, val_og_loss=0.000525]
+
 10%|█         | 2/20 [00:43<05:11, 17.31s/iter, val_loss=0.000524, val_og_loss=0.000525]
+
 15%|█▌        | 3/20 [00:43<03:01, 10.65s/iter, val_loss=0.000524, val_og_loss=0.000525]
+
 15%|█▌        | 3/20 [00:46<03:01, 10.65s/iter, val_loss=0.000525, val_og_loss=0.000525]
+
 20%|██        | 4/20 [00:46<02:00,  7.52s/iter, val_loss=0.000525, val_og_loss=0.000525]
+
 20%|██        | 4/20 [00:49<02:00,  7.52s/iter, val_loss=0.000526, val_og_loss=0.000525]
+
 25%|██▌       | 5/20 [00:49<01:26,  5.79s/iter, val_loss=0.000526, val_og_loss=0.000525]
+
 25%|██▌       | 5/20 [00:51<01:26,  5.79s/iter, val_loss=0.000526, val_og_loss=0.000525]
+
 30%|███       | 6/20 [00:51<01:06,  4.75s/iter, val_loss=0.000526, val_og_loss=0.000525]
+
 30%|███       | 6/20 [00:54<01:06,  4.75s/iter, val_loss=0.000526, val_og_loss=0.000525]
+
 35%|███▌      | 7/20 [00:54<00:53,  4.08s/iter, val_loss=0.000526, val_og_loss=0.000525]
+
 35%|███▌      | 7/20 [00:57<00:53,  4.08s/iter, val_loss=0.000526, val_og_loss=0.000525]
+
 40%|████      | 8/20 [00:57<00:43,  3.65s/iter, val_loss=0.000526, val_og_loss=0.000525]
+
 40%|████      | 8/20 [00:59<00:43,  3.65s/iter, val_loss=0.000525, val_og_loss=0.000525]
+
 45%|████▌     | 9/20 [00:59<00:36,  3.36s/iter, val_loss=0.000525, val_og_loss=0.000525]
+
 45%|████▌     | 9/20 [01:02<00:36,  3.36s/iter, val_loss=0.000524, val_og_loss=0.000525]
+
 50%|█████     | 10/20 [01:02<00:31,  3.16s/iter, val_loss=0.000524, val_og_loss=0.000525]
+
 50%|█████     | 10/20 [01:05<00:31,  3.16s/iter, val_loss=0.000524, val_og_loss=0.000525]
+
 55%|█████▌    | 11/20 [01:05<00:27,  3.03s/iter, val_loss=0.000524, val_og_loss=0.000525]
+
 55%|█████▌    | 11/20 [01:08<00:27,  3.03s/iter, val_loss=0.000524, val_og_loss=0.000525]
+
 60%|██████    | 12/20 [01:08<00:23,  2.93s/iter, val_loss=0.000524, val_og_loss=0.000525]
+
 60%|██████    | 12/20 [01:10<00:23,  2.93s/iter, val_loss=0.000523, val_og_loss=0.000525]
+
 65%|██████▌   | 13/20 [01:10<00:20,  2.87s/iter, val_loss=0.000523, val_og_loss=0.000525]
+
 65%|██████▌   | 13/20 [01:13<00:20,  2.87s/iter, val_loss=0.000523, val_og_loss=0.000525]
+
 70%|███████   | 14/20 [01:13<00:16,  2.82s/iter, val_loss=0.000523, val_og_loss=0.000525]
+
 70%|███████   | 14/20 [01:16<00:16,  2.82s/iter, val_loss=0.000522, val_og_loss=0.000525]
+
 75%|███████▌  | 15/20 [01:16<00:13,  2.79s/iter, val_loss=0.000522, val_og_loss=0.000525]
+
 75%|███████▌  | 15/20 [01:18<00:13,  2.79s/iter, val_loss=0.000522, val_og_loss=0.000525]
+
 80%|████████  | 16/20 [01:18<00:11,  2.77s/iter, val_loss=0.000522, val_og_loss=0.000525]
+
 80%|████████  | 16/20 [01:21<00:11,  2.77s/iter, val_loss=0.000521, val_og_loss=0.000525]
+
 85%|████████▌ | 17/20 [01:21<00:08,  2.76s/iter, val_loss=0.000521, val_og_loss=0.000525]
+
 85%|████████▌ | 17/20 [01:24<00:08,  2.76s/iter, val_loss=0.00052, val_og_loss=0.000525] 
+
 90%|█████████ | 18/20 [01:24<00:05,  2.75s/iter, val_loss=0.00052, val_og_loss=0.000525]
+
 90%|█████████ | 18/20 [01:27<00:05,  2.75s/iter, val_loss=0.00052, val_og_loss=0.000525]
+
 95%|█████████▌| 19/20 [01:27<00:02,  2.74s/iter, val_loss=0.00052, val_og_loss=0.000525]
+
 95%|█████████▌| 19/20 [01:29<00:02,  2.74s/iter, val_loss=0.00052, val_og_loss=0.000525]
+
100%|██████████| 20/20 [01:29<00:00,  2.74s/iter, val_loss=0.00052, val_og_loss=0.000525]
100%|██████████| 20/20 [01:29<00:00,  4.49s/iter, val_loss=0.00052, val_og_loss=0.000525]
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [11:54<14:15, 30.56s/it]INFO:ParoQuant:Best val loss: 0.0005201989552006125, Original val loss: 0.0005252878181636333
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [12:00<14:15, 30.56s/it]2025-12-22 14:42:12,307 - INFO - Best val loss: 0.0005201989552006125, Original val loss: 0.0005252878181636333
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:38<?, ?iter/s, val_loss=0.000269, val_og_loss=0.000272]
+
  5%|▌         | 1/20 [00:38<12:20, 38.97s/iter, val_loss=0.000269, val_og_loss=0.000272]
+
  5%|▌         | 1/20 [00:41<12:20, 38.97s/iter, val_loss=0.00027, val_og_loss=0.000272] 
+
 10%|█         | 2/20 [00:41<05:17, 17.65s/iter, val_loss=0.00027, val_og_loss=0.000272]
+
 10%|█         | 2/20 [00:44<05:17, 17.65s/iter, val_loss=0.000272, val_og_loss=0.000272]
+
 15%|█▌        | 3/20 [00:44<03:04, 10.83s/iter, val_loss=0.000272, val_og_loss=0.000272]
+
 15%|█▌        | 3/20 [00:47<03:04, 10.83s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 20%|██        | 4/20 [00:47<02:02,  7.63s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 20%|██        | 4/20 [00:49<02:02,  7.63s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 25%|██▌       | 5/20 [00:49<01:27,  5.86s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 25%|██▌       | 5/20 [00:52<01:27,  5.86s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 30%|███       | 6/20 [00:52<01:07,  4.81s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 30%|███       | 6/20 [00:55<01:07,  4.81s/iter, val_loss=0.000274, val_og_loss=0.000272]
+
 35%|███▌      | 7/20 [00:55<00:53,  4.13s/iter, val_loss=0.000274, val_og_loss=0.000272]
+
 35%|███▌      | 7/20 [00:58<00:53,  4.13s/iter, val_loss=0.000274, val_og_loss=0.000272]
+
 40%|████      | 8/20 [00:58<00:44,  3.69s/iter, val_loss=0.000274, val_og_loss=0.000272]
+
 40%|████      | 8/20 [01:00<00:44,  3.69s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 45%|████▌     | 9/20 [01:00<00:37,  3.40s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 45%|████▌     | 9/20 [01:03<00:37,  3.40s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 50%|█████     | 10/20 [01:03<00:31,  3.19s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 50%|█████     | 10/20 [01:06<00:31,  3.19s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 55%|█████▌    | 11/20 [01:06<00:27,  3.06s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 55%|█████▌    | 11/20 [01:09<00:27,  3.06s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 60%|██████    | 12/20 [01:09<00:23,  2.97s/iter, val_loss=0.000273, val_og_loss=0.000272]
+
 60%|██████    | 12/20 [01:11<00:23,  2.97s/iter, val_loss=0.000272, val_og_loss=0.000272]
+
 65%|██████▌   | 13/20 [01:11<00:20,  2.90s/iter, val_loss=0.000272, val_og_loss=0.000272]
+
 65%|██████▌   | 13/20 [01:14<00:20,  2.90s/iter, val_loss=0.000272, val_og_loss=0.000272]
+
 70%|███████   | 14/20 [01:14<00:17,  2.85s/iter, val_loss=0.000272, val_og_loss=0.000272]
+
 70%|███████   | 14/20 [01:17<00:17,  2.85s/iter, val_loss=0.000271, val_og_loss=0.000272]
+
 75%|███████▌  | 15/20 [01:17<00:14,  2.82s/iter, val_loss=0.000271, val_og_loss=0.000272]
+
 75%|███████▌  | 15/20 [01:20<00:14,  2.82s/iter, val_loss=0.000271, val_og_loss=0.000272]
+
 80%|████████  | 16/20 [01:20<00:11,  2.79s/iter, val_loss=0.000271, val_og_loss=0.000272]
+
 80%|████████  | 16/20 [01:22<00:11,  2.79s/iter, val_loss=0.000271, val_og_loss=0.000272]
+
 85%|████████▌ | 17/20 [01:22<00:08,  2.79s/iter, val_loss=0.000271, val_og_loss=0.000272]
+
 85%|████████▌ | 17/20 [01:25<00:08,  2.79s/iter, val_loss=0.000271, val_og_loss=0.000272]
+
 90%|█████████ | 18/20 [01:25<00:05,  2.77s/iter, val_loss=0.000271, val_og_loss=0.000272]
+
 90%|█████████ | 18/20 [01:28<00:05,  2.77s/iter, val_loss=0.00027, val_og_loss=0.000272] 
+
 95%|█████████▌| 19/20 [01:28<00:02,  2.76s/iter, val_loss=0.00027, val_og_loss=0.000272]
+
 95%|█████████▌| 19/20 [01:31<00:02,  2.76s/iter, val_loss=0.00027, val_og_loss=0.000272]
+
100%|██████████| 20/20 [01:31<00:00,  2.74s/iter, val_loss=0.00027, val_og_loss=0.000272]
100%|██████████| 20/20 [01:31<00:00,  4.55s/iter, val_loss=0.00027, val_og_loss=0.000272]
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [13:31<14:15, 30.56s/it]INFO:ParoQuant:Best val loss: 0.00026861028163693845, Original val loss: 0.0002715162991080433
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [13:33<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [13:33<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [14:35<14:15, 30.56s/it]2025-12-22 14:43:49,439 - INFO - Best val loss: 0.00026861028163693845, Original val loss: 0.0002715162991080433
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:26<?, ?iter/s, val_loss=1.75e-5, val_og_loss=1.76e-5]
+
  5%|▌         | 1/20 [00:26<08:16, 26.11s/iter, val_loss=1.75e-5, val_og_loss=1.76e-5]
+
  5%|▌         | 1/20 [00:31<08:16, 26.11s/iter, val_loss=1.75e-5, val_og_loss=1.76e-5]
+
 10%|█         | 2/20 [00:31<04:11, 13.95s/iter, val_loss=1.75e-5, val_og_loss=1.76e-5]
+
 10%|█         | 2/20 [00:36<04:11, 13.95s/iter, val_loss=1.75e-5, val_og_loss=1.76e-5]
+
 15%|█▌        | 3/20 [00:36<02:51, 10.07s/iter, val_loss=1.75e-5, val_og_loss=1.76e-5]
+
 15%|█▌        | 3/20 [00:42<02:51, 10.07s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 20%|██        | 4/20 [00:42<02:11,  8.24s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 20%|██        | 4/20 [00:47<02:11,  8.24s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 25%|██▌       | 5/20 [00:47<01:48,  7.22s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 25%|██▌       | 5/20 [00:53<01:48,  7.22s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 30%|███       | 6/20 [00:53<01:32,  6.61s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 30%|███       | 6/20 [00:58<01:32,  6.61s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 35%|███▌      | 7/20 [00:58<01:20,  6.23s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 35%|███▌      | 7/20 [01:04<01:20,  6.23s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 40%|████      | 8/20 [01:04<01:11,  5.98s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 40%|████      | 8/20 [01:09<01:11,  5.98s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 45%|████▌     | 9/20 [01:09<01:03,  5.81s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 45%|████▌     | 9/20 [01:15<01:03,  5.81s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 50%|█████     | 10/20 [01:15<00:56,  5.69s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 50%|█████     | 10/20 [01:20<00:56,  5.69s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 55%|█████▌    | 11/20 [01:20<00:50,  5.61s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 55%|█████▌    | 11/20 [01:25<00:50,  5.61s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 60%|██████    | 12/20 [01:25<00:44,  5.56s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 60%|██████    | 12/20 [01:31<00:44,  5.56s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 65%|██████▌   | 13/20 [01:31<00:38,  5.52s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 65%|██████▌   | 13/20 [01:36<00:38,  5.52s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 70%|███████   | 14/20 [01:36<00:32,  5.49s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 70%|███████   | 14/20 [01:42<00:32,  5.49s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 75%|███████▌  | 15/20 [01:42<00:27,  5.48s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 75%|███████▌  | 15/20 [01:47<00:27,  5.48s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 80%|████████  | 16/20 [01:47<00:21,  5.46s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 80%|████████  | 16/20 [01:53<00:21,  5.46s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 85%|████████▌ | 17/20 [01:53<00:16,  5.45s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 85%|████████▌ | 17/20 [01:58<00:16,  5.45s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 90%|█████████ | 18/20 [01:58<00:10,  5.44s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 90%|█████████ | 18/20 [02:03<00:10,  5.44s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 95%|█████████▌| 19/20 [02:03<00:05,  5.44s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
 95%|█████████▌| 19/20 [02:09<00:05,  5.44s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
100%|██████████| 20/20 [02:09<00:00,  5.43s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
100%|██████████| 20/20 [02:09<00:00,  6.47s/iter, val_loss=1.76e-5, val_og_loss=1.76e-5]
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [16:44<14:15, 30.56s/it]INFO:ParoQuant:Best val loss: 1.7502083210274577e-05, Original val loss: 1.7636453776503913e-05
+
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [16:45<14:15, 30.56s/it]
                                                                 

Processing Blocks:  12%|█▎        | 4/32 [16:55<14:15, 30.56s/it]
Processing Blocks:  16%|█▌        | 5/32 [17:32<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [17:32<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [17:52<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [17:52<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [20:17<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [20:17<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [21:24<2:39:28, 354.39s/it]2025-12-22 14:47:02,819 - INFO - Best val loss: 1.7502083210274577e-05, Original val loss: 1.7636453776503913e-05
+[INFO] Saving Layer 4 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 5...
+[INFO] Starting Sequential Training for Layer 5
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:18<?, ?iter/s, val_loss=0.00155, val_og_loss=0.00187]
+
  5%|▌         | 1/20 [00:18<05:43, 18.07s/iter, val_loss=0.00155, val_og_loss=0.00187]
+
  5%|▌         | 1/20 [00:19<05:43, 18.07s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 10%|█         | 2/20 [00:19<02:31,  8.44s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 10%|█         | 2/20 [00:21<02:31,  8.44s/iter, val_loss=0.00158, val_og_loss=0.00187]
+
 15%|█▌        | 3/20 [00:21<01:31,  5.36s/iter, val_loss=0.00158, val_og_loss=0.00187]
+
 15%|█▌        | 3/20 [00:23<01:31,  5.36s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 20%|██        | 4/20 [00:23<01:02,  3.92s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 20%|██        | 4/20 [00:24<01:02,  3.92s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 25%|██▌       | 5/20 [00:24<00:46,  3.12s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 25%|██▌       | 5/20 [00:26<00:46,  3.12s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 30%|███       | 6/20 [00:26<00:36,  2.64s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 30%|███       | 6/20 [00:28<00:36,  2.64s/iter, val_loss=0.0016, val_og_loss=0.00187] 
+
 35%|███▌      | 7/20 [00:28<00:30,  2.33s/iter, val_loss=0.0016, val_og_loss=0.00187]
+
 35%|███▌      | 7/20 [00:29<00:30,  2.33s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 40%|████      | 8/20 [00:29<00:25,  2.13s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 40%|████      | 8/20 [00:31<00:25,  2.13s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 45%|████▌     | 9/20 [00:31<00:21,  2.00s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 45%|████▌     | 9/20 [00:33<00:21,  2.00s/iter, val_loss=0.00158, val_og_loss=0.00187]
+
 50%|█████     | 10/20 [00:33<00:19,  1.90s/iter, val_loss=0.00158, val_og_loss=0.00187]
+
 50%|█████     | 10/20 [00:35<00:19,  1.90s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.84s/iter, val_loss=0.00159, val_og_loss=0.00187]
+
 55%|█████▌    | 11/20 [00:36<00:16,  1.84s/iter, val_loss=0.00158, val_og_loss=0.00187]
+
 60%|██████    | 12/20 [00:36<00:14,  1.79s/iter, val_loss=0.00158, val_og_loss=0.00187]
+
 60%|██████    | 12/20 [00:38<00:14,  1.79s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.76s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 65%|██████▌   | 13/20 [00:40<00:12,  1.76s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 70%|███████   | 14/20 [00:40<00:10,  1.74s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 70%|███████   | 14/20 [00:41<00:10,  1.74s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.72s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 75%|███████▌  | 15/20 [00:43<00:08,  1.72s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 80%|████████  | 16/20 [00:43<00:06,  1.71s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 80%|████████  | 16/20 [00:45<00:06,  1.71s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.70s/iter, val_loss=0.00157, val_og_loss=0.00187]
+
 85%|████████▌ | 17/20 [00:46<00:05,  1.70s/iter, val_loss=0.00156, val_og_loss=0.00187]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.70s/iter, val_loss=0.00156, val_og_loss=0.00187]
+
 90%|█████████ | 18/20 [00:48<00:03,  1.70s/iter, val_loss=0.00156, val_og_loss=0.00187]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.70s/iter, val_loss=0.00156, val_og_loss=0.00187]
+
 95%|█████████▌| 19/20 [00:50<00:01,  1.70s/iter, val_loss=0.00156, val_og_loss=0.00187]
+
100%|██████████| 20/20 [00:50<00:00,  1.69s/iter, val_loss=0.00156, val_og_loss=0.00187]
100%|██████████| 20/20 [00:50<00:00,  2.51s/iter, val_loss=0.00156, val_og_loss=0.00187]
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [22:15<2:39:28, 354.39s/it]INFO:ParoQuant:Best val loss: 0.0015463074669241905, Original val loss: 0.0018739233491942286
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [22:18<2:39:28, 354.39s/it]2025-12-22 14:52:33,059 - INFO - Best val loss: 0.0015463074669241905, Original val loss: 0.0018739233491942286
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00392, val_og_loss=0.00469]
+
  5%|▌         | 1/20 [00:15<05:02, 15.93s/iter, val_loss=0.00392, val_og_loss=0.00469]
+
  5%|▌         | 1/20 [00:17<05:02, 15.93s/iter, val_loss=0.00395, val_og_loss=0.00469]
+
 10%|█         | 2/20 [00:17<02:12,  7.38s/iter, val_loss=0.00395, val_og_loss=0.00469]
+
 10%|█         | 2/20 [00:18<02:12,  7.38s/iter, val_loss=0.00398, val_og_loss=0.00469]
+
 15%|█▌        | 3/20 [00:18<01:18,  4.65s/iter, val_loss=0.00398, val_og_loss=0.00469]
+
 15%|█▌        | 3/20 [00:20<01:18,  4.65s/iter, val_loss=0.00402, val_og_loss=0.00469]
+
 20%|██        | 4/20 [00:20<00:53,  3.36s/iter, val_loss=0.00402, val_og_loss=0.00469]
+
 20%|██        | 4/20 [00:21<00:53,  3.36s/iter, val_loss=0.004, val_og_loss=0.00469]  
+
 25%|██▌       | 5/20 [00:21<00:39,  2.65s/iter, val_loss=0.004, val_og_loss=0.00469]
+
 25%|██▌       | 5/20 [00:22<00:39,  2.65s/iter, val_loss=0.00399, val_og_loss=0.00469]
+
 30%|███       | 6/20 [00:22<00:31,  2.23s/iter, val_loss=0.00399, val_og_loss=0.00469]
+
 30%|███       | 6/20 [00:24<00:31,  2.23s/iter, val_loss=0.00401, val_og_loss=0.00469]
+
 35%|███▌      | 7/20 [00:24<00:25,  1.96s/iter, val_loss=0.00401, val_og_loss=0.00469]
+
 35%|███▌      | 7/20 [00:25<00:25,  1.96s/iter, val_loss=0.00399, val_og_loss=0.00469]
+
 40%|████      | 8/20 [00:25<00:21,  1.78s/iter, val_loss=0.00399, val_og_loss=0.00469]
+
 40%|████      | 8/20 [00:27<00:21,  1.78s/iter, val_loss=0.00395, val_og_loss=0.00469]
+
 45%|████▌     | 9/20 [00:27<00:18,  1.66s/iter, val_loss=0.00395, val_og_loss=0.00469]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.66s/iter, val_loss=0.00398, val_og_loss=0.00469]
+
 50%|█████     | 10/20 [00:28<00:15,  1.58s/iter, val_loss=0.00398, val_og_loss=0.00469]
+
 50%|█████     | 10/20 [00:29<00:15,  1.58s/iter, val_loss=0.00396, val_og_loss=0.00469]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.52s/iter, val_loss=0.00396, val_og_loss=0.00469]
+
 55%|█████▌    | 11/20 [00:31<00:13,  1.52s/iter, val_loss=0.00398, val_og_loss=0.00469]
+
 60%|██████    | 12/20 [00:31<00:11,  1.48s/iter, val_loss=0.00398, val_og_loss=0.00469]
+
 60%|██████    | 12/20 [00:32<00:11,  1.48s/iter, val_loss=0.00396, val_og_loss=0.00469]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.46s/iter, val_loss=0.00396, val_og_loss=0.00469]
+
 65%|██████▌   | 13/20 [00:34<00:10,  1.46s/iter, val_loss=0.00395, val_og_loss=0.00469]
+
 70%|███████   | 14/20 [00:34<00:08,  1.44s/iter, val_loss=0.00395, val_og_loss=0.00469]
+
 70%|███████   | 14/20 [00:35<00:08,  1.44s/iter, val_loss=0.00394, val_og_loss=0.00469]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.43s/iter, val_loss=0.00394, val_og_loss=0.00469]
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.43s/iter, val_loss=0.00393, val_og_loss=0.00469]
+
 80%|████████  | 16/20 [00:36<00:05,  1.41s/iter, val_loss=0.00393, val_og_loss=0.00469]
+
 80%|████████  | 16/20 [00:38<00:05,  1.41s/iter, val_loss=0.00393, val_og_loss=0.00469]
+
 85%|████████▌ | 17/20 [00:38<00:04,  1.41s/iter, val_loss=0.00393, val_og_loss=0.00469]
+
 85%|████████▌ | 17/20 [00:39<00:04,  1.41s/iter, val_loss=0.00393, val_og_loss=0.00469]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.40s/iter, val_loss=0.00393, val_og_loss=0.00469]
+
 90%|█████████ | 18/20 [00:41<00:02,  1.40s/iter, val_loss=0.00391, val_og_loss=0.00469]
+
 95%|█████████▌| 19/20 [00:41<00:01,  1.40s/iter, val_loss=0.00391, val_og_loss=0.00469]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.40s/iter, val_loss=0.00392, val_og_loss=0.00469]
+
100%|██████████| 20/20 [00:42<00:00,  1.40s/iter, val_loss=0.00392, val_og_loss=0.00469]
100%|██████████| 20/20 [00:42<00:00,  2.12s/iter, val_loss=0.00392, val_og_loss=0.00469]
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [23:00<2:39:28, 354.39s/it]INFO:ParoQuant:Best val loss: 0.003912126179784536, Original val loss: 0.004688845947384834
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [23:04<2:39:28, 354.39s/it]2025-12-22 14:53:18,962 - INFO - Best val loss: 0.003912126179784536, Original val loss: 0.004688845947384834
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:25<?, ?iter/s, val_loss=0.000201, val_og_loss=0.000258]
+
  5%|▌         | 1/20 [00:25<08:00, 25.27s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
  5%|▌         | 1/20 [00:26<08:00, 25.27s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 10%|█         | 2/20 [00:26<03:22, 11.25s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 10%|█         | 2/20 [00:28<03:22, 11.25s/iter, val_loss=0.000202, val_og_loss=0.000258]
+
 15%|█▌        | 3/20 [00:28<01:54,  6.76s/iter, val_loss=0.000202, val_og_loss=0.000258]
+
 15%|█▌        | 3/20 [00:29<01:54,  6.76s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 20%|██        | 4/20 [00:29<01:14,  4.66s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 20%|██        | 4/20 [00:30<01:14,  4.66s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 25%|██▌       | 5/20 [00:30<00:52,  3.49s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 25%|██▌       | 5/20 [00:32<00:52,  3.49s/iter, val_loss=0.000202, val_og_loss=0.000258]
+
 30%|███       | 6/20 [00:32<00:39,  2.80s/iter, val_loss=0.000202, val_og_loss=0.000258]
+
 30%|███       | 6/20 [00:33<00:39,  2.80s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 35%|███▌      | 7/20 [00:33<00:30,  2.35s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 35%|███▌      | 7/20 [00:35<00:30,  2.35s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 40%|████      | 8/20 [00:35<00:24,  2.06s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 40%|████      | 8/20 [00:36<00:24,  2.06s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 45%|████▌     | 9/20 [00:36<00:20,  1.85s/iter, val_loss=0.000201, val_og_loss=0.000258]
+
 45%|████▌     | 9/20 [00:38<00:20,  1.85s/iter, val_loss=0.0002, val_og_loss=0.000258]  
+
 50%|█████     | 10/20 [00:38<00:17,  1.71s/iter, val_loss=0.0002, val_og_loss=0.000258]
+
 50%|█████     | 10/20 [00:39<00:17,  1.71s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 55%|█████▌    | 11/20 [00:39<00:14,  1.61s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 55%|█████▌    | 11/20 [00:40<00:14,  1.61s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 60%|██████    | 12/20 [00:40<00:12,  1.55s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 60%|██████    | 12/20 [00:42<00:12,  1.55s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 65%|██████▌   | 13/20 [00:42<00:10,  1.51s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 65%|██████▌   | 13/20 [00:43<00:10,  1.51s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 70%|███████   | 14/20 [00:43<00:08,  1.48s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 70%|███████   | 14/20 [00:45<00:08,  1.48s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 75%|███████▌  | 15/20 [00:45<00:07,  1.46s/iter, val_loss=0.000199, val_og_loss=0.000258]
+
 75%|███████▌  | 15/20 [00:46<00:07,  1.46s/iter, val_loss=0.000198, val_og_loss=0.000258]
+
 80%|████████  | 16/20 [00:46<00:05,  1.44s/iter, val_loss=0.000198, val_og_loss=0.000258]
+
 80%|████████  | 16/20 [00:47<00:05,  1.44s/iter, val_loss=0.000198, val_og_loss=0.000258]
+
 85%|████████▌ | 17/20 [00:47<00:04,  1.42s/iter, val_loss=0.000198, val_og_loss=0.000258]
+
 85%|████████▌ | 17/20 [00:49<00:04,  1.42s/iter, val_loss=0.000197, val_og_loss=0.000258]
+
 90%|█████████ | 18/20 [00:49<00:02,  1.41s/iter, val_loss=0.000197, val_og_loss=0.000258]
+
 90%|█████████ | 18/20 [00:50<00:02,  1.41s/iter, val_loss=0.000197, val_og_loss=0.000258]
+
 95%|█████████▌| 19/20 [00:50<00:01,  1.41s/iter, val_loss=0.000197, val_og_loss=0.000258]
+
 95%|█████████▌| 19/20 [00:52<00:01,  1.41s/iter, val_loss=0.000197, val_og_loss=0.000258]
+
100%|██████████| 20/20 [00:52<00:00,  1.41s/iter, val_loss=0.000197, val_og_loss=0.000258]
100%|██████████| 20/20 [00:52<00:00,  2.60s/iter, val_loss=0.000197, val_og_loss=0.000258]
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [23:56<2:39:28, 354.39s/it]INFO:ParoQuant:Best val loss: 0.00019698936375789344, Original val loss: 0.00025794250541366637
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [23:59<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [23:59<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [24:32<2:39:28, 354.39s/it]2025-12-22 14:54:14,363 - INFO - Best val loss: 0.00019698936375789344, Original val loss: 0.00025794250541366637
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=6.32e-6, val_og_loss=6.83e-6]
+
  5%|▌         | 1/20 [00:15<05:00, 15.83s/iter, val_loss=6.32e-6, val_og_loss=6.83e-6]
+
  5%|▌         | 1/20 [00:17<05:00, 15.83s/iter, val_loss=6.32e-6, val_og_loss=6.83e-6]
+
 10%|█         | 2/20 [00:17<02:15,  7.51s/iter, val_loss=6.32e-6, val_og_loss=6.83e-6]
+
 10%|█         | 2/20 [00:19<02:15,  7.51s/iter, val_loss=6.35e-6, val_og_loss=6.83e-6]
+
 15%|█▌        | 3/20 [00:19<01:22,  4.85s/iter, val_loss=6.35e-6, val_og_loss=6.83e-6]
+
 15%|█▌        | 3/20 [00:20<01:22,  4.85s/iter, val_loss=6.39e-6, val_og_loss=6.83e-6]
+
 20%|██        | 4/20 [00:20<00:57,  3.61s/iter, val_loss=6.39e-6, val_og_loss=6.83e-6]
+
 20%|██        | 4/20 [00:22<00:57,  3.61s/iter, val_loss=6.44e-6, val_og_loss=6.83e-6]
+
 25%|██▌       | 5/20 [00:22<00:43,  2.92s/iter, val_loss=6.44e-6, val_og_loss=6.83e-6]
+
 25%|██▌       | 5/20 [00:24<00:43,  2.92s/iter, val_loss=6.4e-6, val_og_loss=6.83e-6] 
+
 30%|███       | 6/20 [00:24<00:34,  2.50s/iter, val_loss=6.4e-6, val_og_loss=6.83e-6]
+
 30%|███       | 6/20 [00:25<00:34,  2.50s/iter, val_loss=6.4e-6, val_og_loss=6.83e-6]
+
 35%|███▌      | 7/20 [00:25<00:29,  2.23s/iter, val_loss=6.4e-6, val_og_loss=6.83e-6]
+
 35%|███▌      | 7/20 [00:27<00:29,  2.23s/iter, val_loss=6.39e-6, val_og_loss=6.83e-6]
+
 40%|████      | 8/20 [00:27<00:24,  2.06s/iter, val_loss=6.39e-6, val_og_loss=6.83e-6]
+
 40%|████      | 8/20 [00:29<00:24,  2.06s/iter, val_loss=6.39e-6, val_og_loss=6.83e-6]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.94s/iter, val_loss=6.39e-6, val_og_loss=6.83e-6]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.94s/iter, val_loss=6.4e-6, val_og_loss=6.83e-6] 
+
 50%|█████     | 10/20 [00:31<00:18,  1.86s/iter, val_loss=6.4e-6, val_og_loss=6.83e-6]
+
 50%|█████     | 10/20 [00:32<00:18,  1.86s/iter, val_loss=6.37e-6, val_og_loss=6.83e-6]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.81s/iter, val_loss=6.37e-6, val_og_loss=6.83e-6]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.81s/iter, val_loss=6.37e-6, val_og_loss=6.83e-6]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=6.37e-6, val_og_loss=6.83e-6]
+
 60%|██████    | 12/20 [00:36<00:14,  1.77s/iter, val_loss=6.36e-6, val_og_loss=6.83e-6]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.75s/iter, val_loss=6.36e-6, val_og_loss=6.83e-6]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=6.35e-6, val_og_loss=6.83e-6]
+
 70%|███████   | 14/20 [00:37<00:10,  1.73s/iter, val_loss=6.35e-6, val_og_loss=6.83e-6]
+
 70%|███████   | 14/20 [00:39<00:10,  1.73s/iter, val_loss=6.37e-6, val_og_loss=6.83e-6]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.72s/iter, val_loss=6.37e-6, val_og_loss=6.83e-6]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.72s/iter, val_loss=6.35e-6, val_og_loss=6.83e-6]
+
 80%|████████  | 16/20 [00:41<00:06,  1.71s/iter, val_loss=6.35e-6, val_og_loss=6.83e-6]
+
 80%|████████  | 16/20 [00:42<00:06,  1.71s/iter, val_loss=6.34e-6, val_og_loss=6.83e-6]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=6.34e-6, val_og_loss=6.83e-6]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=6.33e-6, val_og_loss=6.83e-6]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=6.33e-6, val_og_loss=6.83e-6]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.69s/iter, val_loss=6.32e-6, val_og_loss=6.83e-6]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=6.32e-6, val_og_loss=6.83e-6]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=6.32e-6, val_og_loss=6.83e-6]
+
100%|██████████| 20/20 [00:47<00:00,  1.69s/iter, val_loss=6.32e-6, val_og_loss=6.83e-6]
100%|██████████| 20/20 [00:47<00:00,  2.39s/iter, val_loss=6.32e-6, val_og_loss=6.83e-6]
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [25:20<2:39:28, 354.39s/it]INFO:ParoQuant:Best val loss: 6.317984116321895e-06, Original val loss: 6.8322942752274685e-06
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [25:21<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [25:21<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [26:08<2:39:28, 354.39s/it]2025-12-22 14:55:38,439 - INFO - Best val loss: 6.317984116321895e-06, Original val loss: 6.8322942752274685e-06
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:32<?, ?iter/s, val_loss=0.000618, val_og_loss=0.000625]
+
  5%|▌         | 1/20 [00:32<10:15, 32.39s/iter, val_loss=0.000618, val_og_loss=0.000625]
+
  5%|▌         | 1/20 [00:35<10:15, 32.39s/iter, val_loss=0.00062, val_og_loss=0.000625] 
+
 10%|█         | 2/20 [00:35<04:28, 14.94s/iter, val_loss=0.00062, val_og_loss=0.000625]
+
 10%|█         | 2/20 [00:37<04:28, 14.94s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 15%|█▌        | 3/20 [00:37<02:39,  9.37s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 15%|█▌        | 3/20 [00:40<02:39,  9.37s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 20%|██        | 4/20 [00:40<01:47,  6.74s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 20%|██        | 4/20 [00:43<01:47,  6.74s/iter, val_loss=0.000623, val_og_loss=0.000625]
+
 25%|██▌       | 5/20 [00:43<01:19,  5.29s/iter, val_loss=0.000623, val_og_loss=0.000625]
+
 25%|██▌       | 5/20 [00:46<01:19,  5.29s/iter, val_loss=0.000623, val_og_loss=0.000625]
+
 30%|███       | 6/20 [00:46<01:01,  4.42s/iter, val_loss=0.000623, val_og_loss=0.000625]
+
 30%|███       | 6/20 [00:48<01:01,  4.42s/iter, val_loss=0.000624, val_og_loss=0.000625]
+
 35%|███▌      | 7/20 [00:48<00:50,  3.87s/iter, val_loss=0.000624, val_og_loss=0.000625]
+
 35%|███▌      | 7/20 [00:51<00:50,  3.87s/iter, val_loss=0.000624, val_og_loss=0.000625]
+
 40%|████      | 8/20 [00:51<00:42,  3.50s/iter, val_loss=0.000624, val_og_loss=0.000625]
+
 40%|████      | 8/20 [00:54<00:42,  3.50s/iter, val_loss=0.000624, val_og_loss=0.000625]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.26s/iter, val_loss=0.000624, val_og_loss=0.000625]
+
 45%|████▌     | 9/20 [00:56<00:35,  3.26s/iter, val_loss=0.000623, val_og_loss=0.000625]
+
 50%|█████     | 10/20 [00:56<00:30,  3.10s/iter, val_loss=0.000623, val_og_loss=0.000625]
+
 50%|█████     | 10/20 [00:59<00:30,  3.10s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.98s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 55%|█████▌    | 11/20 [01:02<00:26,  2.98s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 60%|██████    | 12/20 [01:02<00:23,  2.90s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 60%|██████    | 12/20 [01:05<00:23,  2.90s/iter, val_loss=0.00062, val_og_loss=0.000625] 
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.85s/iter, val_loss=0.00062, val_og_loss=0.000625]
+
 65%|██████▌   | 13/20 [01:07<00:19,  2.85s/iter, val_loss=0.00062, val_og_loss=0.000625]
+
 70%|███████   | 14/20 [01:07<00:16,  2.81s/iter, val_loss=0.00062, val_og_loss=0.000625]
+
 70%|███████   | 14/20 [01:10<00:16,  2.81s/iter, val_loss=0.00062, val_og_loss=0.000625]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.79s/iter, val_loss=0.00062, val_og_loss=0.000625]
+
 75%|███████▌  | 15/20 [01:13<00:13,  2.79s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 80%|████████  | 16/20 [01:16<00:11,  2.76s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.76s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 85%|████████▌ | 17/20 [01:18<00:08,  2.76s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 90%|█████████ | 18/20 [01:21<00:05,  2.74s/iter, val_loss=0.000618, val_og_loss=0.000625]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.74s/iter, val_loss=0.000618, val_og_loss=0.000625]
+
 95%|█████████▌| 19/20 [01:24<00:02,  2.74s/iter, val_loss=0.000618, val_og_loss=0.000625]
+
100%|██████████| 20/20 [01:24<00:00,  2.73s/iter, val_loss=0.000618, val_og_loss=0.000625]
100%|██████████| 20/20 [01:24<00:00,  4.21s/iter, val_loss=0.000618, val_og_loss=0.000625]
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [27:32<2:39:28, 354.39s/it]INFO:ParoQuant:Best val loss: 0.0006179442862048745, Original val loss: 0.0006252065650187433
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [27:37<2:39:28, 354.39s/it]2025-12-22 14:57:50,825 - INFO - Best val loss: 0.0006179442862048745, Original val loss: 0.0006252065650187433
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:36<?, ?iter/s, val_loss=0.000326, val_og_loss=0.000328]
+
  5%|▌         | 1/20 [00:36<11:36, 36.64s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
  5%|▌         | 1/20 [00:39<11:36, 36.64s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
 10%|█         | 2/20 [00:39<05:00, 16.68s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
 10%|█         | 2/20 [00:42<05:00, 16.68s/iter, val_loss=0.000328, val_og_loss=0.000328]
+
 15%|█▌        | 3/20 [00:42<02:55, 10.30s/iter, val_loss=0.000328, val_og_loss=0.000328]
+
 15%|█▌        | 3/20 [00:44<02:55, 10.30s/iter, val_loss=0.000329, val_og_loss=0.000328]
+
 20%|██        | 4/20 [00:44<01:56,  7.31s/iter, val_loss=0.000329, val_og_loss=0.000328]
+
 20%|██        | 4/20 [00:47<01:56,  7.31s/iter, val_loss=0.00033, val_og_loss=0.000328] 
+
 25%|██▌       | 5/20 [00:47<01:24,  5.66s/iter, val_loss=0.00033, val_og_loss=0.000328]
+
 25%|██▌       | 5/20 [00:50<01:24,  5.66s/iter, val_loss=0.00033, val_og_loss=0.000328]
+
 30%|███       | 6/20 [00:50<01:05,  4.66s/iter, val_loss=0.00033, val_og_loss=0.000328]
+
 30%|███       | 6/20 [00:52<01:05,  4.66s/iter, val_loss=0.00033, val_og_loss=0.000328]
+
 35%|███▌      | 7/20 [00:52<00:52,  4.03s/iter, val_loss=0.00033, val_og_loss=0.000328]
+
 35%|███▌      | 7/20 [00:55<00:52,  4.03s/iter, val_loss=0.00033, val_og_loss=0.000328]
+
 40%|████      | 8/20 [00:55<00:43,  3.61s/iter, val_loss=0.00033, val_og_loss=0.000328]
+
 40%|████      | 8/20 [00:58<00:43,  3.61s/iter, val_loss=0.00033, val_og_loss=0.000328]
+
 45%|████▌     | 9/20 [00:58<00:36,  3.33s/iter, val_loss=0.00033, val_og_loss=0.000328]
+
 45%|████▌     | 9/20 [01:01<00:36,  3.33s/iter, val_loss=0.000329, val_og_loss=0.000328]
+
 50%|█████     | 10/20 [01:01<00:31,  3.15s/iter, val_loss=0.000329, val_og_loss=0.000328]
+
 50%|█████     | 10/20 [01:03<00:31,  3.15s/iter, val_loss=0.000329, val_og_loss=0.000328]
+
 55%|█████▌    | 11/20 [01:03<00:27,  3.02s/iter, val_loss=0.000329, val_og_loss=0.000328]
+
 55%|█████▌    | 11/20 [01:06<00:27,  3.02s/iter, val_loss=0.000328, val_og_loss=0.000328]
+
 60%|██████    | 12/20 [01:06<00:23,  2.93s/iter, val_loss=0.000328, val_og_loss=0.000328]
+
 60%|██████    | 12/20 [01:09<00:23,  2.93s/iter, val_loss=0.000328, val_og_loss=0.000328]
+
 65%|██████▌   | 13/20 [01:09<00:20,  2.87s/iter, val_loss=0.000328, val_og_loss=0.000328]
+
 65%|██████▌   | 13/20 [01:12<00:20,  2.87s/iter, val_loss=0.000327, val_og_loss=0.000328]
+
 70%|███████   | 14/20 [01:12<00:16,  2.82s/iter, val_loss=0.000327, val_og_loss=0.000328]
+
 70%|███████   | 14/20 [01:14<00:16,  2.82s/iter, val_loss=0.000327, val_og_loss=0.000328]
+
 75%|███████▌  | 15/20 [01:14<00:13,  2.80s/iter, val_loss=0.000327, val_og_loss=0.000328]
+
 75%|███████▌  | 15/20 [01:17<00:13,  2.80s/iter, val_loss=0.000327, val_og_loss=0.000328]
+
 80%|████████  | 16/20 [01:17<00:11,  2.77s/iter, val_loss=0.000327, val_og_loss=0.000328]
+
 80%|████████  | 16/20 [01:20<00:11,  2.77s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
 85%|████████▌ | 17/20 [01:20<00:08,  2.76s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
 85%|████████▌ | 17/20 [01:22<00:08,  2.76s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
 90%|█████████ | 18/20 [01:22<00:05,  2.75s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
 90%|█████████ | 18/20 [01:25<00:05,  2.75s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
 95%|█████████▌| 19/20 [01:25<00:02,  2.74s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
 95%|█████████▌| 19/20 [01:28<00:02,  2.74s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
100%|██████████| 20/20 [01:28<00:00,  2.73s/iter, val_loss=0.000326, val_og_loss=0.000328]
100%|██████████| 20/20 [01:28<00:00,  4.42s/iter, val_loss=0.000326, val_og_loss=0.000328]
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [29:06<2:39:28, 354.39s/it]INFO:ParoQuant:Best val loss: 0.000325753353536129, Original val loss: 0.0003284731355961412
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [29:07<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [29:07<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [30:16<2:39:28, 354.39s/it]2025-12-22 14:59:24,094 - INFO - Best val loss: 0.000325753353536129, Original val loss: 0.0003284731355961412
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:36<?, ?iter/s, val_loss=2.55e-5, val_og_loss=2.58e-5]
+
  5%|▌         | 1/20 [00:36<11:25, 36.08s/iter, val_loss=2.55e-5, val_og_loss=2.58e-5]
+
  5%|▌         | 1/20 [00:41<11:25, 36.08s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 10%|█         | 2/20 [00:41<05:25, 18.08s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 10%|█         | 2/20 [00:46<05:25, 18.08s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 15%|█▌        | 3/20 [00:46<03:29, 12.30s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 15%|█▌        | 3/20 [00:52<03:29, 12.30s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 20%|██        | 4/20 [00:52<02:33,  9.58s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 20%|██        | 4/20 [00:57<02:33,  9.58s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 25%|██▌       | 5/20 [00:57<02:01,  8.08s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 25%|██▌       | 5/20 [01:03<02:01,  8.08s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 30%|███       | 6/20 [01:03<01:40,  7.18s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 30%|███       | 6/20 [01:08<01:40,  7.18s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 35%|███▌      | 7/20 [01:08<01:25,  6.60s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 35%|███▌      | 7/20 [01:14<01:25,  6.60s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 40%|████      | 8/20 [01:14<01:14,  6.23s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 40%|████      | 8/20 [01:19<01:14,  6.23s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 45%|████▌     | 9/20 [01:19<01:05,  5.98s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 45%|████▌     | 9/20 [01:24<01:05,  5.98s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 50%|█████     | 10/20 [01:24<00:58,  5.81s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 50%|█████     | 10/20 [01:30<00:58,  5.81s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 55%|█████▌    | 11/20 [01:30<00:51,  5.69s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 55%|█████▌    | 11/20 [01:35<00:51,  5.69s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 60%|██████    | 12/20 [01:35<00:44,  5.61s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 60%|██████    | 12/20 [01:41<00:44,  5.61s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 65%|██████▌   | 13/20 [01:41<00:38,  5.57s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 65%|██████▌   | 13/20 [01:46<00:38,  5.57s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 70%|███████   | 14/20 [01:46<00:33,  5.53s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 70%|███████   | 14/20 [01:52<00:33,  5.53s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 75%|███████▌  | 15/20 [01:52<00:27,  5.51s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 75%|███████▌  | 15/20 [01:57<00:27,  5.51s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 80%|████████  | 16/20 [01:57<00:21,  5.50s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 80%|████████  | 16/20 [02:03<00:21,  5.50s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 85%|████████▌ | 17/20 [02:03<00:16,  5.48s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 85%|████████▌ | 17/20 [02:08<00:16,  5.48s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 90%|█████████ | 18/20 [02:08<00:10,  5.46s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
 90%|█████████ | 18/20 [02:13<00:10,  5.46s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 95%|█████████▌| 19/20 [02:13<00:05,  5.45s/iter, val_loss=2.57e-5, val_og_loss=2.58e-5]
+
 95%|█████████▌| 19/20 [02:19<00:05,  5.45s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
100%|██████████| 20/20 [02:19<00:00,  5.45s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
100%|██████████| 20/20 [02:19<00:00,  6.97s/iter, val_loss=2.56e-5, val_og_loss=2.58e-5]
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [32:36<2:39:28, 354.39s/it]INFO:ParoQuant:Best val loss: 2.5517179892631248e-05, Original val loss: 2.575260805315338e-05
+
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [32:37<2:39:28, 354.39s/it]
                                                                    

Processing Blocks:  16%|█▌        | 5/32 [32:52<2:39:28, 354.39s/it]
Processing Blocks:  19%|█▉        | 6/32 [33:37<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [33:37<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [33:57<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [33:57<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [36:19<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [36:19<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [37:21<4:03:30, 561.93s/it]2025-12-22 15:02:54,003 - INFO - Best val loss: 2.5517179892631248e-05, Original val loss: 2.575260805315338e-05
+[INFO] Saving Layer 5 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 6...
+[INFO] Starting Sequential Training for Layer 6
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00153, val_og_loss=0.00181]
+
  5%|▌         | 1/20 [00:15<04:49, 15.26s/iter, val_loss=0.00153, val_og_loss=0.00181]
+
  5%|▌         | 1/20 [00:16<04:49, 15.26s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 10%|█         | 2/20 [00:16<02:11,  7.28s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 10%|█         | 2/20 [00:18<02:11,  7.28s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 15%|█▌        | 3/20 [00:18<01:20,  4.74s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 15%|█▌        | 3/20 [00:20<01:20,  4.74s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 20%|██        | 4/20 [00:20<00:56,  3.55s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 20%|██        | 4/20 [00:22<00:56,  3.55s/iter, val_loss=0.00157, val_og_loss=0.00181]
+
 25%|██▌       | 5/20 [00:22<00:43,  2.88s/iter, val_loss=0.00157, val_og_loss=0.00181]
+
 25%|██▌       | 5/20 [00:23<00:43,  2.88s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 30%|███       | 6/20 [00:23<00:34,  2.48s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 30%|███       | 6/20 [00:25<00:34,  2.48s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.23s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 35%|███▌      | 7/20 [00:27<00:28,  2.23s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 40%|████      | 8/20 [00:27<00:24,  2.06s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 40%|████      | 8/20 [00:28<00:24,  2.06s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 45%|████▌     | 9/20 [00:28<00:21,  1.94s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.94s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 50%|█████     | 10/20 [00:30<00:18,  1.86s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 50%|█████     | 10/20 [00:32<00:18,  1.86s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.81s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.81s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 60%|██████    | 12/20 [00:33<00:14,  1.77s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 60%|██████    | 12/20 [00:35<00:14,  1.77s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.75s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=0.00154, val_og_loss=0.00181]
+
 70%|███████   | 14/20 [00:37<00:10,  1.73s/iter, val_loss=0.00154, val_og_loss=0.00181]
+
 70%|███████   | 14/20 [00:39<00:10,  1.73s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.72s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.72s/iter, val_loss=0.00154, val_og_loss=0.00181]
+
 80%|████████  | 16/20 [00:40<00:06,  1.71s/iter, val_loss=0.00154, val_og_loss=0.00181]
+
 80%|████████  | 16/20 [00:42<00:06,  1.71s/iter, val_loss=0.00154, val_og_loss=0.00181]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=0.00154, val_og_loss=0.00181]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=0.00154, val_og_loss=0.00181]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.00154, val_og_loss=0.00181]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.69s/iter, val_loss=0.00153, val_og_loss=0.00181]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=0.00153, val_og_loss=0.00181]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=0.00153, val_og_loss=0.00181]
+
100%|██████████| 20/20 [00:47<00:00,  1.68s/iter, val_loss=0.00153, val_og_loss=0.00181]
100%|██████████| 20/20 [00:47<00:00,  2.37s/iter, val_loss=0.00153, val_og_loss=0.00181]
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [38:08<4:03:30, 561.93s/it]INFO:ParoQuant:Best val loss: 0.001531147980131209, Original val loss: 0.0018140203319489956
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [38:12<4:03:30, 561.93s/it]2025-12-22 15:08:26,772 - INFO - Best val loss: 0.001531147980131209, Original val loss: 0.0018140203319489956
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=0.0038, val_og_loss=0.00443]
+
  5%|▌         | 1/20 [00:11<03:43, 11.77s/iter, val_loss=0.0038, val_og_loss=0.00443]
+
  5%|▌         | 1/20 [00:13<03:43, 11.77s/iter, val_loss=0.00382, val_og_loss=0.00443]
+
 10%|█         | 2/20 [00:13<01:42,  5.67s/iter, val_loss=0.00382, val_og_loss=0.00443]
+
 10%|█         | 2/20 [00:14<01:42,  5.67s/iter, val_loss=0.00385, val_og_loss=0.00443]
+
 15%|█▌        | 3/20 [00:14<01:03,  3.72s/iter, val_loss=0.00385, val_og_loss=0.00443]
+
 15%|█▌        | 3/20 [00:15<01:03,  3.72s/iter, val_loss=0.00384, val_og_loss=0.00443]
+
 20%|██        | 4/20 [00:15<00:44,  2.81s/iter, val_loss=0.00384, val_og_loss=0.00443]
+
 20%|██        | 4/20 [00:17<00:44,  2.81s/iter, val_loss=0.00385, val_og_loss=0.00443]
+
 25%|██▌       | 5/20 [00:17<00:34,  2.30s/iter, val_loss=0.00385, val_og_loss=0.00443]
+
 25%|██▌       | 5/20 [00:18<00:34,  2.30s/iter, val_loss=0.00383, val_og_loss=0.00443]
+
 30%|███       | 6/20 [00:18<00:27,  1.99s/iter, val_loss=0.00383, val_og_loss=0.00443]
+
 30%|███       | 6/20 [00:20<00:27,  1.99s/iter, val_loss=0.00384, val_og_loss=0.00443]
+
 35%|███▌      | 7/20 [00:20<00:23,  1.79s/iter, val_loss=0.00384, val_og_loss=0.00443]
+
 35%|███▌      | 7/20 [00:21<00:23,  1.79s/iter, val_loss=0.00385, val_og_loss=0.00443]
+
 40%|████      | 8/20 [00:21<00:19,  1.66s/iter, val_loss=0.00385, val_og_loss=0.00443]
+
 40%|████      | 8/20 [00:22<00:19,  1.66s/iter, val_loss=0.00388, val_og_loss=0.00443]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.58s/iter, val_loss=0.00388, val_og_loss=0.00443]
+
 45%|████▌     | 9/20 [00:24<00:17,  1.58s/iter, val_loss=0.00384, val_og_loss=0.00443]
+
 50%|█████     | 10/20 [00:24<00:15,  1.52s/iter, val_loss=0.00384, val_og_loss=0.00443]
+
 50%|█████     | 10/20 [00:25<00:15,  1.52s/iter, val_loss=0.00381, val_og_loss=0.00443]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.48s/iter, val_loss=0.00381, val_og_loss=0.00443]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.48s/iter, val_loss=0.0038, val_og_loss=0.00443] 
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.0038, val_og_loss=0.00443]
+
 60%|██████    | 12/20 [00:28<00:11,  1.45s/iter, val_loss=0.00382, val_og_loss=0.00443]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.43s/iter, val_loss=0.00382, val_og_loss=0.00443]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.43s/iter, val_loss=0.0038, val_og_loss=0.00443] 
+
 70%|███████   | 14/20 [00:29<00:08,  1.42s/iter, val_loss=0.0038, val_og_loss=0.00443]
+
 70%|███████   | 14/20 [00:31<00:08,  1.42s/iter, val_loss=0.0038, val_og_loss=0.00443]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.42s/iter, val_loss=0.0038, val_og_loss=0.00443]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.42s/iter, val_loss=0.00379, val_og_loss=0.00443]
+
 80%|████████  | 16/20 [00:32<00:05,  1.41s/iter, val_loss=0.00379, val_og_loss=0.00443]
+
 80%|████████  | 16/20 [00:34<00:05,  1.41s/iter, val_loss=0.00377, val_og_loss=0.00443]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00377, val_og_loss=0.00443]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.40s/iter, val_loss=0.00376, val_og_loss=0.00443]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00376, val_og_loss=0.00443]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.00377, val_og_loss=0.00443]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.00377, val_og_loss=0.00443]
+
 95%|█████████▌| 19/20 [00:38<00:01,  1.40s/iter, val_loss=0.00377, val_og_loss=0.00443]
+
100%|██████████| 20/20 [00:38<00:00,  1.40s/iter, val_loss=0.00377, val_og_loss=0.00443]
100%|██████████| 20/20 [00:38<00:00,  1.91s/iter, val_loss=0.00377, val_og_loss=0.00443]
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [38:50<4:03:30, 561.93s/it]INFO:ParoQuant:Best val loss: 0.003759661689400673, Original val loss: 0.004428935702890158
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [38:54<4:03:30, 561.93s/it]2025-12-22 15:09:08,500 - INFO - Best val loss: 0.003759661689400673, Original val loss: 0.004428935702890158
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:12<?, ?iter/s, val_loss=0.000232, val_og_loss=0.000288]
+
  5%|▌         | 1/20 [00:12<04:02, 12.75s/iter, val_loss=0.000232, val_og_loss=0.000288]
+
  5%|▌         | 1/20 [00:14<04:02, 12.75s/iter, val_loss=0.000231, val_og_loss=0.000288]
+
 10%|█         | 2/20 [00:14<01:49,  6.07s/iter, val_loss=0.000231, val_og_loss=0.000288]
+
 10%|█         | 2/20 [00:15<01:49,  6.07s/iter, val_loss=0.000232, val_og_loss=0.000288]
+
 15%|█▌        | 3/20 [00:15<01:06,  3.93s/iter, val_loss=0.000232, val_og_loss=0.000288]
+
 15%|█▌        | 3/20 [00:16<01:06,  3.93s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 20%|██        | 4/20 [00:16<00:46,  2.93s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 20%|██        | 4/20 [00:18<00:46,  2.93s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 25%|██▌       | 5/20 [00:18<00:35,  2.38s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 25%|██▌       | 5/20 [00:19<00:35,  2.38s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 30%|███       | 6/20 [00:19<00:28,  2.05s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 30%|███       | 6/20 [00:21<00:28,  2.05s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 35%|███▌      | 7/20 [00:21<00:23,  1.83s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 35%|███▌      | 7/20 [00:22<00:23,  1.83s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 40%|████      | 8/20 [00:22<00:20,  1.70s/iter, val_loss=0.000233, val_og_loss=0.000288]
+
 40%|████      | 8/20 [00:23<00:20,  1.70s/iter, val_loss=0.000232, val_og_loss=0.000288]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.61s/iter, val_loss=0.000232, val_og_loss=0.000288]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.61s/iter, val_loss=0.000231, val_og_loss=0.000288]
+
 50%|█████     | 10/20 [00:25<00:15,  1.54s/iter, val_loss=0.000231, val_og_loss=0.000288]
+
 50%|█████     | 10/20 [00:26<00:15,  1.54s/iter, val_loss=0.000231, val_og_loss=0.000288]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.50s/iter, val_loss=0.000231, val_og_loss=0.000288]
+
 55%|█████▌    | 11/20 [00:28<00:13,  1.50s/iter, val_loss=0.000232, val_og_loss=0.000288]
+
 60%|██████    | 12/20 [00:28<00:11,  1.47s/iter, val_loss=0.000232, val_og_loss=0.000288]
+
 60%|██████    | 12/20 [00:29<00:11,  1.47s/iter, val_loss=0.00023, val_og_loss=0.000288] 
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.45s/iter, val_loss=0.00023, val_og_loss=0.000288]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.45s/iter, val_loss=0.00023, val_og_loss=0.000288]
+
 70%|███████   | 14/20 [00:30<00:08,  1.44s/iter, val_loss=0.00023, val_og_loss=0.000288]
+
 70%|███████   | 14/20 [00:32<00:08,  1.44s/iter, val_loss=0.00023, val_og_loss=0.000288]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.42s/iter, val_loss=0.00023, val_og_loss=0.000288]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.42s/iter, val_loss=0.00023, val_og_loss=0.000288]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.00023, val_og_loss=0.000288]
+
 80%|████████  | 16/20 [00:35<00:05,  1.41s/iter, val_loss=0.00023, val_og_loss=0.000288]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.41s/iter, val_loss=0.00023, val_og_loss=0.000288]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.41s/iter, val_loss=0.000229, val_og_loss=0.000288]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.000229, val_og_loss=0.000288]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.40s/iter, val_loss=0.000229, val_og_loss=0.000288]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.41s/iter, val_loss=0.000229, val_og_loss=0.000288]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.41s/iter, val_loss=0.000229, val_og_loss=0.000288]
+
100%|██████████| 20/20 [00:39<00:00,  1.40s/iter, val_loss=0.000229, val_og_loss=0.000288]
100%|██████████| 20/20 [00:39<00:00,  1.97s/iter, val_loss=0.000229, val_og_loss=0.000288]
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [39:33<4:03:30, 561.93s/it]INFO:ParoQuant:Best val loss: 0.00022866034123580903, Original val loss: 0.00028849125374108553
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [39:36<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [39:36<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [40:10<4:03:30, 561.93s/it]2025-12-22 15:09:51,886 - INFO - Best val loss: 0.00022866034123580903, Original val loss: 0.00028849125374108553
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:17<?, ?iter/s, val_loss=9.72e-6, val_og_loss=1.02e-5]
+
  5%|▌         | 1/20 [00:17<05:25, 17.15s/iter, val_loss=9.72e-6, val_og_loss=1.02e-5]
+
  5%|▌         | 1/20 [00:18<05:25, 17.15s/iter, val_loss=9.73e-6, val_og_loss=1.02e-5]
+
 10%|█         | 2/20 [00:18<02:25,  8.06s/iter, val_loss=9.73e-6, val_og_loss=1.02e-5]
+
 10%|█         | 2/20 [00:20<02:25,  8.06s/iter, val_loss=9.81e-6, val_og_loss=1.02e-5]
+
 15%|█▌        | 3/20 [00:20<01:27,  5.15s/iter, val_loss=9.81e-6, val_og_loss=1.02e-5]
+
 15%|█▌        | 3/20 [00:22<01:27,  5.15s/iter, val_loss=9.84e-6, val_og_loss=1.02e-5]
+
 20%|██        | 4/20 [00:22<01:00,  3.78s/iter, val_loss=9.84e-6, val_og_loss=1.02e-5]
+
 20%|██        | 4/20 [00:23<01:00,  3.78s/iter, val_loss=9.8e-6, val_og_loss=1.02e-5] 
+
 25%|██▌       | 5/20 [00:23<00:45,  3.03s/iter, val_loss=9.8e-6, val_og_loss=1.02e-5]
+
 25%|██▌       | 5/20 [00:25<00:45,  3.03s/iter, val_loss=9.81e-6, val_og_loss=1.02e-5]
+
 30%|███       | 6/20 [00:25<00:36,  2.58s/iter, val_loss=9.81e-6, val_og_loss=1.02e-5]
+
 30%|███       | 6/20 [00:27<00:36,  2.58s/iter, val_loss=9.84e-6, val_og_loss=1.02e-5]
+
 35%|███▌      | 7/20 [00:27<00:29,  2.28s/iter, val_loss=9.84e-6, val_og_loss=1.02e-5]
+
 35%|███▌      | 7/20 [00:28<00:29,  2.28s/iter, val_loss=9.84e-6, val_og_loss=1.02e-5]
+
 40%|████      | 8/20 [00:28<00:25,  2.09s/iter, val_loss=9.84e-6, val_og_loss=1.02e-5]
+
 40%|████      | 8/20 [00:30<00:25,  2.09s/iter, val_loss=9.83e-6, val_og_loss=1.02e-5]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.97s/iter, val_loss=9.83e-6, val_og_loss=1.02e-5]
+
 45%|████▌     | 9/20 [00:32<00:21,  1.97s/iter, val_loss=9.83e-6, val_og_loss=1.02e-5]
+
 50%|█████     | 10/20 [00:32<00:18,  1.88s/iter, val_loss=9.83e-6, val_og_loss=1.02e-5]
+
 50%|█████     | 10/20 [00:34<00:18,  1.88s/iter, val_loss=9.83e-6, val_og_loss=1.02e-5]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.82s/iter, val_loss=9.83e-6, val_og_loss=1.02e-5]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.82s/iter, val_loss=9.81e-6, val_og_loss=1.02e-5]
+
 60%|██████    | 12/20 [00:35<00:14,  1.78s/iter, val_loss=9.81e-6, val_og_loss=1.02e-5]
+
 60%|██████    | 12/20 [00:37<00:14,  1.78s/iter, val_loss=9.8e-6, val_og_loss=1.02e-5] 
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=9.8e-6, val_og_loss=1.02e-5]
+
 65%|██████▌   | 13/20 [00:39<00:12,  1.75s/iter, val_loss=9.81e-6, val_og_loss=1.02e-5]
+
 70%|███████   | 14/20 [00:39<00:10,  1.73s/iter, val_loss=9.81e-6, val_og_loss=1.02e-5]
+
 70%|███████   | 14/20 [00:40<00:10,  1.73s/iter, val_loss=9.78e-6, val_og_loss=1.02e-5]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.72s/iter, val_loss=9.78e-6, val_og_loss=1.02e-5]
+
 75%|███████▌  | 15/20 [00:42<00:08,  1.72s/iter, val_loss=9.76e-6, val_og_loss=1.02e-5]
+
 80%|████████  | 16/20 [00:42<00:06,  1.71s/iter, val_loss=9.76e-6, val_og_loss=1.02e-5]
+
 80%|████████  | 16/20 [00:44<00:06,  1.71s/iter, val_loss=9.75e-6, val_og_loss=1.02e-5]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=9.75e-6, val_og_loss=1.02e-5]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.70s/iter, val_loss=9.75e-6, val_og_loss=1.02e-5]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.70s/iter, val_loss=9.75e-6, val_og_loss=1.02e-5]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.70s/iter, val_loss=9.74e-6, val_og_loss=1.02e-5]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=9.74e-6, val_og_loss=1.02e-5]
+
 95%|█████████▌| 19/20 [00:49<00:01,  1.69s/iter, val_loss=9.73e-6, val_og_loss=1.02e-5]
+
100%|██████████| 20/20 [00:49<00:00,  1.69s/iter, val_loss=9.73e-6, val_og_loss=1.02e-5]
100%|██████████| 20/20 [00:49<00:00,  2.46s/iter, val_loss=9.73e-6, val_og_loss=1.02e-5]
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [40:59<4:03:30, 561.93s/it]INFO:ParoQuant:Best val loss: 9.723254152049776e-06, Original val loss: 1.0232171007373836e-05
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [41:00<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [41:00<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [41:47<4:03:30, 561.93s/it]2025-12-22 15:11:17,812 - INFO - Best val loss: 9.723254152049776e-06, Original val loss: 1.0232171007373836e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.0007, val_og_loss=0.000711]
+
  5%|▌         | 1/20 [00:31<09:56, 31.38s/iter, val_loss=0.0007, val_og_loss=0.000711]
+
  5%|▌         | 1/20 [00:34<09:56, 31.38s/iter, val_loss=0.000702, val_og_loss=0.000711]
+
 10%|█         | 2/20 [00:34<04:21, 14.52s/iter, val_loss=0.000702, val_og_loss=0.000711]
+
 10%|█         | 2/20 [00:36<04:21, 14.52s/iter, val_loss=0.000703, val_og_loss=0.000711]
+
 15%|█▌        | 3/20 [00:36<02:35,  9.13s/iter, val_loss=0.000703, val_og_loss=0.000711]
+
 15%|█▌        | 3/20 [00:39<02:35,  9.13s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 20%|██        | 4/20 [00:39<01:45,  6.61s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 20%|██        | 4/20 [00:42<01:45,  6.61s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 25%|██▌       | 5/20 [00:42<01:18,  5.21s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 25%|██▌       | 5/20 [00:45<01:18,  5.21s/iter, val_loss=0.000707, val_og_loss=0.000711]
+
 30%|███       | 6/20 [00:45<01:01,  4.36s/iter, val_loss=0.000707, val_og_loss=0.000711]
+
 30%|███       | 6/20 [00:47<01:01,  4.36s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.83s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.83s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 40%|████      | 8/20 [00:50<00:41,  3.48s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 40%|████      | 8/20 [00:53<00:41,  3.48s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.24s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.24s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 50%|█████     | 10/20 [00:55<00:30,  3.08s/iter, val_loss=0.000705, val_og_loss=0.000711]
+
 50%|█████     | 10/20 [00:58<00:30,  3.08s/iter, val_loss=0.000704, val_og_loss=0.000711]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.97s/iter, val_loss=0.000704, val_og_loss=0.000711]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.97s/iter, val_loss=0.000704, val_og_loss=0.000711]
+
 60%|██████    | 12/20 [01:01<00:23,  2.90s/iter, val_loss=0.000704, val_og_loss=0.000711]
+
 60%|██████    | 12/20 [01:04<00:23,  2.90s/iter, val_loss=0.000702, val_og_loss=0.000711]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.85s/iter, val_loss=0.000702, val_og_loss=0.000711]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.85s/iter, val_loss=0.000701, val_og_loss=0.000711]
+
 70%|███████   | 14/20 [01:06<00:16,  2.81s/iter, val_loss=0.000701, val_og_loss=0.000711]
+
 70%|███████   | 14/20 [01:09<00:16,  2.81s/iter, val_loss=0.000701, val_og_loss=0.000711]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.79s/iter, val_loss=0.000701, val_og_loss=0.000711]
+
 75%|███████▌  | 15/20 [01:12<00:13,  2.79s/iter, val_loss=0.0007, val_og_loss=0.000711]  
+
 80%|████████  | 16/20 [01:12<00:11,  2.77s/iter, val_loss=0.0007, val_og_loss=0.000711]
+
 80%|████████  | 16/20 [01:14<00:11,  2.77s/iter, val_loss=0.0007, val_og_loss=0.000711]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.75s/iter, val_loss=0.0007, val_og_loss=0.000711]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.75s/iter, val_loss=0.0007, val_og_loss=0.000711]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.0007, val_og_loss=0.000711]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.74s/iter, val_loss=0.0007, val_og_loss=0.000711]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.0007, val_og_loss=0.000711]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.73s/iter, val_loss=0.000699, val_og_loss=0.000711]
+
100%|██████████| 20/20 [01:23<00:00,  2.73s/iter, val_loss=0.000699, val_og_loss=0.000711]
100%|██████████| 20/20 [01:23<00:00,  4.16s/iter, val_loss=0.000699, val_og_loss=0.000711]
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [43:10<4:03:30, 561.93s/it]INFO:ParoQuant:Best val loss: 0.0006992859416641295, Original val loss: 0.0007109999423846602
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [43:15<4:03:30, 561.93s/it]2025-12-22 15:13:28,237 - INFO - Best val loss: 0.0006992859416641295, Original val loss: 0.0007109999423846602
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.000368, val_og_loss=0.000374]
+
  5%|▌         | 1/20 [00:33<10:33, 33.34s/iter, val_loss=0.000368, val_og_loss=0.000374]
+
  5%|▌         | 1/20 [00:36<10:33, 33.34s/iter, val_loss=0.000369, val_og_loss=0.000374]
+
 10%|█         | 2/20 [00:36<04:35, 15.33s/iter, val_loss=0.000369, val_og_loss=0.000374]
+
 10%|█         | 2/20 [00:38<04:35, 15.33s/iter, val_loss=0.000371, val_og_loss=0.000374]
+
 15%|█▌        | 3/20 [00:38<02:42,  9.56s/iter, val_loss=0.000371, val_og_loss=0.000374]
+
 15%|█▌        | 3/20 [00:41<02:42,  9.56s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 20%|██        | 4/20 [00:41<01:49,  6.86s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 20%|██        | 4/20 [00:44<01:49,  6.86s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 25%|██▌       | 5/20 [00:44<01:20,  5.37s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 25%|██▌       | 5/20 [00:46<01:20,  5.37s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 30%|███       | 6/20 [00:46<01:02,  4.47s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 30%|███       | 6/20 [00:49<01:02,  4.47s/iter, val_loss=0.000374, val_og_loss=0.000374]
+
 35%|███▌      | 7/20 [00:49<00:50,  3.89s/iter, val_loss=0.000374, val_og_loss=0.000374]
+
 35%|███▌      | 7/20 [00:52<00:50,  3.89s/iter, val_loss=0.000374, val_og_loss=0.000374]
+
 40%|████      | 8/20 [00:52<00:42,  3.52s/iter, val_loss=0.000374, val_og_loss=0.000374]
+
 40%|████      | 8/20 [00:55<00:42,  3.52s/iter, val_loss=0.000374, val_og_loss=0.000374]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.27s/iter, val_loss=0.000374, val_og_loss=0.000374]
+
 45%|████▌     | 9/20 [00:57<00:35,  3.27s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 50%|█████     | 10/20 [00:57<00:31,  3.10s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 50%|█████     | 10/20 [01:00<00:31,  3.10s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.98s/iter, val_loss=0.000373, val_og_loss=0.000374]
+
 55%|█████▌    | 11/20 [01:03<00:26,  2.98s/iter, val_loss=0.000372, val_og_loss=0.000374]
+
 60%|██████    | 12/20 [01:03<00:23,  2.91s/iter, val_loss=0.000372, val_og_loss=0.000374]
+
 60%|██████    | 12/20 [01:05<00:23,  2.91s/iter, val_loss=0.000372, val_og_loss=0.000374]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.85s/iter, val_loss=0.000372, val_og_loss=0.000374]
+
 65%|██████▌   | 13/20 [01:08<00:19,  2.85s/iter, val_loss=0.000371, val_og_loss=0.000374]
+
 70%|███████   | 14/20 [01:08<00:16,  2.82s/iter, val_loss=0.000371, val_og_loss=0.000374]
+
 70%|███████   | 14/20 [01:11<00:16,  2.82s/iter, val_loss=0.00037, val_og_loss=0.000374] 
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.80s/iter, val_loss=0.00037, val_og_loss=0.000374]
+
 75%|███████▌  | 15/20 [01:14<00:13,  2.80s/iter, val_loss=0.00037, val_og_loss=0.000374]
+
 80%|████████  | 16/20 [01:14<00:11,  2.78s/iter, val_loss=0.00037, val_og_loss=0.000374]
+
 80%|████████  | 16/20 [01:16<00:11,  2.78s/iter, val_loss=0.00037, val_og_loss=0.000374]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.77s/iter, val_loss=0.00037, val_og_loss=0.000374]
+
 85%|████████▌ | 17/20 [01:19<00:08,  2.77s/iter, val_loss=0.000369, val_og_loss=0.000374]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.77s/iter, val_loss=0.000369, val_og_loss=0.000374]
+
 90%|█████████ | 18/20 [01:22<00:05,  2.77s/iter, val_loss=0.000369, val_og_loss=0.000374]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.76s/iter, val_loss=0.000369, val_og_loss=0.000374]
+
 95%|█████████▌| 19/20 [01:25<00:02,  2.76s/iter, val_loss=0.000369, val_og_loss=0.000374]
+
100%|██████████| 20/20 [01:25<00:00,  2.75s/iter, val_loss=0.000369, val_og_loss=0.000374]
100%|██████████| 20/20 [01:25<00:00,  4.26s/iter, val_loss=0.000369, val_og_loss=0.000374]
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [44:40<4:03:30, 561.93s/it]INFO:ParoQuant:Best val loss: 0.0003679670626297593, Original val loss: 0.00037409079959616065
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [44:42<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [44:42<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [45:54<4:03:30, 561.93s/it]2025-12-22 15:14:58,198 - INFO - Best val loss: 0.0003679670626297593, Original val loss: 0.00037409079959616065
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:34<?, ?iter/s, val_loss=3.24e-5, val_og_loss=3.28e-5]
+
  5%|▌         | 1/20 [00:34<11:04, 34.96s/iter, val_loss=3.24e-5, val_og_loss=3.28e-5]
+
  5%|▌         | 1/20 [00:40<11:04, 34.96s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 10%|█         | 2/20 [00:40<05:16, 17.58s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 10%|█         | 2/20 [00:45<05:16, 17.58s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 15%|█▌        | 3/20 [00:45<03:24, 12.04s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 15%|█▌        | 3/20 [00:51<03:24, 12.04s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 20%|██        | 4/20 [00:51<02:30,  9.43s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 20%|██        | 4/20 [00:56<02:30,  9.43s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 25%|██▌       | 5/20 [00:56<01:59,  7.99s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 25%|██▌       | 5/20 [01:02<01:59,  7.99s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 30%|███       | 6/20 [01:02<01:39,  7.12s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 30%|███       | 6/20 [01:07<01:39,  7.12s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 35%|███▌      | 7/20 [01:07<01:25,  6.57s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 35%|███▌      | 7/20 [01:12<01:25,  6.57s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 40%|████      | 8/20 [01:12<01:14,  6.20s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 40%|████      | 8/20 [01:18<01:14,  6.20s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 45%|████▌     | 9/20 [01:18<01:05,  5.96s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 45%|████▌     | 9/20 [01:23<01:05,  5.96s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 50%|█████     | 10/20 [01:23<00:57,  5.79s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 50%|█████     | 10/20 [01:29<00:57,  5.79s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 55%|█████▌    | 11/20 [01:29<00:51,  5.68s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 55%|█████▌    | 11/20 [01:34<00:51,  5.68s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 60%|██████    | 12/20 [01:34<00:44,  5.61s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 60%|██████    | 12/20 [01:40<00:44,  5.61s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 65%|██████▌   | 13/20 [01:40<00:38,  5.55s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 65%|██████▌   | 13/20 [01:45<00:38,  5.55s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 70%|███████   | 14/20 [01:45<00:33,  5.51s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 70%|███████   | 14/20 [01:50<00:33,  5.51s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 75%|███████▌  | 15/20 [01:50<00:27,  5.49s/iter, val_loss=3.26e-5, val_og_loss=3.28e-5]
+
 75%|███████▌  | 15/20 [01:56<00:27,  5.49s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 80%|████████  | 16/20 [01:56<00:21,  5.47s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 80%|████████  | 16/20 [02:01<00:21,  5.47s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 85%|████████▌ | 17/20 [02:01<00:16,  5.45s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 85%|████████▌ | 17/20 [02:07<00:16,  5.45s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 90%|█████████ | 18/20 [02:07<00:10,  5.45s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 90%|█████████ | 18/20 [02:12<00:10,  5.45s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 95%|█████████▌| 19/20 [02:12<00:05,  5.44s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
 95%|█████████▌| 19/20 [02:18<00:05,  5.44s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
100%|██████████| 20/20 [02:18<00:00,  5.44s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
100%|██████████| 20/20 [02:18<00:00,  6.90s/iter, val_loss=3.25e-5, val_og_loss=3.28e-5]
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [48:12<4:03:30, 561.93s/it]INFO:ParoQuant:Best val loss: 3.2406420359620824e-05, Original val loss: 3.282749094069004e-05
+
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [48:12<4:03:30, 561.93s/it]
                                                                    

Processing Blocks:  19%|█▉        | 6/32 [48:26<4:03:30, 561.93s/it]
Processing Blocks:  22%|██▏       | 7/32 [49:14<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [49:15<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [49:34<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [49:34<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [51:55<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [51:55<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [52:52<4:45:18, 684.72s/it]2025-12-22 15:18:30,260 - INFO - Best val loss: 3.2406420359620824e-05, Original val loss: 3.282749094069004e-05
+[INFO] Saving Layer 6 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 7...
+[INFO] Starting Sequential Training for Layer 7
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00145, val_og_loss=0.00169]
+
  5%|▌         | 1/20 [00:15<04:56, 15.63s/iter, val_loss=0.00145, val_og_loss=0.00169]
+
  5%|▌         | 1/20 [00:17<04:56, 15.63s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 10%|█         | 2/20 [00:17<02:13,  7.43s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 10%|█         | 2/20 [00:19<02:13,  7.43s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 15%|█▌        | 3/20 [00:19<01:21,  4.81s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 15%|█▌        | 3/20 [00:20<01:21,  4.81s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 20%|██        | 4/20 [00:20<00:57,  3.58s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 20%|██        | 4/20 [00:22<00:57,  3.58s/iter, val_loss=0.00149, val_og_loss=0.00169]
+
 25%|██▌       | 5/20 [00:22<00:43,  2.90s/iter, val_loss=0.00149, val_og_loss=0.00169]
+
 25%|██▌       | 5/20 [00:24<00:43,  2.90s/iter, val_loss=0.0015, val_og_loss=0.00169] 
+
 30%|███       | 6/20 [00:24<00:34,  2.49s/iter, val_loss=0.0015, val_og_loss=0.00169]
+
 30%|███       | 6/20 [00:25<00:34,  2.49s/iter, val_loss=0.00149, val_og_loss=0.00169]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.23s/iter, val_loss=0.00149, val_og_loss=0.00169]
+
 35%|███▌      | 7/20 [00:27<00:28,  2.23s/iter, val_loss=0.00149, val_og_loss=0.00169]
+
 40%|████      | 8/20 [00:27<00:24,  2.06s/iter, val_loss=0.00149, val_og_loss=0.00169]
+
 40%|████      | 8/20 [00:29<00:24,  2.06s/iter, val_loss=0.00149, val_og_loss=0.00169]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.94s/iter, val_loss=0.00149, val_og_loss=0.00169]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.94s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 50%|█████     | 10/20 [00:30<00:18,  1.86s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 50%|█████     | 10/20 [00:32<00:18,  1.86s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.81s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.81s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 60%|██████    | 12/20 [00:34<00:14,  1.78s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 60%|██████    | 12/20 [00:35<00:14,  1.78s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.75s/iter, val_loss=0.00148, val_og_loss=0.00169]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 70%|███████   | 14/20 [00:37<00:10,  1.74s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 70%|███████   | 14/20 [00:39<00:10,  1.74s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.72s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.72s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 80%|████████  | 16/20 [00:40<00:06,  1.71s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 80%|████████  | 16/20 [00:42<00:06,  1.71s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.70s/iter, val_loss=0.00147, val_og_loss=0.00169]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.70s/iter, val_loss=0.00146, val_og_loss=0.00169]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.70s/iter, val_loss=0.00146, val_og_loss=0.00169]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.70s/iter, val_loss=0.00146, val_og_loss=0.00169]
+
100%|██████████| 20/20 [00:47<00:00,  1.69s/iter, val_loss=0.00146, val_og_loss=0.00169]
100%|██████████| 20/20 [00:47<00:00,  2.39s/iter, val_loss=0.00146, val_og_loss=0.00169]
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [53:40<4:45:18, 684.72s/it]INFO:ParoQuant:Best val loss: 0.0014494701754301786, Original val loss: 0.0016879246104508638
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [53:42<4:45:18, 684.72s/it]2025-12-22 15:23:58,601 - INFO - Best val loss: 0.0014494701754301786, Original val loss: 0.0016879246104508638
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=0.00387, val_og_loss=0.00448]
+
  5%|▌         | 1/20 [00:11<03:29, 11.03s/iter, val_loss=0.00387, val_og_loss=0.00448]
+
  5%|▌         | 1/20 [00:12<03:29, 11.03s/iter, val_loss=0.00391, val_og_loss=0.00448]
+
 10%|█         | 2/20 [00:12<01:36,  5.36s/iter, val_loss=0.00391, val_og_loss=0.00448]
+
 10%|█         | 2/20 [00:13<01:36,  5.36s/iter, val_loss=0.00391, val_og_loss=0.00448]
+
 15%|█▌        | 3/20 [00:13<01:00,  3.56s/iter, val_loss=0.00391, val_og_loss=0.00448]
+
 15%|█▌        | 3/20 [00:15<01:00,  3.56s/iter, val_loss=0.00395, val_og_loss=0.00448]
+
 20%|██        | 4/20 [00:15<00:43,  2.71s/iter, val_loss=0.00395, val_og_loss=0.00448]
+
 20%|██        | 4/20 [00:16<00:43,  2.71s/iter, val_loss=0.00394, val_og_loss=0.00448]
+
 25%|██▌       | 5/20 [00:16<00:33,  2.23s/iter, val_loss=0.00394, val_og_loss=0.00448]
+
 25%|██▌       | 5/20 [00:18<00:33,  2.23s/iter, val_loss=0.00393, val_og_loss=0.00448]
+
 30%|███       | 6/20 [00:18<00:27,  1.95s/iter, val_loss=0.00393, val_og_loss=0.00448]
+
 30%|███       | 6/20 [00:19<00:27,  1.95s/iter, val_loss=0.00393, val_og_loss=0.00448]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.77s/iter, val_loss=0.00393, val_og_loss=0.00448]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.77s/iter, val_loss=0.00392, val_og_loss=0.00448]
+
 40%|████      | 8/20 [00:20<00:19,  1.65s/iter, val_loss=0.00392, val_og_loss=0.00448]
+
 40%|████      | 8/20 [00:22<00:19,  1.65s/iter, val_loss=0.00396, val_og_loss=0.00448]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.56s/iter, val_loss=0.00396, val_og_loss=0.00448]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.56s/iter, val_loss=0.00395, val_og_loss=0.00448]
+
 50%|█████     | 10/20 [00:23<00:15,  1.51s/iter, val_loss=0.00395, val_og_loss=0.00448]
+
 50%|█████     | 10/20 [00:25<00:15,  1.51s/iter, val_loss=0.00393, val_og_loss=0.00448]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.49s/iter, val_loss=0.00393, val_og_loss=0.00448]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.49s/iter, val_loss=0.00391, val_og_loss=0.00448]
+
 60%|██████    | 12/20 [00:26<00:11,  1.46s/iter, val_loss=0.00391, val_og_loss=0.00448]
+
 60%|██████    | 12/20 [00:27<00:11,  1.46s/iter, val_loss=0.00388, val_og_loss=0.00448]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.44s/iter, val_loss=0.00388, val_og_loss=0.00448]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.44s/iter, val_loss=0.00389, val_og_loss=0.00448]
+
 70%|███████   | 14/20 [00:29<00:08,  1.43s/iter, val_loss=0.00389, val_og_loss=0.00448]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.0039, val_og_loss=0.00448] 
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.42s/iter, val_loss=0.0039, val_og_loss=0.00448]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.42s/iter, val_loss=0.00388, val_og_loss=0.00448]
+
 80%|████████  | 16/20 [00:32<00:05,  1.41s/iter, val_loss=0.00388, val_og_loss=0.00448]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.00387, val_og_loss=0.00448]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.41s/iter, val_loss=0.00387, val_og_loss=0.00448]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.41s/iter, val_loss=0.00388, val_og_loss=0.00448]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.00388, val_og_loss=0.00448]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.00384, val_og_loss=0.00448]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.00384, val_og_loss=0.00448]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00387, val_og_loss=0.00448]
+
100%|██████████| 20/20 [00:37<00:00,  1.41s/iter, val_loss=0.00387, val_og_loss=0.00448]
100%|██████████| 20/20 [00:37<00:00,  1.88s/iter, val_loss=0.00387, val_og_loss=0.00448]
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [54:20<4:45:18, 684.72s/it]INFO:ParoQuant:Best val loss: 0.003843808313831687, Original val loss: 0.004477495327591896
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [54:21<4:45:18, 684.72s/it]2025-12-22 15:24:37,979 - INFO - Best val loss: 0.003843808313831687, Original val loss: 0.004477495327591896
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.00025, val_og_loss=0.0003]
+
  5%|▌         | 1/20 [00:10<03:27, 10.92s/iter, val_loss=0.00025, val_og_loss=0.0003]
+
  5%|▌         | 1/20 [00:12<03:27, 10.92s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 10%|█         | 2/20 [00:12<01:35,  5.32s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 10%|█         | 2/20 [00:13<01:35,  5.32s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 15%|█▌        | 3/20 [00:13<01:00,  3.53s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 15%|█▌        | 3/20 [00:15<01:00,  3.53s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 20%|██        | 4/20 [00:15<00:42,  2.69s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 20%|██        | 4/20 [00:16<00:42,  2.69s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 25%|██▌       | 5/20 [00:16<00:33,  2.23s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 25%|██▌       | 5/20 [00:17<00:33,  2.23s/iter, val_loss=0.000253, val_og_loss=0.0003]
+
 30%|███       | 6/20 [00:17<00:27,  1.94s/iter, val_loss=0.000253, val_og_loss=0.0003]
+
 30%|███       | 6/20 [00:19<00:27,  1.94s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.76s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.76s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 40%|████      | 8/20 [00:20<00:19,  1.64s/iter, val_loss=0.000252, val_og_loss=0.0003]
+
 40%|████      | 8/20 [00:22<00:19,  1.64s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.57s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.57s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 50%|█████     | 10/20 [00:23<00:15,  1.52s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 50%|█████     | 10/20 [00:24<00:15,  1.52s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.48s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.48s/iter, val_loss=0.00025, val_og_loss=0.0003] 
+
 60%|██████    | 12/20 [00:26<00:11,  1.45s/iter, val_loss=0.00025, val_og_loss=0.0003]
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.43s/iter, val_loss=0.000251, val_og_loss=0.0003]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.43s/iter, val_loss=0.00025, val_og_loss=0.0003] 
+
 70%|███████   | 14/20 [00:29<00:08,  1.43s/iter, val_loss=0.00025, val_og_loss=0.0003]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.00025, val_og_loss=0.0003]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.42s/iter, val_loss=0.00025, val_og_loss=0.0003]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.42s/iter, val_loss=0.000249, val_og_loss=0.0003]
+
 80%|████████  | 16/20 [00:31<00:05,  1.41s/iter, val_loss=0.000249, val_og_loss=0.0003]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.000249, val_og_loss=0.0003]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.41s/iter, val_loss=0.000249, val_og_loss=0.0003]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.41s/iter, val_loss=0.000248, val_og_loss=0.0003]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.000248, val_og_loss=0.0003]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.000248, val_og_loss=0.0003]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.000248, val_og_loss=0.0003]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.000248, val_og_loss=0.0003]
+
100%|██████████| 20/20 [00:37<00:00,  1.40s/iter, val_loss=0.000248, val_og_loss=0.0003]
100%|██████████| 20/20 [00:37<00:00,  1.87s/iter, val_loss=0.000248, val_og_loss=0.0003]
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [54:59<4:45:18, 684.72s/it]INFO:ParoQuant:Best val loss: 0.00024755147751420736, Original val loss: 0.00029959995299577713
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [55:01<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [55:01<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [55:35<4:45:18, 684.72s/it]2025-12-22 15:25:17,224 - INFO - Best val loss: 0.00024755147751420736, Original val loss: 0.00029959995299577713
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:17<?, ?iter/s, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
  5%|▌         | 1/20 [00:17<05:27, 17.25s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
  5%|▌         | 1/20 [00:18<05:27, 17.25s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 10%|█         | 2/20 [00:18<02:25,  8.11s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 10%|█         | 2/20 [00:20<02:25,  8.11s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 15%|█▌        | 3/20 [00:20<01:28,  5.18s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 15%|█▌        | 3/20 [00:22<01:28,  5.18s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 20%|██        | 4/20 [00:22<01:00,  3.80s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 20%|██        | 4/20 [00:24<01:00,  3.80s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 25%|██▌       | 5/20 [00:24<00:45,  3.04s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 25%|██▌       | 5/20 [00:25<00:45,  3.04s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 30%|███       | 6/20 [00:25<00:36,  2.59s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 30%|███       | 6/20 [00:27<00:36,  2.59s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 35%|███▌      | 7/20 [00:27<00:29,  2.30s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 35%|███▌      | 7/20 [00:29<00:29,  2.30s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 40%|████      | 8/20 [00:29<00:25,  2.11s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 40%|████      | 8/20 [00:30<00:25,  2.11s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.98s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 45%|████▌     | 9/20 [00:32<00:21,  1.98s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 50%|█████     | 10/20 [00:32<00:18,  1.89s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 50%|█████     | 10/20 [00:34<00:18,  1.89s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.83s/iter, val_loss=1.42e-5, val_og_loss=1.52e-5]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.83s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 60%|██████    | 12/20 [00:35<00:14,  1.79s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 60%|██████    | 12/20 [00:37<00:14,  1.79s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.76s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 65%|██████▌   | 13/20 [00:39<00:12,  1.76s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 70%|███████   | 14/20 [00:39<00:10,  1.74s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 70%|███████   | 14/20 [00:40<00:10,  1.74s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.72s/iter, val_loss=1.41e-5, val_og_loss=1.52e-5]
+
 75%|███████▌  | 15/20 [00:42<00:08,  1.72s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5] 
+
 80%|████████  | 16/20 [00:42<00:06,  1.71s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
+
 80%|████████  | 16/20 [00:44<00:06,  1.71s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.71s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
+
 85%|████████▌ | 17/20 [00:46<00:05,  1.71s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.70s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.70s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
+
 95%|█████████▌| 19/20 [00:49<00:01,  1.69s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
+
100%|██████████| 20/20 [00:49<00:00,  1.69s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
100%|██████████| 20/20 [00:49<00:00,  2.47s/iter, val_loss=1.4e-5, val_og_loss=1.52e-5]
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [56:24<4:45:18, 684.72s/it]INFO:ParoQuant:Best val loss: 1.3963317542220466e-05, Original val loss: 1.5225827155518346e-05
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [56:25<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [56:25<4:45:18, 684.72s/it]
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [57:15<4:45:18, 684.72s/it]2025-12-22 15:26:42,738 - INFO - Best val loss: 1.3963317542220466e-05, Original val loss: 1.5225827155518346e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.000724, val_og_loss=0.000737]
+
  5%|▌         | 1/20 [00:33<10:41, 33.75s/iter, val_loss=0.000724, val_og_loss=0.000737]
+
  5%|▌         | 1/20 [00:36<10:41, 33.75s/iter, val_loss=0.000726, val_og_loss=0.000737]
+
 10%|█         | 2/20 [00:36<04:38, 15.49s/iter, val_loss=0.000726, val_og_loss=0.000737]
+
 10%|█         | 2/20 [00:39<04:38, 15.49s/iter, val_loss=0.00073, val_og_loss=0.000737] 
+
 15%|█▌        | 3/20 [00:39<02:44,  9.66s/iter, val_loss=0.00073, val_og_loss=0.000737]
+
 15%|█▌        | 3/20 [00:41<02:44,  9.66s/iter, val_loss=0.00073, val_og_loss=0.000737]
+
 20%|██        | 4/20 [00:41<01:50,  6.92s/iter, val_loss=0.00073, val_og_loss=0.000737]
+
 20%|██        | 4/20 [00:44<01:50,  6.92s/iter, val_loss=0.000732, val_og_loss=0.000737]
+
 25%|██▌       | 5/20 [00:44<01:21,  5.40s/iter, val_loss=0.000732, val_og_loss=0.000737]
+
 25%|██▌       | 5/20 [00:47<01:21,  5.40s/iter, val_loss=0.000732, val_og_loss=0.000737]
+
 30%|███       | 6/20 [00:47<01:02,  4.49s/iter, val_loss=0.000732, val_og_loss=0.000737]
+
 30%|███       | 6/20 [00:50<01:02,  4.49s/iter, val_loss=0.000731, val_og_loss=0.000737]
+
 35%|███▌      | 7/20 [00:50<00:50,  3.91s/iter, val_loss=0.000731, val_og_loss=0.000737]
+
 35%|███▌      | 7/20 [00:52<00:50,  3.91s/iter, val_loss=0.00073, val_og_loss=0.000737] 
+
 40%|████      | 8/20 [00:52<00:42,  3.53s/iter, val_loss=0.00073, val_og_loss=0.000737]
+
 40%|████      | 8/20 [00:55<00:42,  3.53s/iter, val_loss=0.00073, val_og_loss=0.000737]
+
 45%|████▌     | 9/20 [00:55<00:36,  3.28s/iter, val_loss=0.00073, val_og_loss=0.000737]
+
 45%|████▌     | 9/20 [00:58<00:36,  3.28s/iter, val_loss=0.000729, val_og_loss=0.000737]
+
 50%|█████     | 10/20 [00:58<00:31,  3.11s/iter, val_loss=0.000729, val_og_loss=0.000737]
+
 50%|█████     | 10/20 [01:00<00:31,  3.11s/iter, val_loss=0.000728, val_og_loss=0.000737]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.99s/iter, val_loss=0.000728, val_og_loss=0.000737]
+
 55%|█████▌    | 11/20 [01:03<00:26,  2.99s/iter, val_loss=0.000727, val_og_loss=0.000737]
+
 60%|██████    | 12/20 [01:03<00:23,  2.91s/iter, val_loss=0.000727, val_og_loss=0.000737]
+
 60%|██████    | 12/20 [01:06<00:23,  2.91s/iter, val_loss=0.000728, val_og_loss=0.000737]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.85s/iter, val_loss=0.000728, val_og_loss=0.000737]
+
 65%|██████▌   | 13/20 [01:09<00:19,  2.85s/iter, val_loss=0.000727, val_og_loss=0.000737]
+
 70%|███████   | 14/20 [01:09<00:16,  2.81s/iter, val_loss=0.000727, val_og_loss=0.000737]
+
 70%|███████   | 14/20 [01:11<00:16,  2.81s/iter, val_loss=0.000726, val_og_loss=0.000737]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.000726, val_og_loss=0.000737]
+
 75%|███████▌  | 15/20 [01:14<00:13,  2.78s/iter, val_loss=0.000725, val_og_loss=0.000737]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.000725, val_og_loss=0.000737]
+
 80%|████████  | 16/20 [01:17<00:11,  2.76s/iter, val_loss=0.000724, val_og_loss=0.000737]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.75s/iter, val_loss=0.000724, val_og_loss=0.000737]
+
 85%|████████▌ | 17/20 [01:19<00:08,  2.75s/iter, val_loss=0.000724, val_og_loss=0.000737]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.74s/iter, val_loss=0.000724, val_og_loss=0.000737]
+
 90%|█████████ | 18/20 [01:22<00:05,  2.74s/iter, val_loss=0.000723, val_og_loss=0.000737]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.74s/iter, val_loss=0.000723, val_og_loss=0.000737]
+
 95%|█████████▌| 19/20 [01:25<00:02,  2.74s/iter, val_loss=0.000724, val_og_loss=0.000737]
+
100%|██████████| 20/20 [01:25<00:00,  2.73s/iter, val_loss=0.000724, val_og_loss=0.000737]
100%|██████████| 20/20 [01:25<00:00,  4.27s/iter, val_loss=0.000724, val_og_loss=0.000737]
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [58:40<4:45:18, 684.72s/it]INFO:ParoQuant:Best val loss: 0.0007231888012029231, Original val loss: 0.0007370597450062633
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [58:46<4:45:18, 684.72s/it]2025-12-22 15:28:58,583 - INFO - Best val loss: 0.0007231888012029231, Original val loss: 0.0007370597450062633
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:36<?, ?iter/s, val_loss=0.000409, val_og_loss=0.000416]
+
  5%|▌         | 1/20 [00:36<11:36, 36.65s/iter, val_loss=0.000409, val_og_loss=0.000416]
+
  5%|▌         | 1/20 [00:39<11:36, 36.65s/iter, val_loss=0.000411, val_og_loss=0.000416]
+
 10%|█         | 2/20 [00:39<05:00, 16.71s/iter, val_loss=0.000411, val_og_loss=0.000416]
+
 10%|█         | 2/20 [00:42<05:00, 16.71s/iter, val_loss=0.000413, val_og_loss=0.000416]
+
 15%|█▌        | 3/20 [00:42<02:55, 10.32s/iter, val_loss=0.000413, val_og_loss=0.000416]
+
 15%|█▌        | 3/20 [00:44<02:55, 10.32s/iter, val_loss=0.000415, val_og_loss=0.000416]
+
 20%|██        | 4/20 [00:44<01:57,  7.32s/iter, val_loss=0.000415, val_og_loss=0.000416]
+
 20%|██        | 4/20 [00:47<01:57,  7.32s/iter, val_loss=0.000414, val_og_loss=0.000416]
+
 25%|██▌       | 5/20 [00:47<01:24,  5.66s/iter, val_loss=0.000414, val_og_loss=0.000416]
+
 25%|██▌       | 5/20 [00:50<01:24,  5.66s/iter, val_loss=0.000416, val_og_loss=0.000416]
+
 30%|███       | 6/20 [00:50<01:05,  4.66s/iter, val_loss=0.000416, val_og_loss=0.000416]
+
 30%|███       | 6/20 [00:53<01:05,  4.66s/iter, val_loss=0.000415, val_og_loss=0.000416]
+
 35%|███▌      | 7/20 [00:53<00:52,  4.03s/iter, val_loss=0.000415, val_og_loss=0.000416]
+
 35%|███▌      | 7/20 [00:55<00:52,  4.03s/iter, val_loss=0.000415, val_og_loss=0.000416]
+
 40%|████      | 8/20 [00:55<00:43,  3.62s/iter, val_loss=0.000415, val_og_loss=0.000416]
+
 40%|████      | 8/20 [00:58<00:43,  3.62s/iter, val_loss=0.000415, val_og_loss=0.000416]
+
 45%|████▌     | 9/20 [00:58<00:36,  3.34s/iter, val_loss=0.000415, val_og_loss=0.000416]
+
 45%|████▌     | 9/20 [01:01<00:36,  3.34s/iter, val_loss=0.000414, val_og_loss=0.000416]
+
 50%|█████     | 10/20 [01:01<00:31,  3.16s/iter, val_loss=0.000414, val_og_loss=0.000416]
+
 50%|█████     | 10/20 [01:03<00:31,  3.16s/iter, val_loss=0.000414, val_og_loss=0.000416]
+
 55%|█████▌    | 11/20 [01:03<00:27,  3.03s/iter, val_loss=0.000414, val_og_loss=0.000416]
+
 55%|█████▌    | 11/20 [01:06<00:27,  3.03s/iter, val_loss=0.000412, val_og_loss=0.000416]
+
 60%|██████    | 12/20 [01:06<00:23,  2.94s/iter, val_loss=0.000412, val_og_loss=0.000416]
+
 60%|██████    | 12/20 [01:09<00:23,  2.94s/iter, val_loss=0.000412, val_og_loss=0.000416]
+
 65%|██████▌   | 13/20 [01:09<00:20,  2.87s/iter, val_loss=0.000412, val_og_loss=0.000416]
+
 65%|██████▌   | 13/20 [01:12<00:20,  2.87s/iter, val_loss=0.000412, val_og_loss=0.000416]
+
 70%|███████   | 14/20 [01:12<00:16,  2.83s/iter, val_loss=0.000412, val_og_loss=0.000416]
+
 70%|███████   | 14/20 [01:14<00:16,  2.83s/iter, val_loss=0.000411, val_og_loss=0.000416]
+
 75%|███████▌  | 15/20 [01:14<00:13,  2.80s/iter, val_loss=0.000411, val_og_loss=0.000416]
+
 75%|███████▌  | 15/20 [01:17<00:13,  2.80s/iter, val_loss=0.000411, val_og_loss=0.000416]
+
 80%|████████  | 16/20 [01:17<00:11,  2.78s/iter, val_loss=0.000411, val_og_loss=0.000416]
+
 80%|████████  | 16/20 [01:20<00:11,  2.78s/iter, val_loss=0.000411, val_og_loss=0.000416]
+
 85%|████████▌ | 17/20 [01:20<00:08,  2.76s/iter, val_loss=0.000411, val_og_loss=0.000416]
+
 85%|████████▌ | 17/20 [01:23<00:08,  2.76s/iter, val_loss=0.00041, val_og_loss=0.000416] 
+
 90%|█████████ | 18/20 [01:23<00:05,  2.75s/iter, val_loss=0.00041, val_og_loss=0.000416]
+
 90%|█████████ | 18/20 [01:25<00:05,  2.75s/iter, val_loss=0.00041, val_og_loss=0.000416]
+
 95%|█████████▌| 19/20 [01:25<00:02,  2.74s/iter, val_loss=0.00041, val_og_loss=0.000416]
+
 95%|█████████▌| 19/20 [01:28<00:02,  2.74s/iter, val_loss=0.00041, val_og_loss=0.000416]
+
100%|██████████| 20/20 [01:28<00:00,  2.73s/iter, val_loss=0.00041, val_og_loss=0.000416]
100%|██████████| 20/20 [01:28<00:00,  4.42s/iter, val_loss=0.00041, val_og_loss=0.000416]
+
                                                                    

Processing Blocks:  22%|██▏       | 7/32 [1:00:14<4:45:18, 684.72s/it]INFO:ParoQuant:Best val loss: 0.00040890724631026387, Original val loss: 0.0004163355624768883
+
                                                                      

Processing Blocks:  22%|██▏       | 7/32 [1:00:16<4:45:18, 684.72s/it]
                                                                      

Processing Blocks:  22%|██▏       | 7/32 [1:00:16<4:45:18, 684.72s/it]
                                                                      

Processing Blocks:  22%|██▏       | 7/32 [1:01:29<4:45:18, 684.72s/it]2025-12-22 15:30:32,537 - INFO - Best val loss: 0.00040890724631026387, Original val loss: 0.0004163355624768883
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:36<?, ?iter/s, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
  5%|▌         | 1/20 [00:36<11:32, 36.44s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
  5%|▌         | 1/20 [00:41<11:32, 36.44s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 10%|█         | 2/20 [00:41<05:27, 18.20s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 10%|█         | 2/20 [00:47<05:27, 18.20s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 15%|█▌        | 3/20 [00:47<03:30, 12.37s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 15%|█▌        | 3/20 [00:52<03:30, 12.37s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 20%|██        | 4/20 [00:52<02:34,  9.63s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 20%|██        | 4/20 [00:58<02:34,  9.63s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 25%|██▌       | 5/20 [00:58<02:01,  8.12s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 25%|██▌       | 5/20 [01:03<02:01,  8.12s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 30%|███       | 6/20 [01:03<01:40,  7.21s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 30%|███       | 6/20 [01:09<01:40,  7.21s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 35%|███▌      | 7/20 [01:09<01:26,  6.63s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 35%|███▌      | 7/20 [01:14<01:26,  6.63s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 40%|████      | 8/20 [01:14<01:15,  6.25s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 40%|████      | 8/20 [01:19<01:15,  6.25s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 45%|████▌     | 9/20 [01:19<01:05,  5.99s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 45%|████▌     | 9/20 [01:25<01:05,  5.99s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 50%|█████     | 10/20 [01:25<00:58,  5.82s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 50%|█████     | 10/20 [01:30<00:58,  5.82s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 55%|█████▌    | 11/20 [01:30<00:51,  5.71s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 55%|█████▌    | 11/20 [01:36<00:51,  5.71s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 60%|██████    | 12/20 [01:36<00:45,  5.63s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 60%|██████    | 12/20 [01:41<00:45,  5.63s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 65%|██████▌   | 13/20 [01:41<00:38,  5.57s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 65%|██████▌   | 13/20 [01:47<00:38,  5.57s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 70%|███████   | 14/20 [01:47<00:33,  5.52s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 70%|███████   | 14/20 [01:52<00:33,  5.52s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 75%|███████▌  | 15/20 [01:52<00:27,  5.50s/iter, val_loss=3.84e-5, val_og_loss=3.86e-5]
+
 75%|███████▌  | 15/20 [01:57<00:27,  5.50s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 80%|████████  | 16/20 [01:57<00:21,  5.48s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 80%|████████  | 16/20 [02:03<00:21,  5.48s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 85%|████████▌ | 17/20 [02:03<00:16,  5.46s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 85%|████████▌ | 17/20 [02:08<00:16,  5.46s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 90%|█████████ | 18/20 [02:08<00:10,  5.45s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 90%|█████████ | 18/20 [02:14<00:10,  5.45s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 95%|█████████▌| 19/20 [02:14<00:05,  5.45s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
 95%|█████████▌| 19/20 [02:19<00:05,  5.45s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
100%|██████████| 20/20 [02:19<00:00,  5.44s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
100%|██████████| 20/20 [02:19<00:00,  6.99s/iter, val_loss=3.83e-5, val_og_loss=3.86e-5]
+
                                                                      

Processing Blocks:  22%|██▏       | 7/32 [1:03:48<4:45:18, 684.72s/it]INFO:ParoQuant:Best val loss: 3.825501335086301e-05, Original val loss: 3.862292942358181e-05
+
                                                                      

Processing Blocks:  22%|██▏       | 7/32 [1:03:49<4:45:18, 684.72s/it]
                                                                      

Processing Blocks:  22%|██▏       | 7/32 [1:04:00<4:45:18, 684.72s/it]
Processing Blocks:  25%|██▌       | 8/32 [1:04:39<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:04:40<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:04:59<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:04:59<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:07:24<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:07:24<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:08:21<5:04:30, 761.29s/it]2025-12-22 15:34:06,754 - INFO - Best val loss: 3.825501335086301e-05, Original val loss: 3.862292942358181e-05
+[INFO] Saving Layer 7 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 8...
+[INFO] Starting Sequential Training for Layer 8
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00171, val_og_loss=0.00206]
+
  5%|▌         | 1/20 [00:15<04:52, 15.42s/iter, val_loss=0.00171, val_og_loss=0.00206]
+
  5%|▌         | 1/20 [00:17<04:52, 15.42s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 10%|█         | 2/20 [00:17<02:12,  7.35s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 10%|█         | 2/20 [00:18<02:12,  7.35s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 15%|█▌        | 3/20 [00:18<01:20,  4.76s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 15%|█▌        | 3/20 [00:20<01:20,  4.76s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 20%|██        | 4/20 [00:20<00:56,  3.55s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 20%|██        | 4/20 [00:22<00:56,  3.55s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 25%|██▌       | 5/20 [00:22<00:43,  2.88s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 25%|██▌       | 5/20 [00:23<00:43,  2.88s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 30%|███       | 6/20 [00:23<00:34,  2.47s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 30%|███       | 6/20 [00:25<00:34,  2.47s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.21s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 35%|███▌      | 7/20 [00:27<00:28,  2.21s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 40%|████      | 8/20 [00:27<00:24,  2.05s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 40%|████      | 8/20 [00:28<00:24,  2.05s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 45%|████▌     | 9/20 [00:28<00:21,  1.94s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.94s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 50%|█████     | 10/20 [00:30<00:18,  1.87s/iter, val_loss=0.00176, val_og_loss=0.00206]
+
 50%|█████     | 10/20 [00:32<00:18,  1.87s/iter, val_loss=0.00175, val_og_loss=0.00206]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.81s/iter, val_loss=0.00175, val_og_loss=0.00206]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.81s/iter, val_loss=0.00175, val_og_loss=0.00206]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=0.00175, val_og_loss=0.00206]
+
 60%|██████    | 12/20 [00:35<00:14,  1.77s/iter, val_loss=0.00175, val_og_loss=0.00206]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.75s/iter, val_loss=0.00175, val_og_loss=0.00206]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 70%|███████   | 14/20 [00:37<00:10,  1.73s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 70%|███████   | 14/20 [00:39<00:10,  1.73s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.72s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.72s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 80%|████████  | 16/20 [00:40<00:06,  1.71s/iter, val_loss=0.00174, val_og_loss=0.00206]
+
 80%|████████  | 16/20 [00:42<00:06,  1.71s/iter, val_loss=0.00173, val_og_loss=0.00206]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.71s/iter, val_loss=0.00173, val_og_loss=0.00206]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.71s/iter, val_loss=0.00173, val_og_loss=0.00206]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.70s/iter, val_loss=0.00173, val_og_loss=0.00206]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.70s/iter, val_loss=0.00173, val_og_loss=0.00206]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=0.00173, val_og_loss=0.00206]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=0.00173, val_og_loss=0.00206]
+
100%|██████████| 20/20 [00:47<00:00,  1.69s/iter, val_loss=0.00173, val_og_loss=0.00206]
100%|██████████| 20/20 [00:47<00:00,  2.38s/iter, val_loss=0.00173, val_og_loss=0.00206]
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:09:08<5:04:30, 761.29s/it]INFO:ParoQuant:Best val loss: 0.0017065233550965786, Original val loss: 0.002062107902020216
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:09:10<5:04:30, 761.29s/it]2025-12-22 15:39:26,554 - INFO - Best val loss: 0.0017065233550965786, Original val loss: 0.002062107902020216
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=0.00457, val_og_loss=0.0054]
+
  5%|▌         | 1/20 [00:11<03:29, 11.03s/iter, val_loss=0.00457, val_og_loss=0.0054]
+
  5%|▌         | 1/20 [00:12<03:29, 11.03s/iter, val_loss=0.00465, val_og_loss=0.0054]
+
 10%|█         | 2/20 [00:12<01:36,  5.38s/iter, val_loss=0.00465, val_og_loss=0.0054]
+
 10%|█         | 2/20 [00:13<01:36,  5.38s/iter, val_loss=0.00468, val_og_loss=0.0054]
+
 15%|█▌        | 3/20 [00:13<01:00,  3.56s/iter, val_loss=0.00468, val_og_loss=0.0054]
+
 15%|█▌        | 3/20 [00:15<01:00,  3.56s/iter, val_loss=0.00471, val_og_loss=0.0054]
+
 20%|██        | 4/20 [00:15<00:43,  2.71s/iter, val_loss=0.00471, val_og_loss=0.0054]
+
 20%|██        | 4/20 [00:16<00:43,  2.71s/iter, val_loss=0.00474, val_og_loss=0.0054]
+
 25%|██▌       | 5/20 [00:16<00:33,  2.23s/iter, val_loss=0.00474, val_og_loss=0.0054]
+
 25%|██▌       | 5/20 [00:18<00:33,  2.23s/iter, val_loss=0.00466, val_og_loss=0.0054]
+
 30%|███       | 6/20 [00:18<00:27,  1.95s/iter, val_loss=0.00466, val_og_loss=0.0054]
+
 30%|███       | 6/20 [00:19<00:27,  1.95s/iter, val_loss=0.00469, val_og_loss=0.0054]
+
 35%|███▌      | 7/20 [00:19<00:23,  1.77s/iter, val_loss=0.00469, val_og_loss=0.0054]
+
 35%|███▌      | 7/20 [00:20<00:23,  1.77s/iter, val_loss=0.00467, val_og_loss=0.0054]
+
 40%|████      | 8/20 [00:20<00:19,  1.65s/iter, val_loss=0.00467, val_og_loss=0.0054]
+
 40%|████      | 8/20 [00:22<00:19,  1.65s/iter, val_loss=0.00466, val_og_loss=0.0054]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.57s/iter, val_loss=0.00466, val_og_loss=0.0054]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.57s/iter, val_loss=0.00466, val_og_loss=0.0054]
+
 50%|█████     | 10/20 [00:23<00:15,  1.52s/iter, val_loss=0.00466, val_og_loss=0.0054]
+
 50%|█████     | 10/20 [00:25<00:15,  1.52s/iter, val_loss=0.00468, val_og_loss=0.0054]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.49s/iter, val_loss=0.00468, val_og_loss=0.0054]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.49s/iter, val_loss=0.00464, val_og_loss=0.0054]
+
 60%|██████    | 12/20 [00:26<00:11,  1.46s/iter, val_loss=0.00464, val_og_loss=0.0054]
+
 60%|██████    | 12/20 [00:27<00:11,  1.46s/iter, val_loss=0.00468, val_og_loss=0.0054]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.44s/iter, val_loss=0.00468, val_og_loss=0.0054]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.44s/iter, val_loss=0.00467, val_og_loss=0.0054]
+
 70%|███████   | 14/20 [00:29<00:08,  1.43s/iter, val_loss=0.00467, val_og_loss=0.0054]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.00461, val_og_loss=0.0054]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.42s/iter, val_loss=0.00461, val_og_loss=0.0054]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.42s/iter, val_loss=0.00462, val_og_loss=0.0054]
+
 80%|████████  | 16/20 [00:32<00:05,  1.41s/iter, val_loss=0.00462, val_og_loss=0.0054]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.0046, val_og_loss=0.0054] 
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.42s/iter, val_loss=0.0046, val_og_loss=0.0054]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.42s/iter, val_loss=0.0046, val_og_loss=0.0054]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.41s/iter, val_loss=0.0046, val_og_loss=0.0054]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.41s/iter, val_loss=0.00457, val_og_loss=0.0054]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.41s/iter, val_loss=0.00457, val_og_loss=0.0054]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.41s/iter, val_loss=0.00457, val_og_loss=0.0054]
+
100%|██████████| 20/20 [00:37<00:00,  1.40s/iter, val_loss=0.00457, val_og_loss=0.0054]
100%|██████████| 20/20 [00:37<00:00,  1.88s/iter, val_loss=0.00457, val_og_loss=0.0054]
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:09:48<5:04:30, 761.29s/it]INFO:ParoQuant:Best val loss: 0.004570912569761276, Original val loss: 0.005400912836194038
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:09:49<5:04:30, 761.29s/it]2025-12-22 15:40:06,186 - INFO - Best val loss: 0.004570912569761276, Original val loss: 0.005400912836194038
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:09<?, ?iter/s, val_loss=0.000304, val_og_loss=0.000384]
+
  5%|▌         | 1/20 [00:09<03:08,  9.94s/iter, val_loss=0.000304, val_og_loss=0.000384]
+
  5%|▌         | 1/20 [00:11<03:08,  9.94s/iter, val_loss=0.000306, val_og_loss=0.000384]
+
 10%|█         | 2/20 [00:11<01:28,  4.92s/iter, val_loss=0.000306, val_og_loss=0.000384]
+
 10%|█         | 2/20 [00:12<01:28,  4.92s/iter, val_loss=0.000309, val_og_loss=0.000384]
+
 15%|█▌        | 3/20 [00:12<00:56,  3.31s/iter, val_loss=0.000309, val_og_loss=0.000384]
+
 15%|█▌        | 3/20 [00:14<00:56,  3.31s/iter, val_loss=0.000309, val_og_loss=0.000384]
+
 20%|██        | 4/20 [00:14<00:40,  2.55s/iter, val_loss=0.000309, val_og_loss=0.000384]
+
 20%|██        | 4/20 [00:15<00:40,  2.55s/iter, val_loss=0.000308, val_og_loss=0.000384]
+
 25%|██▌       | 5/20 [00:15<00:32,  2.15s/iter, val_loss=0.000308, val_og_loss=0.000384]
+
 25%|██▌       | 5/20 [00:16<00:32,  2.15s/iter, val_loss=0.000308, val_og_loss=0.000384]
+
 30%|███       | 6/20 [00:16<00:26,  1.89s/iter, val_loss=0.000308, val_og_loss=0.000384]
+
 30%|███       | 6/20 [00:18<00:26,  1.89s/iter, val_loss=0.000308, val_og_loss=0.000384]
+
 35%|███▌      | 7/20 [00:18<00:22,  1.73s/iter, val_loss=0.000308, val_og_loss=0.000384]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.73s/iter, val_loss=0.000309, val_og_loss=0.000384]
+
 40%|████      | 8/20 [00:19<00:19,  1.62s/iter, val_loss=0.000309, val_og_loss=0.000384]
+
 40%|████      | 8/20 [00:21<00:19,  1.62s/iter, val_loss=0.000309, val_og_loss=0.000384]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.55s/iter, val_loss=0.000309, val_og_loss=0.000384]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.55s/iter, val_loss=0.000308, val_og_loss=0.000384]
+
 50%|█████     | 10/20 [00:22<00:15,  1.51s/iter, val_loss=0.000308, val_og_loss=0.000384]
+
 50%|█████     | 10/20 [00:23<00:15,  1.51s/iter, val_loss=0.000307, val_og_loss=0.000384]
+
 55%|█████▌    | 11/20 [00:23<00:13,  1.46s/iter, val_loss=0.000307, val_og_loss=0.000384]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.46s/iter, val_loss=0.000306, val_og_loss=0.000384]
+
 60%|██████    | 12/20 [00:25<00:11,  1.44s/iter, val_loss=0.000306, val_og_loss=0.000384]
+
 60%|██████    | 12/20 [00:26<00:11,  1.44s/iter, val_loss=0.000306, val_og_loss=0.000384]
+
 65%|██████▌   | 13/20 [00:26<00:10,  1.43s/iter, val_loss=0.000306, val_og_loss=0.000384]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.43s/iter, val_loss=0.000306, val_og_loss=0.000384]
+
 70%|███████   | 14/20 [00:28<00:08,  1.42s/iter, val_loss=0.000306, val_og_loss=0.000384]
+
 70%|███████   | 14/20 [00:29<00:08,  1.42s/iter, val_loss=0.000305, val_og_loss=0.000384]
+
 75%|███████▌  | 15/20 [00:29<00:07,  1.42s/iter, val_loss=0.000305, val_og_loss=0.000384]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.42s/iter, val_loss=0.000305, val_og_loss=0.000384]
+
 80%|████████  | 16/20 [00:30<00:05,  1.41s/iter, val_loss=0.000305, val_og_loss=0.000384]
+
 80%|████████  | 16/20 [00:32<00:05,  1.41s/iter, val_loss=0.000304, val_og_loss=0.000384]
+
 85%|████████▌ | 17/20 [00:32<00:04,  1.40s/iter, val_loss=0.000304, val_og_loss=0.000384]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.40s/iter, val_loss=0.000304, val_og_loss=0.000384]
+
 90%|█████████ | 18/20 [00:33<00:02,  1.40s/iter, val_loss=0.000304, val_og_loss=0.000384]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.000304, val_og_loss=0.000384]
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.40s/iter, val_loss=0.000304, val_og_loss=0.000384]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.000304, val_og_loss=0.000384]
+
100%|██████████| 20/20 [00:36<00:00,  1.41s/iter, val_loss=0.000304, val_og_loss=0.000384]
100%|██████████| 20/20 [00:36<00:00,  1.83s/iter, val_loss=0.000304, val_og_loss=0.000384]
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:10:26<5:04:30, 761.29s/it]INFO:ParoQuant:Best val loss: 0.00030388188315555453, Original val loss: 0.00038384884828701615
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:10:28<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:10:28<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:11:01<5:04:30, 761.29s/it]2025-12-22 15:40:44,462 - INFO - Best val loss: 0.00030388188315555453, Original val loss: 0.00038384884828701615
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
  5%|▌         | 1/20 [00:15<04:56, 15.59s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
  5%|▌         | 1/20 [00:17<04:56, 15.59s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
 10%|█         | 2/20 [00:17<02:13,  7.42s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
 10%|█         | 2/20 [00:18<02:13,  7.42s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 15%|█▌        | 3/20 [00:18<01:21,  4.80s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 15%|█▌        | 3/20 [00:20<01:21,  4.80s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 20%|██        | 4/20 [00:20<00:57,  3.57s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 20%|██        | 4/20 [00:22<00:57,  3.57s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 25%|██▌       | 5/20 [00:22<00:43,  2.89s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 25%|██▌       | 5/20 [00:24<00:43,  2.89s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 30%|███       | 6/20 [00:24<00:34,  2.49s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 30%|███       | 6/20 [00:25<00:34,  2.49s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5] 
+
 35%|███▌      | 7/20 [00:25<00:28,  2.22s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
 35%|███▌      | 7/20 [00:27<00:28,  2.22s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 40%|████      | 8/20 [00:27<00:24,  2.05s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 40%|████      | 8/20 [00:29<00:24,  2.05s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5] 
+
 45%|████▌     | 9/20 [00:29<00:21,  1.94s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.94s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.86s/iter, val_loss=1.81e-5, val_og_loss=1.92e-5]
+
 50%|█████     | 10/20 [00:32<00:18,  1.86s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5] 
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.81s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.81s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
 60%|██████    | 12/20 [00:35<00:14,  1.77s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.75s/iter, val_loss=1.8e-5, val_og_loss=1.92e-5]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 70%|███████   | 14/20 [00:39<00:10,  1.72s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.71s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.71s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.71s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.70s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.70s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.70s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.70s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
100%|██████████| 20/20 [00:47<00:00,  1.70s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
100%|██████████| 20/20 [00:47<00:00,  2.38s/iter, val_loss=1.79e-5, val_og_loss=1.92e-5]
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:11:49<5:04:30, 761.29s/it]INFO:ParoQuant:Best val loss: 1.7870219380711205e-05, Original val loss: 1.9217099179513752e-05
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:11:50<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:11:50<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:12:41<5:04:30, 761.29s/it]2025-12-22 15:42:07,473 - INFO - Best val loss: 1.7870219380711205e-05, Original val loss: 1.9217099179513752e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:32<?, ?iter/s, val_loss=0.000788, val_og_loss=0.000805]
+
  5%|▌         | 1/20 [00:32<10:22, 32.75s/iter, val_loss=0.000788, val_og_loss=0.000805]
+
  5%|▌         | 1/20 [00:35<10:22, 32.75s/iter, val_loss=0.000791, val_og_loss=0.000805]
+
 10%|█         | 2/20 [00:35<04:31, 15.09s/iter, val_loss=0.000791, val_og_loss=0.000805]
+
 10%|█         | 2/20 [00:38<04:31, 15.09s/iter, val_loss=0.000794, val_og_loss=0.000805]
+
 15%|█▌        | 3/20 [00:38<02:40,  9.44s/iter, val_loss=0.000794, val_og_loss=0.000805]
+
 15%|█▌        | 3/20 [00:40<02:40,  9.44s/iter, val_loss=0.000795, val_og_loss=0.000805]
+
 20%|██        | 4/20 [00:40<01:48,  6.78s/iter, val_loss=0.000795, val_og_loss=0.000805]
+
 20%|██        | 4/20 [00:43<01:48,  6.78s/iter, val_loss=0.000795, val_og_loss=0.000805]
+
 25%|██▌       | 5/20 [00:43<01:19,  5.32s/iter, val_loss=0.000795, val_og_loss=0.000805]
+
 25%|██▌       | 5/20 [00:46<01:19,  5.32s/iter, val_loss=0.000795, val_og_loss=0.000805]
+
 30%|███       | 6/20 [00:46<01:02,  4.43s/iter, val_loss=0.000795, val_og_loss=0.000805]
+
 30%|███       | 6/20 [00:49<01:02,  4.43s/iter, val_loss=0.000794, val_og_loss=0.000805]
+
 35%|███▌      | 7/20 [00:49<00:50,  3.87s/iter, val_loss=0.000794, val_og_loss=0.000805]
+
 35%|███▌      | 7/20 [00:51<00:50,  3.87s/iter, val_loss=0.000794, val_og_loss=0.000805]
+
 40%|████      | 8/20 [00:51<00:42,  3.50s/iter, val_loss=0.000794, val_og_loss=0.000805]
+
 40%|████      | 8/20 [00:54<00:42,  3.50s/iter, val_loss=0.000794, val_og_loss=0.000805]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.26s/iter, val_loss=0.000794, val_og_loss=0.000805]
+
 45%|████▌     | 9/20 [00:57<00:35,  3.26s/iter, val_loss=0.000793, val_og_loss=0.000805]
+
 50%|█████     | 10/20 [00:57<00:30,  3.10s/iter, val_loss=0.000793, val_og_loss=0.000805]
+
 50%|█████     | 10/20 [00:59<00:30,  3.10s/iter, val_loss=0.000793, val_og_loss=0.000805]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.98s/iter, val_loss=0.000793, val_og_loss=0.000805]
+
 55%|█████▌    | 11/20 [01:02<00:26,  2.98s/iter, val_loss=0.000792, val_og_loss=0.000805]
+
 60%|██████    | 12/20 [01:02<00:23,  2.90s/iter, val_loss=0.000792, val_og_loss=0.000805]
+
 60%|██████    | 12/20 [01:05<00:23,  2.90s/iter, val_loss=0.000791, val_og_loss=0.000805]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.84s/iter, val_loss=0.000791, val_og_loss=0.000805]
+
 65%|██████▌   | 13/20 [01:08<00:19,  2.84s/iter, val_loss=0.00079, val_og_loss=0.000805] 
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.00079, val_og_loss=0.000805]
+
 70%|███████   | 14/20 [01:10<00:16,  2.80s/iter, val_loss=0.000789, val_og_loss=0.000805]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.78s/iter, val_loss=0.000789, val_og_loss=0.000805]
+
 75%|███████▌  | 15/20 [01:13<00:13,  2.78s/iter, val_loss=0.000789, val_og_loss=0.000805]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.000789, val_og_loss=0.000805]
+
 80%|████████  | 16/20 [01:16<00:11,  2.76s/iter, val_loss=0.000789, val_og_loss=0.000805]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.75s/iter, val_loss=0.000789, val_og_loss=0.000805]
+
 85%|████████▌ | 17/20 [01:18<00:08,  2.75s/iter, val_loss=0.000788, val_og_loss=0.000805]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.000788, val_og_loss=0.000805]
+
 90%|█████████ | 18/20 [01:21<00:05,  2.74s/iter, val_loss=0.000788, val_og_loss=0.000805]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.000788, val_og_loss=0.000805]
+
 95%|█████████▌| 19/20 [01:24<00:02,  2.73s/iter, val_loss=0.000788, val_og_loss=0.000805]
+
100%|██████████| 20/20 [01:24<00:00,  2.72s/iter, val_loss=0.000788, val_og_loss=0.000805]
100%|██████████| 20/20 [01:24<00:00,  4.22s/iter, val_loss=0.000788, val_og_loss=0.000805]
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:14:05<5:04:30, 761.29s/it]INFO:ParoQuant:Best val loss: 0.0007876867312006652, Original val loss: 0.0008049422758631408
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:14:11<5:04:30, 761.29s/it]2025-12-22 15:44:23,530 - INFO - Best val loss: 0.0007876867312006652, Original val loss: 0.0008049422758631408
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:38<?, ?iter/s, val_loss=0.000436, val_og_loss=0.000445]
+
  5%|▌         | 1/20 [00:38<12:19, 38.92s/iter, val_loss=0.000436, val_og_loss=0.000445]
+
  5%|▌         | 1/20 [00:41<12:19, 38.92s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 10%|█         | 2/20 [00:41<05:17, 17.63s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 10%|█         | 2/20 [00:44<05:17, 17.63s/iter, val_loss=0.00044, val_og_loss=0.000445] 
+
 15%|█▌        | 3/20 [00:44<03:03, 10.82s/iter, val_loss=0.00044, val_og_loss=0.000445]
+
 15%|█▌        | 3/20 [00:47<03:03, 10.82s/iter, val_loss=0.000441, val_og_loss=0.000445]
+
 20%|██        | 4/20 [00:47<02:01,  7.62s/iter, val_loss=0.000441, val_og_loss=0.000445]
+
 20%|██        | 4/20 [00:49<02:01,  7.62s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 25%|██▌       | 5/20 [00:49<01:27,  5.85s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 25%|██▌       | 5/20 [00:52<01:27,  5.85s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 30%|███       | 6/20 [00:52<01:06,  4.78s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 30%|███       | 6/20 [00:55<01:06,  4.78s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 35%|███▌      | 7/20 [00:55<00:53,  4.12s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 35%|███▌      | 7/20 [00:57<00:53,  4.12s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 40%|████      | 8/20 [00:57<00:44,  3.67s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 40%|████      | 8/20 [01:00<00:44,  3.67s/iter, val_loss=0.000443, val_og_loss=0.000445]
+
 45%|████▌     | 9/20 [01:00<00:37,  3.37s/iter, val_loss=0.000443, val_og_loss=0.000445]
+
 45%|████▌     | 9/20 [01:03<00:37,  3.37s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 50%|█████     | 10/20 [01:03<00:31,  3.17s/iter, val_loss=0.000442, val_og_loss=0.000445]
+
 50%|█████     | 10/20 [01:06<00:31,  3.17s/iter, val_loss=0.000441, val_og_loss=0.000445]
+
 55%|█████▌    | 11/20 [01:06<00:27,  3.03s/iter, val_loss=0.000441, val_og_loss=0.000445]
+
 55%|█████▌    | 11/20 [01:08<00:27,  3.03s/iter, val_loss=0.000441, val_og_loss=0.000445]
+
 60%|██████    | 12/20 [01:08<00:23,  2.93s/iter, val_loss=0.000441, val_og_loss=0.000445]
+
 60%|██████    | 12/20 [01:11<00:23,  2.93s/iter, val_loss=0.00044, val_og_loss=0.000445] 
+
 65%|██████▌   | 13/20 [01:11<00:20,  2.86s/iter, val_loss=0.00044, val_og_loss=0.000445]
+
 65%|██████▌   | 13/20 [01:14<00:20,  2.86s/iter, val_loss=0.000439, val_og_loss=0.000445]
+
 70%|███████   | 14/20 [01:14<00:16,  2.82s/iter, val_loss=0.000439, val_og_loss=0.000445]
+
 70%|███████   | 14/20 [01:16<00:16,  2.82s/iter, val_loss=0.000439, val_og_loss=0.000445]
+
 75%|███████▌  | 15/20 [01:16<00:13,  2.79s/iter, val_loss=0.000439, val_og_loss=0.000445]
+
 75%|███████▌  | 15/20 [01:19<00:13,  2.79s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 80%|████████  | 16/20 [01:19<00:11,  2.76s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 80%|████████  | 16/20 [01:22<00:11,  2.76s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 85%|████████▌ | 17/20 [01:22<00:08,  2.75s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 85%|████████▌ | 17/20 [01:25<00:08,  2.75s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 90%|█████████ | 18/20 [01:25<00:05,  2.74s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 90%|█████████ | 18/20 [01:27<00:05,  2.74s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 95%|█████████▌| 19/20 [01:27<00:02,  2.74s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
 95%|█████████▌| 19/20 [01:30<00:02,  2.74s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
100%|██████████| 20/20 [01:30<00:00,  2.73s/iter, val_loss=0.000438, val_og_loss=0.000445]
100%|██████████| 20/20 [01:30<00:00,  4.53s/iter, val_loss=0.000438, val_og_loss=0.000445]
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:15:42<5:04:30, 761.29s/it]INFO:ParoQuant:Best val loss: 0.000435671943705529, Original val loss: 0.0004447517858352512
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:15:43<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:15:43<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:16:43<5:04:30, 761.29s/it]2025-12-22 15:46:00,135 - INFO - Best val loss: 0.000435671943705529, Original val loss: 0.0004447517858352512
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=4.28e-5, val_og_loss=4.3e-5]
+
  5%|▌         | 1/20 [00:31<10:05, 31.87s/iter, val_loss=4.28e-5, val_og_loss=4.3e-5]
+
  5%|▌         | 1/20 [00:37<10:05, 31.87s/iter, val_loss=4.28e-5, val_og_loss=4.3e-5]
+
 10%|█         | 2/20 [00:37<04:53, 16.31s/iter, val_loss=4.28e-5, val_og_loss=4.3e-5]
+
 10%|█         | 2/20 [00:42<04:53, 16.31s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 15%|█▌        | 3/20 [00:42<03:12, 11.33s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 15%|█▌        | 3/20 [00:48<03:12, 11.33s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5] 
+
 20%|██        | 4/20 [00:48<02:23,  9.00s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 20%|██        | 4/20 [00:53<02:23,  9.00s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 25%|██▌       | 5/20 [00:53<01:55,  7.71s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 25%|██▌       | 5/20 [00:58<01:55,  7.71s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 30%|███       | 6/20 [00:58<01:37,  6.93s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 30%|███       | 6/20 [01:04<01:37,  6.93s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 35%|███▌      | 7/20 [01:04<01:23,  6.44s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 35%|███▌      | 7/20 [01:09<01:23,  6.44s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 40%|████      | 8/20 [01:09<01:13,  6.12s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 40%|████      | 8/20 [01:15<01:13,  6.12s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 45%|████▌     | 9/20 [01:15<01:04,  5.90s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 45%|████▌     | 9/20 [01:20<01:04,  5.90s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 50%|█████     | 10/20 [01:20<00:57,  5.75s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 50%|█████     | 10/20 [01:26<00:57,  5.75s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 55%|█████▌    | 11/20 [01:26<00:50,  5.65s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 55%|█████▌    | 11/20 [01:31<00:50,  5.65s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 60%|██████    | 12/20 [01:31<00:44,  5.58s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 60%|██████    | 12/20 [01:36<00:44,  5.58s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 65%|██████▌   | 13/20 [01:36<00:38,  5.53s/iter, val_loss=4.3e-5, val_og_loss=4.3e-5]
+
 65%|██████▌   | 13/20 [01:42<00:38,  5.53s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 70%|███████   | 14/20 [01:42<00:32,  5.49s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 70%|███████   | 14/20 [01:47<00:32,  5.49s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 75%|███████▌  | 15/20 [01:47<00:27,  5.47s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 75%|███████▌  | 15/20 [01:53<00:27,  5.47s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 80%|████████  | 16/20 [01:53<00:21,  5.46s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 80%|████████  | 16/20 [01:58<00:21,  5.46s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 85%|████████▌ | 17/20 [01:58<00:16,  5.44s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 85%|████████▌ | 17/20 [02:03<00:16,  5.44s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 90%|█████████ | 18/20 [02:03<00:10,  5.44s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 90%|█████████ | 18/20 [02:09<00:10,  5.44s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 95%|█████████▌| 19/20 [02:09<00:05,  5.44s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
 95%|█████████▌| 19/20 [02:14<00:05,  5.44s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
100%|██████████| 20/20 [02:14<00:00,  5.43s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
100%|██████████| 20/20 [02:14<00:00,  6.74s/iter, val_loss=4.29e-5, val_og_loss=4.3e-5]
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:18:58<5:04:30, 761.29s/it]INFO:ParoQuant:Best val loss: 4.2767031118273735e-05, Original val loss: 4.303385503590107e-05
+
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:18:59<5:04:30, 761.29s/it]
                                                                      

Processing Blocks:  25%|██▌       | 8/32 [1:19:08<5:04:30, 761.29s/it]
Processing Blocks:  28%|██▊       | 9/32 [1:19:46<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:19:47<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:20:05<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:20:05<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:22:17<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:22:17<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:23:15<5:09:14, 806.72s/it]2025-12-22 15:49:16,642 - INFO - Best val loss: 4.2767031118273735e-05, Original val loss: 4.303385503590107e-05
+[INFO] Saving Layer 8 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 9...
+[INFO] Starting Sequential Training for Layer 9
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.00171, val_og_loss=0.00207]
+
  5%|▌         | 1/20 [00:14<04:26, 14.01s/iter, val_loss=0.00171, val_og_loss=0.00207]
+
  5%|▌         | 1/20 [00:15<04:26, 14.01s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 10%|█         | 2/20 [00:15<02:01,  6.76s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 10%|█         | 2/20 [00:17<02:01,  6.76s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 15%|█▌        | 3/20 [00:17<01:15,  4.45s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 15%|█▌        | 3/20 [00:19<01:15,  4.45s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 20%|██        | 4/20 [00:19<00:53,  3.37s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 20%|██        | 4/20 [00:20<00:53,  3.37s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 25%|██▌       | 5/20 [00:20<00:41,  2.76s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.76s/iter, val_loss=0.00175, val_og_loss=0.00207]
+
 30%|███       | 6/20 [00:22<00:33,  2.40s/iter, val_loss=0.00175, val_og_loss=0.00207]
+
 30%|███       | 6/20 [00:24<00:33,  2.40s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.16s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.16s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 40%|████      | 8/20 [00:25<00:24,  2.02s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 40%|████      | 8/20 [00:27<00:24,  2.02s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 45%|████▌     | 9/20 [00:27<00:21,  1.92s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.92s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 50%|█████     | 10/20 [00:29<00:18,  1.85s/iter, val_loss=0.00174, val_og_loss=0.00207]
+
 50%|█████     | 10/20 [00:30<00:18,  1.85s/iter, val_loss=0.00175, val_og_loss=0.00207]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.80s/iter, val_loss=0.00175, val_og_loss=0.00207]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.80s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 60%|██████    | 12/20 [00:32<00:14,  1.77s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.75s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.75s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 70%|███████   | 14/20 [00:36<00:10,  1.73s/iter, val_loss=0.00173, val_og_loss=0.00207]
+
 70%|███████   | 14/20 [00:37<00:10,  1.73s/iter, val_loss=0.00172, val_og_loss=0.00207]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.71s/iter, val_loss=0.00172, val_og_loss=0.00207]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.00171, val_og_loss=0.00207]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=0.00171, val_og_loss=0.00207]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=0.00172, val_og_loss=0.00207]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.70s/iter, val_loss=0.00172, val_og_loss=0.00207]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=0.00172, val_og_loss=0.00207]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=0.00172, val_og_loss=0.00207]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.00171, val_og_loss=0.00207]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=0.00171, val_og_loss=0.00207]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=0.00171, val_og_loss=0.00207]
+
100%|██████████| 20/20 [00:46<00:00,  1.68s/iter, val_loss=0.00171, val_og_loss=0.00207]
100%|██████████| 20/20 [00:46<00:00,  2.30s/iter, val_loss=0.00171, val_og_loss=0.00207]
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:24:01<5:09:14, 806.72s/it]INFO:ParoQuant:Best val loss: 0.001710837590508163, Original val loss: 0.0020731359254568815
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:24:03<5:09:14, 806.72s/it]2025-12-22 15:54:19,540 - INFO - Best val loss: 0.001710837590508163, Original val loss: 0.0020731359254568815
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=0.00454, val_og_loss=0.0055]
+
  5%|▌         | 1/20 [00:11<03:34, 11.30s/iter, val_loss=0.00454, val_og_loss=0.0055]
+
  5%|▌         | 1/20 [00:12<03:34, 11.30s/iter, val_loss=0.00465, val_og_loss=0.0055]
+
 10%|█         | 2/20 [00:12<01:38,  5.47s/iter, val_loss=0.00465, val_og_loss=0.0055]
+
 10%|█         | 2/20 [00:14<01:38,  5.47s/iter, val_loss=0.00466, val_og_loss=0.0055]
+
 15%|█▌        | 3/20 [00:14<01:01,  3.61s/iter, val_loss=0.00466, val_og_loss=0.0055]
+
 15%|█▌        | 3/20 [00:15<01:01,  3.61s/iter, val_loss=0.00463, val_og_loss=0.0055]
+
 20%|██        | 4/20 [00:15<00:43,  2.74s/iter, val_loss=0.00463, val_og_loss=0.0055]
+
 20%|██        | 4/20 [00:16<00:43,  2.74s/iter, val_loss=0.00468, val_og_loss=0.0055]
+
 25%|██▌       | 5/20 [00:16<00:33,  2.25s/iter, val_loss=0.00468, val_og_loss=0.0055]
+
 25%|██▌       | 5/20 [00:18<00:33,  2.25s/iter, val_loss=0.00466, val_og_loss=0.0055]
+
 30%|███       | 6/20 [00:18<00:27,  1.96s/iter, val_loss=0.00466, val_og_loss=0.0055]
+
 30%|███       | 6/20 [00:19<00:27,  1.96s/iter, val_loss=0.00466, val_og_loss=0.0055]
+
 35%|███▌      | 7/20 [00:19<00:23,  1.77s/iter, val_loss=0.00466, val_og_loss=0.0055]
+
 35%|███▌      | 7/20 [00:21<00:23,  1.77s/iter, val_loss=0.00465, val_og_loss=0.0055]
+
 40%|████      | 8/20 [00:21<00:19,  1.65s/iter, val_loss=0.00465, val_og_loss=0.0055]
+
 40%|████      | 8/20 [00:22<00:19,  1.65s/iter, val_loss=0.00459, val_og_loss=0.0055]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.57s/iter, val_loss=0.00459, val_og_loss=0.0055]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.57s/iter, val_loss=0.00464, val_og_loss=0.0055]
+
 50%|█████     | 10/20 [00:23<00:15,  1.52s/iter, val_loss=0.00464, val_og_loss=0.0055]
+
 50%|█████     | 10/20 [00:25<00:15,  1.52s/iter, val_loss=0.00461, val_og_loss=0.0055]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.48s/iter, val_loss=0.00461, val_og_loss=0.0055]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.48s/iter, val_loss=0.00465, val_og_loss=0.0055]
+
 60%|██████    | 12/20 [00:26<00:11,  1.46s/iter, val_loss=0.00465, val_og_loss=0.0055]
+
 60%|██████    | 12/20 [00:28<00:11,  1.46s/iter, val_loss=0.00461, val_og_loss=0.0055]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.44s/iter, val_loss=0.00461, val_og_loss=0.0055]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.44s/iter, val_loss=0.00461, val_og_loss=0.0055]
+
 70%|███████   | 14/20 [00:29<00:08,  1.42s/iter, val_loss=0.00461, val_og_loss=0.0055]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.00455, val_og_loss=0.0055]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.00455, val_og_loss=0.0055]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.41s/iter, val_loss=0.00458, val_og_loss=0.0055]
+
 80%|████████  | 16/20 [00:32<00:05,  1.42s/iter, val_loss=0.00458, val_og_loss=0.0055]
+
 80%|████████  | 16/20 [00:33<00:05,  1.42s/iter, val_loss=0.00455, val_og_loss=0.0055]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.41s/iter, val_loss=0.00455, val_og_loss=0.0055]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.41s/iter, val_loss=0.00457, val_og_loss=0.0055]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00457, val_og_loss=0.0055]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.00453, val_og_loss=0.0055]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.00453, val_og_loss=0.0055]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00456, val_og_loss=0.0055]
+
100%|██████████| 20/20 [00:37<00:00,  1.41s/iter, val_loss=0.00456, val_og_loss=0.0055]
100%|██████████| 20/20 [00:37<00:00,  1.89s/iter, val_loss=0.00456, val_og_loss=0.0055]
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:24:41<5:09:14, 806.72s/it]INFO:ParoQuant:Best val loss: 0.0045298198238015175, Original val loss: 0.00549833569675684
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:24:42<5:09:14, 806.72s/it]2025-12-22 15:54:59,173 - INFO - Best val loss: 0.0045298198238015175, Original val loss: 0.00549833569675684
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=0.000412, val_og_loss=0.00052]
+
  5%|▌         | 1/20 [00:11<03:45, 11.89s/iter, val_loss=0.000412, val_og_loss=0.00052]
+
  5%|▌         | 1/20 [00:13<03:45, 11.89s/iter, val_loss=0.000411, val_og_loss=0.00052]
+
 10%|█         | 2/20 [00:13<01:42,  5.71s/iter, val_loss=0.000411, val_og_loss=0.00052]
+
 10%|█         | 2/20 [00:14<01:42,  5.71s/iter, val_loss=0.000414, val_og_loss=0.00052]
+
 15%|█▌        | 3/20 [00:14<01:03,  3.74s/iter, val_loss=0.000414, val_og_loss=0.00052]
+
 15%|█▌        | 3/20 [00:16<01:03,  3.74s/iter, val_loss=0.000415, val_og_loss=0.00052]
+
 20%|██        | 4/20 [00:16<00:45,  2.81s/iter, val_loss=0.000415, val_og_loss=0.00052]
+
 20%|██        | 4/20 [00:17<00:45,  2.81s/iter, val_loss=0.000415, val_og_loss=0.00052]
+
 25%|██▌       | 5/20 [00:17<00:34,  2.30s/iter, val_loss=0.000415, val_og_loss=0.00052]
+
 25%|██▌       | 5/20 [00:18<00:34,  2.30s/iter, val_loss=0.000416, val_og_loss=0.00052]
+
 30%|███       | 6/20 [00:18<00:27,  2.00s/iter, val_loss=0.000416, val_og_loss=0.00052]
+
 30%|███       | 6/20 [00:20<00:27,  2.00s/iter, val_loss=0.000416, val_og_loss=0.00052]
+
 35%|███▌      | 7/20 [00:20<00:23,  1.80s/iter, val_loss=0.000416, val_og_loss=0.00052]
+
 35%|███▌      | 7/20 [00:21<00:23,  1.80s/iter, val_loss=0.000415, val_og_loss=0.00052]
+
 40%|████      | 8/20 [00:21<00:20,  1.67s/iter, val_loss=0.000415, val_og_loss=0.00052]
+
 40%|████      | 8/20 [00:23<00:20,  1.67s/iter, val_loss=0.000416, val_og_loss=0.00052]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.59s/iter, val_loss=0.000416, val_og_loss=0.00052]
+
 45%|████▌     | 9/20 [00:24<00:17,  1.59s/iter, val_loss=0.000413, val_og_loss=0.00052]
+
 50%|█████     | 10/20 [00:24<00:15,  1.53s/iter, val_loss=0.000413, val_og_loss=0.00052]
+
 50%|█████     | 10/20 [00:25<00:15,  1.53s/iter, val_loss=0.000412, val_og_loss=0.00052]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.49s/iter, val_loss=0.000412, val_og_loss=0.00052]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.49s/iter, val_loss=0.000412, val_og_loss=0.00052]
+
 60%|██████    | 12/20 [00:27<00:11,  1.46s/iter, val_loss=0.000412, val_og_loss=0.00052]
+
 60%|██████    | 12/20 [00:28<00:11,  1.46s/iter, val_loss=0.000412, val_og_loss=0.00052]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.44s/iter, val_loss=0.000412, val_og_loss=0.00052]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.44s/iter, val_loss=0.000411, val_og_loss=0.00052]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.000411, val_og_loss=0.00052]
+
 70%|███████   | 14/20 [00:31<00:08,  1.43s/iter, val_loss=0.000411, val_og_loss=0.00052]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.42s/iter, val_loss=0.000411, val_og_loss=0.00052]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.42s/iter, val_loss=0.000411, val_og_loss=0.00052]
+
 80%|████████  | 16/20 [00:32<00:05,  1.42s/iter, val_loss=0.000411, val_og_loss=0.00052]
+
 80%|████████  | 16/20 [00:34<00:05,  1.42s/iter, val_loss=0.00041, val_og_loss=0.00052] 
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.41s/iter, val_loss=0.00041, val_og_loss=0.00052]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.41s/iter, val_loss=0.000408, val_og_loss=0.00052]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.41s/iter, val_loss=0.000408, val_og_loss=0.00052]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.41s/iter, val_loss=0.000408, val_og_loss=0.00052]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.000408, val_og_loss=0.00052]
+
 95%|█████████▌| 19/20 [00:38<00:01,  1.40s/iter, val_loss=0.000408, val_og_loss=0.00052]
+
100%|██████████| 20/20 [00:38<00:00,  1.40s/iter, val_loss=0.000408, val_og_loss=0.00052]
100%|██████████| 20/20 [00:38<00:00,  1.92s/iter, val_loss=0.000408, val_og_loss=0.00052]
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:25:21<5:09:14, 806.72s/it]INFO:ParoQuant:Best val loss: 0.0004082032828591764, Original val loss: 0.0005195991834625602
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:25:24<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:25:24<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:25:55<5:09:14, 806.72s/it]2025-12-22 15:55:39,345 - INFO - Best val loss: 0.0004082032828591764, Original val loss: 0.0005195991834625602
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
  5%|▌         | 1/20 [00:15<04:48, 15.18s/iter, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
  5%|▌         | 1/20 [00:16<04:48, 15.18s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 10%|█         | 2/20 [00:16<02:10,  7.24s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 10%|█         | 2/20 [00:18<02:10,  7.24s/iter, val_loss=2.3e-5, val_og_loss=2.44e-5] 
+
 15%|█▌        | 3/20 [00:18<01:19,  4.70s/iter, val_loss=2.3e-5, val_og_loss=2.44e-5]
+
 15%|█▌        | 3/20 [00:20<01:19,  4.70s/iter, val_loss=2.3e-5, val_og_loss=2.44e-5]
+
 20%|██        | 4/20 [00:20<00:56,  3.51s/iter, val_loss=2.3e-5, val_og_loss=2.44e-5]
+
 20%|██        | 4/20 [00:21<00:56,  3.51s/iter, val_loss=2.3e-5, val_og_loss=2.44e-5]
+
 25%|██▌       | 5/20 [00:21<00:42,  2.86s/iter, val_loss=2.3e-5, val_og_loss=2.44e-5]
+
 25%|██▌       | 5/20 [00:23<00:42,  2.86s/iter, val_loss=2.3e-5, val_og_loss=2.44e-5]
+
 30%|███       | 6/20 [00:23<00:34,  2.46s/iter, val_loss=2.3e-5, val_og_loss=2.44e-5]
+
 30%|███       | 6/20 [00:25<00:34,  2.46s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.20s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 35%|███▌      | 7/20 [00:26<00:28,  2.20s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 40%|████      | 8/20 [00:26<00:24,  2.03s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 40%|████      | 8/20 [00:28<00:24,  2.03s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 45%|████▌     | 9/20 [00:28<00:21,  1.93s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.93s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.85s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 50%|█████     | 10/20 [00:32<00:18,  1.85s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.80s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.80s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 60%|██████    | 12/20 [00:33<00:14,  1.76s/iter, val_loss=2.29e-5, val_og_loss=2.44e-5]
+
 60%|██████    | 12/20 [00:35<00:14,  1.76s/iter, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.75s/iter, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.73s/iter, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
 70%|███████   | 14/20 [00:38<00:10,  1.73s/iter, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.71s/iter, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=2.28e-5, val_og_loss=2.44e-5]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=2.27e-5, val_og_loss=2.44e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=2.27e-5, val_og_loss=2.44e-5]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.70s/iter, val_loss=2.26e-5, val_og_loss=2.44e-5]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=2.26e-5, val_og_loss=2.44e-5]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.69s/iter, val_loss=2.27e-5, val_og_loss=2.44e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.68s/iter, val_loss=2.27e-5, val_og_loss=2.44e-5]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.68s/iter, val_loss=2.27e-5, val_og_loss=2.44e-5]
+
100%|██████████| 20/20 [00:47<00:00,  1.68s/iter, val_loss=2.27e-5, val_og_loss=2.44e-5]
100%|██████████| 20/20 [00:47<00:00,  2.36s/iter, val_loss=2.27e-5, val_og_loss=2.44e-5]
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:26:42<5:09:14, 806.72s/it]INFO:ParoQuant:Best val loss: 2.2637863366981037e-05, Original val loss: 2.4378012312809005e-05
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:26:42<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:26:42<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:27:26<5:09:14, 806.72s/it]2025-12-22 15:57:00,192 - INFO - Best val loss: 2.2637863366981037e-05, Original val loss: 2.4378012312809005e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00085, val_og_loss=0.000867]
+
  5%|▌         | 1/20 [00:29<09:25, 29.75s/iter, val_loss=0.00085, val_og_loss=0.000867]
+
  5%|▌         | 1/20 [00:32<09:25, 29.75s/iter, val_loss=0.000852, val_og_loss=0.000867]
+
 10%|█         | 2/20 [00:32<04:09, 13.84s/iter, val_loss=0.000852, val_og_loss=0.000867]
+
 10%|█         | 2/20 [00:35<04:09, 13.84s/iter, val_loss=0.000855, val_og_loss=0.000867]
+
 15%|█▌        | 3/20 [00:35<02:28,  8.76s/iter, val_loss=0.000855, val_og_loss=0.000867]
+
 15%|█▌        | 3/20 [00:37<02:28,  8.76s/iter, val_loss=0.000856, val_og_loss=0.000867]
+
 20%|██        | 4/20 [00:37<01:41,  6.37s/iter, val_loss=0.000856, val_og_loss=0.000867]
+
 20%|██        | 4/20 [00:40<01:41,  6.37s/iter, val_loss=0.000858, val_og_loss=0.000867]
+
 25%|██▌       | 5/20 [00:40<01:15,  5.05s/iter, val_loss=0.000858, val_og_loss=0.000867]
+
 25%|██▌       | 5/20 [00:43<01:15,  5.05s/iter, val_loss=0.000858, val_og_loss=0.000867]
+
 30%|███       | 6/20 [00:43<00:59,  4.26s/iter, val_loss=0.000858, val_og_loss=0.000867]
+
 30%|███       | 6/20 [00:46<00:59,  4.26s/iter, val_loss=0.000856, val_og_loss=0.000867]
+
 35%|███▌      | 7/20 [00:46<00:48,  3.75s/iter, val_loss=0.000856, val_og_loss=0.000867]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.75s/iter, val_loss=0.000857, val_og_loss=0.000867]
+
 40%|████      | 8/20 [00:48<00:41,  3.42s/iter, val_loss=0.000857, val_og_loss=0.000867]
+
 40%|████      | 8/20 [00:51<00:41,  3.42s/iter, val_loss=0.000856, val_og_loss=0.000867]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.20s/iter, val_loss=0.000856, val_og_loss=0.000867]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.20s/iter, val_loss=0.000856, val_og_loss=0.000867]
+
 50%|█████     | 10/20 [00:54<00:30,  3.05s/iter, val_loss=0.000856, val_og_loss=0.000867]
+
 50%|█████     | 10/20 [00:56<00:30,  3.05s/iter, val_loss=0.000855, val_og_loss=0.000867]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.95s/iter, val_loss=0.000855, val_og_loss=0.000867]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.95s/iter, val_loss=0.000854, val_og_loss=0.000867]
+
 60%|██████    | 12/20 [00:59<00:23,  2.88s/iter, val_loss=0.000854, val_og_loss=0.000867]
+
 60%|██████    | 12/20 [01:02<00:23,  2.88s/iter, val_loss=0.000853, val_og_loss=0.000867]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.83s/iter, val_loss=0.000853, val_og_loss=0.000867]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.83s/iter, val_loss=0.000853, val_og_loss=0.000867]
+
 70%|███████   | 14/20 [01:05<00:16,  2.79s/iter, val_loss=0.000853, val_og_loss=0.000867]
+
 70%|███████   | 14/20 [01:07<00:16,  2.79s/iter, val_loss=0.000852, val_og_loss=0.000867]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.000852, val_og_loss=0.000867]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.000852, val_og_loss=0.000867]
+
 80%|████████  | 16/20 [01:10<00:11,  2.75s/iter, val_loss=0.000852, val_og_loss=0.000867]
+
 80%|████████  | 16/20 [01:13<00:11,  2.75s/iter, val_loss=0.000852, val_og_loss=0.000867]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.000852, val_og_loss=0.000867]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.000851, val_og_loss=0.000867]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.73s/iter, val_loss=0.000851, val_og_loss=0.000867]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.00085, val_og_loss=0.000867] 
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.72s/iter, val_loss=0.00085, val_og_loss=0.000867]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.72s/iter, val_loss=0.000849, val_og_loss=0.000867]
+
100%|██████████| 20/20 [01:21<00:00,  2.72s/iter, val_loss=0.000849, val_og_loss=0.000867]
100%|██████████| 20/20 [01:21<00:00,  4.06s/iter, val_loss=0.000849, val_og_loss=0.000867]
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:28:47<5:09:14, 806.72s/it]INFO:ParoQuant:Best val loss: 0.000849416886921972, Original val loss: 0.0008667724905535579
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:28:52<5:09:14, 806.72s/it]2025-12-22 15:59:05,851 - INFO - Best val loss: 0.000849416886921972, Original val loss: 0.0008667724905535579
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.000472, val_og_loss=0.000481]
+
  5%|▌         | 1/20 [00:31<10:05, 31.88s/iter, val_loss=0.000472, val_og_loss=0.000481]
+
  5%|▌         | 1/20 [00:34<10:05, 31.88s/iter, val_loss=0.000474, val_og_loss=0.000481]
+
 10%|█         | 2/20 [00:34<04:24, 14.72s/iter, val_loss=0.000474, val_og_loss=0.000481]
+
 10%|█         | 2/20 [00:37<04:24, 14.72s/iter, val_loss=0.000476, val_og_loss=0.000481]
+
 15%|█▌        | 3/20 [00:37<02:37,  9.24s/iter, val_loss=0.000476, val_og_loss=0.000481]
+
 15%|█▌        | 3/20 [00:40<02:37,  9.24s/iter, val_loss=0.000477, val_og_loss=0.000481]
+
 20%|██        | 4/20 [00:40<01:46,  6.66s/iter, val_loss=0.000477, val_og_loss=0.000481]
+
 20%|██        | 4/20 [00:42<01:46,  6.66s/iter, val_loss=0.000478, val_og_loss=0.000481]
+
 25%|██▌       | 5/20 [00:42<01:18,  5.24s/iter, val_loss=0.000478, val_og_loss=0.000481]
+
 25%|██▌       | 5/20 [00:45<01:18,  5.24s/iter, val_loss=0.000479, val_og_loss=0.000481]
+
 30%|███       | 6/20 [00:45<01:01,  4.38s/iter, val_loss=0.000479, val_og_loss=0.000481]
+
 30%|███       | 6/20 [00:48<01:01,  4.38s/iter, val_loss=0.000479, val_og_loss=0.000481]
+
 35%|███▌      | 7/20 [00:48<00:49,  3.83s/iter, val_loss=0.000479, val_og_loss=0.000481]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.83s/iter, val_loss=0.000478, val_og_loss=0.000481]
+
 40%|████      | 8/20 [00:50<00:41,  3.48s/iter, val_loss=0.000478, val_og_loss=0.000481]
+
 40%|████      | 8/20 [00:53<00:41,  3.48s/iter, val_loss=0.000478, val_og_loss=0.000481]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.24s/iter, val_loss=0.000478, val_og_loss=0.000481]
+
 45%|████▌     | 9/20 [00:56<00:35,  3.24s/iter, val_loss=0.000477, val_og_loss=0.000481]
+
 50%|█████     | 10/20 [00:56<00:30,  3.08s/iter, val_loss=0.000477, val_og_loss=0.000481]
+
 50%|█████     | 10/20 [00:59<00:30,  3.08s/iter, val_loss=0.000477, val_og_loss=0.000481]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.97s/iter, val_loss=0.000477, val_og_loss=0.000481]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.97s/iter, val_loss=0.000476, val_og_loss=0.000481]
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.000476, val_og_loss=0.000481]
+
 60%|██████    | 12/20 [01:04<00:23,  2.89s/iter, val_loss=0.000476, val_og_loss=0.000481]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.000476, val_og_loss=0.000481]
+
 65%|██████▌   | 13/20 [01:07<00:19,  2.84s/iter, val_loss=0.000475, val_og_loss=0.000481]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.000475, val_og_loss=0.000481]
+
 70%|███████   | 14/20 [01:09<00:16,  2.80s/iter, val_loss=0.000474, val_og_loss=0.000481]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.000474, val_og_loss=0.000481]
+
 75%|███████▌  | 15/20 [01:12<00:13,  2.78s/iter, val_loss=0.000474, val_og_loss=0.000481]
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.000474, val_og_loss=0.000481]
+
 80%|████████  | 16/20 [01:15<00:11,  2.76s/iter, val_loss=0.000473, val_og_loss=0.000481]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.000473, val_og_loss=0.000481]
+
 85%|████████▌ | 17/20 [01:18<00:08,  2.74s/iter, val_loss=0.000473, val_og_loss=0.000481]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.000473, val_og_loss=0.000481]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.73s/iter, val_loss=0.000473, val_og_loss=0.000481]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.72s/iter, val_loss=0.000473, val_og_loss=0.000481]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.72s/iter, val_loss=0.000473, val_og_loss=0.000481]
+
100%|██████████| 20/20 [01:23<00:00,  2.73s/iter, val_loss=0.000473, val_og_loss=0.000481]
100%|██████████| 20/20 [01:23<00:00,  4.17s/iter, val_loss=0.000473, val_og_loss=0.000481]
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:30:16<5:09:14, 806.72s/it]INFO:ParoQuant:Best val loss: 0.0004722562152892351, Original val loss: 0.00048090197378769517
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:30:17<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:30:17<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:31:17<5:09:14, 806.72s/it]2025-12-22 16:00:34,121 - INFO - Best val loss: 0.0004722562152892351, Original val loss: 0.00048090197378769517
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=4.89e-5, val_og_loss=4.94e-5]
+
  5%|▌         | 1/20 [00:33<10:43, 33.89s/iter, val_loss=4.89e-5, val_og_loss=4.94e-5]
+
  5%|▌         | 1/20 [00:39<10:43, 33.89s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5] 
+
 10%|█         | 2/20 [00:39<05:08, 17.14s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 10%|█         | 2/20 [00:44<05:08, 17.14s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 15%|█▌        | 3/20 [00:44<03:20, 11.79s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 15%|█▌        | 3/20 [00:50<03:20, 11.79s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 20%|██        | 4/20 [00:50<02:28,  9.27s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 20%|██        | 4/20 [00:55<02:28,  9.27s/iter, val_loss=4.91e-5, val_og_loss=4.94e-5]
+
 25%|██▌       | 5/20 [00:55<01:58,  7.88s/iter, val_loss=4.91e-5, val_og_loss=4.94e-5]
+
 25%|██▌       | 5/20 [01:00<01:58,  7.88s/iter, val_loss=4.91e-5, val_og_loss=4.94e-5]
+
 30%|███       | 6/20 [01:00<01:38,  7.04s/iter, val_loss=4.91e-5, val_og_loss=4.94e-5]
+
 30%|███       | 6/20 [01:06<01:38,  7.04s/iter, val_loss=4.91e-5, val_og_loss=4.94e-5]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.51s/iter, val_loss=4.91e-5, val_og_loss=4.94e-5]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.51s/iter, val_loss=4.91e-5, val_og_loss=4.94e-5]
+
 40%|████      | 8/20 [01:11<01:13,  6.17s/iter, val_loss=4.91e-5, val_og_loss=4.94e-5]
+
 40%|████      | 8/20 [01:17<01:13,  6.17s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5] 
+
 45%|████▌     | 9/20 [01:17<01:05,  5.93s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.93s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 50%|█████     | 10/20 [01:22<00:57,  5.77s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 50%|█████     | 10/20 [01:28<00:57,  5.77s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 55%|█████▌    | 11/20 [01:28<00:51,  5.67s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 55%|█████▌    | 11/20 [01:33<00:51,  5.67s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 60%|██████    | 12/20 [01:33<00:44,  5.59s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 60%|██████    | 12/20 [01:38<00:44,  5.59s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.54s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.54s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 70%|███████   | 14/20 [01:44<00:32,  5.50s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 70%|███████   | 14/20 [01:49<00:32,  5.50s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.47s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 75%|███████▌  | 15/20 [01:55<00:27,  5.47s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 80%|████████  | 16/20 [01:55<00:21,  5.46s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 80%|████████  | 16/20 [02:00<00:21,  5.46s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.44s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.44s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.44s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.43s/iter, val_loss=4.9e-5, val_og_loss=4.94e-5]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.43s/iter, val_loss=4.89e-5, val_og_loss=4.94e-5]
+
100%|██████████| 20/20 [02:16<00:00,  5.43s/iter, val_loss=4.89e-5, val_og_loss=4.94e-5]
100%|██████████| 20/20 [02:16<00:00,  6.84s/iter, val_loss=4.89e-5, val_og_loss=4.94e-5]
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:33:34<5:09:14, 806.72s/it]INFO:ParoQuant:Best val loss: 4.8899848479777575e-05, Original val loss: 4.943561725667678e-05
+
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:33:35<5:09:14, 806.72s/it]
                                                                      

Processing Blocks:  28%|██▊       | 9/32 [1:33:45<5:09:14, 806.72s/it]
Processing Blocks:  31%|███▏      | 10/32 [1:34:23<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:34:24<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:34:42<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:34:42<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:36:57<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:36:57<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:38:03<5:03:44, 828.40s/it]2025-12-22 16:03:52,655 - INFO - Best val loss: 4.8899848479777575e-05, Original val loss: 4.943561725667678e-05
+[INFO] Saving Layer 9 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 10...
+[INFO] Starting Sequential Training for Layer 10
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:17<?, ?iter/s, val_loss=0.00193, val_og_loss=0.00241]
+
  5%|▌         | 1/20 [00:17<05:24, 17.08s/iter, val_loss=0.00193, val_og_loss=0.00241]
+
  5%|▌         | 1/20 [00:18<05:24, 17.08s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 10%|█         | 2/20 [00:18<02:24,  8.03s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 10%|█         | 2/20 [00:20<02:24,  8.03s/iter, val_loss=0.00199, val_og_loss=0.00241]
+
 15%|█▌        | 3/20 [00:20<01:27,  5.15s/iter, val_loss=0.00199, val_og_loss=0.00241]
+
 15%|█▌        | 3/20 [00:22<01:27,  5.15s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 20%|██        | 4/20 [00:22<01:00,  3.79s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 20%|██        | 4/20 [00:23<01:00,  3.79s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 25%|██▌       | 5/20 [00:23<00:45,  3.05s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 25%|██▌       | 5/20 [00:25<00:45,  3.05s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 30%|███       | 6/20 [00:25<00:36,  2.59s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 30%|███       | 6/20 [00:27<00:36,  2.59s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 35%|███▌      | 7/20 [00:27<00:29,  2.29s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 35%|███▌      | 7/20 [00:28<00:29,  2.29s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 40%|████      | 8/20 [00:28<00:25,  2.10s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 40%|████      | 8/20 [00:30<00:25,  2.10s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.98s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 45%|████▌     | 9/20 [00:32<00:21,  1.98s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 50%|█████     | 10/20 [00:32<00:18,  1.89s/iter, val_loss=0.00198, val_og_loss=0.00241]
+
 50%|█████     | 10/20 [00:34<00:18,  1.89s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.82s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.82s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 60%|██████    | 12/20 [00:35<00:14,  1.78s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 60%|██████    | 12/20 [00:37<00:14,  1.78s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=0.00197, val_og_loss=0.00241]
+
 65%|██████▌   | 13/20 [00:39<00:12,  1.75s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 70%|███████   | 14/20 [00:39<00:10,  1.72s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 70%|███████   | 14/20 [00:40<00:10,  1.72s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.71s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 75%|███████▌  | 15/20 [00:42<00:08,  1.71s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 80%|████████  | 16/20 [00:44<00:06,  1.70s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=0.00196, val_og_loss=0.00241]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.70s/iter, val_loss=0.00195, val_og_loss=0.00241]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.69s/iter, val_loss=0.00195, val_og_loss=0.00241]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.69s/iter, val_loss=0.00195, val_og_loss=0.00241]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=0.00195, val_og_loss=0.00241]
+
 95%|█████████▌| 19/20 [00:49<00:01,  1.69s/iter, val_loss=0.00194, val_og_loss=0.00241]
+
100%|██████████| 20/20 [00:49<00:00,  1.69s/iter, val_loss=0.00194, val_og_loss=0.00241]
100%|██████████| 20/20 [00:49<00:00,  2.46s/iter, val_loss=0.00194, val_og_loss=0.00241]
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:38:52<5:03:44, 828.40s/it]INFO:ParoQuant:Best val loss: 0.001934447675012052, Original val loss: 0.0024101994931697845
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:38:56<5:03:44, 828.40s/it]2025-12-22 16:09:10,652 - INFO - Best val loss: 0.001934447675012052, Original val loss: 0.0024101994931697845
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.00521, val_og_loss=0.00636]
+
  5%|▌         | 1/20 [00:16<05:22, 16.97s/iter, val_loss=0.00521, val_og_loss=0.00636]
+
  5%|▌         | 1/20 [00:18<05:22, 16.97s/iter, val_loss=0.0053, val_og_loss=0.00636] 
+
 10%|█         | 2/20 [00:18<02:20,  7.82s/iter, val_loss=0.0053, val_og_loss=0.00636]
+
 10%|█         | 2/20 [00:19<02:20,  7.82s/iter, val_loss=0.0053, val_og_loss=0.00636]
+
 15%|█▌        | 3/20 [00:19<01:22,  4.88s/iter, val_loss=0.0053, val_og_loss=0.00636]
+
 15%|█▌        | 3/20 [00:21<01:22,  4.88s/iter, val_loss=0.00528, val_og_loss=0.00636]
+
 20%|██        | 4/20 [00:21<00:56,  3.50s/iter, val_loss=0.00528, val_og_loss=0.00636]
+
 20%|██        | 4/20 [00:22<00:56,  3.50s/iter, val_loss=0.00534, val_og_loss=0.00636]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.74s/iter, val_loss=0.00534, val_og_loss=0.00636]
+
 25%|██▌       | 5/20 [00:23<00:41,  2.74s/iter, val_loss=0.00531, val_og_loss=0.00636]
+
 30%|███       | 6/20 [00:23<00:31,  2.28s/iter, val_loss=0.00531, val_og_loss=0.00636]
+
 30%|███       | 6/20 [00:25<00:31,  2.28s/iter, val_loss=0.00529, val_og_loss=0.00636]
+
 35%|███▌      | 7/20 [00:25<00:25,  1.99s/iter, val_loss=0.00529, val_og_loss=0.00636]
+
 35%|███▌      | 7/20 [00:26<00:25,  1.99s/iter, val_loss=0.00529, val_og_loss=0.00636]
+
 40%|████      | 8/20 [00:26<00:21,  1.80s/iter, val_loss=0.00529, val_og_loss=0.00636]
+
 40%|████      | 8/20 [00:28<00:21,  1.80s/iter, val_loss=0.00528, val_og_loss=0.00636]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.67s/iter, val_loss=0.00528, val_og_loss=0.00636]
+
 45%|████▌     | 9/20 [00:29<00:18,  1.67s/iter, val_loss=0.00529, val_og_loss=0.00636]
+
 50%|█████     | 10/20 [00:29<00:15,  1.58s/iter, val_loss=0.00529, val_og_loss=0.00636]
+
 50%|█████     | 10/20 [00:30<00:15,  1.58s/iter, val_loss=0.00528, val_og_loss=0.00636]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.52s/iter, val_loss=0.00528, val_og_loss=0.00636]
+
 55%|█████▌    | 11/20 [00:32<00:13,  1.52s/iter, val_loss=0.00528, val_og_loss=0.00636]
+
 60%|██████    | 12/20 [00:32<00:11,  1.49s/iter, val_loss=0.00528, val_og_loss=0.00636]
+
 60%|██████    | 12/20 [00:33<00:11,  1.49s/iter, val_loss=0.00524, val_og_loss=0.00636]
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.46s/iter, val_loss=0.00524, val_og_loss=0.00636]
+
 65%|██████▌   | 13/20 [00:35<00:10,  1.46s/iter, val_loss=0.00525, val_og_loss=0.00636]
+
 70%|███████   | 14/20 [00:35<00:08,  1.44s/iter, val_loss=0.00525, val_og_loss=0.00636]
+
 70%|███████   | 14/20 [00:36<00:08,  1.44s/iter, val_loss=0.00524, val_og_loss=0.00636]
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.42s/iter, val_loss=0.00524, val_og_loss=0.00636]
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.42s/iter, val_loss=0.00523, val_og_loss=0.00636]
+
 80%|████████  | 16/20 [00:37<00:05,  1.41s/iter, val_loss=0.00523, val_og_loss=0.00636]
+
 80%|████████  | 16/20 [00:39<00:05,  1.41s/iter, val_loss=0.00525, val_og_loss=0.00636]
+
 85%|████████▌ | 17/20 [00:39<00:04,  1.41s/iter, val_loss=0.00525, val_og_loss=0.00636]
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.41s/iter, val_loss=0.00523, val_og_loss=0.00636]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.40s/iter, val_loss=0.00523, val_og_loss=0.00636]
+
 90%|█████████ | 18/20 [00:42<00:02,  1.40s/iter, val_loss=0.00521, val_og_loss=0.00636]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.40s/iter, val_loss=0.00521, val_og_loss=0.00636]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.40s/iter, val_loss=0.0052, val_og_loss=0.00636] 
+
100%|██████████| 20/20 [00:43<00:00,  1.39s/iter, val_loss=0.0052, val_og_loss=0.00636]
100%|██████████| 20/20 [00:43<00:00,  2.17s/iter, val_loss=0.0052, val_og_loss=0.00636]
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:39:39<5:03:44, 828.40s/it]INFO:ParoQuant:Best val loss: 0.00519849918782711, Original val loss: 0.006356095429509878
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:39:42<5:03:44, 828.40s/it]2025-12-22 16:09:57,643 - INFO - Best val loss: 0.00519849918782711, Original val loss: 0.006356095429509878
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00032, val_og_loss=0.00042]
+
  5%|▌         | 1/20 [00:15<04:55, 15.53s/iter, val_loss=0.00032, val_og_loss=0.00042]
+
  5%|▌         | 1/20 [00:16<04:55, 15.53s/iter, val_loss=0.000322, val_og_loss=0.00042]
+
 10%|█         | 2/20 [00:16<02:09,  7.21s/iter, val_loss=0.000322, val_og_loss=0.00042]
+
 10%|█         | 2/20 [00:18<02:09,  7.21s/iter, val_loss=0.000324, val_og_loss=0.00042]
+
 15%|█▌        | 3/20 [00:18<01:17,  4.56s/iter, val_loss=0.000324, val_og_loss=0.00042]
+
 15%|█▌        | 3/20 [00:19<01:17,  4.56s/iter, val_loss=0.000325, val_og_loss=0.00042]
+
 20%|██        | 4/20 [00:19<00:52,  3.31s/iter, val_loss=0.000325, val_og_loss=0.00042]
+
 20%|██        | 4/20 [00:21<00:52,  3.31s/iter, val_loss=0.000325, val_og_loss=0.00042]
+
 25%|██▌       | 5/20 [00:21<00:39,  2.62s/iter, val_loss=0.000325, val_og_loss=0.00042]
+
 25%|██▌       | 5/20 [00:22<00:39,  2.62s/iter, val_loss=0.000326, val_og_loss=0.00042]
+
 30%|███       | 6/20 [00:22<00:30,  2.20s/iter, val_loss=0.000326, val_og_loss=0.00042]
+
 30%|███       | 6/20 [00:23<00:30,  2.20s/iter, val_loss=0.000326, val_og_loss=0.00042]
+
 35%|███▌      | 7/20 [00:23<00:25,  1.94s/iter, val_loss=0.000326, val_og_loss=0.00042]
+
 35%|███▌      | 7/20 [00:25<00:25,  1.94s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 40%|████      | 8/20 [00:25<00:21,  1.77s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 40%|████      | 8/20 [00:26<00:21,  1.77s/iter, val_loss=0.000324, val_og_loss=0.00042]
+
 45%|████▌     | 9/20 [00:26<00:18,  1.65s/iter, val_loss=0.000324, val_og_loss=0.00042]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.65s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 50%|█████     | 10/20 [00:28<00:15,  1.57s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 50%|█████     | 10/20 [00:29<00:15,  1.57s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.52s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.52s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 60%|██████    | 12/20 [00:30<00:11,  1.48s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 60%|██████    | 12/20 [00:32<00:11,  1.48s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.45s/iter, val_loss=0.000323, val_og_loss=0.00042]
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.45s/iter, val_loss=0.000321, val_og_loss=0.00042]
+
 70%|███████   | 14/20 [00:33<00:08,  1.43s/iter, val_loss=0.000321, val_og_loss=0.00042]
+
 70%|███████   | 14/20 [00:35<00:08,  1.43s/iter, val_loss=0.000322, val_og_loss=0.00042]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.42s/iter, val_loss=0.000322, val_og_loss=0.00042]
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.42s/iter, val_loss=0.000321, val_og_loss=0.00042]
+
 80%|████████  | 16/20 [00:36<00:05,  1.41s/iter, val_loss=0.000321, val_og_loss=0.00042]
+
 80%|████████  | 16/20 [00:37<00:05,  1.41s/iter, val_loss=0.000321, val_og_loss=0.00042]
+
 85%|████████▌ | 17/20 [00:37<00:04,  1.41s/iter, val_loss=0.000321, val_og_loss=0.00042]
+
 85%|████████▌ | 17/20 [00:39<00:04,  1.41s/iter, val_loss=0.00032, val_og_loss=0.00042] 
+
 90%|█████████ | 18/20 [00:39<00:02,  1.41s/iter, val_loss=0.00032, val_og_loss=0.00042]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.41s/iter, val_loss=0.00032, val_og_loss=0.00042]
+
 95%|█████████▌| 19/20 [00:40<00:01,  1.40s/iter, val_loss=0.00032, val_og_loss=0.00042]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.40s/iter, val_loss=0.000319, val_og_loss=0.00042]
+
100%|██████████| 20/20 [00:42<00:00,  1.40s/iter, val_loss=0.000319, val_og_loss=0.00042]
100%|██████████| 20/20 [00:42<00:00,  2.10s/iter, val_loss=0.000319, val_og_loss=0.00042]
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:40:24<5:03:44, 828.40s/it]INFO:ParoQuant:Best val loss: 0.0003194150049239397, Original val loss: 0.00042046571616083384
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:40:27<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:40:27<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:40:58<5:03:44, 828.40s/it]2025-12-22 16:10:42,964 - INFO - Best val loss: 0.0003194150049239397, Original val loss: 0.00042046571616083384
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
  5%|▌         | 1/20 [00:13<04:17, 13.56s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
  5%|▌         | 1/20 [00:15<04:17, 13.56s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 10%|█         | 2/20 [00:15<01:58,  6.57s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 10%|█         | 2/20 [00:16<01:58,  6.57s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 15%|█▌        | 3/20 [00:16<01:13,  4.34s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 15%|█▌        | 3/20 [00:18<01:13,  4.34s/iter, val_loss=2.13e-5, val_og_loss=2.28e-5]
+
 20%|██        | 4/20 [00:18<00:52,  3.29s/iter, val_loss=2.13e-5, val_og_loss=2.28e-5]
+
 20%|██        | 4/20 [00:20<00:52,  3.29s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 25%|██▌       | 5/20 [00:20<00:40,  2.72s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.72s/iter, val_loss=2.13e-5, val_og_loss=2.28e-5]
+
 30%|███       | 6/20 [00:21<00:33,  2.36s/iter, val_loss=2.13e-5, val_og_loss=2.28e-5]
+
 30%|███       | 6/20 [00:23<00:33,  2.36s/iter, val_loss=2.13e-5, val_og_loss=2.28e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.14s/iter, val_loss=2.13e-5, val_og_loss=2.28e-5]
+
 35%|███▌      | 7/20 [00:25<00:27,  2.14s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 40%|████      | 8/20 [00:25<00:23,  1.99s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 40%|████      | 8/20 [00:27<00:23,  1.99s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.90s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.90s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 50%|█████     | 10/20 [00:28<00:18,  1.83s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.83s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.78s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.78s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 60%|██████    | 12/20 [00:32<00:13,  1.75s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 60%|██████    | 12/20 [00:33<00:13,  1.75s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.73s/iter, val_loss=2.12e-5, val_og_loss=2.28e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=2.11e-5, val_og_loss=2.28e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.71s/iter, val_loss=2.11e-5, val_og_loss=2.28e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.71s/iter, val_loss=2.11e-5, val_og_loss=2.28e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.70s/iter, val_loss=2.11e-5, val_og_loss=2.28e-5]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.70s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5] 
+
 80%|████████  | 16/20 [00:38<00:06,  1.69s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.69s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.70s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.69s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
+
100%|██████████| 20/20 [00:45<00:00,  1.68s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
100%|██████████| 20/20 [00:45<00:00,  2.28s/iter, val_loss=2.1e-5, val_og_loss=2.28e-5]
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:41:43<5:03:44, 828.40s/it]INFO:ParoQuant:Best val loss: 2.0985373339499347e-05, Original val loss: 2.275828410347458e-05
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:41:43<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:41:43<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:42:27<5:03:44, 828.40s/it]2025-12-22 16:12:01,530 - INFO - Best val loss: 2.0985373339499347e-05, Original val loss: 2.275828410347458e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.000764, val_og_loss=0.000778]
+
  5%|▌         | 1/20 [00:29<09:15, 29.25s/iter, val_loss=0.000764, val_og_loss=0.000778]
+
  5%|▌         | 1/20 [00:31<09:15, 29.25s/iter, val_loss=0.000766, val_og_loss=0.000778]
+
 10%|█         | 2/20 [00:31<04:05, 13.65s/iter, val_loss=0.000766, val_og_loss=0.000778]
+
 10%|█         | 2/20 [00:34<04:05, 13.65s/iter, val_loss=0.000768, val_og_loss=0.000778]
+
 15%|█▌        | 3/20 [00:34<02:27,  8.65s/iter, val_loss=0.000768, val_og_loss=0.000778]
+
 15%|█▌        | 3/20 [00:37<02:27,  8.65s/iter, val_loss=0.000771, val_og_loss=0.000778]
+
 20%|██        | 4/20 [00:37<01:40,  6.31s/iter, val_loss=0.000771, val_og_loss=0.000778]
+
 20%|██        | 4/20 [00:40<01:40,  6.31s/iter, val_loss=0.000771, val_og_loss=0.000778]
+
 25%|██▌       | 5/20 [00:40<01:15,  5.01s/iter, val_loss=0.000771, val_og_loss=0.000778]
+
 25%|██▌       | 5/20 [00:42<01:15,  5.01s/iter, val_loss=0.000772, val_og_loss=0.000778]
+
 30%|███       | 6/20 [00:42<00:59,  4.23s/iter, val_loss=0.000772, val_og_loss=0.000778]
+
 30%|███       | 6/20 [00:45<00:59,  4.23s/iter, val_loss=0.000772, val_og_loss=0.000778]
+
 35%|███▌      | 7/20 [00:45<00:48,  3.73s/iter, val_loss=0.000772, val_og_loss=0.000778]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.73s/iter, val_loss=0.000771, val_og_loss=0.000778]
+
 40%|████      | 8/20 [00:48<00:40,  3.41s/iter, val_loss=0.000771, val_og_loss=0.000778]
+
 40%|████      | 8/20 [00:50<00:40,  3.41s/iter, val_loss=0.000771, val_og_loss=0.000778]
+
 45%|████▌     | 9/20 [00:50<00:35,  3.19s/iter, val_loss=0.000771, val_og_loss=0.000778]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.19s/iter, val_loss=0.000769, val_og_loss=0.000778]
+
 50%|█████     | 10/20 [00:53<00:30,  3.05s/iter, val_loss=0.000769, val_og_loss=0.000778]
+
 50%|█████     | 10/20 [00:56<00:30,  3.05s/iter, val_loss=0.000768, val_og_loss=0.000778]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.95s/iter, val_loss=0.000768, val_og_loss=0.000778]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.95s/iter, val_loss=0.000768, val_og_loss=0.000778]
+
 60%|██████    | 12/20 [00:59<00:23,  2.88s/iter, val_loss=0.000768, val_og_loss=0.000778]
+
 60%|██████    | 12/20 [01:01<00:23,  2.88s/iter, val_loss=0.000767, val_og_loss=0.000778]
+
 65%|██████▌   | 13/20 [01:01<00:19,  2.83s/iter, val_loss=0.000767, val_og_loss=0.000778]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.83s/iter, val_loss=0.000767, val_og_loss=0.000778]
+
 70%|███████   | 14/20 [01:04<00:16,  2.80s/iter, val_loss=0.000767, val_og_loss=0.000778]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.000766, val_og_loss=0.000778]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.78s/iter, val_loss=0.000766, val_og_loss=0.000778]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.78s/iter, val_loss=0.000765, val_og_loss=0.000778]
+
 80%|████████  | 16/20 [01:10<00:11,  2.76s/iter, val_loss=0.000765, val_og_loss=0.000778]
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.000765, val_og_loss=0.000778]
+
 85%|████████▌ | 17/20 [01:12<00:08,  2.74s/iter, val_loss=0.000765, val_og_loss=0.000778]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.000765, val_og_loss=0.000778]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.74s/iter, val_loss=0.000765, val_og_loss=0.000778]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.000764, val_og_loss=0.000778]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.000764, val_og_loss=0.000778]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.000764, val_og_loss=0.000778]
+
100%|██████████| 20/20 [01:20<00:00,  2.73s/iter, val_loss=0.000764, val_og_loss=0.000778]
100%|██████████| 20/20 [01:20<00:00,  4.05s/iter, val_loss=0.000764, val_og_loss=0.000778]
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:43:48<5:03:44, 828.40s/it]INFO:ParoQuant:Best val loss: 0.0007635744987055659, Original val loss: 0.0007783721666783094
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:43:52<5:03:44, 828.40s/it]2025-12-22 16:14:06,249 - INFO - Best val loss: 0.0007635744987055659, Original val loss: 0.0007783721666783094
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.000466, val_og_loss=0.000475]
+
  5%|▌         | 1/20 [00:31<09:57, 31.46s/iter, val_loss=0.000466, val_og_loss=0.000475]
+
  5%|▌         | 1/20 [00:34<09:57, 31.46s/iter, val_loss=0.000469, val_og_loss=0.000475]
+
 10%|█         | 2/20 [00:34<04:21, 14.55s/iter, val_loss=0.000469, val_og_loss=0.000475]
+
 10%|█         | 2/20 [00:36<04:21, 14.55s/iter, val_loss=0.000471, val_og_loss=0.000475]
+
 15%|█▌        | 3/20 [00:36<02:35,  9.15s/iter, val_loss=0.000471, val_og_loss=0.000475]
+
 15%|█▌        | 3/20 [00:39<02:35,  9.15s/iter, val_loss=0.000473, val_og_loss=0.000475]
+
 20%|██        | 4/20 [00:39<01:45,  6.61s/iter, val_loss=0.000473, val_og_loss=0.000475]
+
 20%|██        | 4/20 [00:42<01:45,  6.61s/iter, val_loss=0.000474, val_og_loss=0.000475]
+
 25%|██▌       | 5/20 [00:42<01:18,  5.20s/iter, val_loss=0.000474, val_og_loss=0.000475]
+
 25%|██▌       | 5/20 [00:45<01:18,  5.20s/iter, val_loss=0.000474, val_og_loss=0.000475]
+
 30%|███       | 6/20 [00:45<01:01,  4.36s/iter, val_loss=0.000474, val_og_loss=0.000475]
+
 30%|███       | 6/20 [00:47<01:01,  4.36s/iter, val_loss=0.000475, val_og_loss=0.000475]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.82s/iter, val_loss=0.000475, val_og_loss=0.000475]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.82s/iter, val_loss=0.000474, val_og_loss=0.000475]
+
 40%|████      | 8/20 [00:50<00:41,  3.47s/iter, val_loss=0.000474, val_og_loss=0.000475]
+
 40%|████      | 8/20 [00:53<00:41,  3.47s/iter, val_loss=0.000473, val_og_loss=0.000475]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.23s/iter, val_loss=0.000473, val_og_loss=0.000475]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.23s/iter, val_loss=0.000473, val_og_loss=0.000475]
+
 50%|█████     | 10/20 [00:55<00:30,  3.07s/iter, val_loss=0.000473, val_og_loss=0.000475]
+
 50%|█████     | 10/20 [00:58<00:30,  3.07s/iter, val_loss=0.000473, val_og_loss=0.000475]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.97s/iter, val_loss=0.000473, val_og_loss=0.000475]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.97s/iter, val_loss=0.000472, val_og_loss=0.000475]
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.000472, val_og_loss=0.000475]
+
 60%|██████    | 12/20 [01:04<00:23,  2.89s/iter, val_loss=0.000472, val_og_loss=0.000475]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.000472, val_og_loss=0.000475]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000471, val_og_loss=0.000475]
+
 70%|███████   | 14/20 [01:06<00:16,  2.81s/iter, val_loss=0.000471, val_og_loss=0.000475]
+
 70%|███████   | 14/20 [01:09<00:16,  2.81s/iter, val_loss=0.000471, val_og_loss=0.000475]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.000471, val_og_loss=0.000475]
+
 75%|███████▌  | 15/20 [01:12<00:13,  2.78s/iter, val_loss=0.00047, val_og_loss=0.000475] 
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.00047, val_og_loss=0.000475]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.00047, val_og_loss=0.000475]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.75s/iter, val_loss=0.00047, val_og_loss=0.000475]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.75s/iter, val_loss=0.000469, val_og_loss=0.000475]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.000469, val_og_loss=0.000475]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.74s/iter, val_loss=0.000469, val_og_loss=0.000475]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.74s/iter, val_loss=0.000469, val_og_loss=0.000475]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.74s/iter, val_loss=0.000469, val_og_loss=0.000475]
+
100%|██████████| 20/20 [01:23<00:00,  2.73s/iter, val_loss=0.000469, val_og_loss=0.000475]
100%|██████████| 20/20 [01:23<00:00,  4.16s/iter, val_loss=0.000469, val_og_loss=0.000475]
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:45:16<5:03:44, 828.40s/it]INFO:ParoQuant:Best val loss: 0.00046648248098790646, Original val loss: 0.0004754799883812666
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:45:17<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:45:17<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:46:18<5:03:44, 828.40s/it]2025-12-22 16:15:34,088 - INFO - Best val loss: 0.00046648248098790646, Original val loss: 0.0004754799883812666
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:35<?, ?iter/s, val_loss=4.94e-5, val_og_loss=5e-5]
+
  5%|▌         | 1/20 [00:35<11:21, 35.85s/iter, val_loss=4.94e-5, val_og_loss=5e-5]
+
  5%|▌         | 1/20 [00:41<11:21, 35.85s/iter, val_loss=4.94e-5, val_og_loss=5e-5]
+
 10%|█         | 2/20 [00:41<05:23, 17.96s/iter, val_loss=4.94e-5, val_og_loss=5e-5]
+
 10%|█         | 2/20 [00:46<05:23, 17.96s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 15%|█▌        | 3/20 [00:46<03:27, 12.23s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 15%|█▌        | 3/20 [00:52<03:27, 12.23s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 20%|██        | 4/20 [00:52<02:32,  9.54s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 20%|██        | 4/20 [00:57<02:32,  9.54s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 25%|██▌       | 5/20 [00:57<02:00,  8.06s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 25%|██▌       | 5/20 [01:02<02:00,  8.06s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 30%|███       | 6/20 [01:02<01:40,  7.17s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 30%|███       | 6/20 [01:08<01:40,  7.17s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 35%|███▌      | 7/20 [01:08<01:25,  6.60s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 35%|███▌      | 7/20 [01:13<01:25,  6.60s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 40%|████      | 8/20 [01:13<01:14,  6.22s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 40%|████      | 8/20 [01:19<01:14,  6.22s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 45%|████▌     | 9/20 [01:19<01:05,  5.97s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 45%|████▌     | 9/20 [01:24<01:05,  5.97s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 50%|█████     | 10/20 [01:24<00:58,  5.80s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 50%|█████     | 10/20 [01:30<00:58,  5.80s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 55%|█████▌    | 11/20 [01:30<00:51,  5.68s/iter, val_loss=4.96e-5, val_og_loss=5e-5]
+
 55%|█████▌    | 11/20 [01:35<00:51,  5.68s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 60%|██████    | 12/20 [01:35<00:44,  5.60s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 60%|██████    | 12/20 [01:40<00:44,  5.60s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 65%|██████▌   | 13/20 [01:40<00:38,  5.55s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 65%|██████▌   | 13/20 [01:46<00:38,  5.55s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 70%|███████   | 14/20 [01:46<00:33,  5.51s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 70%|███████   | 14/20 [01:51<00:33,  5.51s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 75%|███████▌  | 15/20 [01:51<00:27,  5.48s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 75%|███████▌  | 15/20 [01:57<00:27,  5.48s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 80%|████████  | 16/20 [01:57<00:21,  5.46s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 80%|████████  | 16/20 [02:02<00:21,  5.46s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 85%|████████▌ | 17/20 [02:02<00:16,  5.45s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
 85%|████████▌ | 17/20 [02:07<00:16,  5.45s/iter, val_loss=4.94e-5, val_og_loss=5e-5]
+
 90%|█████████ | 18/20 [02:07<00:10,  5.44s/iter, val_loss=4.94e-5, val_og_loss=5e-5]
+
 90%|█████████ | 18/20 [02:13<00:10,  5.44s/iter, val_loss=4.94e-5, val_og_loss=5e-5]
+
 95%|█████████▌| 19/20 [02:13<00:05,  5.43s/iter, val_loss=4.94e-5, val_og_loss=5e-5]
+
 95%|█████████▌| 19/20 [02:18<00:05,  5.43s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
100%|██████████| 20/20 [02:18<00:00,  5.43s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
100%|██████████| 20/20 [02:18<00:00,  6.94s/iter, val_loss=4.95e-5, val_og_loss=5e-5]
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:48:36<5:03:44, 828.40s/it]INFO:ParoQuant:Best val loss: 4.9403512093704194e-05, Original val loss: 4.9954171117860824e-05
+
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:48:37<5:03:44, 828.40s/it]
                                                                       

Processing Blocks:  31%|███▏      | 10/32 [1:48:47<5:03:44, 828.40s/it]
Processing Blocks:  34%|███▍      | 11/32 [1:49:24<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:49:25<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:49:42<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:49:42<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:51:55<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:51:55<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:53:01<4:57:45, 850.73s/it]2025-12-22 16:18:54,832 - INFO - Best val loss: 4.9403512093704194e-05, Original val loss: 4.9954171117860824e-05
+[INFO] Saving Layer 10 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 11...
+[INFO] Starting Sequential Training for Layer 11
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.00174, val_og_loss=0.00212]
+
  5%|▌         | 1/20 [00:16<05:09, 16.30s/iter, val_loss=0.00174, val_og_loss=0.00212]
+
  5%|▌         | 1/20 [00:17<05:09, 16.30s/iter, val_loss=0.00175, val_og_loss=0.00212]
+
 10%|█         | 2/20 [00:17<02:18,  7.71s/iter, val_loss=0.00175, val_og_loss=0.00212]
+
 10%|█         | 2/20 [00:19<02:18,  7.71s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 15%|█▌        | 3/20 [00:19<01:24,  4.97s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 15%|█▌        | 3/20 [00:21<01:24,  4.97s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 20%|██        | 4/20 [00:21<00:58,  3.68s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 20%|██        | 4/20 [00:23<00:58,  3.68s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 25%|██▌       | 5/20 [00:23<00:44,  2.96s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 25%|██▌       | 5/20 [00:24<00:44,  2.96s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 30%|███       | 6/20 [00:24<00:35,  2.53s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 30%|███       | 6/20 [00:26<00:35,  2.53s/iter, val_loss=0.00176, val_og_loss=0.00212]
+
 35%|███▌      | 7/20 [00:26<00:29,  2.25s/iter, val_loss=0.00176, val_og_loss=0.00212]
+
 35%|███▌      | 7/20 [00:28<00:29,  2.25s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 40%|████      | 8/20 [00:28<00:24,  2.07s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 40%|████      | 8/20 [00:29<00:24,  2.07s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.95s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.95s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 50%|█████     | 10/20 [00:31<00:18,  1.87s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 50%|█████     | 10/20 [00:33<00:18,  1.87s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.81s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.81s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=0.00177, val_og_loss=0.00212]
+
 60%|██████    | 12/20 [00:36<00:14,  1.77s/iter, val_loss=0.00176, val_og_loss=0.00212]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.75s/iter, val_loss=0.00176, val_og_loss=0.00212]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.75s/iter, val_loss=0.00175, val_og_loss=0.00212]
+
 70%|███████   | 14/20 [00:38<00:10,  1.73s/iter, val_loss=0.00175, val_og_loss=0.00212]
+
 70%|███████   | 14/20 [00:39<00:10,  1.73s/iter, val_loss=0.00175, val_og_loss=0.00212]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.00175, val_og_loss=0.00212]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.71s/iter, val_loss=0.00175, val_og_loss=0.00212]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=0.00175, val_og_loss=0.00212]
+
 80%|████████  | 16/20 [00:43<00:06,  1.70s/iter, val_loss=0.00174, val_og_loss=0.00212]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.70s/iter, val_loss=0.00174, val_og_loss=0.00212]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=0.00174, val_og_loss=0.00212]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.00174, val_og_loss=0.00212]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.69s/iter, val_loss=0.00173, val_og_loss=0.00212]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=0.00173, val_og_loss=0.00212]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.69s/iter, val_loss=0.00173, val_og_loss=0.00212]
+
100%|██████████| 20/20 [00:48<00:00,  1.69s/iter, val_loss=0.00173, val_og_loss=0.00212]
100%|██████████| 20/20 [00:48<00:00,  2.42s/iter, val_loss=0.00173, val_og_loss=0.00212]
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:53:49<4:57:45, 850.73s/it]INFO:ParoQuant:Best val loss: 0.001731746830046177, Original val loss: 0.0021236767061054707
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:53:53<4:57:45, 850.73s/it]2025-12-22 16:24:07,805 - INFO - Best val loss: 0.001731746830046177, Original val loss: 0.0021236767061054707
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:17<?, ?iter/s, val_loss=0.00486, val_og_loss=0.00582]
+
  5%|▌         | 1/20 [00:17<05:41, 17.99s/iter, val_loss=0.00486, val_og_loss=0.00582]
+
  5%|▌         | 1/20 [00:19<05:41, 17.99s/iter, val_loss=0.00494, val_og_loss=0.00582]
+
 10%|█         | 2/20 [00:19<02:28,  8.23s/iter, val_loss=0.00494, val_og_loss=0.00582]
+
 10%|█         | 2/20 [00:20<02:28,  8.23s/iter, val_loss=0.00495, val_og_loss=0.00582]
+
 15%|█▌        | 3/20 [00:20<01:26,  5.11s/iter, val_loss=0.00495, val_og_loss=0.00582]
+
 15%|█▌        | 3/20 [00:22<01:26,  5.11s/iter, val_loss=0.00493, val_og_loss=0.00582]
+
 20%|██        | 4/20 [00:22<00:58,  3.64s/iter, val_loss=0.00493, val_og_loss=0.00582]
+
 20%|██        | 4/20 [00:23<00:58,  3.64s/iter, val_loss=0.00498, val_og_loss=0.00582]
+
 25%|██▌       | 5/20 [00:23<00:42,  2.84s/iter, val_loss=0.00498, val_og_loss=0.00582]
+
 25%|██▌       | 5/20 [00:25<00:42,  2.84s/iter, val_loss=0.00498, val_og_loss=0.00582]
+
 30%|███       | 6/20 [00:25<00:33,  2.36s/iter, val_loss=0.00498, val_og_loss=0.00582]
+
 30%|███       | 6/20 [00:26<00:33,  2.36s/iter, val_loss=0.00495, val_og_loss=0.00582]
+
 35%|███▌      | 7/20 [00:26<00:26,  2.05s/iter, val_loss=0.00495, val_og_loss=0.00582]
+
 35%|███▌      | 7/20 [00:27<00:26,  2.05s/iter, val_loss=0.00493, val_og_loss=0.00582]
+
 40%|████      | 8/20 [00:27<00:22,  1.84s/iter, val_loss=0.00493, val_og_loss=0.00582]
+
 40%|████      | 8/20 [00:29<00:22,  1.84s/iter, val_loss=0.00494, val_og_loss=0.00582]
+
 45%|████▌     | 9/20 [00:29<00:18,  1.70s/iter, val_loss=0.00494, val_og_loss=0.00582]
+
 45%|████▌     | 9/20 [00:30<00:18,  1.70s/iter, val_loss=0.00496, val_og_loss=0.00582]
+
 50%|█████     | 10/20 [00:30<00:16,  1.60s/iter, val_loss=0.00496, val_og_loss=0.00582]
+
 50%|█████     | 10/20 [00:32<00:16,  1.60s/iter, val_loss=0.00496, val_og_loss=0.00582]
+
 55%|█████▌    | 11/20 [00:32<00:13,  1.54s/iter, val_loss=0.00496, val_og_loss=0.00582]
+
 55%|█████▌    | 11/20 [00:33<00:13,  1.54s/iter, val_loss=0.00495, val_og_loss=0.00582]
+
 60%|██████    | 12/20 [00:33<00:11,  1.50s/iter, val_loss=0.00495, val_og_loss=0.00582]
+
 60%|██████    | 12/20 [00:34<00:11,  1.50s/iter, val_loss=0.00495, val_og_loss=0.00582]
+
 65%|██████▌   | 13/20 [00:34<00:10,  1.46s/iter, val_loss=0.00495, val_og_loss=0.00582]
+
 65%|██████▌   | 13/20 [00:36<00:10,  1.46s/iter, val_loss=0.00489, val_og_loss=0.00582]
+
 70%|███████   | 14/20 [00:36<00:08,  1.44s/iter, val_loss=0.00489, val_og_loss=0.00582]
+
 70%|███████   | 14/20 [00:37<00:08,  1.44s/iter, val_loss=0.0049, val_og_loss=0.00582] 
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.43s/iter, val_loss=0.0049, val_og_loss=0.00582]
+
 75%|███████▌  | 15/20 [00:38<00:07,  1.43s/iter, val_loss=0.0049, val_og_loss=0.00582]
+
 80%|████████  | 16/20 [00:38<00:05,  1.42s/iter, val_loss=0.0049, val_og_loss=0.00582]
+
 80%|████████  | 16/20 [00:40<00:05,  1.42s/iter, val_loss=0.00487, val_og_loss=0.00582]
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.41s/iter, val_loss=0.00487, val_og_loss=0.00582]
+
 85%|████████▌ | 17/20 [00:41<00:04,  1.41s/iter, val_loss=0.00484, val_og_loss=0.00582]
+
 90%|█████████ | 18/20 [00:41<00:02,  1.41s/iter, val_loss=0.00484, val_og_loss=0.00582]
+
 90%|█████████ | 18/20 [00:43<00:02,  1.41s/iter, val_loss=0.00483, val_og_loss=0.00582]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.40s/iter, val_loss=0.00483, val_og_loss=0.00582]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.40s/iter, val_loss=0.00485, val_og_loss=0.00582]
+
100%|██████████| 20/20 [00:44<00:00,  1.40s/iter, val_loss=0.00485, val_og_loss=0.00582]
100%|██████████| 20/20 [00:44<00:00,  2.23s/iter, val_loss=0.00485, val_og_loss=0.00582]
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:54:38<4:57:45, 850.73s/it]INFO:ParoQuant:Best val loss: 0.004831013735383749, Original val loss: 0.005822835490107536
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:54:42<4:57:45, 850.73s/it]2025-12-22 16:24:56,005 - INFO - Best val loss: 0.004831013735383749, Original val loss: 0.005822835490107536
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.000339, val_og_loss=0.000421]
+
  5%|▌         | 1/20 [00:16<05:10, 16.37s/iter, val_loss=0.000339, val_og_loss=0.000421]
+
  5%|▌         | 1/20 [00:17<05:10, 16.37s/iter, val_loss=0.00034, val_og_loss=0.000421] 
+
 10%|█         | 2/20 [00:17<02:16,  7.56s/iter, val_loss=0.00034, val_og_loss=0.000421]
+
 10%|█         | 2/20 [00:19<02:16,  7.56s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 15%|█▌        | 3/20 [00:19<01:20,  4.75s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 15%|█▌        | 3/20 [00:20<01:20,  4.75s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 20%|██        | 4/20 [00:20<00:54,  3.42s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 20%|██        | 4/20 [00:21<00:54,  3.42s/iter, val_loss=0.000342, val_og_loss=0.000421]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.69s/iter, val_loss=0.000342, val_og_loss=0.000421]
+
 25%|██▌       | 5/20 [00:23<00:40,  2.69s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 30%|███       | 6/20 [00:23<00:31,  2.26s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 30%|███       | 6/20 [00:24<00:31,  2.26s/iter, val_loss=0.000345, val_og_loss=0.000421]
+
 35%|███▌      | 7/20 [00:24<00:25,  1.97s/iter, val_loss=0.000345, val_og_loss=0.000421]
+
 35%|███▌      | 7/20 [00:26<00:25,  1.97s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 40%|████      | 8/20 [00:26<00:21,  1.79s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 40%|████      | 8/20 [00:27<00:21,  1.79s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 45%|████▌     | 9/20 [00:27<00:18,  1.66s/iter, val_loss=0.000343, val_og_loss=0.000421]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.66s/iter, val_loss=0.000342, val_og_loss=0.000421]
+
 50%|█████     | 10/20 [00:28<00:15,  1.59s/iter, val_loss=0.000342, val_og_loss=0.000421]
+
 50%|█████     | 10/20 [00:30<00:15,  1.59s/iter, val_loss=0.000342, val_og_loss=0.000421]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.53s/iter, val_loss=0.000342, val_og_loss=0.000421]
+
 55%|█████▌    | 11/20 [00:31<00:13,  1.53s/iter, val_loss=0.000341, val_og_loss=0.000421]
+
 60%|██████    | 12/20 [00:31<00:11,  1.48s/iter, val_loss=0.000341, val_og_loss=0.000421]
+
 60%|██████    | 12/20 [00:33<00:11,  1.48s/iter, val_loss=0.00034, val_og_loss=0.000421] 
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.45s/iter, val_loss=0.00034, val_og_loss=0.000421]
+
 65%|██████▌   | 13/20 [00:34<00:10,  1.45s/iter, val_loss=0.000341, val_og_loss=0.000421]
+
 70%|███████   | 14/20 [00:34<00:08,  1.43s/iter, val_loss=0.000341, val_og_loss=0.000421]
+
 70%|███████   | 14/20 [00:35<00:08,  1.43s/iter, val_loss=0.000339, val_og_loss=0.000421]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.42s/iter, val_loss=0.000339, val_og_loss=0.000421]
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.42s/iter, val_loss=0.000337, val_og_loss=0.000421]
+
 80%|████████  | 16/20 [00:37<00:05,  1.41s/iter, val_loss=0.000337, val_og_loss=0.000421]
+
 80%|████████  | 16/20 [00:38<00:05,  1.41s/iter, val_loss=0.000338, val_og_loss=0.000421]
+
 85%|████████▌ | 17/20 [00:38<00:04,  1.41s/iter, val_loss=0.000338, val_og_loss=0.000421]
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.41s/iter, val_loss=0.000338, val_og_loss=0.000421]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.40s/iter, val_loss=0.000338, val_og_loss=0.000421]
+
 90%|█████████ | 18/20 [00:41<00:02,  1.40s/iter, val_loss=0.000338, val_og_loss=0.000421]
+
 95%|█████████▌| 19/20 [00:41<00:01,  1.40s/iter, val_loss=0.000338, val_og_loss=0.000421]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.40s/iter, val_loss=0.000337, val_og_loss=0.000421]
+
100%|██████████| 20/20 [00:42<00:00,  1.40s/iter, val_loss=0.000337, val_og_loss=0.000421]
100%|██████████| 20/20 [00:42<00:00,  2.14s/iter, val_loss=0.000337, val_og_loss=0.000421]
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:55:25<4:57:45, 850.73s/it]INFO:ParoQuant:Best val loss: 0.00033696211175993085, Original val loss: 0.000421342410845682
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:55:28<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:55:28<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:55:58<4:57:45, 850.73s/it]2025-12-22 16:25:43,563 - INFO - Best val loss: 0.00033696211175993085, Original val loss: 0.000421342410845682
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=2.5e-5, val_og_loss=2.81e-5]
+
  5%|▌         | 1/20 [00:14<04:30, 14.23s/iter, val_loss=2.5e-5, val_og_loss=2.81e-5]
+
  5%|▌         | 1/20 [00:15<04:30, 14.23s/iter, val_loss=2.51e-5, val_og_loss=2.81e-5]
+
 10%|█         | 2/20 [00:15<02:03,  6.86s/iter, val_loss=2.51e-5, val_og_loss=2.81e-5]
+
 10%|█         | 2/20 [00:17<02:03,  6.86s/iter, val_loss=2.51e-5, val_og_loss=2.81e-5]
+
 15%|█▌        | 3/20 [00:17<01:16,  4.50s/iter, val_loss=2.51e-5, val_og_loss=2.81e-5]
+
 15%|█▌        | 3/20 [00:19<01:16,  4.50s/iter, val_loss=2.51e-5, val_og_loss=2.81e-5]
+
 20%|██        | 4/20 [00:19<00:54,  3.38s/iter, val_loss=2.51e-5, val_og_loss=2.81e-5]
+
 20%|██        | 4/20 [00:20<00:54,  3.38s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 25%|██▌       | 5/20 [00:20<00:41,  2.77s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.77s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 30%|███       | 6/20 [00:22<00:33,  2.40s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 30%|███       | 6/20 [00:24<00:33,  2.40s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.17s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 35%|███▌      | 7/20 [00:26<00:28,  2.17s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 40%|████      | 8/20 [00:26<00:24,  2.01s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 40%|████      | 8/20 [00:27<00:24,  2.01s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 45%|████▌     | 9/20 [00:27<00:21,  1.91s/iter, val_loss=2.52e-5, val_og_loss=2.81e-5]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.91s/iter, val_loss=2.51e-5, val_og_loss=2.81e-5]
+
 50%|█████     | 10/20 [00:29<00:18,  1.84s/iter, val_loss=2.51e-5, val_og_loss=2.81e-5]
+
 50%|█████     | 10/20 [00:31<00:18,  1.84s/iter, val_loss=2.5e-5, val_og_loss=2.81e-5] 
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.80s/iter, val_loss=2.5e-5, val_og_loss=2.81e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.80s/iter, val_loss=2.5e-5, val_og_loss=2.81e-5]
+
 60%|██████    | 12/20 [00:32<00:14,  1.77s/iter, val_loss=2.5e-5, val_og_loss=2.81e-5]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=2.5e-5, val_og_loss=2.81e-5]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.74s/iter, val_loss=2.5e-5, val_og_loss=2.81e-5]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.74s/iter, val_loss=2.5e-5, val_og_loss=2.81e-5]
+
 70%|███████   | 14/20 [00:36<00:10,  1.73s/iter, val_loss=2.5e-5, val_og_loss=2.81e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.73s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.72s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.72s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
 80%|████████  | 16/20 [00:39<00:06,  1.71s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
 80%|████████  | 16/20 [00:41<00:06,  1.71s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.70s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=2.48e-5, val_og_loss=2.81e-5]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=2.48e-5, val_og_loss=2.81e-5]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
100%|██████████| 20/20 [00:46<00:00,  1.69s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
100%|██████████| 20/20 [00:46<00:00,  2.31s/iter, val_loss=2.49e-5, val_og_loss=2.81e-5]
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:56:44<4:57:45, 850.73s/it]INFO:ParoQuant:Best val loss: 2.480992588971276e-05, Original val loss: 2.814852268784307e-05
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:56:45<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:56:45<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:57:28<4:57:45, 850.73s/it]2025-12-22 16:27:02,897 - INFO - Best val loss: 2.480992588971276e-05, Original val loss: 2.814852268784307e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.000782, val_og_loss=0.000798]
+
  5%|▌         | 1/20 [00:30<09:34, 30.22s/iter, val_loss=0.000782, val_og_loss=0.000798]
+
  5%|▌         | 1/20 [00:32<09:34, 30.22s/iter, val_loss=0.000784, val_og_loss=0.000798]
+
 10%|█         | 2/20 [00:32<04:12, 14.04s/iter, val_loss=0.000784, val_og_loss=0.000798]
+
 10%|█         | 2/20 [00:35<04:12, 14.04s/iter, val_loss=0.000788, val_og_loss=0.000798]
+
 15%|█▌        | 3/20 [00:35<02:30,  8.86s/iter, val_loss=0.000788, val_og_loss=0.000798]
+
 15%|█▌        | 3/20 [00:38<02:30,  8.86s/iter, val_loss=0.000789, val_og_loss=0.000798]
+
 20%|██        | 4/20 [00:38<01:42,  6.43s/iter, val_loss=0.000789, val_og_loss=0.000798]
+
 20%|██        | 4/20 [00:41<01:42,  6.43s/iter, val_loss=0.00079, val_og_loss=0.000798] 
+
 25%|██▌       | 5/20 [00:41<01:16,  5.09s/iter, val_loss=0.00079, val_og_loss=0.000798]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.09s/iter, val_loss=0.000788, val_og_loss=0.000798]
+
 30%|███       | 6/20 [00:43<01:00,  4.29s/iter, val_loss=0.000788, val_og_loss=0.000798]
+
 30%|███       | 6/20 [00:46<01:00,  4.29s/iter, val_loss=0.000789, val_og_loss=0.000798]
+
 35%|███▌      | 7/20 [00:46<00:49,  3.77s/iter, val_loss=0.000789, val_og_loss=0.000798]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.77s/iter, val_loss=0.000787, val_og_loss=0.000798]
+
 40%|████      | 8/20 [00:49<00:41,  3.44s/iter, val_loss=0.000787, val_og_loss=0.000798]
+
 40%|████      | 8/20 [00:51<00:41,  3.44s/iter, val_loss=0.000788, val_og_loss=0.000798]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.21s/iter, val_loss=0.000788, val_og_loss=0.000798]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.21s/iter, val_loss=0.000787, val_og_loss=0.000798]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.000787, val_og_loss=0.000798]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.000786, val_og_loss=0.000798]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.000786, val_og_loss=0.000798]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.000785, val_og_loss=0.000798]
+
 60%|██████    | 12/20 [01:00<00:23,  2.88s/iter, val_loss=0.000785, val_og_loss=0.000798]
+
 60%|██████    | 12/20 [01:02<00:23,  2.88s/iter, val_loss=0.000784, val_og_loss=0.000798]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.84s/iter, val_loss=0.000784, val_og_loss=0.000798]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.84s/iter, val_loss=0.000783, val_og_loss=0.000798]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.000783, val_og_loss=0.000798]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.000783, val_og_loss=0.000798]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.77s/iter, val_loss=0.000783, val_og_loss=0.000798]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.000783, val_og_loss=0.000798]
+
 80%|████████  | 16/20 [01:10<00:11,  2.76s/iter, val_loss=0.000783, val_og_loss=0.000798]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.000781, val_og_loss=0.000798]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.000781, val_og_loss=0.000798]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.000781, val_og_loss=0.000798]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.73s/iter, val_loss=0.000781, val_og_loss=0.000798]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.73s/iter, val_loss=0.000781, val_og_loss=0.000798]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.000781, val_og_loss=0.000798]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.00078, val_og_loss=0.000798] 
+
100%|██████████| 20/20 [01:21<00:00,  2.72s/iter, val_loss=0.00078, val_og_loss=0.000798]
100%|██████████| 20/20 [01:21<00:00,  4.09s/iter, val_loss=0.00078, val_og_loss=0.000798]
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:58:50<4:57:45, 850.73s/it]INFO:ParoQuant:Best val loss: 0.0007799750310368836, Original val loss: 0.0007981387316249311
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [1:58:55<4:57:45, 850.73s/it]2025-12-22 16:29:08,609 - INFO - Best val loss: 0.0007799750310368836, Original val loss: 0.0007981387316249311
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:32<?, ?iter/s, val_loss=0.000496, val_og_loss=0.000508]
+
  5%|▌         | 1/20 [00:32<10:20, 32.67s/iter, val_loss=0.000496, val_og_loss=0.000508]
+
  5%|▌         | 1/20 [00:35<10:20, 32.67s/iter, val_loss=0.000498, val_og_loss=0.000508]
+
 10%|█         | 2/20 [00:35<04:30, 15.05s/iter, val_loss=0.000498, val_og_loss=0.000508]
+
 10%|█         | 2/20 [00:38<04:30, 15.05s/iter, val_loss=0.0005, val_og_loss=0.000508]  
+
 15%|█▌        | 3/20 [00:38<02:40,  9.41s/iter, val_loss=0.0005, val_og_loss=0.000508]
+
 15%|█▌        | 3/20 [00:40<02:40,  9.41s/iter, val_loss=0.000502, val_og_loss=0.000508]
+
 20%|██        | 4/20 [00:40<01:48,  6.77s/iter, val_loss=0.000502, val_og_loss=0.000508]
+
 20%|██        | 4/20 [00:43<01:48,  6.77s/iter, val_loss=0.000503, val_og_loss=0.000508]
+
 25%|██▌       | 5/20 [00:43<01:19,  5.31s/iter, val_loss=0.000503, val_og_loss=0.000508]
+
 25%|██▌       | 5/20 [00:46<01:19,  5.31s/iter, val_loss=0.000504, val_og_loss=0.000508]
+
 30%|███       | 6/20 [00:46<01:01,  4.42s/iter, val_loss=0.000504, val_og_loss=0.000508]
+
 30%|███       | 6/20 [00:48<01:01,  4.42s/iter, val_loss=0.000502, val_og_loss=0.000508]
+
 35%|███▌      | 7/20 [00:48<00:50,  3.87s/iter, val_loss=0.000502, val_og_loss=0.000508]
+
 35%|███▌      | 7/20 [00:51<00:50,  3.87s/iter, val_loss=0.000503, val_og_loss=0.000508]
+
 40%|████      | 8/20 [00:51<00:41,  3.50s/iter, val_loss=0.000503, val_og_loss=0.000508]
+
 40%|████      | 8/20 [00:54<00:41,  3.50s/iter, val_loss=0.000502, val_og_loss=0.000508]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.26s/iter, val_loss=0.000502, val_og_loss=0.000508]
+
 45%|████▌     | 9/20 [00:57<00:35,  3.26s/iter, val_loss=0.000501, val_og_loss=0.000508]
+
 50%|█████     | 10/20 [00:57<00:30,  3.09s/iter, val_loss=0.000501, val_og_loss=0.000508]
+
 50%|█████     | 10/20 [00:59<00:30,  3.09s/iter, val_loss=0.0005, val_og_loss=0.000508]  
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.98s/iter, val_loss=0.0005, val_og_loss=0.000508]
+
 55%|█████▌    | 11/20 [01:02<00:26,  2.98s/iter, val_loss=0.000499, val_og_loss=0.000508]
+
 60%|██████    | 12/20 [01:02<00:23,  2.90s/iter, val_loss=0.000499, val_og_loss=0.000508]
+
 60%|██████    | 12/20 [01:05<00:23,  2.90s/iter, val_loss=0.000499, val_og_loss=0.000508]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.85s/iter, val_loss=0.000499, val_og_loss=0.000508]
+
 65%|██████▌   | 13/20 [01:07<00:19,  2.85s/iter, val_loss=0.000498, val_og_loss=0.000508]
+
 70%|███████   | 14/20 [01:07<00:16,  2.81s/iter, val_loss=0.000498, val_og_loss=0.000508]
+
 70%|███████   | 14/20 [01:10<00:16,  2.81s/iter, val_loss=0.000498, val_og_loss=0.000508]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.78s/iter, val_loss=0.000498, val_og_loss=0.000508]
+
 75%|███████▌  | 15/20 [01:13<00:13,  2.78s/iter, val_loss=0.000497, val_og_loss=0.000508]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.000497, val_og_loss=0.000508]
+
 80%|████████  | 16/20 [01:16<00:11,  2.76s/iter, val_loss=0.000497, val_og_loss=0.000508]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.000497, val_og_loss=0.000508]
+
 85%|████████▌ | 17/20 [01:18<00:08,  2.74s/iter, val_loss=0.000496, val_og_loss=0.000508]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.000496, val_og_loss=0.000508]
+
 90%|█████████ | 18/20 [01:21<00:05,  2.73s/iter, val_loss=0.000496, val_og_loss=0.000508]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.72s/iter, val_loss=0.000496, val_og_loss=0.000508]
+
 95%|█████████▌| 19/20 [01:24<00:02,  2.72s/iter, val_loss=0.000496, val_og_loss=0.000508]
+
100%|██████████| 20/20 [01:24<00:00,  2.72s/iter, val_loss=0.000496, val_og_loss=0.000508]
100%|██████████| 20/20 [01:24<00:00,  4.21s/iter, val_loss=0.000496, val_og_loss=0.000508]
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [2:00:19<4:57:45, 850.73s/it]INFO:ParoQuant:Best val loss: 0.0004957395140081644, Original val loss: 0.0005076070665381849
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [2:00:20<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [2:00:20<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [2:01:21<4:57:45, 850.73s/it]2025-12-22 16:30:37,511 - INFO - Best val loss: 0.0004957395140081644, Original val loss: 0.0005076070665381849
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:34<?, ?iter/s, val_loss=5.24e-5, val_og_loss=5.33e-5]
+
  5%|▌         | 1/20 [00:34<11:01, 34.82s/iter, val_loss=5.24e-5, val_og_loss=5.33e-5]
+
  5%|▌         | 1/20 [00:40<11:01, 34.82s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 10%|█         | 2/20 [00:40<05:15, 17.52s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 10%|█         | 2/20 [00:45<05:15, 17.52s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 15%|█▌        | 3/20 [00:45<03:23, 11.99s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 15%|█▌        | 3/20 [00:51<03:23, 11.99s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 20%|██        | 4/20 [00:51<02:30,  9.40s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 20%|██        | 4/20 [00:56<02:30,  9.40s/iter, val_loss=5.26e-5, val_og_loss=5.33e-5]
+
 25%|██▌       | 5/20 [00:56<01:59,  7.96s/iter, val_loss=5.26e-5, val_og_loss=5.33e-5]
+
 25%|██▌       | 5/20 [01:01<01:59,  7.96s/iter, val_loss=5.26e-5, val_og_loss=5.33e-5]
+
 30%|███       | 6/20 [01:01<01:39,  7.10s/iter, val_loss=5.26e-5, val_og_loss=5.33e-5]
+
 30%|███       | 6/20 [01:07<01:39,  7.10s/iter, val_loss=5.26e-5, val_og_loss=5.33e-5]
+
 35%|███▌      | 7/20 [01:07<01:25,  6.55s/iter, val_loss=5.26e-5, val_og_loss=5.33e-5]
+
 35%|███▌      | 7/20 [01:12<01:25,  6.55s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 40%|████      | 8/20 [01:12<01:14,  6.20s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 40%|████      | 8/20 [01:18<01:14,  6.20s/iter, val_loss=5.26e-5, val_og_loss=5.33e-5]
+
 45%|████▌     | 9/20 [01:18<01:05,  5.96s/iter, val_loss=5.26e-5, val_og_loss=5.33e-5]
+
 45%|████▌     | 9/20 [01:23<01:05,  5.96s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 50%|█████     | 10/20 [01:23<00:57,  5.79s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 50%|█████     | 10/20 [01:29<00:57,  5.79s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 55%|█████▌    | 11/20 [01:29<00:51,  5.68s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 55%|█████▌    | 11/20 [01:34<00:51,  5.68s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 60%|██████    | 12/20 [01:34<00:44,  5.60s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 60%|██████    | 12/20 [01:39<00:44,  5.60s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 65%|██████▌   | 13/20 [01:39<00:38,  5.55s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 65%|██████▌   | 13/20 [01:45<00:38,  5.55s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 70%|███████   | 14/20 [01:45<00:33,  5.51s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 70%|███████   | 14/20 [01:50<00:33,  5.51s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 75%|███████▌  | 15/20 [01:50<00:27,  5.48s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 75%|███████▌  | 15/20 [01:56<00:27,  5.48s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 80%|████████  | 16/20 [01:56<00:21,  5.46s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 80%|████████  | 16/20 [02:01<00:21,  5.46s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 85%|████████▌ | 17/20 [02:01<00:16,  5.45s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 85%|████████▌ | 17/20 [02:06<00:16,  5.45s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 90%|█████████ | 18/20 [02:06<00:10,  5.44s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 90%|█████████ | 18/20 [02:12<00:10,  5.44s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 95%|█████████▌| 19/20 [02:12<00:05,  5.43s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
 95%|█████████▌| 19/20 [02:17<00:05,  5.43s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
100%|██████████| 20/20 [02:17<00:00,  5.43s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
100%|██████████| 20/20 [02:17<00:00,  6.89s/iter, val_loss=5.25e-5, val_og_loss=5.33e-5]
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [2:03:39<4:57:45, 850.73s/it]INFO:ParoQuant:Best val loss: 5.244166823104024e-05, Original val loss: 5.329932173481211e-05
+
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [2:03:39<4:57:45, 850.73s/it]
                                                                       

Processing Blocks:  34%|███▍      | 11/32 [2:03:49<4:57:45, 850.73s/it]
Processing Blocks:  38%|███▊      | 12/32 [2:04:27<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:04:28<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:04:45<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:04:45<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:07:02<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:07:02<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:08:09<4:48:51, 866.58s/it]2025-12-22 16:33:57,161 - INFO - Best val loss: 5.244166823104024e-05, Original val loss: 5.329932173481211e-05
+[INFO] Saving Layer 11 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 12...
+[INFO] Starting Sequential Training for Layer 12
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:22<?, ?iter/s, val_loss=0.00155, val_og_loss=0.00181]
+
  5%|▌         | 1/20 [00:22<07:07, 22.51s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
  5%|▌         | 1/20 [00:24<07:07, 22.51s/iter, val_loss=0.00158, val_og_loss=0.00181]
+
 10%|█         | 2/20 [00:24<03:04, 10.27s/iter, val_loss=0.00158, val_og_loss=0.00181]
+
 10%|█         | 2/20 [00:25<03:04, 10.27s/iter, val_loss=0.00158, val_og_loss=0.00181]
+
 15%|█▌        | 3/20 [00:25<01:47,  6.35s/iter, val_loss=0.00158, val_og_loss=0.00181]
+
 15%|█▌        | 3/20 [00:27<01:47,  6.35s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 20%|██        | 4/20 [00:27<01:12,  4.51s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 20%|██        | 4/20 [00:29<01:12,  4.51s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 25%|██▌       | 5/20 [00:29<00:52,  3.49s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 25%|██▌       | 5/20 [00:30<00:52,  3.49s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 30%|███       | 6/20 [00:30<00:40,  2.88s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 30%|███       | 6/20 [00:32<00:40,  2.88s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 35%|███▌      | 7/20 [00:32<00:32,  2.49s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 35%|███▌      | 7/20 [00:34<00:32,  2.49s/iter, val_loss=0.00158, val_og_loss=0.00181]
+
 40%|████      | 8/20 [00:34<00:26,  2.23s/iter, val_loss=0.00158, val_og_loss=0.00181]
+
 40%|████      | 8/20 [00:36<00:26,  2.23s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 45%|████▌     | 9/20 [00:36<00:22,  2.06s/iter, val_loss=0.00159, val_og_loss=0.00181]
+
 45%|████▌     | 9/20 [00:37<00:22,  2.06s/iter, val_loss=0.00158, val_og_loss=0.00181]
+
 50%|█████     | 10/20 [00:37<00:19,  1.94s/iter, val_loss=0.00158, val_og_loss=0.00181]
+
 50%|█████     | 10/20 [00:39<00:19,  1.94s/iter, val_loss=0.00157, val_og_loss=0.00181]
+
 55%|█████▌    | 11/20 [00:39<00:16,  1.86s/iter, val_loss=0.00157, val_og_loss=0.00181]
+
 55%|█████▌    | 11/20 [00:41<00:16,  1.86s/iter, val_loss=0.00157, val_og_loss=0.00181]
+
 60%|██████    | 12/20 [00:41<00:14,  1.81s/iter, val_loss=0.00157, val_og_loss=0.00181]
+
 60%|██████    | 12/20 [00:42<00:14,  1.81s/iter, val_loss=0.00157, val_og_loss=0.00181]
+
 65%|██████▌   | 13/20 [00:42<00:12,  1.77s/iter, val_loss=0.00157, val_og_loss=0.00181]
+
 65%|██████▌   | 13/20 [00:44<00:12,  1.77s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 70%|███████   | 14/20 [00:44<00:10,  1.74s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 70%|███████   | 14/20 [00:46<00:10,  1.74s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 75%|███████▌  | 15/20 [00:46<00:08,  1.73s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 75%|███████▌  | 15/20 [00:47<00:08,  1.73s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 80%|████████  | 16/20 [00:47<00:06,  1.72s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 80%|████████  | 16/20 [00:49<00:06,  1.72s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 85%|████████▌ | 17/20 [00:49<00:05,  1.70s/iter, val_loss=0.00156, val_og_loss=0.00181]
+
 85%|████████▌ | 17/20 [00:51<00:05,  1.70s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 90%|█████████ | 18/20 [00:51<00:03,  1.70s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 90%|█████████ | 18/20 [00:52<00:03,  1.70s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 95%|█████████▌| 19/20 [00:52<00:01,  1.70s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
 95%|█████████▌| 19/20 [00:54<00:01,  1.70s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
100%|██████████| 20/20 [00:54<00:00,  1.69s/iter, val_loss=0.00155, val_og_loss=0.00181]
100%|██████████| 20/20 [00:54<00:00,  2.73s/iter, val_loss=0.00155, val_og_loss=0.00181]
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:09:04<4:48:51, 866.58s/it]INFO:ParoQuant:Best val loss: 0.00154593912884593, Original val loss: 0.0018130040261894464
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:09:07<4:48:51, 866.58s/it]2025-12-22 16:39:22,135 - INFO - Best val loss: 0.00154593912884593, Original val loss: 0.0018130040261894464
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.00388, val_og_loss=0.00458]
+
  5%|▌         | 1/20 [00:16<05:16, 16.67s/iter, val_loss=0.00388, val_og_loss=0.00458]
+
  5%|▌         | 1/20 [00:18<05:16, 16.67s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 10%|█         | 2/20 [00:18<02:18,  7.68s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 10%|█         | 2/20 [00:19<02:18,  7.68s/iter, val_loss=0.00396, val_og_loss=0.00458]
+
 15%|█▌        | 3/20 [00:19<01:21,  4.81s/iter, val_loss=0.00396, val_og_loss=0.00458]
+
 15%|█▌        | 3/20 [00:20<01:21,  4.81s/iter, val_loss=0.00394, val_og_loss=0.00458]
+
 20%|██        | 4/20 [00:20<00:55,  3.46s/iter, val_loss=0.00394, val_og_loss=0.00458]
+
 20%|██        | 4/20 [00:22<00:55,  3.46s/iter, val_loss=0.00396, val_og_loss=0.00458]
+
 25%|██▌       | 5/20 [00:22<00:40,  2.72s/iter, val_loss=0.00396, val_og_loss=0.00458]
+
 25%|██▌       | 5/20 [00:23<00:40,  2.72s/iter, val_loss=0.00401, val_og_loss=0.00458]
+
 30%|███       | 6/20 [00:23<00:31,  2.27s/iter, val_loss=0.00401, val_og_loss=0.00458]
+
 30%|███       | 6/20 [00:25<00:31,  2.27s/iter, val_loss=0.00396, val_og_loss=0.00458]
+
 35%|███▌      | 7/20 [00:25<00:25,  1.98s/iter, val_loss=0.00396, val_og_loss=0.00458]
+
 35%|███▌      | 7/20 [00:26<00:25,  1.98s/iter, val_loss=0.00394, val_og_loss=0.00458]
+
 40%|████      | 8/20 [00:26<00:21,  1.79s/iter, val_loss=0.00394, val_og_loss=0.00458]
+
 40%|████      | 8/20 [00:27<00:21,  1.79s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 45%|████▌     | 9/20 [00:27<00:18,  1.67s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 45%|████▌     | 9/20 [00:29<00:18,  1.67s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 50%|█████     | 10/20 [00:29<00:15,  1.59s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 50%|█████     | 10/20 [00:30<00:15,  1.59s/iter, val_loss=0.00396, val_og_loss=0.00458]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.53s/iter, val_loss=0.00396, val_og_loss=0.00458]
+
 55%|█████▌    | 11/20 [00:31<00:13,  1.53s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 60%|██████    | 12/20 [00:31<00:11,  1.48s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 60%|██████    | 12/20 [00:33<00:11,  1.48s/iter, val_loss=0.0039, val_og_loss=0.00458] 
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.45s/iter, val_loss=0.0039, val_og_loss=0.00458]
+
 65%|██████▌   | 13/20 [00:34<00:10,  1.45s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 70%|███████   | 14/20 [00:34<00:08,  1.44s/iter, val_loss=0.00395, val_og_loss=0.00458]
+
 70%|███████   | 14/20 [00:36<00:08,  1.44s/iter, val_loss=0.0039, val_og_loss=0.00458] 
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.43s/iter, val_loss=0.0039, val_og_loss=0.00458]
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.43s/iter, val_loss=0.00391, val_og_loss=0.00458]
+
 80%|████████  | 16/20 [00:37<00:05,  1.42s/iter, val_loss=0.00391, val_og_loss=0.00458]
+
 80%|████████  | 16/20 [00:38<00:05,  1.42s/iter, val_loss=0.0039, val_og_loss=0.00458] 
+
 85%|████████▌ | 17/20 [00:38<00:04,  1.41s/iter, val_loss=0.0039, val_og_loss=0.00458]
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.41s/iter, val_loss=0.00391, val_og_loss=0.00458]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.40s/iter, val_loss=0.00391, val_og_loss=0.00458]
+
 90%|█████████ | 18/20 [00:41<00:02,  1.40s/iter, val_loss=0.00387, val_og_loss=0.00458]
+
 95%|█████████▌| 19/20 [00:41<00:01,  1.41s/iter, val_loss=0.00387, val_og_loss=0.00458]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.41s/iter, val_loss=0.00385, val_og_loss=0.00458]
+
100%|██████████| 20/20 [00:43<00:00,  1.40s/iter, val_loss=0.00385, val_og_loss=0.00458]
100%|██████████| 20/20 [00:43<00:00,  2.16s/iter, val_loss=0.00385, val_og_loss=0.00458]
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:09:50<4:48:51, 866.58s/it]INFO:ParoQuant:Best val loss: 0.0038483559619635344, Original val loss: 0.004583181347697973
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:09:54<4:48:51, 866.58s/it]2025-12-22 16:40:08,751 - INFO - Best val loss: 0.0038483559619635344, Original val loss: 0.004583181347697973
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.000402, val_og_loss=0.000482]
+
  5%|▌         | 1/20 [00:16<05:09, 16.30s/iter, val_loss=0.000402, val_og_loss=0.000482]
+
  5%|▌         | 1/20 [00:17<05:09, 16.30s/iter, val_loss=0.000403, val_og_loss=0.000482]
+
 10%|█         | 2/20 [00:17<02:15,  7.53s/iter, val_loss=0.000403, val_og_loss=0.000482]
+
 10%|█         | 2/20 [00:19<02:15,  7.53s/iter, val_loss=0.000406, val_og_loss=0.000482]
+
 15%|█▌        | 3/20 [00:19<01:20,  4.72s/iter, val_loss=0.000406, val_og_loss=0.000482]
+
 15%|█▌        | 3/20 [00:20<01:20,  4.72s/iter, val_loss=0.000406, val_og_loss=0.000482]
+
 20%|██        | 4/20 [00:20<00:54,  3.41s/iter, val_loss=0.000406, val_og_loss=0.000482]
+
 20%|██        | 4/20 [00:21<00:54,  3.41s/iter, val_loss=0.000405, val_og_loss=0.000482]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.68s/iter, val_loss=0.000405, val_og_loss=0.000482]
+
 25%|██▌       | 5/20 [00:23<00:40,  2.68s/iter, val_loss=0.000406, val_og_loss=0.000482]
+
 30%|███       | 6/20 [00:23<00:31,  2.25s/iter, val_loss=0.000406, val_og_loss=0.000482]
+
 30%|███       | 6/20 [00:24<00:31,  2.25s/iter, val_loss=0.000405, val_og_loss=0.000482]
+
 35%|███▌      | 7/20 [00:24<00:25,  1.97s/iter, val_loss=0.000405, val_og_loss=0.000482]
+
 35%|███▌      | 7/20 [00:26<00:25,  1.97s/iter, val_loss=0.000404, val_og_loss=0.000482]
+
 40%|████      | 8/20 [00:26<00:21,  1.78s/iter, val_loss=0.000404, val_og_loss=0.000482]
+
 40%|████      | 8/20 [00:27<00:21,  1.78s/iter, val_loss=0.000404, val_og_loss=0.000482]
+
 45%|████▌     | 9/20 [00:27<00:18,  1.66s/iter, val_loss=0.000404, val_og_loss=0.000482]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.66s/iter, val_loss=0.000405, val_og_loss=0.000482]
+
 50%|█████     | 10/20 [00:28<00:15,  1.58s/iter, val_loss=0.000405, val_og_loss=0.000482]
+
 50%|█████     | 10/20 [00:30<00:15,  1.58s/iter, val_loss=0.000405, val_og_loss=0.000482]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.53s/iter, val_loss=0.000405, val_og_loss=0.000482]
+
 55%|█████▌    | 11/20 [00:31<00:13,  1.53s/iter, val_loss=0.000403, val_og_loss=0.000482]
+
 60%|██████    | 12/20 [00:31<00:11,  1.49s/iter, val_loss=0.000403, val_og_loss=0.000482]
+
 60%|██████    | 12/20 [00:33<00:11,  1.49s/iter, val_loss=0.000402, val_og_loss=0.000482]
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.46s/iter, val_loss=0.000402, val_og_loss=0.000482]
+
 65%|██████▌   | 13/20 [00:34<00:10,  1.46s/iter, val_loss=0.000402, val_og_loss=0.000482]
+
 70%|███████   | 14/20 [00:34<00:08,  1.44s/iter, val_loss=0.000402, val_og_loss=0.000482]
+
 70%|███████   | 14/20 [00:35<00:08,  1.44s/iter, val_loss=0.000401, val_og_loss=0.000482]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.43s/iter, val_loss=0.000401, val_og_loss=0.000482]
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.43s/iter, val_loss=0.0004, val_og_loss=0.000482]  
+
 80%|████████  | 16/20 [00:37<00:05,  1.42s/iter, val_loss=0.0004, val_og_loss=0.000482]
+
 80%|████████  | 16/20 [00:38<00:05,  1.42s/iter, val_loss=0.0004, val_og_loss=0.000482]
+
 85%|████████▌ | 17/20 [00:38<00:04,  1.41s/iter, val_loss=0.0004, val_og_loss=0.000482]
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.41s/iter, val_loss=0.000399, val_og_loss=0.000482]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.41s/iter, val_loss=0.000399, val_og_loss=0.000482]
+
 90%|█████████ | 18/20 [00:41<00:02,  1.41s/iter, val_loss=0.0004, val_og_loss=0.000482]  
+
 95%|█████████▌| 19/20 [00:41<00:01,  1.41s/iter, val_loss=0.0004, val_og_loss=0.000482]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.41s/iter, val_loss=0.000398, val_og_loss=0.000482]
+
100%|██████████| 20/20 [00:42<00:00,  1.41s/iter, val_loss=0.000398, val_og_loss=0.000482]
100%|██████████| 20/20 [00:42<00:00,  2.14s/iter, val_loss=0.000398, val_og_loss=0.000482]
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:10:36<4:48:51, 866.58s/it]INFO:ParoQuant:Best val loss: 0.00039822427788749337, Original val loss: 0.0004817236331291497
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:10:39<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:10:39<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:11:11<4:48:51, 866.58s/it]2025-12-22 16:40:54,854 - INFO - Best val loss: 0.00039822427788749337, Original val loss: 0.0004817236331291497
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:12<?, ?iter/s, val_loss=2.88e-5, val_og_loss=3.04e-5]
+
  5%|▌         | 1/20 [00:12<03:51, 12.16s/iter, val_loss=2.88e-5, val_og_loss=3.04e-5]
+
  5%|▌         | 1/20 [00:13<03:51, 12.16s/iter, val_loss=2.9e-5, val_og_loss=3.04e-5] 
+
 10%|█         | 2/20 [00:13<01:47,  6.00s/iter, val_loss=2.9e-5, val_og_loss=3.04e-5]
+
 10%|█         | 2/20 [00:15<01:47,  6.00s/iter, val_loss=2.88e-5, val_og_loss=3.04e-5]
+
 15%|█▌        | 3/20 [00:15<01:08,  4.03s/iter, val_loss=2.88e-5, val_og_loss=3.04e-5]
+
 15%|█▌        | 3/20 [00:17<01:08,  4.03s/iter, val_loss=2.89e-5, val_og_loss=3.04e-5]
+
 20%|██        | 4/20 [00:17<00:49,  3.11s/iter, val_loss=2.89e-5, val_og_loss=3.04e-5]
+
 20%|██        | 4/20 [00:18<00:49,  3.11s/iter, val_loss=2.89e-5, val_og_loss=3.04e-5]
+
 25%|██▌       | 5/20 [00:18<00:38,  2.59s/iter, val_loss=2.89e-5, val_og_loss=3.04e-5]
+
 25%|██▌       | 5/20 [00:20<00:38,  2.59s/iter, val_loss=2.9e-5, val_og_loss=3.04e-5] 
+
 30%|███       | 6/20 [00:20<00:31,  2.28s/iter, val_loss=2.9e-5, val_og_loss=3.04e-5]
+
 30%|███       | 6/20 [00:22<00:31,  2.28s/iter, val_loss=2.91e-5, val_og_loss=3.04e-5]
+
 35%|███▌      | 7/20 [00:22<00:27,  2.09s/iter, val_loss=2.91e-5, val_og_loss=3.04e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.09s/iter, val_loss=2.89e-5, val_og_loss=3.04e-5]
+
 40%|████      | 8/20 [00:23<00:23,  1.96s/iter, val_loss=2.89e-5, val_og_loss=3.04e-5]
+
 40%|████      | 8/20 [00:25<00:23,  1.96s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 45%|████▌     | 9/20 [00:25<00:20,  1.87s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.87s/iter, val_loss=2.88e-5, val_og_loss=3.04e-5]
+
 50%|█████     | 10/20 [00:27<00:18,  1.81s/iter, val_loss=2.88e-5, val_og_loss=3.04e-5]
+
 50%|█████     | 10/20 [00:29<00:18,  1.81s/iter, val_loss=2.88e-5, val_og_loss=3.04e-5]
+
 55%|█████▌    | 11/20 [00:29<00:15,  1.78s/iter, val_loss=2.88e-5, val_og_loss=3.04e-5]
+
 55%|█████▌    | 11/20 [00:30<00:15,  1.78s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 60%|██████    | 12/20 [00:30<00:13,  1.75s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 60%|██████    | 12/20 [00:32<00:13,  1.75s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 65%|██████▌   | 13/20 [00:32<00:12,  1.73s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 70%|███████   | 14/20 [00:34<00:10,  1.71s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.71s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 75%|███████▌  | 15/20 [00:35<00:08,  1.71s/iter, val_loss=2.87e-5, val_og_loss=3.04e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.71s/iter, val_loss=2.86e-5, val_og_loss=3.04e-5]
+
 80%|████████  | 16/20 [00:37<00:06,  1.70s/iter, val_loss=2.86e-5, val_og_loss=3.04e-5]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=2.86e-5, val_og_loss=3.04e-5]
+
 85%|████████▌ | 17/20 [00:39<00:05,  1.69s/iter, val_loss=2.86e-5, val_og_loss=3.04e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=2.85e-5, val_og_loss=3.04e-5]
+
 90%|█████████ | 18/20 [00:40<00:03,  1.69s/iter, val_loss=2.85e-5, val_og_loss=3.04e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=2.85e-5, val_og_loss=3.04e-5]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.69s/iter, val_loss=2.85e-5, val_og_loss=3.04e-5]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=2.85e-5, val_og_loss=3.04e-5]
+
100%|██████████| 20/20 [00:44<00:00,  1.68s/iter, val_loss=2.85e-5, val_og_loss=3.04e-5]
100%|██████████| 20/20 [00:44<00:00,  2.21s/iter, val_loss=2.85e-5, val_og_loss=3.04e-5]
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:11:55<4:48:51, 866.58s/it]INFO:ParoQuant:Best val loss: 2.8485837901826017e-05, Original val loss: 3.0395151043194346e-05
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:11:56<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:11:56<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:12:39<4:48:51, 866.58s/it]2025-12-22 16:42:13,665 - INFO - Best val loss: 2.8485837901826017e-05, Original val loss: 3.0395151043194346e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.000752, val_og_loss=0.000774]
+
  5%|▌         | 1/20 [00:30<09:45, 30.84s/iter, val_loss=0.000752, val_og_loss=0.000774]
+
  5%|▌         | 1/20 [00:33<09:45, 30.84s/iter, val_loss=0.000754, val_og_loss=0.000774]
+
 10%|█         | 2/20 [00:33<04:17, 14.29s/iter, val_loss=0.000754, val_og_loss=0.000774]
+
 10%|█         | 2/20 [00:36<04:17, 14.29s/iter, val_loss=0.000756, val_og_loss=0.000774]
+
 15%|█▌        | 3/20 [00:36<02:33,  9.01s/iter, val_loss=0.000756, val_og_loss=0.000774]
+
 15%|█▌        | 3/20 [00:38<02:33,  9.01s/iter, val_loss=0.000759, val_og_loss=0.000774]
+
 20%|██        | 4/20 [00:38<01:44,  6.52s/iter, val_loss=0.000759, val_og_loss=0.000774]
+
 20%|██        | 4/20 [00:41<01:44,  6.52s/iter, val_loss=0.000759, val_og_loss=0.000774]
+
 25%|██▌       | 5/20 [00:41<01:17,  5.15s/iter, val_loss=0.000759, val_og_loss=0.000774]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.15s/iter, val_loss=0.00076, val_og_loss=0.000774] 
+
 30%|███       | 6/20 [00:44<01:00,  4.32s/iter, val_loss=0.00076, val_og_loss=0.000774]
+
 30%|███       | 6/20 [00:47<01:00,  4.32s/iter, val_loss=0.00076, val_og_loss=0.000774]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.80s/iter, val_loss=0.00076, val_og_loss=0.000774]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.80s/iter, val_loss=0.00076, val_og_loss=0.000774]
+
 40%|████      | 8/20 [00:49<00:41,  3.45s/iter, val_loss=0.00076, val_og_loss=0.000774]
+
 40%|████      | 8/20 [00:52<00:41,  3.45s/iter, val_loss=0.000759, val_og_loss=0.000774]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.22s/iter, val_loss=0.000759, val_og_loss=0.000774]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.22s/iter, val_loss=0.000759, val_og_loss=0.000774]
+
 50%|█████     | 10/20 [00:55<00:30,  3.07s/iter, val_loss=0.000759, val_og_loss=0.000774]
+
 50%|█████     | 10/20 [00:58<00:30,  3.07s/iter, val_loss=0.000758, val_og_loss=0.000774]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.96s/iter, val_loss=0.000758, val_og_loss=0.000774]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.000758, val_og_loss=0.000774]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.000758, val_og_loss=0.000774]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.000757, val_og_loss=0.000774]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.84s/iter, val_loss=0.000757, val_og_loss=0.000774]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000756, val_og_loss=0.000774]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.000756, val_og_loss=0.000774]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.000755, val_og_loss=0.000774]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.78s/iter, val_loss=0.000755, val_og_loss=0.000774]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.000754, val_og_loss=0.000774]
+
 80%|████████  | 16/20 [01:11<00:11,  2.75s/iter, val_loss=0.000754, val_og_loss=0.000774]
+
 80%|████████  | 16/20 [01:14<00:11,  2.75s/iter, val_loss=0.000754, val_og_loss=0.000774]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.74s/iter, val_loss=0.000754, val_og_loss=0.000774]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.74s/iter, val_loss=0.000753, val_og_loss=0.000774]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.73s/iter, val_loss=0.000753, val_og_loss=0.000774]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.73s/iter, val_loss=0.000753, val_og_loss=0.000774]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.000753, val_og_loss=0.000774]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.000752, val_og_loss=0.000774]
+
100%|██████████| 20/20 [01:22<00:00,  2.72s/iter, val_loss=0.000752, val_og_loss=0.000774]
100%|██████████| 20/20 [01:22<00:00,  4.12s/iter, val_loss=0.000752, val_og_loss=0.000774]
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:14:01<4:48:51, 866.58s/it]INFO:ParoQuant:Best val loss: 0.0007520844228565693, Original val loss: 0.0007743178866803646
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:14:06<4:48:51, 866.58s/it]2025-12-22 16:44:19,934 - INFO - Best val loss: 0.0007520844228565693, Original val loss: 0.0007743178866803646
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.000508, val_og_loss=0.000523]
+
  5%|▌         | 1/20 [00:33<10:32, 33.28s/iter, val_loss=0.000508, val_og_loss=0.000523]
+
  5%|▌         | 1/20 [00:35<10:32, 33.28s/iter, val_loss=0.00051, val_og_loss=0.000523] 
+
 10%|█         | 2/20 [00:35<04:35, 15.30s/iter, val_loss=0.00051, val_og_loss=0.000523]
+
 10%|█         | 2/20 [00:38<04:35, 15.30s/iter, val_loss=0.000513, val_og_loss=0.000523]
+
 15%|█▌        | 3/20 [00:38<02:42,  9.56s/iter, val_loss=0.000513, val_og_loss=0.000523]
+
 15%|█▌        | 3/20 [00:41<02:42,  9.56s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 20%|██        | 4/20 [00:41<01:49,  6.85s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 20%|██        | 4/20 [00:44<01:49,  6.85s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 25%|██▌       | 5/20 [00:44<01:20,  5.36s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 25%|██▌       | 5/20 [00:46<01:20,  5.36s/iter, val_loss=0.000516, val_og_loss=0.000523]
+
 30%|███       | 6/20 [00:46<01:02,  4.46s/iter, val_loss=0.000516, val_og_loss=0.000523]
+
 30%|███       | 6/20 [00:49<01:02,  4.46s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 35%|███▌      | 7/20 [00:49<00:50,  3.89s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 35%|███▌      | 7/20 [00:52<00:50,  3.89s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 40%|████      | 8/20 [00:52<00:42,  3.52s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 40%|████      | 8/20 [00:54<00:42,  3.52s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.27s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 45%|████▌     | 9/20 [00:57<00:35,  3.27s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 50%|█████     | 10/20 [00:57<00:31,  3.10s/iter, val_loss=0.000515, val_og_loss=0.000523]
+
 50%|█████     | 10/20 [01:00<00:31,  3.10s/iter, val_loss=0.000514, val_og_loss=0.000523]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.98s/iter, val_loss=0.000514, val_og_loss=0.000523]
+
 55%|█████▌    | 11/20 [01:03<00:26,  2.98s/iter, val_loss=0.000513, val_og_loss=0.000523]
+
 60%|██████    | 12/20 [01:03<00:23,  2.90s/iter, val_loss=0.000513, val_og_loss=0.000523]
+
 60%|██████    | 12/20 [01:05<00:23,  2.90s/iter, val_loss=0.000513, val_og_loss=0.000523]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.85s/iter, val_loss=0.000513, val_og_loss=0.000523]
+
 65%|██████▌   | 13/20 [01:08<00:19,  2.85s/iter, val_loss=0.000512, val_og_loss=0.000523]
+
 70%|███████   | 14/20 [01:08<00:16,  2.81s/iter, val_loss=0.000512, val_og_loss=0.000523]
+
 70%|███████   | 14/20 [01:11<00:16,  2.81s/iter, val_loss=0.000511, val_og_loss=0.000523]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.000511, val_og_loss=0.000523]
+
 75%|███████▌  | 15/20 [01:14<00:13,  2.78s/iter, val_loss=0.000511, val_og_loss=0.000523]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.000511, val_og_loss=0.000523]
+
 80%|████████  | 16/20 [01:16<00:11,  2.76s/iter, val_loss=0.000511, val_og_loss=0.000523]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.75s/iter, val_loss=0.000511, val_og_loss=0.000523]
+
 85%|████████▌ | 17/20 [01:19<00:08,  2.75s/iter, val_loss=0.000511, val_og_loss=0.000523]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.73s/iter, val_loss=0.000511, val_og_loss=0.000523]
+
 90%|█████████ | 18/20 [01:22<00:05,  2.73s/iter, val_loss=0.00051, val_og_loss=0.000523] 
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.00051, val_og_loss=0.000523]
+
 95%|█████████▌| 19/20 [01:24<00:02,  2.73s/iter, val_loss=0.00051, val_og_loss=0.000523]
+
100%|██████████| 20/20 [01:24<00:00,  2.73s/iter, val_loss=0.00051, val_og_loss=0.000523]
100%|██████████| 20/20 [01:24<00:00,  4.24s/iter, val_loss=0.00051, val_og_loss=0.000523]
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:15:31<4:48:51, 866.58s/it]INFO:ParoQuant:Best val loss: 0.0005076294764876366, Original val loss: 0.0005225102650001645
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:15:32<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:15:32<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:16:33<4:48:51, 866.58s/it]2025-12-22 16:45:49,560 - INFO - Best val loss: 0.0005076294764876366, Original val loss: 0.0005225102650001645
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
  5%|▌         | 1/20 [00:33<10:36, 33.53s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
  5%|▌         | 1/20 [00:38<10:36, 33.53s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 10%|█         | 2/20 [00:38<05:05, 16.99s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 10%|█         | 2/20 [00:44<05:05, 16.99s/iter, val_loss=5.79e-5, val_og_loss=5.88e-5]
+
 15%|█▌        | 3/20 [00:44<03:19, 11.71s/iter, val_loss=5.79e-5, val_og_loss=5.88e-5]
+
 15%|█▌        | 3/20 [00:49<03:19, 11.71s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 20%|██        | 4/20 [00:49<02:27,  9.23s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 20%|██        | 4/20 [00:55<02:27,  9.23s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 25%|██▌       | 5/20 [00:55<01:57,  7.85s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 25%|██▌       | 5/20 [01:00<01:57,  7.85s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 30%|███       | 6/20 [01:00<01:38,  7.02s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 30%|███       | 6/20 [01:06<01:38,  7.02s/iter, val_loss=5.79e-5, val_og_loss=5.88e-5]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.50s/iter, val_loss=5.79e-5, val_og_loss=5.88e-5]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.50s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 40%|████      | 8/20 [01:11<01:13,  6.15s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 40%|████      | 8/20 [01:16<01:13,  6.15s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 45%|████▌     | 9/20 [01:16<01:05,  5.93s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.93s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 50%|█████     | 10/20 [01:22<00:57,  5.77s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 50%|█████     | 10/20 [01:27<00:57,  5.77s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 55%|█████▌    | 11/20 [01:27<00:50,  5.66s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 55%|█████▌    | 11/20 [01:33<00:50,  5.66s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 60%|██████    | 12/20 [01:33<00:44,  5.59s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 60%|██████    | 12/20 [01:38<00:44,  5.59s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.53s/iter, val_loss=5.78e-5, val_og_loss=5.88e-5]
+
 65%|██████▌   | 13/20 [01:43<00:38,  5.53s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 70%|███████   | 14/20 [01:43<00:32,  5.50s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 70%|███████   | 14/20 [01:49<00:32,  5.50s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.47s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 75%|███████▌  | 15/20 [01:54<00:27,  5.47s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 80%|████████  | 16/20 [01:54<00:21,  5.46s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 80%|████████  | 16/20 [02:00<00:21,  5.46s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.44s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.44s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.44s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.43s/iter, val_loss=5.77e-5, val_og_loss=5.88e-5]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.43s/iter, val_loss=5.76e-5, val_og_loss=5.88e-5]
+
100%|██████████| 20/20 [02:16<00:00,  5.43s/iter, val_loss=5.76e-5, val_og_loss=5.88e-5]
100%|██████████| 20/20 [02:16<00:00,  6.82s/iter, val_loss=5.76e-5, val_og_loss=5.88e-5]
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:18:49<4:48:51, 866.58s/it]INFO:ParoQuant:Best val loss: 5.764181332779117e-05, Original val loss: 5.878332376596518e-05
+
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:18:50<4:48:51, 866.58s/it]
                                                                       

Processing Blocks:  38%|███▊      | 12/32 [2:19:00<4:48:51, 866.58s/it]
Processing Blocks:  41%|████      | 13/32 [2:19:38<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:19:39<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:19:56<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:19:56<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:22:10<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:22:10<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:23:18<4:38:39, 879.99s/it]2025-12-22 16:49:07,574 - INFO - Best val loss: 5.764181332779117e-05, Original val loss: 5.878332376596518e-05
+[INFO] Saving Layer 12 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 13...
+[INFO] Starting Sequential Training for Layer 13
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00197, val_og_loss=0.00244]
+
  5%|▌         | 1/20 [00:15<04:53, 15.43s/iter, val_loss=0.00197, val_og_loss=0.00244]
+
  5%|▌         | 1/20 [00:17<04:53, 15.43s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 10%|█         | 2/20 [00:17<02:12,  7.37s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 10%|█         | 2/20 [00:18<02:12,  7.37s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 15%|█▌        | 3/20 [00:18<01:21,  4.78s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 15%|█▌        | 3/20 [00:20<01:21,  4.78s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 20%|██        | 4/20 [00:20<00:56,  3.56s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 20%|██        | 4/20 [00:22<00:56,  3.56s/iter, val_loss=0.002, val_og_loss=0.00244]  
+
 25%|██▌       | 5/20 [00:22<00:43,  2.89s/iter, val_loss=0.002, val_og_loss=0.00244]
+
 25%|██▌       | 5/20 [00:23<00:43,  2.89s/iter, val_loss=0.002, val_og_loss=0.00244]
+
 30%|███       | 6/20 [00:23<00:34,  2.48s/iter, val_loss=0.002, val_og_loss=0.00244]
+
 30%|███       | 6/20 [00:25<00:34,  2.48s/iter, val_loss=0.002, val_og_loss=0.00244]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.22s/iter, val_loss=0.002, val_og_loss=0.00244]
+
 35%|███▌      | 7/20 [00:27<00:28,  2.22s/iter, val_loss=0.002, val_og_loss=0.00244]
+
 40%|████      | 8/20 [00:27<00:24,  2.06s/iter, val_loss=0.002, val_og_loss=0.00244]
+
 40%|████      | 8/20 [00:29<00:24,  2.06s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.94s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.94s/iter, val_loss=0.00198, val_og_loss=0.00244]
+
 50%|█████     | 10/20 [00:30<00:18,  1.86s/iter, val_loss=0.00198, val_og_loss=0.00244]
+
 50%|█████     | 10/20 [00:32<00:18,  1.86s/iter, val_loss=0.002, val_og_loss=0.00244]  
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.80s/iter, val_loss=0.002, val_og_loss=0.00244]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.80s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=0.00199, val_og_loss=0.00244]
+
 60%|██████    | 12/20 [00:35<00:14,  1.77s/iter, val_loss=0.00198, val_og_loss=0.00244]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.74s/iter, val_loss=0.00198, val_og_loss=0.00244]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.74s/iter, val_loss=0.00197, val_og_loss=0.00244]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=0.00197, val_og_loss=0.00244]
+
 70%|███████   | 14/20 [00:39<00:10,  1.72s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.71s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=0.00197, val_og_loss=0.00244]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=0.00197, val_og_loss=0.00244]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.69s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
100%|██████████| 20/20 [00:47<00:00,  1.68s/iter, val_loss=0.00196, val_og_loss=0.00244]
100%|██████████| 20/20 [00:47<00:00,  2.37s/iter, val_loss=0.00196, val_og_loss=0.00244]
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:24:06<4:38:39, 879.99s/it]INFO:ParoQuant:Best val loss: 0.001956447958946228, Original val loss: 0.0024391962215304375
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:24:09<4:38:39, 879.99s/it]2025-12-22 16:54:24,434 - INFO - Best val loss: 0.001956447958946228, Original val loss: 0.0024391962215304375
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:18<?, ?iter/s, val_loss=0.00574, val_og_loss=0.00684]
+
  5%|▌         | 1/20 [00:18<05:48, 18.37s/iter, val_loss=0.00574, val_og_loss=0.00684]
+
  5%|▌         | 1/20 [00:19<05:48, 18.37s/iter, val_loss=0.0059, val_og_loss=0.00684] 
+
 10%|█         | 2/20 [00:19<02:30,  8.38s/iter, val_loss=0.0059, val_og_loss=0.00684]
+
 10%|█         | 2/20 [00:21<02:30,  8.38s/iter, val_loss=0.0059, val_og_loss=0.00684]
+
 15%|█▌        | 3/20 [00:21<01:28,  5.19s/iter, val_loss=0.0059, val_og_loss=0.00684]
+
 15%|█▌        | 3/20 [00:22<01:28,  5.19s/iter, val_loss=0.00589, val_og_loss=0.00684]
+
 20%|██        | 4/20 [00:22<00:59,  3.70s/iter, val_loss=0.00589, val_og_loss=0.00684]
+
 20%|██        | 4/20 [00:23<00:59,  3.70s/iter, val_loss=0.00588, val_og_loss=0.00684]
+
 25%|██▌       | 5/20 [00:23<00:42,  2.86s/iter, val_loss=0.00588, val_og_loss=0.00684]
+
 25%|██▌       | 5/20 [00:25<00:42,  2.86s/iter, val_loss=0.00589, val_og_loss=0.00684]
+
 30%|███       | 6/20 [00:25<00:33,  2.36s/iter, val_loss=0.00589, val_og_loss=0.00684]
+
 30%|███       | 6/20 [00:26<00:33,  2.36s/iter, val_loss=0.00589, val_og_loss=0.00684]
+
 35%|███▌      | 7/20 [00:26<00:26,  2.05s/iter, val_loss=0.00589, val_og_loss=0.00684]
+
 35%|███▌      | 7/20 [00:28<00:26,  2.05s/iter, val_loss=0.00586, val_og_loss=0.00684]
+
 40%|████      | 8/20 [00:28<00:22,  1.84s/iter, val_loss=0.00586, val_og_loss=0.00684]
+
 40%|████      | 8/20 [00:29<00:22,  1.84s/iter, val_loss=0.00581, val_og_loss=0.00684]
+
 45%|████▌     | 9/20 [00:29<00:18,  1.71s/iter, val_loss=0.00581, val_og_loss=0.00684]
+
 45%|████▌     | 9/20 [00:30<00:18,  1.71s/iter, val_loss=0.00586, val_og_loss=0.00684]
+
 50%|█████     | 10/20 [00:30<00:16,  1.61s/iter, val_loss=0.00586, val_og_loss=0.00684]
+
 50%|█████     | 10/20 [00:32<00:16,  1.61s/iter, val_loss=0.00584, val_og_loss=0.00684]
+
 55%|█████▌    | 11/20 [00:32<00:13,  1.54s/iter, val_loss=0.00584, val_og_loss=0.00684]
+
 55%|█████▌    | 11/20 [00:33<00:13,  1.54s/iter, val_loss=0.00583, val_og_loss=0.00684]
+
 60%|██████    | 12/20 [00:33<00:11,  1.50s/iter, val_loss=0.00583, val_og_loss=0.00684]
+
 60%|██████    | 12/20 [00:35<00:11,  1.50s/iter, val_loss=0.00582, val_og_loss=0.00684]
+
 65%|██████▌   | 13/20 [00:35<00:10,  1.47s/iter, val_loss=0.00582, val_og_loss=0.00684]
+
 65%|██████▌   | 13/20 [00:36<00:10,  1.47s/iter, val_loss=0.00582, val_og_loss=0.00684]
+
 70%|███████   | 14/20 [00:36<00:08,  1.45s/iter, val_loss=0.00582, val_og_loss=0.00684]
+
 70%|███████   | 14/20 [00:37<00:08,  1.45s/iter, val_loss=0.00584, val_og_loss=0.00684]
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.43s/iter, val_loss=0.00584, val_og_loss=0.00684]
+
 75%|███████▌  | 15/20 [00:39<00:07,  1.43s/iter, val_loss=0.00584, val_og_loss=0.00684]
+
 80%|████████  | 16/20 [00:39<00:05,  1.42s/iter, val_loss=0.00584, val_og_loss=0.00684]
+
 80%|████████  | 16/20 [00:40<00:05,  1.42s/iter, val_loss=0.0058, val_og_loss=0.00684] 
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.41s/iter, val_loss=0.0058, val_og_loss=0.00684]
+
 85%|████████▌ | 17/20 [00:42<00:04,  1.41s/iter, val_loss=0.0058, val_og_loss=0.00684]
+
 90%|█████████ | 18/20 [00:42<00:02,  1.42s/iter, val_loss=0.0058, val_og_loss=0.00684]
+
 90%|█████████ | 18/20 [00:43<00:02,  1.42s/iter, val_loss=0.00578, val_og_loss=0.00684]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.41s/iter, val_loss=0.00578, val_og_loss=0.00684]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.41s/iter, val_loss=0.00579, val_og_loss=0.00684]
+
100%|██████████| 20/20 [00:44<00:00,  1.41s/iter, val_loss=0.00579, val_og_loss=0.00684]
100%|██████████| 20/20 [00:44<00:00,  2.25s/iter, val_loss=0.00579, val_og_loss=0.00684]
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:24:54<4:38:39, 879.99s/it]INFO:ParoQuant:Best val loss: 0.005736387800425291, Original val loss: 0.006838664878159761
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:24:58<4:38:39, 879.99s/it]2025-12-22 16:55:12,871 - INFO - Best val loss: 0.005736387800425291, Original val loss: 0.006838664878159761
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:20<?, ?iter/s, val_loss=0.000434, val_og_loss=0.000544]
+
  5%|▌         | 1/20 [00:20<06:23, 20.16s/iter, val_loss=0.000434, val_og_loss=0.000544]
+
  5%|▌         | 1/20 [00:21<06:23, 20.16s/iter, val_loss=0.000436, val_og_loss=0.000544]
+
 10%|█         | 2/20 [00:21<02:44,  9.12s/iter, val_loss=0.000436, val_og_loss=0.000544]
+
 10%|█         | 2/20 [00:22<02:44,  9.12s/iter, val_loss=0.000436, val_og_loss=0.000544]
+
 15%|█▌        | 3/20 [00:22<01:35,  5.59s/iter, val_loss=0.000436, val_og_loss=0.000544]
+
 15%|█▌        | 3/20 [00:24<01:35,  5.59s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 20%|██        | 4/20 [00:24<01:02,  3.94s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 20%|██        | 4/20 [00:25<01:02,  3.94s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 25%|██▌       | 5/20 [00:25<00:45,  3.02s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 25%|██▌       | 5/20 [00:27<00:45,  3.02s/iter, val_loss=0.000438, val_og_loss=0.000544]
+
 30%|███       | 6/20 [00:27<00:34,  2.47s/iter, val_loss=0.000438, val_og_loss=0.000544]
+
 30%|███       | 6/20 [00:28<00:34,  2.47s/iter, val_loss=0.000439, val_og_loss=0.000544]
+
 35%|███▌      | 7/20 [00:28<00:27,  2.12s/iter, val_loss=0.000439, val_og_loss=0.000544]
+
 35%|███▌      | 7/20 [00:29<00:27,  2.12s/iter, val_loss=0.000439, val_og_loss=0.000544]
+
 40%|████      | 8/20 [00:29<00:22,  1.89s/iter, val_loss=0.000439, val_og_loss=0.000544]
+
 40%|████      | 8/20 [00:31<00:22,  1.89s/iter, val_loss=0.000438, val_og_loss=0.000544]
+
 45%|████▌     | 9/20 [00:31<00:19,  1.73s/iter, val_loss=0.000438, val_og_loss=0.000544]
+
 45%|████▌     | 9/20 [00:32<00:19,  1.73s/iter, val_loss=0.000439, val_og_loss=0.000544]
+
 50%|█████     | 10/20 [00:32<00:16,  1.63s/iter, val_loss=0.000439, val_og_loss=0.000544]
+
 50%|█████     | 10/20 [00:34<00:16,  1.63s/iter, val_loss=0.000438, val_og_loss=0.000544]
+
 55%|█████▌    | 11/20 [00:34<00:14,  1.56s/iter, val_loss=0.000438, val_og_loss=0.000544]
+
 55%|█████▌    | 11/20 [00:35<00:14,  1.56s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 60%|██████    | 12/20 [00:35<00:12,  1.51s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 60%|██████    | 12/20 [00:36<00:12,  1.51s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 65%|██████▌   | 13/20 [00:36<00:10,  1.47s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 65%|██████▌   | 13/20 [00:38<00:10,  1.47s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 70%|███████   | 14/20 [00:38<00:08,  1.45s/iter, val_loss=0.000437, val_og_loss=0.000544]
+
 70%|███████   | 14/20 [00:39<00:08,  1.45s/iter, val_loss=0.000436, val_og_loss=0.000544]
+
 75%|███████▌  | 15/20 [00:39<00:07,  1.44s/iter, val_loss=0.000436, val_og_loss=0.000544]
+
 75%|███████▌  | 15/20 [00:41<00:07,  1.44s/iter, val_loss=0.000435, val_og_loss=0.000544]
+
 80%|████████  | 16/20 [00:41<00:05,  1.42s/iter, val_loss=0.000435, val_og_loss=0.000544]
+
 80%|████████  | 16/20 [00:42<00:05,  1.42s/iter, val_loss=0.000436, val_og_loss=0.000544]
+
 85%|████████▌ | 17/20 [00:42<00:04,  1.41s/iter, val_loss=0.000436, val_og_loss=0.000544]
+
 85%|████████▌ | 17/20 [00:43<00:04,  1.41s/iter, val_loss=0.000434, val_og_loss=0.000544]
+
 90%|█████████ | 18/20 [00:43<00:02,  1.41s/iter, val_loss=0.000434, val_og_loss=0.000544]
+
 90%|█████████ | 18/20 [00:45<00:02,  1.41s/iter, val_loss=0.000435, val_og_loss=0.000544]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.41s/iter, val_loss=0.000435, val_og_loss=0.000544]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.41s/iter, val_loss=0.000434, val_og_loss=0.000544]
+
100%|██████████| 20/20 [00:46<00:00,  1.40s/iter, val_loss=0.000434, val_og_loss=0.000544]
100%|██████████| 20/20 [00:46<00:00,  2.33s/iter, val_loss=0.000434, val_og_loss=0.000544]
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:25:44<4:38:39, 879.99s/it]INFO:ParoQuant:Best val loss: 0.00043427228229120374, Original val loss: 0.0005440061213448644
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:25:47<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:25:47<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:26:19<4:38:39, 879.99s/it]2025-12-22 16:56:02,750 - INFO - Best val loss: 0.00043427228229120374, Original val loss: 0.0005440061213448644
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
  5%|▌         | 1/20 [00:13<04:12, 13.30s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
  5%|▌         | 1/20 [00:14<04:12, 13.30s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 10%|█         | 2/20 [00:14<01:56,  6.47s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 10%|█         | 2/20 [00:16<01:56,  6.47s/iter, val_loss=3.74e-5, val_og_loss=4.04e-5]
+
 15%|█▌        | 3/20 [00:16<01:12,  4.29s/iter, val_loss=3.74e-5, val_og_loss=4.04e-5]
+
 15%|█▌        | 3/20 [00:18<01:12,  4.29s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 20%|██        | 4/20 [00:18<00:52,  3.26s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 20%|██        | 4/20 [00:20<00:52,  3.26s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 25%|██▌       | 5/20 [00:20<00:40,  2.69s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.69s/iter, val_loss=3.74e-5, val_og_loss=4.04e-5]
+
 30%|███       | 6/20 [00:21<00:32,  2.35s/iter, val_loss=3.74e-5, val_og_loss=4.04e-5]
+
 30%|███       | 6/20 [00:23<00:32,  2.35s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.13s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 35%|███▌      | 7/20 [00:25<00:27,  2.13s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 40%|████      | 8/20 [00:25<00:23,  1.99s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 40%|████      | 8/20 [00:26<00:23,  1.99s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 45%|████▌     | 9/20 [00:26<00:20,  1.89s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.89s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 50%|█████     | 10/20 [00:28<00:18,  1.83s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.83s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.79s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.79s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 60%|██████    | 12/20 [00:31<00:14,  1.75s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 60%|██████    | 12/20 [00:33<00:14,  1.75s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.73s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.71s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 70%|███████   | 14/20 [00:36<00:10,  1.71s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 75%|███████▌  | 15/20 [00:36<00:08,  1.70s/iter, val_loss=3.73e-5, val_og_loss=4.04e-5]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.70s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 80%|████████  | 16/20 [00:38<00:06,  1.70s/iter, val_loss=3.72e-5, val_og_loss=4.04e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=3.71e-5, val_og_loss=4.04e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=3.71e-5, val_og_loss=4.04e-5]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.69s/iter, val_loss=3.7e-5, val_og_loss=4.04e-5] 
+
 90%|█████████ | 18/20 [00:41<00:03,  1.69s/iter, val_loss=3.7e-5, val_og_loss=4.04e-5]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=3.7e-5, val_og_loss=4.04e-5]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.68s/iter, val_loss=3.7e-5, val_og_loss=4.04e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.68s/iter, val_loss=3.7e-5, val_og_loss=4.04e-5]
+
100%|██████████| 20/20 [00:45<00:00,  1.69s/iter, val_loss=3.7e-5, val_og_loss=4.04e-5]
100%|██████████| 20/20 [00:45<00:00,  2.26s/iter, val_loss=3.7e-5, val_og_loss=4.04e-5]
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:27:04<4:38:39, 879.99s/it]INFO:ParoQuant:Best val loss: 3.702418689499609e-05, Original val loss: 4.035399979329668e-05
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:27:05<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:27:05<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:27:48<4:38:39, 879.99s/it]2025-12-22 16:57:22,511 - INFO - Best val loss: 3.702418689499609e-05, Original val loss: 4.035399979329668e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:26<?, ?iter/s, val_loss=0.000839, val_og_loss=0.00086]
+
  5%|▌         | 1/20 [00:26<08:21, 26.40s/iter, val_loss=0.000839, val_og_loss=0.00086]
+
  5%|▌         | 1/20 [00:29<08:21, 26.40s/iter, val_loss=0.000842, val_og_loss=0.00086]
+
 10%|█         | 2/20 [00:29<03:44, 12.47s/iter, val_loss=0.000842, val_og_loss=0.00086]
+
 10%|█         | 2/20 [00:31<03:44, 12.47s/iter, val_loss=0.000846, val_og_loss=0.00086]
+
 15%|█▌        | 3/20 [00:31<02:16,  8.02s/iter, val_loss=0.000846, val_og_loss=0.00086]
+
 15%|█▌        | 3/20 [00:34<02:16,  8.02s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 20%|██        | 4/20 [00:34<01:34,  5.92s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 20%|██        | 4/20 [00:37<01:34,  5.92s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 25%|██▌       | 5/20 [00:37<01:11,  4.76s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 25%|██▌       | 5/20 [00:39<01:11,  4.76s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 30%|███       | 6/20 [00:39<00:56,  4.07s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 30%|███       | 6/20 [00:42<00:56,  4.07s/iter, val_loss=0.000848, val_og_loss=0.00086]
+
 35%|███▌      | 7/20 [00:42<00:47,  3.63s/iter, val_loss=0.000848, val_og_loss=0.00086]
+
 35%|███▌      | 7/20 [00:45<00:47,  3.63s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 40%|████      | 8/20 [00:45<00:40,  3.34s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 40%|████      | 8/20 [00:48<00:40,  3.34s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 45%|████▌     | 9/20 [00:48<00:34,  3.15s/iter, val_loss=0.000847, val_og_loss=0.00086]
+
 45%|████▌     | 9/20 [00:50<00:34,  3.15s/iter, val_loss=0.000846, val_og_loss=0.00086]
+
 50%|█████     | 10/20 [00:50<00:30,  3.02s/iter, val_loss=0.000846, val_og_loss=0.00086]
+
 50%|█████     | 10/20 [00:53<00:30,  3.02s/iter, val_loss=0.000845, val_og_loss=0.00086]
+
 55%|█████▌    | 11/20 [00:53<00:26,  2.93s/iter, val_loss=0.000845, val_og_loss=0.00086]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.93s/iter, val_loss=0.000844, val_og_loss=0.00086]
+
 60%|██████    | 12/20 [00:56<00:22,  2.86s/iter, val_loss=0.000844, val_og_loss=0.00086]
+
 60%|██████    | 12/20 [00:59<00:22,  2.86s/iter, val_loss=0.000843, val_og_loss=0.00086]
+
 65%|██████▌   | 13/20 [00:59<00:19,  2.82s/iter, val_loss=0.000843, val_og_loss=0.00086]
+
 65%|██████▌   | 13/20 [01:01<00:19,  2.82s/iter, val_loss=0.000842, val_og_loss=0.00086]
+
 70%|███████   | 14/20 [01:01<00:16,  2.79s/iter, val_loss=0.000842, val_og_loss=0.00086]
+
 70%|███████   | 14/20 [01:04<00:16,  2.79s/iter, val_loss=0.000841, val_og_loss=0.00086]
+
 75%|███████▌  | 15/20 [01:04<00:13,  2.77s/iter, val_loss=0.000841, val_og_loss=0.00086]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.000841, val_og_loss=0.00086]
+
 80%|████████  | 16/20 [01:07<00:11,  2.76s/iter, val_loss=0.000841, val_og_loss=0.00086]
+
 80%|████████  | 16/20 [01:09<00:11,  2.76s/iter, val_loss=0.00084, val_og_loss=0.00086] 
+
 85%|████████▌ | 17/20 [01:09<00:08,  2.75s/iter, val_loss=0.00084, val_og_loss=0.00086]
+
 85%|████████▌ | 17/20 [01:12<00:08,  2.75s/iter, val_loss=0.00084, val_og_loss=0.00086]
+
 90%|█████████ | 18/20 [01:12<00:05,  2.74s/iter, val_loss=0.00084, val_og_loss=0.00086]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.74s/iter, val_loss=0.000839, val_og_loss=0.00086]
+
 95%|█████████▌| 19/20 [01:15<00:02,  2.74s/iter, val_loss=0.000839, val_og_loss=0.00086]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.74s/iter, val_loss=0.000839, val_og_loss=0.00086]
+
100%|██████████| 20/20 [01:18<00:00,  2.73s/iter, val_loss=0.000839, val_og_loss=0.00086]
100%|██████████| 20/20 [01:18<00:00,  3.91s/iter, val_loss=0.000839, val_og_loss=0.00086]
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:29:06<4:38:39, 879.99s/it]INFO:ParoQuant:Best val loss: 0.0008387299021705985, Original val loss: 0.0008600431028753519
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:29:11<4:38:39, 879.99s/it]2025-12-22 16:59:24,885 - INFO - Best val loss: 0.0008387299021705985, Original val loss: 0.0008600431028753519
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.000574, val_og_loss=0.000591]
+
  5%|▌         | 1/20 [00:31<09:58, 31.50s/iter, val_loss=0.000574, val_og_loss=0.000591]
+
  5%|▌         | 1/20 [00:34<09:58, 31.50s/iter, val_loss=0.000577, val_og_loss=0.000591]
+
 10%|█         | 2/20 [00:34<04:22, 14.57s/iter, val_loss=0.000577, val_og_loss=0.000591]
+
 10%|█         | 2/20 [00:36<04:22, 14.57s/iter, val_loss=0.00058, val_og_loss=0.000591] 
+
 15%|█▌        | 3/20 [00:36<02:35,  9.16s/iter, val_loss=0.00058, val_og_loss=0.000591]
+
 15%|█▌        | 3/20 [00:39<02:35,  9.16s/iter, val_loss=0.000581, val_og_loss=0.000591]
+
 20%|██        | 4/20 [00:39<01:45,  6.62s/iter, val_loss=0.000581, val_og_loss=0.000591]
+
 20%|██        | 4/20 [00:42<01:45,  6.62s/iter, val_loss=0.000582, val_og_loss=0.000591]
+
 25%|██▌       | 5/20 [00:42<01:18,  5.21s/iter, val_loss=0.000582, val_og_loss=0.000591]
+
 25%|██▌       | 5/20 [00:45<01:18,  5.21s/iter, val_loss=0.000582, val_og_loss=0.000591]
+
 30%|███       | 6/20 [00:45<01:01,  4.36s/iter, val_loss=0.000582, val_og_loss=0.000591]
+
 30%|███       | 6/20 [00:47<01:01,  4.36s/iter, val_loss=0.000583, val_og_loss=0.000591]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.83s/iter, val_loss=0.000583, val_og_loss=0.000591]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.83s/iter, val_loss=0.000583, val_og_loss=0.000591]
+
 40%|████      | 8/20 [00:50<00:41,  3.48s/iter, val_loss=0.000583, val_og_loss=0.000591]
+
 40%|████      | 8/20 [00:53<00:41,  3.48s/iter, val_loss=0.000583, val_og_loss=0.000591]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.24s/iter, val_loss=0.000583, val_og_loss=0.000591]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.24s/iter, val_loss=0.000582, val_og_loss=0.000591]
+
 50%|█████     | 10/20 [00:55<00:30,  3.08s/iter, val_loss=0.000582, val_og_loss=0.000591]
+
 50%|█████     | 10/20 [00:58<00:30,  3.08s/iter, val_loss=0.000581, val_og_loss=0.000591]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.97s/iter, val_loss=0.000581, val_og_loss=0.000591]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.97s/iter, val_loss=0.00058, val_og_loss=0.000591] 
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.00058, val_og_loss=0.000591]
+
 60%|██████    | 12/20 [01:04<00:23,  2.89s/iter, val_loss=0.000579, val_og_loss=0.000591]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.000579, val_og_loss=0.000591]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000579, val_og_loss=0.000591]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.000579, val_og_loss=0.000591]
+
 70%|███████   | 14/20 [01:09<00:16,  2.80s/iter, val_loss=0.000578, val_og_loss=0.000591]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.000578, val_og_loss=0.000591]
+
 75%|███████▌  | 15/20 [01:12<00:13,  2.78s/iter, val_loss=0.000577, val_og_loss=0.000591]
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.000577, val_og_loss=0.000591]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.000576, val_og_loss=0.000591]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.74s/iter, val_loss=0.000576, val_og_loss=0.000591]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.74s/iter, val_loss=0.000576, val_og_loss=0.000591]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.000576, val_og_loss=0.000591]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.74s/iter, val_loss=0.000575, val_og_loss=0.000591]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.000575, val_og_loss=0.000591]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.73s/iter, val_loss=0.000575, val_og_loss=0.000591]
+
100%|██████████| 20/20 [01:23<00:00,  2.72s/iter, val_loss=0.000575, val_og_loss=0.000591]
100%|██████████| 20/20 [01:23<00:00,  4.16s/iter, val_loss=0.000575, val_og_loss=0.000591]
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:30:34<4:38:39, 879.99s/it]INFO:ParoQuant:Best val loss: 0.0005744157242588699, Original val loss: 0.0005914687062613666
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:30:35<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:30:35<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:31:35<4:38:39, 879.99s/it]2025-12-22 17:00:52,441 - INFO - Best val loss: 0.0005744157242588699, Original val loss: 0.0005914687062613666
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:34<?, ?iter/s, val_loss=7e-5, val_og_loss=7.11e-5]
+
  5%|▌         | 1/20 [00:34<10:48, 34.16s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
  5%|▌         | 1/20 [00:39<10:48, 34.16s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 10%|█         | 2/20 [00:39<05:10, 17.25s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 10%|█         | 2/20 [00:44<05:10, 17.25s/iter, val_loss=7.02e-5, val_og_loss=7.11e-5]
+
 15%|█▌        | 3/20 [00:44<03:21, 11.84s/iter, val_loss=7.02e-5, val_og_loss=7.11e-5]
+
 15%|█▌        | 3/20 [00:50<03:21, 11.84s/iter, val_loss=7.02e-5, val_og_loss=7.11e-5]
+
 20%|██        | 4/20 [00:50<02:28,  9.31s/iter, val_loss=7.02e-5, val_og_loss=7.11e-5]
+
 20%|██        | 4/20 [00:55<02:28,  9.31s/iter, val_loss=7.02e-5, val_og_loss=7.11e-5]
+
 25%|██▌       | 5/20 [00:55<01:58,  7.91s/iter, val_loss=7.02e-5, val_og_loss=7.11e-5]
+
 25%|██▌       | 5/20 [01:01<01:58,  7.91s/iter, val_loss=7.01e-5, val_og_loss=7.11e-5]
+
 30%|███       | 6/20 [01:01<01:38,  7.06s/iter, val_loss=7.01e-5, val_og_loss=7.11e-5]
+
 30%|███       | 6/20 [01:06<01:38,  7.06s/iter, val_loss=7.01e-5, val_og_loss=7.11e-5]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.53s/iter, val_loss=7.01e-5, val_og_loss=7.11e-5]
+
 35%|███▌      | 7/20 [01:12<01:24,  6.53s/iter, val_loss=7.01e-5, val_og_loss=7.11e-5]
+
 40%|████      | 8/20 [01:12<01:14,  6.17s/iter, val_loss=7.01e-5, val_og_loss=7.11e-5]
+
 40%|████      | 8/20 [01:17<01:14,  6.17s/iter, val_loss=7.01e-5, val_og_loss=7.11e-5]
+
 45%|████▌     | 9/20 [01:17<01:05,  5.94s/iter, val_loss=7.01e-5, val_og_loss=7.11e-5]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.94s/iter, val_loss=7.02e-5, val_og_loss=7.11e-5]
+
 50%|█████     | 10/20 [01:22<00:57,  5.78s/iter, val_loss=7.02e-5, val_og_loss=7.11e-5]
+
 50%|█████     | 10/20 [01:28<00:57,  5.78s/iter, val_loss=7e-5, val_og_loss=7.11e-5]   
+
 55%|█████▌    | 11/20 [01:28<00:51,  5.67s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 55%|█████▌    | 11/20 [01:33<00:51,  5.67s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 60%|██████    | 12/20 [01:33<00:44,  5.60s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 60%|██████    | 12/20 [01:39<00:44,  5.60s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 65%|██████▌   | 13/20 [01:39<00:38,  5.54s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.54s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 70%|███████   | 14/20 [01:44<00:33,  5.50s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 70%|███████   | 14/20 [01:50<00:33,  5.50s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 75%|███████▌  | 15/20 [01:50<00:27,  5.48s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 75%|███████▌  | 15/20 [01:55<00:27,  5.48s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 80%|████████  | 16/20 [01:55<00:21,  5.47s/iter, val_loss=7e-5, val_og_loss=7.11e-5]
+
 80%|████████  | 16/20 [02:00<00:21,  5.47s/iter, val_loss=6.99e-5, val_og_loss=7.11e-5]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.45s/iter, val_loss=6.99e-5, val_og_loss=7.11e-5]
+
 85%|████████▌ | 17/20 [02:06<00:16,  5.45s/iter, val_loss=6.99e-5, val_og_loss=7.11e-5]
+
 90%|█████████ | 18/20 [02:06<00:10,  5.44s/iter, val_loss=6.99e-5, val_og_loss=7.11e-5]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=6.99e-5, val_og_loss=7.11e-5]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.43s/iter, val_loss=6.99e-5, val_og_loss=7.11e-5]
+
 95%|█████████▌| 19/20 [02:17<00:05,  5.43s/iter, val_loss=6.99e-5, val_og_loss=7.11e-5]
+
100%|██████████| 20/20 [02:17<00:00,  5.43s/iter, val_loss=6.99e-5, val_og_loss=7.11e-5]
100%|██████████| 20/20 [02:17<00:00,  6.86s/iter, val_loss=6.99e-5, val_og_loss=7.11e-5]
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:33:52<4:38:39, 879.99s/it]INFO:ParoQuant:Best val loss: 6.988506356719881e-05, Original val loss: 7.110537262633443e-05
+
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:33:52<4:38:39, 879.99s/it]
                                                                       

Processing Blocks:  41%|████      | 13/32 [2:34:02<4:38:39, 879.99s/it]
Processing Blocks:  44%|████▍     | 14/32 [2:34:40<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:34:41<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:34:59<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:34:59<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:37:16<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:37:16<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:38:24<4:26:00, 886.67s/it]2025-12-22 17:04:10,132 - INFO - Best val loss: 6.988506356719881e-05, Original val loss: 7.110537262633443e-05
+[INFO] Saving Layer 13 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 14...
+[INFO] Starting Sequential Training for Layer 14
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:18<?, ?iter/s, val_loss=0.0019, val_og_loss=0.00242]
+
  5%|▌         | 1/20 [00:18<05:54, 18.66s/iter, val_loss=0.0019, val_og_loss=0.00242]
+
  5%|▌         | 1/20 [00:20<05:54, 18.66s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 10%|█         | 2/20 [00:20<02:36,  8.69s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 10%|█         | 2/20 [00:22<02:36,  8.69s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 15%|█▌        | 3/20 [00:22<01:33,  5.50s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 15%|█▌        | 3/20 [00:23<01:33,  5.50s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 20%|██        | 4/20 [00:23<01:04,  4.00s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 20%|██        | 4/20 [00:25<01:04,  4.00s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 25%|██▌       | 5/20 [00:25<00:47,  3.17s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 25%|██▌       | 5/20 [00:27<00:47,  3.17s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 30%|███       | 6/20 [00:27<00:37,  2.66s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 30%|███       | 6/20 [00:28<00:37,  2.66s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 35%|███▌      | 7/20 [00:28<00:30,  2.34s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 35%|███▌      | 7/20 [00:30<00:30,  2.34s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 40%|████      | 8/20 [00:30<00:25,  2.13s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 40%|████      | 8/20 [00:32<00:25,  2.13s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 45%|████▌     | 9/20 [00:32<00:21,  2.00s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 45%|████▌     | 9/20 [00:33<00:21,  2.00s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 50%|█████     | 10/20 [00:33<00:18,  1.90s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 50%|█████     | 10/20 [00:35<00:18,  1.90s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.83s/iter, val_loss=0.00194, val_og_loss=0.00242]
+
 55%|█████▌    | 11/20 [00:37<00:16,  1.83s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 60%|██████    | 12/20 [00:37<00:14,  1.78s/iter, val_loss=0.00193, val_og_loss=0.00242]
+
 60%|██████    | 12/20 [00:38<00:14,  1.78s/iter, val_loss=0.00192, val_og_loss=0.00242]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.76s/iter, val_loss=0.00192, val_og_loss=0.00242]
+
 65%|██████▌   | 13/20 [00:40<00:12,  1.76s/iter, val_loss=0.00191, val_og_loss=0.00242]
+
 70%|███████   | 14/20 [00:40<00:10,  1.73s/iter, val_loss=0.00191, val_og_loss=0.00242]
+
 70%|███████   | 14/20 [00:42<00:10,  1.73s/iter, val_loss=0.00192, val_og_loss=0.00242]
+
 75%|███████▌  | 15/20 [00:42<00:08,  1.71s/iter, val_loss=0.00192, val_og_loss=0.00242]
+
 75%|███████▌  | 15/20 [00:43<00:08,  1.71s/iter, val_loss=0.00191, val_og_loss=0.00242]
+
 80%|████████  | 16/20 [00:43<00:06,  1.70s/iter, val_loss=0.00191, val_og_loss=0.00242]
+
 80%|████████  | 16/20 [00:45<00:06,  1.70s/iter, val_loss=0.00191, val_og_loss=0.00242]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.70s/iter, val_loss=0.00191, val_og_loss=0.00242]
+
 85%|████████▌ | 17/20 [00:47<00:05,  1.70s/iter, val_loss=0.0019, val_og_loss=0.00242] 
+
 90%|█████████ | 18/20 [00:47<00:03,  1.70s/iter, val_loss=0.0019, val_og_loss=0.00242]
+
 90%|█████████ | 18/20 [00:49<00:03,  1.70s/iter, val_loss=0.0019, val_og_loss=0.00242]
+
 95%|█████████▌| 19/20 [00:49<00:01,  1.69s/iter, val_loss=0.0019, val_og_loss=0.00242]
+
 95%|█████████▌| 19/20 [00:50<00:01,  1.69s/iter, val_loss=0.0019, val_og_loss=0.00242]
+
100%|██████████| 20/20 [00:50<00:00,  1.69s/iter, val_loss=0.0019, val_og_loss=0.00242]
100%|██████████| 20/20 [00:50<00:00,  2.54s/iter, val_loss=0.0019, val_og_loss=0.00242]
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:39:15<4:26:00, 886.67s/it]INFO:ParoQuant:Best val loss: 0.0018954029073938727, Original val loss: 0.002421849174425006
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:39:19<4:26:00, 886.67s/it]2025-12-22 17:09:33,046 - INFO - Best val loss: 0.0018954029073938727, Original val loss: 0.002421849174425006
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:18<?, ?iter/s, val_loss=0.00578, val_og_loss=0.00695]
+
  5%|▌         | 1/20 [00:18<05:45, 18.21s/iter, val_loss=0.00578, val_og_loss=0.00695]
+
  5%|▌         | 1/20 [00:19<05:45, 18.21s/iter, val_loss=0.00587, val_og_loss=0.00695]
+
 10%|█         | 2/20 [00:19<02:29,  8.31s/iter, val_loss=0.00587, val_og_loss=0.00695]
+
 10%|█         | 2/20 [00:20<02:29,  8.31s/iter, val_loss=0.00592, val_og_loss=0.00695]
+
 15%|█▌        | 3/20 [00:20<01:27,  5.16s/iter, val_loss=0.00592, val_og_loss=0.00695]
+
 15%|█▌        | 3/20 [00:22<01:27,  5.16s/iter, val_loss=0.00588, val_og_loss=0.00695]
+
 20%|██        | 4/20 [00:22<00:58,  3.67s/iter, val_loss=0.00588, val_og_loss=0.00695]
+
 20%|██        | 4/20 [00:23<00:58,  3.67s/iter, val_loss=0.00593, val_og_loss=0.00695]
+
 25%|██▌       | 5/20 [00:23<00:42,  2.86s/iter, val_loss=0.00593, val_og_loss=0.00695]
+
 25%|██▌       | 5/20 [00:25<00:42,  2.86s/iter, val_loss=0.00588, val_og_loss=0.00695]
+
 30%|███       | 6/20 [00:25<00:33,  2.36s/iter, val_loss=0.00588, val_og_loss=0.00695]
+
 30%|███       | 6/20 [00:26<00:33,  2.36s/iter, val_loss=0.0059, val_og_loss=0.00695] 
+
 35%|███▌      | 7/20 [00:26<00:26,  2.04s/iter, val_loss=0.0059, val_og_loss=0.00695]
+
 35%|███▌      | 7/20 [00:27<00:26,  2.04s/iter, val_loss=0.00593, val_og_loss=0.00695]
+
 40%|████      | 8/20 [00:27<00:22,  1.84s/iter, val_loss=0.00593, val_og_loss=0.00695]
+
 40%|████      | 8/20 [00:29<00:22,  1.84s/iter, val_loss=0.00593, val_og_loss=0.00695]
+
 45%|████▌     | 9/20 [00:29<00:18,  1.70s/iter, val_loss=0.00593, val_og_loss=0.00695]
+
 45%|████▌     | 9/20 [00:30<00:18,  1.70s/iter, val_loss=0.00584, val_og_loss=0.00695]
+
 50%|█████     | 10/20 [00:30<00:16,  1.60s/iter, val_loss=0.00584, val_og_loss=0.00695]
+
 50%|█████     | 10/20 [00:32<00:16,  1.60s/iter, val_loss=0.00588, val_og_loss=0.00695]
+
 55%|█████▌    | 11/20 [00:32<00:13,  1.54s/iter, val_loss=0.00588, val_og_loss=0.00695]
+
 55%|█████▌    | 11/20 [00:33<00:13,  1.54s/iter, val_loss=0.00586, val_og_loss=0.00695]
+
 60%|██████    | 12/20 [00:33<00:11,  1.49s/iter, val_loss=0.00586, val_og_loss=0.00695]
+
 60%|██████    | 12/20 [00:34<00:11,  1.49s/iter, val_loss=0.0059, val_og_loss=0.00695] 
+
 65%|██████▌   | 13/20 [00:34<00:10,  1.46s/iter, val_loss=0.0059, val_og_loss=0.00695]
+
 65%|██████▌   | 13/20 [00:36<00:10,  1.46s/iter, val_loss=0.00585, val_og_loss=0.00695]
+
 70%|███████   | 14/20 [00:36<00:08,  1.45s/iter, val_loss=0.00585, val_og_loss=0.00695]
+
 70%|███████   | 14/20 [00:37<00:08,  1.45s/iter, val_loss=0.00585, val_og_loss=0.00695]
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.43s/iter, val_loss=0.00585, val_og_loss=0.00695]
+
 75%|███████▌  | 15/20 [00:39<00:07,  1.43s/iter, val_loss=0.00582, val_og_loss=0.00695]
+
 80%|████████  | 16/20 [00:39<00:05,  1.42s/iter, val_loss=0.00582, val_og_loss=0.00695]
+
 80%|████████  | 16/20 [00:40<00:05,  1.42s/iter, val_loss=0.00584, val_og_loss=0.00695]
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.41s/iter, val_loss=0.00584, val_og_loss=0.00695]
+
 85%|████████▌ | 17/20 [00:41<00:04,  1.41s/iter, val_loss=0.00578, val_og_loss=0.00695]
+
 90%|█████████ | 18/20 [00:41<00:02,  1.40s/iter, val_loss=0.00578, val_og_loss=0.00695]
+
 90%|█████████ | 18/20 [00:43<00:02,  1.40s/iter, val_loss=0.0058, val_og_loss=0.00695] 
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.41s/iter, val_loss=0.0058, val_og_loss=0.00695]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.41s/iter, val_loss=0.00579, val_og_loss=0.00695]
+
100%|██████████| 20/20 [00:44<00:00,  1.40s/iter, val_loss=0.00579, val_og_loss=0.00695]
100%|██████████| 20/20 [00:44<00:00,  2.24s/iter, val_loss=0.00579, val_og_loss=0.00695]
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:40:04<4:26:00, 886.67s/it]INFO:ParoQuant:Best val loss: 0.005782202817499638, Original val loss: 0.006945238448679447
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:40:08<4:26:00, 886.67s/it]2025-12-22 17:10:21,985 - INFO - Best val loss: 0.005782202817499638, Original val loss: 0.006945238448679447
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:20<?, ?iter/s, val_loss=0.000434, val_og_loss=0.000556]
+
  5%|▌         | 1/20 [00:20<06:24, 20.21s/iter, val_loss=0.000434, val_og_loss=0.000556]
+
  5%|▌         | 1/20 [00:21<06:24, 20.21s/iter, val_loss=0.000437, val_og_loss=0.000556]
+
 10%|█         | 2/20 [00:21<02:44,  9.14s/iter, val_loss=0.000437, val_og_loss=0.000556]
+
 10%|█         | 2/20 [00:22<02:44,  9.14s/iter, val_loss=0.000439, val_og_loss=0.000556]
+
 15%|█▌        | 3/20 [00:22<01:35,  5.60s/iter, val_loss=0.000439, val_og_loss=0.000556]
+
 15%|█▌        | 3/20 [00:24<01:35,  5.60s/iter, val_loss=0.000441, val_og_loss=0.000556]
+
 20%|██        | 4/20 [00:24<01:03,  3.94s/iter, val_loss=0.000441, val_og_loss=0.000556]
+
 20%|██        | 4/20 [00:25<01:03,  3.94s/iter, val_loss=0.00044, val_og_loss=0.000556] 
+
 25%|██▌       | 5/20 [00:25<00:45,  3.03s/iter, val_loss=0.00044, val_og_loss=0.000556]
+
 25%|██▌       | 5/20 [00:27<00:45,  3.03s/iter, val_loss=0.000441, val_og_loss=0.000556]
+
 30%|███       | 6/20 [00:27<00:34,  2.47s/iter, val_loss=0.000441, val_og_loss=0.000556]
+
 30%|███       | 6/20 [00:28<00:34,  2.47s/iter, val_loss=0.000439, val_og_loss=0.000556]
+
 35%|███▌      | 7/20 [00:28<00:27,  2.12s/iter, val_loss=0.000439, val_og_loss=0.000556]
+
 35%|███▌      | 7/20 [00:29<00:27,  2.12s/iter, val_loss=0.00044, val_og_loss=0.000556] 
+
 40%|████      | 8/20 [00:29<00:22,  1.89s/iter, val_loss=0.00044, val_og_loss=0.000556]
+
 40%|████      | 8/20 [00:31<00:22,  1.89s/iter, val_loss=0.000439, val_og_loss=0.000556]
+
 45%|████▌     | 9/20 [00:31<00:19,  1.73s/iter, val_loss=0.000439, val_og_loss=0.000556]
+
 45%|████▌     | 9/20 [00:32<00:19,  1.73s/iter, val_loss=0.000438, val_og_loss=0.000556]
+
 50%|█████     | 10/20 [00:32<00:16,  1.64s/iter, val_loss=0.000438, val_og_loss=0.000556]
+
 50%|█████     | 10/20 [00:34<00:16,  1.64s/iter, val_loss=0.000437, val_og_loss=0.000556]
+
 55%|█████▌    | 11/20 [00:34<00:14,  1.56s/iter, val_loss=0.000437, val_og_loss=0.000556]
+
 55%|█████▌    | 11/20 [00:35<00:14,  1.56s/iter, val_loss=0.000437, val_og_loss=0.000556]
+
 60%|██████    | 12/20 [00:35<00:12,  1.51s/iter, val_loss=0.000437, val_og_loss=0.000556]
+
 60%|██████    | 12/20 [00:36<00:12,  1.51s/iter, val_loss=0.000436, val_og_loss=0.000556]
+
 65%|██████▌   | 13/20 [00:36<00:10,  1.47s/iter, val_loss=0.000436, val_og_loss=0.000556]
+
 65%|██████▌   | 13/20 [00:38<00:10,  1.47s/iter, val_loss=0.000434, val_og_loss=0.000556]
+
 70%|███████   | 14/20 [00:38<00:08,  1.46s/iter, val_loss=0.000434, val_og_loss=0.000556]
+
 70%|███████   | 14/20 [00:39<00:08,  1.46s/iter, val_loss=0.000434, val_og_loss=0.000556]
+
 75%|███████▌  | 15/20 [00:39<00:07,  1.44s/iter, val_loss=0.000434, val_og_loss=0.000556]
+
 75%|███████▌  | 15/20 [00:41<00:07,  1.44s/iter, val_loss=0.000435, val_og_loss=0.000556]
+
 80%|████████  | 16/20 [00:41<00:05,  1.42s/iter, val_loss=0.000435, val_og_loss=0.000556]
+
 80%|████████  | 16/20 [00:42<00:05,  1.42s/iter, val_loss=0.000433, val_og_loss=0.000556]
+
 85%|████████▌ | 17/20 [00:42<00:04,  1.41s/iter, val_loss=0.000433, val_og_loss=0.000556]
+
 85%|████████▌ | 17/20 [00:43<00:04,  1.41s/iter, val_loss=0.000432, val_og_loss=0.000556]
+
 90%|█████████ | 18/20 [00:43<00:02,  1.40s/iter, val_loss=0.000432, val_og_loss=0.000556]
+
 90%|█████████ | 18/20 [00:45<00:02,  1.40s/iter, val_loss=0.000432, val_og_loss=0.000556]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.41s/iter, val_loss=0.000432, val_og_loss=0.000556]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.41s/iter, val_loss=0.000432, val_og_loss=0.000556]
+
100%|██████████| 20/20 [00:46<00:00,  1.40s/iter, val_loss=0.000432, val_og_loss=0.000556]
100%|██████████| 20/20 [00:46<00:00,  2.34s/iter, val_loss=0.000432, val_og_loss=0.000556]
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:40:54<4:26:00, 886.67s/it]INFO:ParoQuant:Best val loss: 0.00043164833914488554, Original val loss: 0.0005563765298575163
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:40:57<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:40:57<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:41:29<4:26:00, 886.67s/it]2025-12-22 17:11:12,744 - INFO - Best val loss: 0.00043164833914488554, Original val loss: 0.0005563765298575163
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
  5%|▌         | 1/20 [00:13<04:17, 13.55s/iter, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
  5%|▌         | 1/20 [00:15<04:17, 13.55s/iter, val_loss=3.7e-5, val_og_loss=3.92e-5] 
+
 10%|█         | 2/20 [00:15<01:58,  6.57s/iter, val_loss=3.7e-5, val_og_loss=3.92e-5]
+
 10%|█         | 2/20 [00:16<01:58,  6.57s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 15%|█▌        | 3/20 [00:16<01:13,  4.34s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 15%|█▌        | 3/20 [00:18<01:13,  4.34s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 20%|██        | 4/20 [00:18<00:52,  3.30s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 20%|██        | 4/20 [00:20<00:52,  3.30s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 25%|██▌       | 5/20 [00:20<00:40,  2.71s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.71s/iter, val_loss=3.73e-5, val_og_loss=3.92e-5]
+
 30%|███       | 6/20 [00:21<00:33,  2.36s/iter, val_loss=3.73e-5, val_og_loss=3.92e-5]
+
 30%|███       | 6/20 [00:23<00:33,  2.36s/iter, val_loss=3.73e-5, val_og_loss=3.92e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.14s/iter, val_loss=3.73e-5, val_og_loss=3.92e-5]
+
 35%|███▌      | 7/20 [00:25<00:27,  2.14s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 40%|████      | 8/20 [00:25<00:23,  1.99s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 40%|████      | 8/20 [00:27<00:23,  1.99s/iter, val_loss=3.73e-5, val_og_loss=3.92e-5]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.90s/iter, val_loss=3.73e-5, val_og_loss=3.92e-5]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.90s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 50%|█████     | 10/20 [00:28<00:18,  1.83s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.83s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.79s/iter, val_loss=3.72e-5, val_og_loss=3.92e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=3.71e-5, val_og_loss=3.92e-5]
+
 60%|██████    | 12/20 [00:32<00:14,  1.75s/iter, val_loss=3.71e-5, val_og_loss=3.92e-5]
+
 60%|██████    | 12/20 [00:33<00:14,  1.75s/iter, val_loss=3.7e-5, val_og_loss=3.92e-5] 
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.74s/iter, val_loss=3.7e-5, val_og_loss=3.92e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.74s/iter, val_loss=3.7e-5, val_og_loss=3.92e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.72s/iter, val_loss=3.7e-5, val_og_loss=3.92e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=3.69e-5, val_og_loss=3.92e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.71s/iter, val_loss=3.69e-5, val_og_loss=3.92e-5]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
 80%|████████  | 16/20 [00:38<00:06,  1.70s/iter, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.70s/iter, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.69s/iter, val_loss=3.68e-5, val_og_loss=3.92e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=3.67e-5, val_og_loss=3.92e-5]
+
100%|██████████| 20/20 [00:45<00:00,  1.68s/iter, val_loss=3.67e-5, val_og_loss=3.92e-5]
100%|██████████| 20/20 [00:45<00:00,  2.28s/iter, val_loss=3.67e-5, val_og_loss=3.92e-5]
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:42:14<4:26:00, 886.67s/it]INFO:ParoQuant:Best val loss: 3.673634637380019e-05, Original val loss: 3.918275979231112e-05
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:42:15<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:42:15<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:42:59<4:26:00, 886.67s/it]2025-12-22 17:12:32,584 - INFO - Best val loss: 3.673634637380019e-05, Original val loss: 3.918275979231112e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.000891, val_og_loss=0.000914]
+
  5%|▌         | 1/20 [00:29<09:27, 29.89s/iter, val_loss=0.000891, val_og_loss=0.000914]
+
  5%|▌         | 1/20 [00:32<09:27, 29.89s/iter, val_loss=0.000898, val_og_loss=0.000914]
+
 10%|█         | 2/20 [00:32<04:10, 13.91s/iter, val_loss=0.000898, val_og_loss=0.000914]
+
 10%|█         | 2/20 [00:35<04:10, 13.91s/iter, val_loss=0.000897, val_og_loss=0.000914]
+
 15%|█▌        | 3/20 [00:35<02:29,  8.80s/iter, val_loss=0.000897, val_og_loss=0.000914]
+
 15%|█▌        | 3/20 [00:38<02:29,  8.80s/iter, val_loss=0.000899, val_og_loss=0.000914]
+
 20%|██        | 4/20 [00:38<01:42,  6.40s/iter, val_loss=0.000899, val_og_loss=0.000914]
+
 20%|██        | 4/20 [00:40<01:42,  6.40s/iter, val_loss=0.000901, val_og_loss=0.000914]
+
 25%|██▌       | 5/20 [00:40<01:16,  5.07s/iter, val_loss=0.000901, val_og_loss=0.000914]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.07s/iter, val_loss=0.000902, val_og_loss=0.000914]
+
 30%|███       | 6/20 [00:43<00:59,  4.27s/iter, val_loss=0.000902, val_og_loss=0.000914]
+
 30%|███       | 6/20 [00:46<00:59,  4.27s/iter, val_loss=0.000901, val_og_loss=0.000914]
+
 35%|███▌      | 7/20 [00:46<00:48,  3.76s/iter, val_loss=0.000901, val_og_loss=0.000914]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.76s/iter, val_loss=0.0009, val_og_loss=0.000914]  
+
 40%|████      | 8/20 [00:48<00:41,  3.43s/iter, val_loss=0.0009, val_og_loss=0.000914]
+
 40%|████      | 8/20 [00:51<00:41,  3.43s/iter, val_loss=0.0009, val_og_loss=0.000914]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.20s/iter, val_loss=0.0009, val_og_loss=0.000914]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.20s/iter, val_loss=0.000899, val_og_loss=0.000914]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.000899, val_og_loss=0.000914]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.000897, val_og_loss=0.000914]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.95s/iter, val_loss=0.000897, val_og_loss=0.000914]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.95s/iter, val_loss=0.000897, val_og_loss=0.000914]
+
 60%|██████    | 12/20 [00:59<00:23,  2.88s/iter, val_loss=0.000897, val_og_loss=0.000914]
+
 60%|██████    | 12/20 [01:02<00:23,  2.88s/iter, val_loss=0.000896, val_og_loss=0.000914]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.83s/iter, val_loss=0.000896, val_og_loss=0.000914]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.83s/iter, val_loss=0.000895, val_og_loss=0.000914]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.000895, val_og_loss=0.000914]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.000894, val_og_loss=0.000914]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.000894, val_og_loss=0.000914]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.000893, val_og_loss=0.000914]
+
 80%|████████  | 16/20 [01:10<00:11,  2.76s/iter, val_loss=0.000893, val_og_loss=0.000914]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.000892, val_og_loss=0.000914]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.000892, val_og_loss=0.000914]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.000892, val_og_loss=0.000914]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.74s/iter, val_loss=0.000892, val_og_loss=0.000914]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.000892, val_og_loss=0.000914]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.000892, val_og_loss=0.000914]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.000891, val_og_loss=0.000914]
+
100%|██████████| 20/20 [01:21<00:00,  2.73s/iter, val_loss=0.000891, val_og_loss=0.000914]
100%|██████████| 20/20 [01:21<00:00,  4.08s/iter, val_loss=0.000891, val_og_loss=0.000914]
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:44:21<4:26:00, 886.67s/it]INFO:ParoQuant:Best val loss: 0.0008910683682188392, Original val loss: 0.000913522788323462
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:44:25<4:26:00, 886.67s/it]2025-12-22 17:14:39,155 - INFO - Best val loss: 0.0008910683682188392, Original val loss: 0.000913522788323462
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.000571, val_og_loss=0.00059]
+
  5%|▌         | 1/20 [00:31<09:51, 31.14s/iter, val_loss=0.000571, val_og_loss=0.00059]
+
  5%|▌         | 1/20 [00:33<09:51, 31.14s/iter, val_loss=0.000574, val_og_loss=0.00059]
+
 10%|█         | 2/20 [00:33<04:19, 14.43s/iter, val_loss=0.000574, val_og_loss=0.00059]
+
 10%|█         | 2/20 [00:36<04:19, 14.43s/iter, val_loss=0.000578, val_og_loss=0.00059]
+
 15%|█▌        | 3/20 [00:36<02:34,  9.08s/iter, val_loss=0.000578, val_og_loss=0.00059]
+
 15%|█▌        | 3/20 [00:39<02:34,  9.08s/iter, val_loss=0.00058, val_og_loss=0.00059] 
+
 20%|██        | 4/20 [00:39<01:45,  6.57s/iter, val_loss=0.00058, val_og_loss=0.00059]
+
 20%|██        | 4/20 [00:42<01:45,  6.57s/iter, val_loss=0.000581, val_og_loss=0.00059]
+
 25%|██▌       | 5/20 [00:42<01:17,  5.18s/iter, val_loss=0.000581, val_og_loss=0.00059]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.18s/iter, val_loss=0.000582, val_og_loss=0.00059]
+
 30%|███       | 6/20 [00:44<01:00,  4.34s/iter, val_loss=0.000582, val_og_loss=0.00059]
+
 30%|███       | 6/20 [00:47<01:00,  4.34s/iter, val_loss=0.000581, val_og_loss=0.00059]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.82s/iter, val_loss=0.000581, val_og_loss=0.00059]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.82s/iter, val_loss=0.000583, val_og_loss=0.00059]
+
 40%|████      | 8/20 [00:50<00:41,  3.46s/iter, val_loss=0.000583, val_og_loss=0.00059]
+
 40%|████      | 8/20 [00:52<00:41,  3.46s/iter, val_loss=0.000579, val_og_loss=0.00059]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.23s/iter, val_loss=0.000579, val_og_loss=0.00059]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.23s/iter, val_loss=0.000578, val_og_loss=0.00059]
+
 50%|█████     | 10/20 [00:55<00:30,  3.07s/iter, val_loss=0.000578, val_og_loss=0.00059]
+
 50%|█████     | 10/20 [00:58<00:30,  3.07s/iter, val_loss=0.000578, val_og_loss=0.00059]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.96s/iter, val_loss=0.000578, val_og_loss=0.00059]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.96s/iter, val_loss=0.000578, val_og_loss=0.00059]
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.000578, val_og_loss=0.00059]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.000577, val_og_loss=0.00059]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.84s/iter, val_loss=0.000577, val_og_loss=0.00059]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000577, val_og_loss=0.00059]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.000577, val_og_loss=0.00059]
+
 70%|███████   | 14/20 [01:09<00:16,  2.80s/iter, val_loss=0.000577, val_og_loss=0.00059]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.000577, val_og_loss=0.00059]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.000576, val_og_loss=0.00059]
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.000576, val_og_loss=0.00059]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.000575, val_og_loss=0.00059]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.75s/iter, val_loss=0.000575, val_og_loss=0.00059]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.75s/iter, val_loss=0.000575, val_og_loss=0.00059]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.000575, val_og_loss=0.00059]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.74s/iter, val_loss=0.000575, val_og_loss=0.00059]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.000575, val_og_loss=0.00059]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.000575, val_og_loss=0.00059]
+
100%|██████████| 20/20 [01:22<00:00,  2.73s/iter, val_loss=0.000575, val_og_loss=0.00059]
100%|██████████| 20/20 [01:22<00:00,  4.14s/iter, val_loss=0.000575, val_og_loss=0.00059]
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:45:48<4:26:00, 886.67s/it]INFO:ParoQuant:Best val loss: 0.0005710569093935192, Original val loss: 0.0005903878482058644
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:45:50<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:45:50<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:46:50<4:26:00, 886.67s/it]2025-12-22 17:16:06,573 - INFO - Best val loss: 0.0005710569093935192, Original val loss: 0.0005903878482058644
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:34<?, ?iter/s, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
  5%|▌         | 1/20 [00:34<10:52, 34.36s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
  5%|▌         | 1/20 [00:39<10:52, 34.36s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 10%|█         | 2/20 [00:39<05:12, 17.35s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 10%|█         | 2/20 [00:45<05:12, 17.35s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 15%|█▌        | 3/20 [00:45<03:22, 11.90s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 15%|█▌        | 3/20 [00:50<03:22, 11.90s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 20%|██        | 4/20 [00:50<02:29,  9.34s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 20%|██        | 4/20 [00:56<02:29,  9.34s/iter, val_loss=7.58e-5, val_og_loss=7.69e-5]
+
 25%|██▌       | 5/20 [00:56<01:58,  7.93s/iter, val_loss=7.58e-5, val_og_loss=7.69e-5]
+
 25%|██▌       | 5/20 [01:01<01:58,  7.93s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 30%|███       | 6/20 [01:01<01:39,  7.08s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 30%|███       | 6/20 [01:06<01:39,  7.08s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.54s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 35%|███▌      | 7/20 [01:12<01:24,  6.54s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 40%|████      | 8/20 [01:12<01:14,  6.18s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 40%|████      | 8/20 [01:17<01:14,  6.18s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 45%|████▌     | 9/20 [01:17<01:05,  5.95s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 45%|████▌     | 9/20 [01:23<01:05,  5.95s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 50%|█████     | 10/20 [01:23<00:57,  5.78s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 50%|█████     | 10/20 [01:28<00:57,  5.78s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 55%|█████▌    | 11/20 [01:28<00:51,  5.67s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 55%|█████▌    | 11/20 [01:34<00:51,  5.67s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 60%|██████    | 12/20 [01:34<00:44,  5.60s/iter, val_loss=7.57e-5, val_og_loss=7.69e-5]
+
 60%|██████    | 12/20 [01:39<00:44,  5.60s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 65%|██████▌   | 13/20 [01:39<00:38,  5.54s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.54s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 70%|███████   | 14/20 [01:44<00:33,  5.51s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 70%|███████   | 14/20 [01:50<00:33,  5.51s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 75%|███████▌  | 15/20 [01:50<00:27,  5.48s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 75%|███████▌  | 15/20 [01:55<00:27,  5.48s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 80%|████████  | 16/20 [01:55<00:21,  5.46s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 80%|████████  | 16/20 [02:01<00:21,  5.46s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 85%|████████▌ | 17/20 [02:01<00:16,  5.45s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 85%|████████▌ | 17/20 [02:06<00:16,  5.45s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 90%|█████████ | 18/20 [02:06<00:10,  5.44s/iter, val_loss=7.56e-5, val_og_loss=7.69e-5]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=7.55e-5, val_og_loss=7.69e-5]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=7.55e-5, val_og_loss=7.69e-5]
+
 95%|█████████▌| 19/20 [02:17<00:05,  5.44s/iter, val_loss=7.55e-5, val_og_loss=7.69e-5]
+
100%|██████████| 20/20 [02:17<00:00,  5.43s/iter, val_loss=7.55e-5, val_og_loss=7.69e-5]
100%|██████████| 20/20 [02:17<00:00,  6.87s/iter, val_loss=7.55e-5, val_og_loss=7.69e-5]
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:49:07<4:26:00, 886.67s/it]INFO:ParoQuant:Best val loss: 7.551294402219355e-05, Original val loss: 7.690541679039598e-05
+
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:49:07<4:26:00, 886.67s/it]
                                                                       

Processing Blocks:  44%|████▍     | 14/32 [2:49:17<4:26:00, 886.67s/it]
Processing Blocks:  47%|████▋     | 15/32 [2:49:55<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:49:56<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:50:15<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:50:15<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:52:31<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:52:31<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:53:32<4:13:39, 895.28s/it]2025-12-22 17:19:25,388 - INFO - Best val loss: 7.551294402219355e-05, Original val loss: 7.690541679039598e-05
+[INFO] Saving Layer 14 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 15...
+[INFO] Starting Sequential Training for Layer 15
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.0023, val_og_loss=0.00279]
+
  5%|▌         | 1/20 [00:14<04:29, 14.21s/iter, val_loss=0.0023, val_og_loss=0.00279]
+
  5%|▌         | 1/20 [00:15<04:29, 14.21s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 10%|█         | 2/20 [00:15<02:03,  6.84s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 10%|█         | 2/20 [00:17<02:03,  6.84s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 15%|█▌        | 3/20 [00:17<01:16,  4.49s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 15%|█▌        | 3/20 [00:19<01:16,  4.49s/iter, val_loss=0.00236, val_og_loss=0.00279]
+
 20%|██        | 4/20 [00:19<00:54,  3.38s/iter, val_loss=0.00236, val_og_loss=0.00279]
+
 20%|██        | 4/20 [00:20<00:54,  3.38s/iter, val_loss=0.00237, val_og_loss=0.00279]
+
 25%|██▌       | 5/20 [00:20<00:41,  2.76s/iter, val_loss=0.00237, val_og_loss=0.00279]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.76s/iter, val_loss=0.00236, val_og_loss=0.00279]
+
 30%|███       | 6/20 [00:22<00:33,  2.40s/iter, val_loss=0.00236, val_og_loss=0.00279]
+
 30%|███       | 6/20 [00:24<00:33,  2.40s/iter, val_loss=0.00236, val_og_loss=0.00279]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.17s/iter, val_loss=0.00236, val_og_loss=0.00279]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.17s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 40%|████      | 8/20 [00:25<00:24,  2.01s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 40%|████      | 8/20 [00:27<00:24,  2.01s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.91s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 45%|████▌     | 9/20 [00:29<00:20,  1.91s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 50%|█████     | 10/20 [00:29<00:18,  1.84s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 50%|█████     | 10/20 [00:31<00:18,  1.84s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.79s/iter, val_loss=0.00235, val_og_loss=0.00279]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=0.00234, val_og_loss=0.00279]
+
 60%|██████    | 12/20 [00:32<00:14,  1.76s/iter, val_loss=0.00234, val_og_loss=0.00279]
+
 60%|██████    | 12/20 [00:34<00:14,  1.76s/iter, val_loss=0.00233, val_og_loss=0.00279]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=0.00233, val_og_loss=0.00279]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.73s/iter, val_loss=0.00234, val_og_loss=0.00279]
+
 70%|███████   | 14/20 [00:36<00:10,  1.72s/iter, val_loss=0.00234, val_og_loss=0.00279]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=0.00233, val_og_loss=0.00279]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.71s/iter, val_loss=0.00233, val_og_loss=0.00279]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.00232, val_og_loss=0.00279]
+
 80%|████████  | 16/20 [00:39<00:06,  1.71s/iter, val_loss=0.00232, val_og_loss=0.00279]
+
 80%|████████  | 16/20 [00:41<00:06,  1.71s/iter, val_loss=0.00232, val_og_loss=0.00279]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.70s/iter, val_loss=0.00232, val_og_loss=0.00279]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=0.00231, val_og_loss=0.00279]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.70s/iter, val_loss=0.00231, val_og_loss=0.00279]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.70s/iter, val_loss=0.00231, val_og_loss=0.00279]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=0.00231, val_og_loss=0.00279]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=0.00231, val_og_loss=0.00279]
+
100%|██████████| 20/20 [00:46<00:00,  1.69s/iter, val_loss=0.00231, val_og_loss=0.00279]
100%|██████████| 20/20 [00:46<00:00,  2.31s/iter, val_loss=0.00231, val_og_loss=0.00279]
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:54:18<4:13:39, 895.28s/it]INFO:ParoQuant:Best val loss: 0.0023027865681797266, Original val loss: 0.002792125102132559
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:54:22<4:13:39, 895.28s/it]2025-12-22 17:24:36,490 - INFO - Best val loss: 0.0023027865681797266, Original val loss: 0.002792125102132559
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.00525, val_og_loss=0.0063]
+
  5%|▌         | 1/20 [00:10<03:26, 10.87s/iter, val_loss=0.00525, val_og_loss=0.0063]
+
  5%|▌         | 1/20 [00:12<03:26, 10.87s/iter, val_loss=0.00542, val_og_loss=0.0063]
+
 10%|█         | 2/20 [00:12<01:35,  5.30s/iter, val_loss=0.00542, val_og_loss=0.0063]
+
 10%|█         | 2/20 [00:13<01:35,  5.30s/iter, val_loss=0.00541, val_og_loss=0.0063]
+
 15%|█▌        | 3/20 [00:13<00:59,  3.51s/iter, val_loss=0.00541, val_og_loss=0.0063]
+
 15%|█▌        | 3/20 [00:15<00:59,  3.51s/iter, val_loss=0.00545, val_og_loss=0.0063]
+
 20%|██        | 4/20 [00:15<00:42,  2.68s/iter, val_loss=0.00545, val_og_loss=0.0063]
+
 20%|██        | 4/20 [00:16<00:42,  2.68s/iter, val_loss=0.00544, val_og_loss=0.0063]
+
 25%|██▌       | 5/20 [00:16<00:33,  2.22s/iter, val_loss=0.00544, val_og_loss=0.0063]
+
 25%|██▌       | 5/20 [00:17<00:33,  2.22s/iter, val_loss=0.00543, val_og_loss=0.0063]
+
 30%|███       | 6/20 [00:17<00:27,  1.94s/iter, val_loss=0.00543, val_og_loss=0.0063]
+
 30%|███       | 6/20 [00:19<00:27,  1.94s/iter, val_loss=0.00538, val_og_loss=0.0063]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.76s/iter, val_loss=0.00538, val_og_loss=0.0063]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.76s/iter, val_loss=0.00539, val_og_loss=0.0063]
+
 40%|████      | 8/20 [00:20<00:19,  1.64s/iter, val_loss=0.00539, val_og_loss=0.0063]
+
 40%|████      | 8/20 [00:22<00:19,  1.64s/iter, val_loss=0.00544, val_og_loss=0.0063]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.56s/iter, val_loss=0.00544, val_og_loss=0.0063]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.56s/iter, val_loss=0.00544, val_og_loss=0.0063]
+
 50%|█████     | 10/20 [00:23<00:15,  1.52s/iter, val_loss=0.00544, val_og_loss=0.0063]
+
 50%|█████     | 10/20 [00:24<00:15,  1.52s/iter, val_loss=0.00539, val_og_loss=0.0063]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.47s/iter, val_loss=0.00539, val_og_loss=0.0063]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.47s/iter, val_loss=0.00537, val_og_loss=0.0063]
+
 60%|██████    | 12/20 [00:26<00:11,  1.44s/iter, val_loss=0.00537, val_og_loss=0.0063]
+
 60%|██████    | 12/20 [00:27<00:11,  1.44s/iter, val_loss=0.00536, val_og_loss=0.0063]
+
 65%|██████▌   | 13/20 [00:27<00:09,  1.42s/iter, val_loss=0.00536, val_og_loss=0.0063]
+
 65%|██████▌   | 13/20 [00:28<00:09,  1.42s/iter, val_loss=0.00536, val_og_loss=0.0063]
+
 70%|███████   | 14/20 [00:28<00:08,  1.41s/iter, val_loss=0.00536, val_og_loss=0.0063]
+
 70%|███████   | 14/20 [00:30<00:08,  1.41s/iter, val_loss=0.00536, val_og_loss=0.0063]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.40s/iter, val_loss=0.00536, val_og_loss=0.0063]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.40s/iter, val_loss=0.00535, val_og_loss=0.0063]
+
 80%|████████  | 16/20 [00:31<00:05,  1.39s/iter, val_loss=0.00535, val_og_loss=0.0063]
+
 80%|████████  | 16/20 [00:33<00:05,  1.39s/iter, val_loss=0.00537, val_og_loss=0.0063]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.39s/iter, val_loss=0.00537, val_og_loss=0.0063]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.39s/iter, val_loss=0.00531, val_og_loss=0.0063]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.38s/iter, val_loss=0.00531, val_og_loss=0.0063]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.38s/iter, val_loss=0.0053, val_og_loss=0.0063] 
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.38s/iter, val_loss=0.0053, val_og_loss=0.0063]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.38s/iter, val_loss=0.0053, val_og_loss=0.0063]
+
100%|██████████| 20/20 [00:37<00:00,  1.39s/iter, val_loss=0.0053, val_og_loss=0.0063]
100%|██████████| 20/20 [00:37<00:00,  1.86s/iter, val_loss=0.0053, val_og_loss=0.0063]
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:54:59<4:13:39, 895.28s/it]INFO:ParoQuant:Best val loss: 0.005251689348369837, Original val loss: 0.00629781186580658
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:55:02<4:13:39, 895.28s/it]2025-12-22 17:25:17,209 - INFO - Best val loss: 0.005251689348369837, Original val loss: 0.00629781186580658
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:12<?, ?iter/s, val_loss=0.000494, val_og_loss=0.000604]
+
  5%|▌         | 1/20 [00:12<03:53, 12.28s/iter, val_loss=0.000494, val_og_loss=0.000604]
+
  5%|▌         | 1/20 [00:13<03:53, 12.28s/iter, val_loss=0.000496, val_og_loss=0.000604]
+
 10%|█         | 2/20 [00:13<01:45,  5.88s/iter, val_loss=0.000496, val_og_loss=0.000604]
+
 10%|█         | 2/20 [00:15<01:45,  5.88s/iter, val_loss=0.000498, val_og_loss=0.000604]
+
 15%|█▌        | 3/20 [00:15<01:05,  3.84s/iter, val_loss=0.000498, val_og_loss=0.000604]
+
 15%|█▌        | 3/20 [00:16<01:05,  3.84s/iter, val_loss=0.000498, val_og_loss=0.000604]
+
 20%|██        | 4/20 [00:16<00:45,  2.87s/iter, val_loss=0.000498, val_og_loss=0.000604]
+
 20%|██        | 4/20 [00:17<00:45,  2.87s/iter, val_loss=0.0005, val_og_loss=0.000604]  
+
 25%|██▌       | 5/20 [00:17<00:35,  2.34s/iter, val_loss=0.0005, val_og_loss=0.000604]
+
 25%|██▌       | 5/20 [00:19<00:35,  2.34s/iter, val_loss=0.0005, val_og_loss=0.000604]
+
 30%|███       | 6/20 [00:19<00:28,  2.02s/iter, val_loss=0.0005, val_og_loss=0.000604]
+
 30%|███       | 6/20 [00:20<00:28,  2.02s/iter, val_loss=0.000499, val_og_loss=0.000604]
+
 35%|███▌      | 7/20 [00:20<00:23,  1.82s/iter, val_loss=0.000499, val_og_loss=0.000604]
+
 35%|███▌      | 7/20 [00:22<00:23,  1.82s/iter, val_loss=0.0005, val_og_loss=0.000604]  
+
 40%|████      | 8/20 [00:22<00:20,  1.68s/iter, val_loss=0.0005, val_og_loss=0.000604]
+
 40%|████      | 8/20 [00:23<00:20,  1.68s/iter, val_loss=0.000498, val_og_loss=0.000604]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.59s/iter, val_loss=0.000498, val_og_loss=0.000604]
+
 45%|████▌     | 9/20 [00:24<00:17,  1.59s/iter, val_loss=0.000498, val_og_loss=0.000604]
+
 50%|█████     | 10/20 [00:24<00:15,  1.53s/iter, val_loss=0.000498, val_og_loss=0.000604]
+
 50%|█████     | 10/20 [00:26<00:15,  1.53s/iter, val_loss=0.000497, val_og_loss=0.000604]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.49s/iter, val_loss=0.000497, val_og_loss=0.000604]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.49s/iter, val_loss=0.000496, val_og_loss=0.000604]
+
 60%|██████    | 12/20 [00:27<00:11,  1.47s/iter, val_loss=0.000496, val_og_loss=0.000604]
+
 60%|██████    | 12/20 [00:29<00:11,  1.47s/iter, val_loss=0.000495, val_og_loss=0.000604]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.44s/iter, val_loss=0.000495, val_og_loss=0.000604]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.44s/iter, val_loss=0.000493, val_og_loss=0.000604]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.000493, val_og_loss=0.000604]
+
 70%|███████   | 14/20 [00:31<00:08,  1.43s/iter, val_loss=0.000493, val_og_loss=0.000604]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.42s/iter, val_loss=0.000493, val_og_loss=0.000604]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.42s/iter, val_loss=0.000492, val_og_loss=0.000604]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.000492, val_og_loss=0.000604]
+
 80%|████████  | 16/20 [00:34<00:05,  1.41s/iter, val_loss=0.000493, val_og_loss=0.000604]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.42s/iter, val_loss=0.000493, val_og_loss=0.000604]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.42s/iter, val_loss=0.000492, val_og_loss=0.000604]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.41s/iter, val_loss=0.000492, val_og_loss=0.000604]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.41s/iter, val_loss=0.00049, val_og_loss=0.000604] 
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.41s/iter, val_loss=0.00049, val_og_loss=0.000604]
+
 95%|█████████▌| 19/20 [00:38<00:01,  1.41s/iter, val_loss=0.000491, val_og_loss=0.000604]
+
100%|██████████| 20/20 [00:38<00:00,  1.40s/iter, val_loss=0.000491, val_og_loss=0.000604]
100%|██████████| 20/20 [00:38<00:00,  1.94s/iter, val_loss=0.000491, val_og_loss=0.000604]
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:55:41<4:13:39, 895.28s/it]INFO:ParoQuant:Best val loss: 0.0004904529196210206, Original val loss: 0.0006038984283804893
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:55:44<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:55:44<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:56:15<4:13:39, 895.28s/it]2025-12-22 17:25:59,547 - INFO - Best val loss: 0.0004904529196210206, Original val loss: 0.0006038984283804893
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=3.74e-5, val_og_loss=3.85e-5]
+
  5%|▌         | 1/20 [00:14<04:39, 14.69s/iter, val_loss=3.74e-5, val_og_loss=3.85e-5]
+
  5%|▌         | 1/20 [00:16<04:39, 14.69s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 10%|█         | 2/20 [00:16<02:06,  7.05s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 10%|█         | 2/20 [00:18<02:06,  7.05s/iter, val_loss=3.76e-5, val_og_loss=3.85e-5]
+
 15%|█▌        | 3/20 [00:18<01:18,  4.60s/iter, val_loss=3.76e-5, val_og_loss=3.85e-5]
+
 15%|█▌        | 3/20 [00:19<01:18,  4.60s/iter, val_loss=3.76e-5, val_og_loss=3.85e-5]
+
 20%|██        | 4/20 [00:19<00:55,  3.45s/iter, val_loss=3.76e-5, val_og_loss=3.85e-5]
+
 20%|██        | 4/20 [00:21<00:55,  3.45s/iter, val_loss=3.77e-5, val_og_loss=3.85e-5]
+
 25%|██▌       | 5/20 [00:21<00:42,  2.81s/iter, val_loss=3.77e-5, val_og_loss=3.85e-5]
+
 25%|██▌       | 5/20 [00:23<00:42,  2.81s/iter, val_loss=3.77e-5, val_og_loss=3.85e-5]
+
 30%|███       | 6/20 [00:23<00:34,  2.43s/iter, val_loss=3.77e-5, val_og_loss=3.85e-5]
+
 30%|███       | 6/20 [00:24<00:34,  2.43s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.19s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 35%|███▌      | 7/20 [00:26<00:28,  2.19s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 40%|████      | 8/20 [00:26<00:24,  2.03s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 40%|████      | 8/20 [00:28<00:24,  2.03s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 45%|████▌     | 9/20 [00:28<00:21,  1.92s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.92s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 50%|█████     | 10/20 [00:29<00:18,  1.85s/iter, val_loss=3.75e-5, val_og_loss=3.85e-5]
+
 50%|█████     | 10/20 [00:31<00:18,  1.85s/iter, val_loss=3.74e-5, val_og_loss=3.85e-5]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.80s/iter, val_loss=3.74e-5, val_og_loss=3.85e-5]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.80s/iter, val_loss=3.74e-5, val_og_loss=3.85e-5]
+
 60%|██████    | 12/20 [00:33<00:14,  1.76s/iter, val_loss=3.74e-5, val_og_loss=3.85e-5]
+
 60%|██████    | 12/20 [00:34<00:14,  1.76s/iter, val_loss=3.74e-5, val_og_loss=3.85e-5]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.74s/iter, val_loss=3.74e-5, val_og_loss=3.85e-5]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.74s/iter, val_loss=3.73e-5, val_og_loss=3.85e-5]
+
 70%|███████   | 14/20 [00:36<00:10,  1.73s/iter, val_loss=3.73e-5, val_og_loss=3.85e-5]
+
 70%|███████   | 14/20 [00:38<00:10,  1.73s/iter, val_loss=3.73e-5, val_og_loss=3.85e-5]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=3.73e-5, val_og_loss=3.85e-5]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=3.73e-5, val_og_loss=3.85e-5]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=3.73e-5, val_og_loss=3.85e-5]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=3.72e-5, val_og_loss=3.85e-5]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.70s/iter, val_loss=3.72e-5, val_og_loss=3.85e-5]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.70s/iter, val_loss=3.72e-5, val_og_loss=3.85e-5]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=3.72e-5, val_og_loss=3.85e-5]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.69s/iter, val_loss=3.72e-5, val_og_loss=3.85e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=3.72e-5, val_og_loss=3.85e-5]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=3.72e-5, val_og_loss=3.85e-5]
+
100%|██████████| 20/20 [00:46<00:00,  1.69s/iter, val_loss=3.72e-5, val_og_loss=3.85e-5]
100%|██████████| 20/20 [00:46<00:00,  2.34s/iter, val_loss=3.72e-5, val_og_loss=3.85e-5]
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:57:02<4:13:39, 895.28s/it]INFO:ParoQuant:Best val loss: 3.717768413480371e-05, Original val loss: 3.8510002923430875e-05
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:57:02<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:57:02<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:57:47<4:13:39, 895.28s/it]2025-12-22 17:27:20,474 - INFO - Best val loss: 3.717768413480371e-05, Original val loss: 3.8510002923430875e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.000981, val_og_loss=0.00101]
+
  5%|▌         | 1/20 [00:29<09:28, 29.93s/iter, val_loss=0.000981, val_og_loss=0.00101]
+
  5%|▌         | 1/20 [00:32<09:28, 29.93s/iter, val_loss=0.000986, val_og_loss=0.00101]
+
 10%|█         | 2/20 [00:32<04:10, 13.92s/iter, val_loss=0.000986, val_og_loss=0.00101]
+
 10%|█         | 2/20 [00:35<04:10, 13.92s/iter, val_loss=0.000987, val_og_loss=0.00101]
+
 15%|█▌        | 3/20 [00:35<02:29,  8.80s/iter, val_loss=0.000987, val_og_loss=0.00101]
+
 15%|█▌        | 3/20 [00:38<02:29,  8.80s/iter, val_loss=0.000989, val_og_loss=0.00101]
+
 20%|██        | 4/20 [00:38<01:42,  6.39s/iter, val_loss=0.000989, val_og_loss=0.00101]
+
 20%|██        | 4/20 [00:40<01:42,  6.39s/iter, val_loss=0.00099, val_og_loss=0.00101] 
+
 25%|██▌       | 5/20 [00:40<01:16,  5.07s/iter, val_loss=0.00099, val_og_loss=0.00101]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.07s/iter, val_loss=0.000991, val_og_loss=0.00101]
+
 30%|███       | 6/20 [00:43<00:59,  4.27s/iter, val_loss=0.000991, val_og_loss=0.00101]
+
 30%|███       | 6/20 [00:46<00:59,  4.27s/iter, val_loss=0.00099, val_og_loss=0.00101] 
+
 35%|███▌      | 7/20 [00:46<00:48,  3.76s/iter, val_loss=0.00099, val_og_loss=0.00101]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.76s/iter, val_loss=0.000988, val_og_loss=0.00101]
+
 40%|████      | 8/20 [00:48<00:41,  3.43s/iter, val_loss=0.000988, val_og_loss=0.00101]
+
 40%|████      | 8/20 [00:51<00:41,  3.43s/iter, val_loss=0.000987, val_og_loss=0.00101]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.21s/iter, val_loss=0.000987, val_og_loss=0.00101]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.21s/iter, val_loss=0.000987, val_og_loss=0.00101]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.000987, val_og_loss=0.00101]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.000986, val_og_loss=0.00101]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.000986, val_og_loss=0.00101]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.96s/iter, val_loss=0.000985, val_og_loss=0.00101]
+
 60%|██████    | 12/20 [00:59<00:23,  2.88s/iter, val_loss=0.000985, val_og_loss=0.00101]
+
 60%|██████    | 12/20 [01:02<00:23,  2.88s/iter, val_loss=0.000983, val_og_loss=0.00101]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.83s/iter, val_loss=0.000983, val_og_loss=0.00101]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.83s/iter, val_loss=0.000983, val_og_loss=0.00101]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.000983, val_og_loss=0.00101]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.000982, val_og_loss=0.00101]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.000982, val_og_loss=0.00101]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.000981, val_og_loss=0.00101]
+
 80%|████████  | 16/20 [01:10<00:11,  2.76s/iter, val_loss=0.000981, val_og_loss=0.00101]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.000981, val_og_loss=0.00101]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.000981, val_og_loss=0.00101]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.000979, val_og_loss=0.00101]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.73s/iter, val_loss=0.000979, val_og_loss=0.00101]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.000979, val_og_loss=0.00101]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.000979, val_og_loss=0.00101]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.000979, val_og_loss=0.00101]
+
100%|██████████| 20/20 [01:21<00:00,  2.72s/iter, val_loss=0.000979, val_og_loss=0.00101]
100%|██████████| 20/20 [01:21<00:00,  4.08s/iter, val_loss=0.000979, val_og_loss=0.00101]
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:59:09<4:13:39, 895.28s/it]INFO:ParoQuant:Best val loss: 0.000978607451543212, Original val loss: 0.0010092316661030054
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [2:59:14<4:13:39, 895.28s/it]2025-12-22 17:29:27,433 - INFO - Best val loss: 0.000978607451543212, Original val loss: 0.0010092316661030054
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:32<?, ?iter/s, val_loss=0.000579, val_og_loss=0.000599]
+
  5%|▌         | 1/20 [00:32<10:15, 32.39s/iter, val_loss=0.000579, val_og_loss=0.000599]
+
  5%|▌         | 1/20 [00:35<10:15, 32.39s/iter, val_loss=0.000583, val_og_loss=0.000599]
+
 10%|█         | 2/20 [00:35<04:28, 14.93s/iter, val_loss=0.000583, val_og_loss=0.000599]
+
 10%|█         | 2/20 [00:37<04:28, 14.93s/iter, val_loss=0.000585, val_og_loss=0.000599]
+
 15%|█▌        | 3/20 [00:37<02:39,  9.36s/iter, val_loss=0.000585, val_og_loss=0.000599]
+
 15%|█▌        | 3/20 [00:40<02:39,  9.36s/iter, val_loss=0.000587, val_og_loss=0.000599]
+
 20%|██        | 4/20 [00:40<01:47,  6.74s/iter, val_loss=0.000587, val_og_loss=0.000599]
+
 20%|██        | 4/20 [00:43<01:47,  6.74s/iter, val_loss=0.000588, val_og_loss=0.000599]
+
 25%|██▌       | 5/20 [00:43<01:19,  5.29s/iter, val_loss=0.000588, val_og_loss=0.000599]
+
 25%|██▌       | 5/20 [00:45<01:19,  5.29s/iter, val_loss=0.000589, val_og_loss=0.000599]
+
 30%|███       | 6/20 [00:45<01:01,  4.41s/iter, val_loss=0.000589, val_og_loss=0.000599]
+
 30%|███       | 6/20 [00:48<01:01,  4.41s/iter, val_loss=0.00059, val_og_loss=0.000599] 
+
 35%|███▌      | 7/20 [00:48<00:50,  3.85s/iter, val_loss=0.00059, val_og_loss=0.000599]
+
 35%|███▌      | 7/20 [00:51<00:50,  3.85s/iter, val_loss=0.000588, val_og_loss=0.000599]
+
 40%|████      | 8/20 [00:51<00:41,  3.49s/iter, val_loss=0.000588, val_og_loss=0.000599]
+
 40%|████      | 8/20 [00:54<00:41,  3.49s/iter, val_loss=0.000587, val_og_loss=0.000599]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.25s/iter, val_loss=0.000587, val_og_loss=0.000599]
+
 45%|████▌     | 9/20 [00:56<00:35,  3.25s/iter, val_loss=0.000587, val_og_loss=0.000599]
+
 50%|█████     | 10/20 [00:56<00:30,  3.08s/iter, val_loss=0.000587, val_og_loss=0.000599]
+
 50%|█████     | 10/20 [00:59<00:30,  3.08s/iter, val_loss=0.000585, val_og_loss=0.000599]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.97s/iter, val_loss=0.000585, val_og_loss=0.000599]
+
 55%|█████▌    | 11/20 [01:02<00:26,  2.97s/iter, val_loss=0.000586, val_og_loss=0.000599]
+
 60%|██████    | 12/20 [01:02<00:23,  2.90s/iter, val_loss=0.000586, val_og_loss=0.000599]
+
 60%|██████    | 12/20 [01:04<00:23,  2.90s/iter, val_loss=0.000586, val_og_loss=0.000599]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.000586, val_og_loss=0.000599]
+
 65%|██████▌   | 13/20 [01:07<00:19,  2.84s/iter, val_loss=0.000585, val_og_loss=0.000599]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.000585, val_og_loss=0.000599]
+
 70%|███████   | 14/20 [01:10<00:16,  2.80s/iter, val_loss=0.000583, val_og_loss=0.000599]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.78s/iter, val_loss=0.000583, val_og_loss=0.000599]
+
 75%|███████▌  | 15/20 [01:13<00:13,  2.78s/iter, val_loss=0.000583, val_og_loss=0.000599]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.000583, val_og_loss=0.000599]
+
 80%|████████  | 16/20 [01:15<00:11,  2.76s/iter, val_loss=0.000583, val_og_loss=0.000599]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.75s/iter, val_loss=0.000583, val_og_loss=0.000599]
+
 85%|████████▌ | 17/20 [01:18<00:08,  2.75s/iter, val_loss=0.000582, val_og_loss=0.000599]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.000582, val_og_loss=0.000599]
+
 90%|█████████ | 18/20 [01:21<00:05,  2.74s/iter, val_loss=0.000582, val_og_loss=0.000599]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.000582, val_og_loss=0.000599]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.73s/iter, val_loss=0.000581, val_og_loss=0.000599]
+
100%|██████████| 20/20 [01:23<00:00,  2.73s/iter, val_loss=0.000581, val_og_loss=0.000599]
100%|██████████| 20/20 [01:23<00:00,  4.20s/iter, val_loss=0.000581, val_og_loss=0.000599]
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [3:00:38<4:13:39, 895.28s/it]INFO:ParoQuant:Best val loss: 0.0005788736743852496, Original val loss: 0.0005990131758153439
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [3:00:39<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [3:00:39<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [3:01:40<4:13:39, 895.28s/it]2025-12-22 17:30:55,988 - INFO - Best val loss: 0.0005788736743852496, Original val loss: 0.0005990131758153439
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:34<?, ?iter/s, val_loss=8.86e-5, val_og_loss=8.99e-5]
+
  5%|▌         | 1/20 [00:34<10:52, 34.35s/iter, val_loss=8.86e-5, val_og_loss=8.99e-5]
+
  5%|▌         | 1/20 [00:39<10:52, 34.35s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 10%|█         | 2/20 [00:39<05:12, 17.33s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 10%|█         | 2/20 [00:45<05:12, 17.33s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 15%|█▌        | 3/20 [00:45<03:22, 11.89s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 15%|█▌        | 3/20 [00:50<03:22, 11.89s/iter, val_loss=8.9e-5, val_og_loss=8.99e-5] 
+
 20%|██        | 4/20 [00:50<02:29,  9.34s/iter, val_loss=8.9e-5, val_og_loss=8.99e-5]
+
 20%|██        | 4/20 [00:56<02:29,  9.34s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 25%|██▌       | 5/20 [00:56<01:58,  7.92s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 25%|██▌       | 5/20 [01:01<01:58,  7.92s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 30%|███       | 6/20 [01:01<01:39,  7.07s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 30%|███       | 6/20 [01:06<01:39,  7.07s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.53s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 35%|███▌      | 7/20 [01:12<01:24,  6.53s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 40%|████      | 8/20 [01:12<01:14,  6.18s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 40%|████      | 8/20 [01:17<01:14,  6.18s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 45%|████▌     | 9/20 [01:17<01:05,  5.94s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 45%|████▌     | 9/20 [01:23<01:05,  5.94s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 50%|█████     | 10/20 [01:23<00:57,  5.79s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 50%|█████     | 10/20 [01:28<00:57,  5.79s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 55%|█████▌    | 11/20 [01:28<00:51,  5.67s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 55%|█████▌    | 11/20 [01:33<00:51,  5.67s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 60%|██████    | 12/20 [01:33<00:44,  5.60s/iter, val_loss=8.89e-5, val_og_loss=8.99e-5]
+
 60%|██████    | 12/20 [01:39<00:44,  5.60s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 65%|██████▌   | 13/20 [01:39<00:38,  5.54s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.54s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 70%|███████   | 14/20 [01:44<00:33,  5.51s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 70%|███████   | 14/20 [01:50<00:33,  5.51s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 75%|███████▌  | 15/20 [01:50<00:27,  5.49s/iter, val_loss=8.88e-5, val_og_loss=8.99e-5]
+
 75%|███████▌  | 15/20 [01:55<00:27,  5.49s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
 80%|████████  | 16/20 [01:55<00:21,  5.47s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
 80%|████████  | 16/20 [02:01<00:21,  5.47s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
 85%|████████▌ | 17/20 [02:01<00:16,  5.46s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
 85%|████████▌ | 17/20 [02:06<00:16,  5.46s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
 90%|█████████ | 18/20 [02:06<00:10,  5.44s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
 95%|█████████▌| 19/20 [02:17<00:05,  5.44s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
100%|██████████| 20/20 [02:17<00:00,  5.43s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
100%|██████████| 20/20 [02:17<00:00,  6.87s/iter, val_loss=8.87e-5, val_og_loss=8.99e-5]
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [3:03:57<4:13:39, 895.28s/it]INFO:ParoQuant:Best val loss: 8.858332876116037e-05, Original val loss: 8.98561702342704e-05
+
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [3:03:58<4:13:39, 895.28s/it]
                                                                       

Processing Blocks:  47%|████▋     | 15/32 [3:04:08<4:13:39, 895.28s/it]
Processing Blocks:  50%|█████     | 16/32 [3:04:46<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:04:46<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:05:04<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:05:04<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:07:22<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:07:22<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:08:21<3:58:22, 893.88s/it]2025-12-22 17:34:15,377 - INFO - Best val loss: 8.858332876116037e-05, Original val loss: 8.98561702342704e-05
+[INFO] Saving Layer 15 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 16...
+[INFO] Starting Sequential Training for Layer 16
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00204, val_og_loss=0.00261]
+
  5%|▌         | 1/20 [00:15<04:57, 15.64s/iter, val_loss=0.00204, val_og_loss=0.00261]
+
  5%|▌         | 1/20 [00:17<04:57, 15.64s/iter, val_loss=0.00206, val_og_loss=0.00261]
+
 10%|█         | 2/20 [00:17<02:13,  7.44s/iter, val_loss=0.00206, val_og_loss=0.00261]
+
 10%|█         | 2/20 [00:19<02:13,  7.44s/iter, val_loss=0.00208, val_og_loss=0.00261]
+
 15%|█▌        | 3/20 [00:19<01:21,  4.81s/iter, val_loss=0.00208, val_og_loss=0.00261]
+
 15%|█▌        | 3/20 [00:20<01:21,  4.81s/iter, val_loss=0.00208, val_og_loss=0.00261]
+
 20%|██        | 4/20 [00:20<00:57,  3.59s/iter, val_loss=0.00208, val_og_loss=0.00261]
+
 20%|██        | 4/20 [00:22<00:57,  3.59s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 25%|██▌       | 5/20 [00:22<00:43,  2.91s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 25%|██▌       | 5/20 [00:24<00:43,  2.91s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 30%|███       | 6/20 [00:24<00:35,  2.50s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 30%|███       | 6/20 [00:25<00:35,  2.50s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 35%|███▌      | 7/20 [00:25<00:29,  2.24s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 35%|███▌      | 7/20 [00:27<00:29,  2.24s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 40%|████      | 8/20 [00:27<00:24,  2.06s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 40%|████      | 8/20 [00:29<00:24,  2.06s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.95s/iter, val_loss=0.00209, val_og_loss=0.00261]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.95s/iter, val_loss=0.00208, val_og_loss=0.00261]
+
 50%|█████     | 10/20 [00:30<00:18,  1.87s/iter, val_loss=0.00208, val_og_loss=0.00261]
+
 50%|█████     | 10/20 [00:32<00:18,  1.87s/iter, val_loss=0.00207, val_og_loss=0.00261]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.81s/iter, val_loss=0.00207, val_og_loss=0.00261]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.81s/iter, val_loss=0.00207, val_og_loss=0.00261]
+
 60%|██████    | 12/20 [00:34<00:14,  1.78s/iter, val_loss=0.00207, val_og_loss=0.00261]
+
 60%|██████    | 12/20 [00:35<00:14,  1.78s/iter, val_loss=0.00206, val_og_loss=0.00261]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.75s/iter, val_loss=0.00206, val_og_loss=0.00261]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=0.00206, val_og_loss=0.00261]
+
 70%|███████   | 14/20 [00:37<00:10,  1.73s/iter, val_loss=0.00206, val_og_loss=0.00261]
+
 70%|███████   | 14/20 [00:39<00:10,  1.73s/iter, val_loss=0.00205, val_og_loss=0.00261]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.00205, val_og_loss=0.00261]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.71s/iter, val_loss=0.00205, val_og_loss=0.00261]
+
 80%|████████  | 16/20 [00:41<00:06,  1.71s/iter, val_loss=0.00205, val_og_loss=0.00261]
+
 80%|████████  | 16/20 [00:42<00:06,  1.71s/iter, val_loss=0.00205, val_og_loss=0.00261]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=0.00205, val_og_loss=0.00261]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=0.00205, val_og_loss=0.00261]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.00205, val_og_loss=0.00261]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.69s/iter, val_loss=0.00204, val_og_loss=0.00261]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=0.00204, val_og_loss=0.00261]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=0.00204, val_og_loss=0.00261]
+
100%|██████████| 20/20 [00:47<00:00,  1.69s/iter, val_loss=0.00204, val_og_loss=0.00261]
100%|██████████| 20/20 [00:47<00:00,  2.39s/iter, val_loss=0.00204, val_og_loss=0.00261]
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:09:09<3:58:22, 893.88s/it]INFO:ParoQuant:Best val loss: 0.0020380106288939714, Original val loss: 0.002612569835036993
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:09:12<3:58:22, 893.88s/it]2025-12-22 17:39:27,121 - INFO - Best val loss: 0.0020380106288939714, Original val loss: 0.002612569835036993
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.00534, val_og_loss=0.00655]
+
  5%|▌         | 1/20 [00:10<03:24, 10.76s/iter, val_loss=0.00534, val_og_loss=0.00655]
+
  5%|▌         | 1/20 [00:12<03:24, 10.76s/iter, val_loss=0.00546, val_og_loss=0.00655]
+
 10%|█         | 2/20 [00:12<01:34,  5.25s/iter, val_loss=0.00546, val_og_loss=0.00655]
+
 10%|█         | 2/20 [00:13<01:34,  5.25s/iter, val_loss=0.00548, val_og_loss=0.00655]
+
 15%|█▌        | 3/20 [00:13<00:59,  3.48s/iter, val_loss=0.00548, val_og_loss=0.00655]
+
 15%|█▌        | 3/20 [00:14<00:59,  3.48s/iter, val_loss=0.00548, val_og_loss=0.00655]
+
 20%|██        | 4/20 [00:14<00:42,  2.66s/iter, val_loss=0.00548, val_og_loss=0.00655]
+
 20%|██        | 4/20 [00:16<00:42,  2.66s/iter, val_loss=0.00548, val_og_loss=0.00655]
+
 25%|██▌       | 5/20 [00:16<00:33,  2.21s/iter, val_loss=0.00548, val_og_loss=0.00655]
+
 25%|██▌       | 5/20 [00:17<00:33,  2.21s/iter, val_loss=0.00547, val_og_loss=0.00655]
+
 30%|███       | 6/20 [00:17<00:27,  1.93s/iter, val_loss=0.00547, val_og_loss=0.00655]
+
 30%|███       | 6/20 [00:19<00:27,  1.93s/iter, val_loss=0.0055, val_og_loss=0.00655] 
+
 35%|███▌      | 7/20 [00:19<00:22,  1.75s/iter, val_loss=0.0055, val_og_loss=0.00655]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.75s/iter, val_loss=0.00545, val_og_loss=0.00655]
+
 40%|████      | 8/20 [00:20<00:19,  1.64s/iter, val_loss=0.00545, val_og_loss=0.00655]
+
 40%|████      | 8/20 [00:21<00:19,  1.64s/iter, val_loss=0.00546, val_og_loss=0.00655]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.56s/iter, val_loss=0.00546, val_og_loss=0.00655]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.56s/iter, val_loss=0.00546, val_og_loss=0.00655]
+
 50%|█████     | 10/20 [00:23<00:15,  1.51s/iter, val_loss=0.00546, val_og_loss=0.00655]
+
 50%|█████     | 10/20 [00:24<00:15,  1.51s/iter, val_loss=0.00548, val_og_loss=0.00655]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.47s/iter, val_loss=0.00548, val_og_loss=0.00655]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.47s/iter, val_loss=0.00544, val_og_loss=0.00655]
+
 60%|██████    | 12/20 [00:26<00:11,  1.44s/iter, val_loss=0.00544, val_og_loss=0.00655]
+
 60%|██████    | 12/20 [00:27<00:11,  1.44s/iter, val_loss=0.00541, val_og_loss=0.00655]
+
 65%|██████▌   | 13/20 [00:27<00:09,  1.43s/iter, val_loss=0.00541, val_og_loss=0.00655]
+
 65%|██████▌   | 13/20 [00:28<00:09,  1.43s/iter, val_loss=0.00541, val_og_loss=0.00655]
+
 70%|███████   | 14/20 [00:28<00:08,  1.42s/iter, val_loss=0.00541, val_og_loss=0.00655]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.00539, val_og_loss=0.00655]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.00539, val_og_loss=0.00655]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.41s/iter, val_loss=0.00546, val_og_loss=0.00655]
+
 80%|████████  | 16/20 [00:31<00:05,  1.40s/iter, val_loss=0.00546, val_og_loss=0.00655]
+
 80%|████████  | 16/20 [00:33<00:05,  1.40s/iter, val_loss=0.0054, val_og_loss=0.00655] 
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.40s/iter, val_loss=0.0054, val_og_loss=0.00655]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00537, val_og_loss=0.00655]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.00537, val_og_loss=0.00655]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00535, val_og_loss=0.00655]
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.40s/iter, val_loss=0.00535, val_og_loss=0.00655]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00535, val_og_loss=0.00655]
+
100%|██████████| 20/20 [00:37<00:00,  1.40s/iter, val_loss=0.00535, val_og_loss=0.00655]
100%|██████████| 20/20 [00:37<00:00,  1.86s/iter, val_loss=0.00535, val_og_loss=0.00655]
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:09:49<3:58:22, 893.88s/it]INFO:ParoQuant:Best val loss: 0.005339955911040306, Original val loss: 0.006553352810442448
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:09:52<3:58:22, 893.88s/it]2025-12-22 17:40:07,755 - INFO - Best val loss: 0.005339955911040306, Original val loss: 0.006553352810442448
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=0.000438, val_og_loss=0.000559]
+
  5%|▌         | 1/20 [00:11<03:32, 11.18s/iter, val_loss=0.000438, val_og_loss=0.000559]
+
  5%|▌         | 1/20 [00:12<03:32, 11.18s/iter, val_loss=0.00044, val_og_loss=0.000559] 
+
 10%|█         | 2/20 [00:12<01:37,  5.41s/iter, val_loss=0.00044, val_og_loss=0.000559]
+
 10%|█         | 2/20 [00:13<01:37,  5.41s/iter, val_loss=0.00044, val_og_loss=0.000559]
+
 15%|█▌        | 3/20 [00:13<01:00,  3.58s/iter, val_loss=0.00044, val_og_loss=0.000559]
+
 15%|█▌        | 3/20 [00:15<01:00,  3.58s/iter, val_loss=0.000441, val_og_loss=0.000559]
+
 20%|██        | 4/20 [00:15<00:43,  2.71s/iter, val_loss=0.000441, val_og_loss=0.000559]
+
 20%|██        | 4/20 [00:16<00:43,  2.71s/iter, val_loss=0.000441, val_og_loss=0.000559]
+
 25%|██▌       | 5/20 [00:16<00:33,  2.24s/iter, val_loss=0.000441, val_og_loss=0.000559]
+
 25%|██▌       | 5/20 [00:18<00:33,  2.24s/iter, val_loss=0.000442, val_og_loss=0.000559]
+
 30%|███       | 6/20 [00:18<00:27,  1.95s/iter, val_loss=0.000442, val_og_loss=0.000559]
+
 30%|███       | 6/20 [00:19<00:27,  1.95s/iter, val_loss=0.000442, val_og_loss=0.000559]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.77s/iter, val_loss=0.000442, val_og_loss=0.000559]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.77s/iter, val_loss=0.00044, val_og_loss=0.000559] 
+
 40%|████      | 8/20 [00:20<00:19,  1.65s/iter, val_loss=0.00044, val_og_loss=0.000559]
+
 40%|████      | 8/20 [00:22<00:19,  1.65s/iter, val_loss=0.00044, val_og_loss=0.000559]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.57s/iter, val_loss=0.00044, val_og_loss=0.000559]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.57s/iter, val_loss=0.000439, val_og_loss=0.000559]
+
 50%|█████     | 10/20 [00:23<00:15,  1.52s/iter, val_loss=0.000439, val_og_loss=0.000559]
+
 50%|█████     | 10/20 [00:25<00:15,  1.52s/iter, val_loss=0.000438, val_og_loss=0.000559]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.48s/iter, val_loss=0.000438, val_og_loss=0.000559]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.48s/iter, val_loss=0.000438, val_og_loss=0.000559]
+
 60%|██████    | 12/20 [00:26<00:11,  1.46s/iter, val_loss=0.000438, val_og_loss=0.000559]
+
 60%|██████    | 12/20 [00:27<00:11,  1.46s/iter, val_loss=0.000437, val_og_loss=0.000559]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.44s/iter, val_loss=0.000437, val_og_loss=0.000559]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.44s/iter, val_loss=0.000436, val_og_loss=0.000559]
+
 70%|███████   | 14/20 [00:29<00:08,  1.42s/iter, val_loss=0.000436, val_og_loss=0.000559]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.000435, val_og_loss=0.000559]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.42s/iter, val_loss=0.000435, val_og_loss=0.000559]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.42s/iter, val_loss=0.000434, val_og_loss=0.000559]
+
 80%|████████  | 16/20 [00:32<00:05,  1.41s/iter, val_loss=0.000434, val_og_loss=0.000559]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.000434, val_og_loss=0.000559]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.41s/iter, val_loss=0.000434, val_og_loss=0.000559]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.41s/iter, val_loss=0.000434, val_og_loss=0.000559]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.000434, val_og_loss=0.000559]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.000433, val_og_loss=0.000559]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.000433, val_og_loss=0.000559]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.000434, val_og_loss=0.000559]
+
100%|██████████| 20/20 [00:37<00:00,  1.41s/iter, val_loss=0.000434, val_og_loss=0.000559]
100%|██████████| 20/20 [00:37<00:00,  1.89s/iter, val_loss=0.000434, val_og_loss=0.000559]
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:10:30<3:58:22, 893.88s/it]INFO:ParoQuant:Best val loss: 0.0004328708746470511, Original val loss: 0.0005587997147813439
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:10:32<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:10:32<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:11:03<3:58:22, 893.88s/it]2025-12-22 17:40:48,496 - INFO - Best val loss: 0.0004328708746470511, Original val loss: 0.0005587997147813439
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
  5%|▌         | 1/20 [00:16<05:07, 16.19s/iter, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
  5%|▌         | 1/20 [00:17<05:07, 16.19s/iter, val_loss=3.39e-5, val_og_loss=3.5e-5]
+
 10%|█         | 2/20 [00:17<02:17,  7.66s/iter, val_loss=3.39e-5, val_og_loss=3.5e-5]
+
 10%|█         | 2/20 [00:19<02:17,  7.66s/iter, val_loss=3.38e-5, val_og_loss=3.5e-5]
+
 15%|█▌        | 3/20 [00:19<01:23,  4.93s/iter, val_loss=3.38e-5, val_og_loss=3.5e-5]
+
 15%|█▌        | 3/20 [00:21<01:23,  4.93s/iter, val_loss=3.38e-5, val_og_loss=3.5e-5]
+
 20%|██        | 4/20 [00:21<00:58,  3.65s/iter, val_loss=3.38e-5, val_og_loss=3.5e-5]
+
 20%|██        | 4/20 [00:22<00:58,  3.65s/iter, val_loss=3.39e-5, val_og_loss=3.5e-5]
+
 25%|██▌       | 5/20 [00:22<00:44,  2.94s/iter, val_loss=3.39e-5, val_og_loss=3.5e-5]
+
 25%|██▌       | 5/20 [00:24<00:44,  2.94s/iter, val_loss=3.38e-5, val_og_loss=3.5e-5]
+
 30%|███       | 6/20 [00:24<00:35,  2.51s/iter, val_loss=3.38e-5, val_og_loss=3.5e-5]
+
 30%|███       | 6/20 [00:26<00:35,  2.51s/iter, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
 35%|███▌      | 7/20 [00:26<00:29,  2.24s/iter, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
 35%|███▌      | 7/20 [00:27<00:29,  2.24s/iter, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
 40%|████      | 8/20 [00:27<00:24,  2.06s/iter, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
 40%|████      | 8/20 [00:29<00:24,  2.06s/iter, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.94s/iter, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.94s/iter, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
 50%|█████     | 10/20 [00:31<00:18,  1.86s/iter, val_loss=3.37e-5, val_og_loss=3.5e-5]
+
 50%|█████     | 10/20 [00:33<00:18,  1.86s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.81s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.81s/iter, val_loss=3.35e-5, val_og_loss=3.5e-5]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=3.35e-5, val_og_loss=3.5e-5]
+
 60%|██████    | 12/20 [00:36<00:14,  1.77s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.75s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.75s/iter, val_loss=3.35e-5, val_og_loss=3.5e-5]
+
 70%|███████   | 14/20 [00:38<00:10,  1.73s/iter, val_loss=3.35e-5, val_og_loss=3.5e-5]
+
 70%|███████   | 14/20 [00:39<00:10,  1.73s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.72s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.72s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 80%|████████  | 16/20 [00:43<00:06,  1.70s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.69s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.69s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=3.36e-5, val_og_loss=3.5e-5]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.69s/iter, val_loss=3.35e-5, val_og_loss=3.5e-5]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=3.35e-5, val_og_loss=3.5e-5]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.69s/iter, val_loss=3.35e-5, val_og_loss=3.5e-5]
+
100%|██████████| 20/20 [00:48<00:00,  1.68s/iter, val_loss=3.35e-5, val_og_loss=3.5e-5]
100%|██████████| 20/20 [00:48<00:00,  2.41s/iter, val_loss=3.35e-5, val_og_loss=3.5e-5]
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:11:51<3:58:22, 893.88s/it]INFO:ParoQuant:Best val loss: 3.347724486957304e-05, Original val loss: 3.4959408367285505e-05
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:11:52<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:11:52<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:12:35<3:58:22, 893.88s/it]2025-12-22 17:42:09,872 - INFO - Best val loss: 3.347724486957304e-05, Original val loss: 3.4959408367285505e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.00101, val_og_loss=0.00103]
+
  5%|▌         | 1/20 [00:30<09:48, 30.98s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
  5%|▌         | 1/20 [00:33<09:48, 30.98s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 10%|█         | 2/20 [00:33<04:18, 14.36s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 10%|█         | 2/20 [00:36<04:18, 14.36s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 15%|█▌        | 3/20 [00:36<02:33,  9.04s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 15%|█▌        | 3/20 [00:39<02:33,  9.04s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 20%|██        | 4/20 [00:39<01:44,  6.54s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 20%|██        | 4/20 [00:41<01:44,  6.54s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 25%|██▌       | 5/20 [00:41<01:17,  5.16s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.16s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 30%|███       | 6/20 [00:44<01:00,  4.33s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 30%|███       | 6/20 [00:47<01:00,  4.33s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.80s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.80s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 40%|████      | 8/20 [00:49<00:41,  3.46s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 40%|████      | 8/20 [00:52<00:41,  3.46s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.23s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.23s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 50%|█████     | 10/20 [00:55<00:30,  3.08s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 50%|█████     | 10/20 [00:58<00:30,  3.08s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.96s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.00102, val_og_loss=0.00103]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.84s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 70%|███████   | 14/20 [01:06<00:16,  2.81s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 70%|███████   | 14/20 [01:09<00:16,  2.81s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.75s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.75s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.74s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
100%|██████████| 20/20 [01:22<00:00,  2.73s/iter, val_loss=0.00101, val_og_loss=0.00103]
100%|██████████| 20/20 [01:22<00:00,  4.13s/iter, val_loss=0.00101, val_og_loss=0.00103]
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:13:58<3:58:22, 893.88s/it]INFO:ParoQuant:Best val loss: 0.0010071407305076718, Original val loss: 0.0010339174186810851
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:14:02<3:58:22, 893.88s/it]2025-12-22 17:44:16,549 - INFO - Best val loss: 0.0010071407305076718, Original val loss: 0.0010339174186810851
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.000563, val_og_loss=0.000579]
+
  5%|▌         | 1/20 [00:31<09:53, 31.21s/iter, val_loss=0.000563, val_og_loss=0.000579]
+
  5%|▌         | 1/20 [00:33<09:53, 31.21s/iter, val_loss=0.000565, val_og_loss=0.000579]
+
 10%|█         | 2/20 [00:33<04:20, 14.45s/iter, val_loss=0.000565, val_og_loss=0.000579]
+
 10%|█         | 2/20 [00:36<04:20, 14.45s/iter, val_loss=0.000568, val_og_loss=0.000579]
+
 15%|█▌        | 3/20 [00:36<02:34,  9.10s/iter, val_loss=0.000568, val_og_loss=0.000579]
+
 15%|█▌        | 3/20 [00:39<02:34,  9.10s/iter, val_loss=0.00057, val_og_loss=0.000579] 
+
 20%|██        | 4/20 [00:39<01:45,  6.58s/iter, val_loss=0.00057, val_og_loss=0.000579]
+
 20%|██        | 4/20 [00:42<01:45,  6.58s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 25%|██▌       | 5/20 [00:42<01:17,  5.19s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.19s/iter, val_loss=0.000572, val_og_loss=0.000579]
+
 30%|███       | 6/20 [00:44<01:00,  4.35s/iter, val_loss=0.000572, val_og_loss=0.000579]
+
 30%|███       | 6/20 [00:47<01:00,  4.35s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.81s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.81s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 40%|████      | 8/20 [00:50<00:41,  3.47s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 40%|████      | 8/20 [00:52<00:41,  3.47s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.23s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.23s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 50%|█████     | 10/20 [00:55<00:30,  3.07s/iter, val_loss=0.000571, val_og_loss=0.000579]
+
 50%|█████     | 10/20 [00:58<00:30,  3.07s/iter, val_loss=0.00057, val_og_loss=0.000579] 
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.96s/iter, val_loss=0.00057, val_og_loss=0.000579]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.96s/iter, val_loss=0.000569, val_og_loss=0.000579]
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.000569, val_og_loss=0.000579]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.000569, val_og_loss=0.000579]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.84s/iter, val_loss=0.000569, val_og_loss=0.000579]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000568, val_og_loss=0.000579]
+
 70%|███████   | 14/20 [01:06<00:16,  2.81s/iter, val_loss=0.000568, val_og_loss=0.000579]
+
 70%|███████   | 14/20 [01:09<00:16,  2.81s/iter, val_loss=0.000567, val_og_loss=0.000579]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.000567, val_og_loss=0.000579]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.000567, val_og_loss=0.000579]
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.000567, val_og_loss=0.000579]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.000566, val_og_loss=0.000579]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.74s/iter, val_loss=0.000566, val_og_loss=0.000579]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.74s/iter, val_loss=0.000566, val_og_loss=0.000579]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.73s/iter, val_loss=0.000566, val_og_loss=0.000579]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.73s/iter, val_loss=0.000566, val_og_loss=0.000579]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.000566, val_og_loss=0.000579]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.000566, val_og_loss=0.000579]
+
100%|██████████| 20/20 [01:22<00:00,  2.73s/iter, val_loss=0.000566, val_og_loss=0.000579]
100%|██████████| 20/20 [01:22<00:00,  4.14s/iter, val_loss=0.000566, val_og_loss=0.000579]
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:15:25<3:58:22, 893.88s/it]INFO:ParoQuant:Best val loss: 0.0005626765196211636, Original val loss: 0.0005792698939330876
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:15:26<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:15:26<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:16:26<3:58:22, 893.88s/it]2025-12-22 17:45:43,487 - INFO - Best val loss: 0.0005626765196211636, Original val loss: 0.0005792698939330876
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=8.75e-5, val_og_loss=8.87e-5]
+
  5%|▌         | 1/20 [00:33<10:38, 33.58s/iter, val_loss=8.75e-5, val_og_loss=8.87e-5]
+
  5%|▌         | 1/20 [00:39<10:38, 33.58s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 10%|█         | 2/20 [00:39<05:06, 17.02s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 10%|█         | 2/20 [00:44<05:06, 17.02s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 15%|█▌        | 3/20 [00:44<03:19, 11.74s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 15%|█▌        | 3/20 [00:49<03:19, 11.74s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 20%|██        | 4/20 [00:49<02:27,  9.25s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 20%|██        | 4/20 [00:55<02:27,  9.25s/iter, val_loss=8.79e-5, val_og_loss=8.87e-5]
+
 25%|██▌       | 5/20 [00:55<01:58,  7.87s/iter, val_loss=8.79e-5, val_og_loss=8.87e-5]
+
 25%|██▌       | 5/20 [01:00<01:58,  7.87s/iter, val_loss=8.79e-5, val_og_loss=8.87e-5]
+
 30%|███       | 6/20 [01:00<01:38,  7.04s/iter, val_loss=8.79e-5, val_og_loss=8.87e-5]
+
 30%|███       | 6/20 [01:06<01:38,  7.04s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.51s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.51s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 40%|████      | 8/20 [01:11<01:13,  6.16s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 40%|████      | 8/20 [01:16<01:13,  6.16s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 45%|████▌     | 9/20 [01:16<01:05,  5.93s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.93s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 50%|█████     | 10/20 [01:22<00:57,  5.77s/iter, val_loss=8.78e-5, val_og_loss=8.87e-5]
+
 50%|█████     | 10/20 [01:27<00:57,  5.77s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 55%|█████▌    | 11/20 [01:27<00:51,  5.67s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 55%|█████▌    | 11/20 [01:33<00:51,  5.67s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 60%|██████    | 12/20 [01:33<00:44,  5.60s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 60%|██████    | 12/20 [01:38<00:44,  5.60s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.54s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.54s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 70%|███████   | 14/20 [01:44<00:33,  5.50s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 70%|███████   | 14/20 [01:49<00:33,  5.50s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.48s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 75%|███████▌  | 15/20 [01:54<00:27,  5.48s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 80%|████████  | 16/20 [01:54<00:21,  5.46s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 80%|████████  | 16/20 [02:00<00:21,  5.46s/iter, val_loss=8.76e-5, val_og_loss=8.87e-5]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.45s/iter, val_loss=8.76e-5, val_og_loss=8.87e-5]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.45s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.44s/iter, val_loss=8.77e-5, val_og_loss=8.87e-5]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=8.76e-5, val_og_loss=8.87e-5]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=8.76e-5, val_og_loss=8.87e-5]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.44s/iter, val_loss=8.75e-5, val_og_loss=8.87e-5]
+
100%|██████████| 20/20 [02:16<00:00,  5.43s/iter, val_loss=8.75e-5, val_og_loss=8.87e-5]
100%|██████████| 20/20 [02:16<00:00,  6.83s/iter, val_loss=8.75e-5, val_og_loss=8.87e-5]
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:18:43<3:58:22, 893.88s/it]INFO:ParoQuant:Best val loss: 8.751574205234647e-05, Original val loss: 8.87324204086326e-05
+
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:18:44<3:58:22, 893.88s/it]
                                                                       

Processing Blocks:  50%|█████     | 16/32 [3:18:55<3:58:22, 893.88s/it]
Processing Blocks:  53%|█████▎    | 17/32 [3:19:33<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:19:33<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:19:52<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:19:52<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:22:08<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:22:08<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:23:05<3:42:56, 891.76s/it]2025-12-22 17:49:01,378 - INFO - Best val loss: 8.751574205234647e-05, Original val loss: 8.87324204086326e-05
+[INFO] Saving Layer 16 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 17...
+[INFO] Starting Sequential Training for Layer 17
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.00222, val_og_loss=0.00269]
+
  5%|▌         | 1/20 [00:16<05:08, 16.24s/iter, val_loss=0.00222, val_og_loss=0.00269]
+
  5%|▌         | 1/20 [00:17<05:08, 16.24s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 10%|█         | 2/20 [00:17<02:18,  7.67s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 10%|█         | 2/20 [00:19<02:18,  7.67s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 15%|█▌        | 3/20 [00:19<01:23,  4.94s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 15%|█▌        | 3/20 [00:21<01:23,  4.94s/iter, val_loss=0.00226, val_og_loss=0.00269]
+
 20%|██        | 4/20 [00:21<00:58,  3.65s/iter, val_loss=0.00226, val_og_loss=0.00269]
+
 20%|██        | 4/20 [00:22<00:58,  3.65s/iter, val_loss=0.00226, val_og_loss=0.00269]
+
 25%|██▌       | 5/20 [00:22<00:44,  2.94s/iter, val_loss=0.00226, val_og_loss=0.00269]
+
 25%|██▌       | 5/20 [00:24<00:44,  2.94s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 30%|███       | 6/20 [00:24<00:35,  2.51s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 30%|███       | 6/20 [00:26<00:35,  2.51s/iter, val_loss=0.00226, val_og_loss=0.00269]
+
 35%|███▌      | 7/20 [00:26<00:29,  2.24s/iter, val_loss=0.00226, val_og_loss=0.00269]
+
 35%|███▌      | 7/20 [00:27<00:29,  2.24s/iter, val_loss=0.00226, val_og_loss=0.00269]
+
 40%|████      | 8/20 [00:27<00:24,  2.06s/iter, val_loss=0.00226, val_og_loss=0.00269]
+
 40%|████      | 8/20 [00:29<00:24,  2.06s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.94s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.94s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 50%|█████     | 10/20 [00:31<00:18,  1.86s/iter, val_loss=0.00225, val_og_loss=0.00269]
+
 50%|█████     | 10/20 [00:33<00:18,  1.86s/iter, val_loss=0.00224, val_og_loss=0.00269]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.80s/iter, val_loss=0.00224, val_og_loss=0.00269]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.80s/iter, val_loss=0.00224, val_og_loss=0.00269]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=0.00224, val_og_loss=0.00269]
+
 60%|██████    | 12/20 [00:36<00:14,  1.77s/iter, val_loss=0.00223, val_og_loss=0.00269]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.74s/iter, val_loss=0.00223, val_og_loss=0.00269]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.74s/iter, val_loss=0.00223, val_og_loss=0.00269]
+
 70%|███████   | 14/20 [00:38<00:10,  1.72s/iter, val_loss=0.00223, val_og_loss=0.00269]
+
 70%|███████   | 14/20 [00:39<00:10,  1.72s/iter, val_loss=0.00223, val_og_loss=0.00269]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.00223, val_og_loss=0.00269]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.71s/iter, val_loss=0.00223, val_og_loss=0.00269]
+
 80%|████████  | 16/20 [00:41<00:06,  1.71s/iter, val_loss=0.00223, val_og_loss=0.00269]
+
 80%|████████  | 16/20 [00:43<00:06,  1.71s/iter, val_loss=0.00222, val_og_loss=0.00269]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.70s/iter, val_loss=0.00222, val_og_loss=0.00269]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=0.00221, val_og_loss=0.00269]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.00221, val_og_loss=0.00269]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.69s/iter, val_loss=0.00221, val_og_loss=0.00269]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=0.00221, val_og_loss=0.00269]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.69s/iter, val_loss=0.00221, val_og_loss=0.00269]
+
100%|██████████| 20/20 [00:48<00:00,  1.69s/iter, val_loss=0.00221, val_og_loss=0.00269]
100%|██████████| 20/20 [00:48<00:00,  2.41s/iter, val_loss=0.00221, val_og_loss=0.00269]
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:23:53<3:42:56, 891.76s/it]INFO:ParoQuant:Best val loss: 0.0022082380019128323, Original val loss: 0.002694537630304694
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:23:54<3:42:56, 891.76s/it]2025-12-22 17:54:11,281 - INFO - Best val loss: 0.0022082380019128323, Original val loss: 0.002694537630304694
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.00578, val_og_loss=0.00671]
+
  5%|▌         | 1/20 [00:10<03:11, 10.10s/iter, val_loss=0.00578, val_og_loss=0.00671]
+
  5%|▌         | 1/20 [00:11<03:11, 10.10s/iter, val_loss=0.00583, val_og_loss=0.00671]
+
 10%|█         | 2/20 [00:11<01:29,  4.97s/iter, val_loss=0.00583, val_og_loss=0.00671]
+
 10%|█         | 2/20 [00:12<01:29,  4.97s/iter, val_loss=0.00584, val_og_loss=0.00671]
+
 15%|█▌        | 3/20 [00:12<00:56,  3.35s/iter, val_loss=0.00584, val_og_loss=0.00671]
+
 15%|█▌        | 3/20 [00:14<00:56,  3.35s/iter, val_loss=0.00583, val_og_loss=0.00671]
+
 20%|██        | 4/20 [00:14<00:41,  2.58s/iter, val_loss=0.00583, val_og_loss=0.00671]
+
 20%|██        | 4/20 [00:15<00:41,  2.58s/iter, val_loss=0.00587, val_og_loss=0.00671]
+
 25%|██▌       | 5/20 [00:15<00:32,  2.15s/iter, val_loss=0.00587, val_og_loss=0.00671]
+
 25%|██▌       | 5/20 [00:17<00:32,  2.15s/iter, val_loss=0.00589, val_og_loss=0.00671]
+
 30%|███       | 6/20 [00:17<00:26,  1.89s/iter, val_loss=0.00589, val_og_loss=0.00671]
+
 30%|███       | 6/20 [00:18<00:26,  1.89s/iter, val_loss=0.00585, val_og_loss=0.00671]
+
 35%|███▌      | 7/20 [00:18<00:22,  1.72s/iter, val_loss=0.00585, val_og_loss=0.00671]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.72s/iter, val_loss=0.00584, val_og_loss=0.00671]
+
 40%|████      | 8/20 [00:19<00:19,  1.62s/iter, val_loss=0.00584, val_og_loss=0.00671]
+
 40%|████      | 8/20 [00:21<00:19,  1.62s/iter, val_loss=0.00585, val_og_loss=0.00671]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.55s/iter, val_loss=0.00585, val_og_loss=0.00671]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.55s/iter, val_loss=0.00584, val_og_loss=0.00671]
+
 50%|█████     | 10/20 [00:22<00:14,  1.50s/iter, val_loss=0.00584, val_og_loss=0.00671]
+
 50%|█████     | 10/20 [00:24<00:14,  1.50s/iter, val_loss=0.00583, val_og_loss=0.00671]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.46s/iter, val_loss=0.00583, val_og_loss=0.00671]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.46s/iter, val_loss=0.00582, val_og_loss=0.00671]
+
 60%|██████    | 12/20 [00:25<00:11,  1.44s/iter, val_loss=0.00582, val_og_loss=0.00671]
+
 60%|██████    | 12/20 [00:26<00:11,  1.44s/iter, val_loss=0.00586, val_og_loss=0.00671]
+
 65%|██████▌   | 13/20 [00:26<00:09,  1.43s/iter, val_loss=0.00586, val_og_loss=0.00671]
+
 65%|██████▌   | 13/20 [00:28<00:09,  1.43s/iter, val_loss=0.00581, val_og_loss=0.00671]
+
 70%|███████   | 14/20 [00:28<00:08,  1.42s/iter, val_loss=0.00581, val_og_loss=0.00671]
+
 70%|███████   | 14/20 [00:29<00:08,  1.42s/iter, val_loss=0.00582, val_og_loss=0.00671]
+
 75%|███████▌  | 15/20 [00:29<00:07,  1.41s/iter, val_loss=0.00582, val_og_loss=0.00671]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.00581, val_og_loss=0.00671]
+
 80%|████████  | 16/20 [00:30<00:05,  1.40s/iter, val_loss=0.00581, val_og_loss=0.00671]
+
 80%|████████  | 16/20 [00:32<00:05,  1.40s/iter, val_loss=0.0058, val_og_loss=0.00671] 
+
 85%|████████▌ | 17/20 [00:32<00:04,  1.40s/iter, val_loss=0.0058, val_og_loss=0.00671]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.40s/iter, val_loss=0.00579, val_og_loss=0.00671]
+
 90%|█████████ | 18/20 [00:33<00:02,  1.40s/iter, val_loss=0.00579, val_og_loss=0.00671]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00578, val_og_loss=0.00671]
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.40s/iter, val_loss=0.00578, val_og_loss=0.00671]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.00578, val_og_loss=0.00671]
+
100%|██████████| 20/20 [00:36<00:00,  1.40s/iter, val_loss=0.00578, val_og_loss=0.00671]
100%|██████████| 20/20 [00:36<00:00,  1.83s/iter, val_loss=0.00578, val_og_loss=0.00671]
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:24:31<3:42:56, 891.76s/it]INFO:ParoQuant:Best val loss: 0.0057770730927586555, Original val loss: 0.0067121172323822975
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:24:33<3:42:56, 891.76s/it]2025-12-22 17:54:49,220 - INFO - Best val loss: 0.0057770730927586555, Original val loss: 0.0067121172323822975
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:12<?, ?iter/s, val_loss=0.000526, val_og_loss=0.000652]
+
  5%|▌         | 1/20 [00:12<03:52, 12.22s/iter, val_loss=0.000526, val_og_loss=0.000652]
+
  5%|▌         | 1/20 [00:13<03:52, 12.22s/iter, val_loss=0.000527, val_og_loss=0.000652]
+
 10%|█         | 2/20 [00:13<01:45,  5.85s/iter, val_loss=0.000527, val_og_loss=0.000652]
+
 10%|█         | 2/20 [00:15<01:45,  5.85s/iter, val_loss=0.00053, val_og_loss=0.000652] 
+
 15%|█▌        | 3/20 [00:15<01:04,  3.82s/iter, val_loss=0.00053, val_og_loss=0.000652]
+
 15%|█▌        | 3/20 [00:16<01:04,  3.82s/iter, val_loss=0.000528, val_og_loss=0.000652]
+
 20%|██        | 4/20 [00:16<00:45,  2.86s/iter, val_loss=0.000528, val_og_loss=0.000652]
+
 20%|██        | 4/20 [00:17<00:45,  2.86s/iter, val_loss=0.00053, val_og_loss=0.000652] 
+
 25%|██▌       | 5/20 [00:17<00:34,  2.33s/iter, val_loss=0.00053, val_og_loss=0.000652]
+
 25%|██▌       | 5/20 [00:19<00:34,  2.33s/iter, val_loss=0.000532, val_og_loss=0.000652]
+
 30%|███       | 6/20 [00:19<00:28,  2.02s/iter, val_loss=0.000532, val_og_loss=0.000652]
+
 30%|███       | 6/20 [00:20<00:28,  2.02s/iter, val_loss=0.000532, val_og_loss=0.000652]
+
 35%|███▌      | 7/20 [00:20<00:23,  1.81s/iter, val_loss=0.000532, val_og_loss=0.000652]
+
 35%|███▌      | 7/20 [00:21<00:23,  1.81s/iter, val_loss=0.00053, val_og_loss=0.000652] 
+
 40%|████      | 8/20 [00:21<00:20,  1.68s/iter, val_loss=0.00053, val_og_loss=0.000652]
+
 40%|████      | 8/20 [00:23<00:20,  1.68s/iter, val_loss=0.000529, val_og_loss=0.000652]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.59s/iter, val_loss=0.000529, val_og_loss=0.000652]
+
 45%|████▌     | 9/20 [00:24<00:17,  1.59s/iter, val_loss=0.000528, val_og_loss=0.000652]
+
 50%|█████     | 10/20 [00:24<00:15,  1.53s/iter, val_loss=0.000528, val_og_loss=0.000652]
+
 50%|█████     | 10/20 [00:26<00:15,  1.53s/iter, val_loss=0.000527, val_og_loss=0.000652]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.49s/iter, val_loss=0.000527, val_og_loss=0.000652]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.49s/iter, val_loss=0.000527, val_og_loss=0.000652]
+
 60%|██████    | 12/20 [00:27<00:11,  1.46s/iter, val_loss=0.000527, val_og_loss=0.000652]
+
 60%|██████    | 12/20 [00:28<00:11,  1.46s/iter, val_loss=0.000526, val_og_loss=0.000652]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.44s/iter, val_loss=0.000526, val_og_loss=0.000652]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.44s/iter, val_loss=0.000524, val_og_loss=0.000652]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.000524, val_og_loss=0.000652]
+
 70%|███████   | 14/20 [00:31<00:08,  1.42s/iter, val_loss=0.000524, val_og_loss=0.000652]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.42s/iter, val_loss=0.000524, val_og_loss=0.000652]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.42s/iter, val_loss=0.000524, val_og_loss=0.000652]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.000524, val_og_loss=0.000652]
+
 80%|████████  | 16/20 [00:34<00:05,  1.41s/iter, val_loss=0.000523, val_og_loss=0.000652]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.000523, val_og_loss=0.000652]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.40s/iter, val_loss=0.000522, val_og_loss=0.000652]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.000522, val_og_loss=0.000652]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.40s/iter, val_loss=0.000521, val_og_loss=0.000652]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.000521, val_og_loss=0.000652]
+
 95%|█████████▌| 19/20 [00:38<00:01,  1.40s/iter, val_loss=0.000521, val_og_loss=0.000652]
+
100%|██████████| 20/20 [00:38<00:00,  1.40s/iter, val_loss=0.000521, val_og_loss=0.000652]
100%|██████████| 20/20 [00:38<00:00,  1.93s/iter, val_loss=0.000521, val_og_loss=0.000652]
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:25:12<3:42:56, 891.76s/it]INFO:ParoQuant:Best val loss: 0.0005209114751778543, Original val loss: 0.0006517681176774204
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:25:14<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:25:14<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:25:44<3:42:56, 891.76s/it]2025-12-22 17:55:30,161 - INFO - Best val loss: 0.0005209114751778543, Original val loss: 0.0006517681176774204
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
  5%|▌         | 1/20 [00:14<04:26, 14.01s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
  5%|▌         | 1/20 [00:15<04:26, 14.01s/iter, val_loss=2.94e-5, val_og_loss=2.99e-5]
+
 10%|█         | 2/20 [00:15<02:01,  6.77s/iter, val_loss=2.94e-5, val_og_loss=2.99e-5]
+
 10%|█         | 2/20 [00:17<02:01,  6.77s/iter, val_loss=2.95e-5, val_og_loss=2.99e-5]
+
 15%|█▌        | 3/20 [00:17<01:15,  4.45s/iter, val_loss=2.95e-5, val_og_loss=2.99e-5]
+
 15%|█▌        | 3/20 [00:19<01:15,  4.45s/iter, val_loss=2.95e-5, val_og_loss=2.99e-5]
+
 20%|██        | 4/20 [00:19<00:53,  3.36s/iter, val_loss=2.95e-5, val_og_loss=2.99e-5]
+
 20%|██        | 4/20 [00:20<00:53,  3.36s/iter, val_loss=2.93e-5, val_og_loss=2.99e-5]
+
 25%|██▌       | 5/20 [00:20<00:41,  2.75s/iter, val_loss=2.93e-5, val_og_loss=2.99e-5]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.75s/iter, val_loss=2.95e-5, val_og_loss=2.99e-5]
+
 30%|███       | 6/20 [00:22<00:33,  2.39s/iter, val_loss=2.95e-5, val_og_loss=2.99e-5]
+
 30%|███       | 6/20 [00:24<00:33,  2.39s/iter, val_loss=2.93e-5, val_og_loss=2.99e-5]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.15s/iter, val_loss=2.93e-5, val_og_loss=2.99e-5]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.15s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 40%|████      | 8/20 [00:25<00:24,  2.00s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 40%|████      | 8/20 [00:27<00:24,  2.00s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.90s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 45%|████▌     | 9/20 [00:29<00:20,  1.90s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 50%|█████     | 10/20 [00:29<00:18,  1.83s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.83s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.79s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 60%|██████    | 12/20 [00:32<00:14,  1.75s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 60%|██████    | 12/20 [00:34<00:14,  1.75s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.71s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.71s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.71s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=2.9e-5, val_og_loss=2.99e-5] 
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=2.9e-5, val_og_loss=2.99e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.69s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=2.92e-5, val_og_loss=2.99e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
100%|██████████| 20/20 [00:45<00:00,  1.68s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
100%|██████████| 20/20 [00:45<00:00,  2.30s/iter, val_loss=2.91e-5, val_og_loss=2.99e-5]
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:26:30<3:42:56, 891.76s/it]INFO:ParoQuant:Best val loss: 2.901597326854244e-05, Original val loss: 2.985208629979752e-05
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:26:31<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:26:31<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:27:14<3:42:56, 891.76s/it]2025-12-22 17:56:48,625 - INFO - Best val loss: 2.901597326854244e-05, Original val loss: 2.985208629979752e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00105, val_og_loss=0.00108]
+
  5%|▌         | 1/20 [00:29<09:29, 29.96s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
  5%|▌         | 1/20 [00:32<09:29, 29.96s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 10%|█         | 2/20 [00:32<04:10, 13.94s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 10%|█         | 2/20 [00:35<04:10, 13.94s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 15%|█▌        | 3/20 [00:35<02:29,  8.81s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 15%|█▌        | 3/20 [00:38<02:29,  8.81s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 20%|██        | 4/20 [00:38<01:42,  6.40s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 20%|██        | 4/20 [00:40<01:42,  6.40s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 25%|██▌       | 5/20 [00:40<01:16,  5.07s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.07s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 30%|███       | 6/20 [00:43<00:59,  4.27s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 30%|███       | 6/20 [00:46<00:59,  4.27s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 35%|███▌      | 7/20 [00:46<00:48,  3.76s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.76s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 40%|████      | 8/20 [00:48<00:41,  3.43s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 40%|████      | 8/20 [00:51<00:41,  3.43s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.21s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.21s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.95s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.95s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 60%|██████    | 12/20 [00:59<00:23,  2.89s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 60%|██████    | 12/20 [01:02<00:23,  2.89s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.84s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.84s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 70%|███████   | 14/20 [01:05<00:16,  2.81s/iter, val_loss=0.00106, val_og_loss=0.00108]
+
 70%|███████   | 14/20 [01:08<00:16,  2.81s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.78s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.78s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 80%|████████  | 16/20 [01:10<00:11,  2.76s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.75s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.75s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.74s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.74s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.74s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
100%|██████████| 20/20 [01:21<00:00,  2.73s/iter, val_loss=0.00105, val_og_loss=0.00108]
100%|██████████| 20/20 [01:21<00:00,  4.08s/iter, val_loss=0.00105, val_og_loss=0.00108]
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:28:36<3:42:56, 891.76s/it]INFO:ParoQuant:Best val loss: 0.0010514466557651758, Original val loss: 0.0010781434830278158
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:28:40<3:42:56, 891.76s/it]2025-12-22 17:58:54,057 - INFO - Best val loss: 0.0010514466557651758, Original val loss: 0.0010781434830278158
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.000568, val_og_loss=0.000584]
+
  5%|▌         | 1/20 [00:31<10:03, 31.75s/iter, val_loss=0.000568, val_og_loss=0.000584]
+
  5%|▌         | 1/20 [00:34<10:03, 31.75s/iter, val_loss=0.000571, val_og_loss=0.000584]
+
 10%|█         | 2/20 [00:34<04:24, 14.67s/iter, val_loss=0.000571, val_og_loss=0.000584]
+
 10%|█         | 2/20 [00:37<04:24, 14.67s/iter, val_loss=0.000576, val_og_loss=0.000584]
+
 15%|█▌        | 3/20 [00:37<02:36,  9.21s/iter, val_loss=0.000576, val_og_loss=0.000584]
+
 15%|█▌        | 3/20 [00:39<02:36,  9.21s/iter, val_loss=0.000578, val_og_loss=0.000584]
+
 20%|██        | 4/20 [00:39<01:46,  6.65s/iter, val_loss=0.000578, val_og_loss=0.000584]
+
 20%|██        | 4/20 [00:42<01:46,  6.65s/iter, val_loss=0.000581, val_og_loss=0.000584]
+
 25%|██▌       | 5/20 [00:42<01:18,  5.23s/iter, val_loss=0.000581, val_og_loss=0.000584]
+
 25%|██▌       | 5/20 [00:45<01:18,  5.23s/iter, val_loss=0.000581, val_og_loss=0.000584]
+
 30%|███       | 6/20 [00:45<01:01,  4.37s/iter, val_loss=0.000581, val_og_loss=0.000584]
+
 30%|███       | 6/20 [00:48<01:01,  4.37s/iter, val_loss=0.000581, val_og_loss=0.000584]
+
 35%|███▌      | 7/20 [00:48<00:49,  3.83s/iter, val_loss=0.000581, val_og_loss=0.000584]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.83s/iter, val_loss=0.000582, val_og_loss=0.000584]
+
 40%|████      | 8/20 [00:50<00:41,  3.48s/iter, val_loss=0.000582, val_og_loss=0.000584]
+
 40%|████      | 8/20 [00:53<00:41,  3.48s/iter, val_loss=0.00058, val_og_loss=0.000584] 
+
 45%|████▌     | 9/20 [00:53<00:35,  3.25s/iter, val_loss=0.00058, val_og_loss=0.000584]
+
 45%|████▌     | 9/20 [00:56<00:35,  3.25s/iter, val_loss=0.000579, val_og_loss=0.000584]
+
 50%|█████     | 10/20 [00:56<00:30,  3.08s/iter, val_loss=0.000579, val_og_loss=0.000584]
+
 50%|█████     | 10/20 [00:58<00:30,  3.08s/iter, val_loss=0.000577, val_og_loss=0.000584]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.97s/iter, val_loss=0.000577, val_og_loss=0.000584]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.97s/iter, val_loss=0.000577, val_og_loss=0.000584]
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.000577, val_og_loss=0.000584]
+
 60%|██████    | 12/20 [01:04<00:23,  2.89s/iter, val_loss=0.000575, val_og_loss=0.000584]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.000575, val_og_loss=0.000584]
+
 65%|██████▌   | 13/20 [01:07<00:19,  2.84s/iter, val_loss=0.000575, val_og_loss=0.000584]
+
 70%|███████   | 14/20 [01:07<00:16,  2.81s/iter, val_loss=0.000575, val_og_loss=0.000584]
+
 70%|███████   | 14/20 [01:09<00:16,  2.81s/iter, val_loss=0.000574, val_og_loss=0.000584]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.000574, val_og_loss=0.000584]
+
 75%|███████▌  | 15/20 [01:12<00:13,  2.78s/iter, val_loss=0.000574, val_og_loss=0.000584]
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.000574, val_og_loss=0.000584]
+
 80%|████████  | 16/20 [01:15<00:11,  2.76s/iter, val_loss=0.000573, val_og_loss=0.000584]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.75s/iter, val_loss=0.000573, val_og_loss=0.000584]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.75s/iter, val_loss=0.000573, val_og_loss=0.000584]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.000573, val_og_loss=0.000584]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.74s/iter, val_loss=0.000572, val_og_loss=0.000584]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.74s/iter, val_loss=0.000572, val_og_loss=0.000584]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.74s/iter, val_loss=0.000572, val_og_loss=0.000584]
+
100%|██████████| 20/20 [01:23<00:00,  2.73s/iter, val_loss=0.000572, val_og_loss=0.000584]
100%|██████████| 20/20 [01:23<00:00,  4.17s/iter, val_loss=0.000572, val_og_loss=0.000584]
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:30:04<3:42:56, 891.76s/it]INFO:ParoQuant:Best val loss: 0.000568371033295989, Original val loss: 0.000584146473556757
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:30:05<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:30:05<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:31:05<3:42:56, 891.76s/it]2025-12-22 18:00:22,122 - INFO - Best val loss: 0.000568371033295989, Original val loss: 0.000584146473556757
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:32<?, ?iter/s, val_loss=9.89e-5, val_og_loss=0.0001]
+
  5%|▌         | 1/20 [00:32<10:19, 32.60s/iter, val_loss=9.89e-5, val_og_loss=0.0001]
+
  5%|▌         | 1/20 [00:38<10:19, 32.60s/iter, val_loss=9.92e-5, val_og_loss=0.0001]
+
 10%|█         | 2/20 [00:38<04:59, 16.62s/iter, val_loss=9.92e-5, val_og_loss=0.0001]
+
 10%|█         | 2/20 [00:43<04:59, 16.62s/iter, val_loss=9.94e-5, val_og_loss=0.0001]
+
 15%|█▌        | 3/20 [00:43<03:15, 11.51s/iter, val_loss=9.94e-5, val_og_loss=0.0001]
+
 15%|█▌        | 3/20 [00:48<03:15, 11.51s/iter, val_loss=9.94e-5, val_og_loss=0.0001]
+
 20%|██        | 4/20 [00:48<02:25,  9.10s/iter, val_loss=9.94e-5, val_og_loss=0.0001]
+
 20%|██        | 4/20 [00:54<02:25,  9.10s/iter, val_loss=9.96e-5, val_og_loss=0.0001]
+
 25%|██▌       | 5/20 [00:54<01:56,  7.77s/iter, val_loss=9.96e-5, val_og_loss=0.0001]
+
 25%|██▌       | 5/20 [00:59<01:56,  7.77s/iter, val_loss=9.96e-5, val_og_loss=0.0001]
+
 30%|███       | 6/20 [00:59<01:37,  6.97s/iter, val_loss=9.96e-5, val_og_loss=0.0001]
+
 30%|███       | 6/20 [01:05<01:37,  6.97s/iter, val_loss=9.96e-5, val_og_loss=0.0001]
+
 35%|███▌      | 7/20 [01:05<01:24,  6.46s/iter, val_loss=9.96e-5, val_og_loss=0.0001]
+
 35%|███▌      | 7/20 [01:10<01:24,  6.46s/iter, val_loss=9.95e-5, val_og_loss=0.0001]
+
 40%|████      | 8/20 [01:10<01:13,  6.14s/iter, val_loss=9.95e-5, val_og_loss=0.0001]
+
 40%|████      | 8/20 [01:15<01:13,  6.14s/iter, val_loss=9.95e-5, val_og_loss=0.0001]
+
 45%|████▌     | 9/20 [01:15<01:05,  5.91s/iter, val_loss=9.95e-5, val_og_loss=0.0001]
+
 45%|████▌     | 9/20 [01:21<01:05,  5.91s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 50%|█████     | 10/20 [01:21<00:57,  5.76s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 50%|█████     | 10/20 [01:26<00:57,  5.76s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 55%|█████▌    | 11/20 [01:26<00:50,  5.66s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 55%|█████▌    | 11/20 [01:32<00:50,  5.66s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 60%|██████    | 12/20 [01:32<00:44,  5.58s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 60%|██████    | 12/20 [01:37<00:44,  5.58s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 65%|██████▌   | 13/20 [01:37<00:38,  5.55s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 65%|██████▌   | 13/20 [01:43<00:38,  5.55s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 70%|███████   | 14/20 [01:43<00:33,  5.51s/iter, val_loss=9.93e-5, val_og_loss=0.0001]
+
 70%|███████   | 14/20 [01:48<00:33,  5.51s/iter, val_loss=9.92e-5, val_og_loss=0.0001]
+
 75%|███████▌  | 15/20 [01:48<00:27,  5.48s/iter, val_loss=9.92e-5, val_og_loss=0.0001]
+
 75%|███████▌  | 15/20 [01:53<00:27,  5.48s/iter, val_loss=9.92e-5, val_og_loss=0.0001]
+
 80%|████████  | 16/20 [01:53<00:21,  5.46s/iter, val_loss=9.92e-5, val_og_loss=0.0001]
+
 80%|████████  | 16/20 [01:59<00:21,  5.46s/iter, val_loss=9.91e-5, val_og_loss=0.0001]
+
 85%|████████▌ | 17/20 [01:59<00:16,  5.45s/iter, val_loss=9.91e-5, val_og_loss=0.0001]
+
 85%|████████▌ | 17/20 [02:04<00:16,  5.45s/iter, val_loss=9.91e-5, val_og_loss=0.0001]
+
 90%|█████████ | 18/20 [02:04<00:10,  5.44s/iter, val_loss=9.91e-5, val_og_loss=0.0001]
+
 90%|█████████ | 18/20 [02:10<00:10,  5.44s/iter, val_loss=9.9e-5, val_og_loss=0.0001] 
+
 95%|█████████▌| 19/20 [02:10<00:05,  5.44s/iter, val_loss=9.9e-5, val_og_loss=0.0001]
+
 95%|█████████▌| 19/20 [02:15<00:05,  5.44s/iter, val_loss=9.9e-5, val_og_loss=0.0001]
+
100%|██████████| 20/20 [02:15<00:00,  5.43s/iter, val_loss=9.9e-5, val_og_loss=0.0001]
100%|██████████| 20/20 [02:15<00:00,  6.78s/iter, val_loss=9.9e-5, val_og_loss=0.0001]
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:33:21<3:42:56, 891.76s/it]INFO:ParoQuant:Best val loss: 9.885794861475006e-05, Original val loss: 0.00010008425306295976
+
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:33:21<3:42:56, 891.76s/it]
                                                                       

Processing Blocks:  53%|█████▎    | 17/32 [3:33:32<3:42:56, 891.76s/it]
Processing Blocks:  56%|█████▋    | 18/32 [3:34:10<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:34:10<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:34:28<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:34:28<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:36:44<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:36:44<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:37:41<3:27:02, 887.34s/it]2025-12-22 18:03:39,231 - INFO - Best val loss: 9.885794861475006e-05, Original val loss: 0.00010008425306295976
+[INFO] Saving Layer 17 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 18...
+[INFO] Starting Sequential Training for Layer 18
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.00228, val_og_loss=0.00273]
+
  5%|▌         | 1/20 [00:14<04:28, 14.15s/iter, val_loss=0.00228, val_og_loss=0.00273]
+
  5%|▌         | 1/20 [00:15<04:28, 14.15s/iter, val_loss=0.00229, val_og_loss=0.00273]
+
 10%|█         | 2/20 [00:15<02:02,  6.82s/iter, val_loss=0.00229, val_og_loss=0.00273]
+
 10%|█         | 2/20 [00:17<02:02,  6.82s/iter, val_loss=0.0023, val_og_loss=0.00273] 
+
 15%|█▌        | 3/20 [00:17<01:15,  4.47s/iter, val_loss=0.0023, val_og_loss=0.00273]
+
 15%|█▌        | 3/20 [00:19<01:15,  4.47s/iter, val_loss=0.0023, val_og_loss=0.00273]
+
 20%|██        | 4/20 [00:19<00:53,  3.37s/iter, val_loss=0.0023, val_og_loss=0.00273]
+
 20%|██        | 4/20 [00:20<00:53,  3.37s/iter, val_loss=0.00233, val_og_loss=0.00273]
+
 25%|██▌       | 5/20 [00:20<00:41,  2.76s/iter, val_loss=0.00233, val_og_loss=0.00273]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.76s/iter, val_loss=0.00232, val_og_loss=0.00273]
+
 30%|███       | 6/20 [00:22<00:33,  2.39s/iter, val_loss=0.00232, val_og_loss=0.00273]
+
 30%|███       | 6/20 [00:24<00:33,  2.39s/iter, val_loss=0.00231, val_og_loss=0.00273]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.16s/iter, val_loss=0.00231, val_og_loss=0.00273]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.16s/iter, val_loss=0.00231, val_og_loss=0.00273]
+
 40%|████      | 8/20 [00:25<00:24,  2.01s/iter, val_loss=0.00231, val_og_loss=0.00273]
+
 40%|████      | 8/20 [00:27<00:24,  2.01s/iter, val_loss=0.00231, val_og_loss=0.00273]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.90s/iter, val_loss=0.00231, val_og_loss=0.00273]
+
 45%|████▌     | 9/20 [00:29<00:20,  1.90s/iter, val_loss=0.0023, val_og_loss=0.00273] 
+
 50%|█████     | 10/20 [00:29<00:18,  1.84s/iter, val_loss=0.0023, val_og_loss=0.00273]
+
 50%|█████     | 10/20 [00:30<00:18,  1.84s/iter, val_loss=0.0023, val_og_loss=0.00273]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.79s/iter, val_loss=0.0023, val_og_loss=0.00273]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=0.0023, val_og_loss=0.00273]
+
 60%|██████    | 12/20 [00:32<00:14,  1.76s/iter, val_loss=0.0023, val_og_loss=0.00273]
+
 60%|██████    | 12/20 [00:34<00:14,  1.76s/iter, val_loss=0.00229, val_og_loss=0.00273]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=0.00229, val_og_loss=0.00273]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.73s/iter, val_loss=0.00229, val_og_loss=0.00273]
+
 70%|███████   | 14/20 [00:36<00:10,  1.72s/iter, val_loss=0.00229, val_og_loss=0.00273]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=0.00228, val_og_loss=0.00273]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.71s/iter, val_loss=0.00228, val_og_loss=0.00273]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.00228, val_og_loss=0.00273]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=0.00228, val_og_loss=0.00273]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=0.00227, val_og_loss=0.00273]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.69s/iter, val_loss=0.00227, val_og_loss=0.00273]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.69s/iter, val_loss=0.00227, val_og_loss=0.00273]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=0.00227, val_og_loss=0.00273]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.00227, val_og_loss=0.00273]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=0.00227, val_og_loss=0.00273]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=0.00228, val_og_loss=0.00273]
+
100%|██████████| 20/20 [00:46<00:00,  1.68s/iter, val_loss=0.00228, val_og_loss=0.00273]
100%|██████████| 20/20 [00:46<00:00,  2.30s/iter, val_loss=0.00228, val_og_loss=0.00273]
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:38:27<3:27:02, 887.34s/it]INFO:ParoQuant:Best val loss: 0.0022694002836942673, Original val loss: 0.0027254403103142977
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:38:29<3:27:02, 887.34s/it]2025-12-22 18:08:45,378 - INFO - Best val loss: 0.0022694002836942673, Original val loss: 0.0027254403103142977
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=0.0064, val_og_loss=0.0074]
+
  5%|▌         | 1/20 [00:11<03:47, 11.97s/iter, val_loss=0.0064, val_og_loss=0.0074]
+
  5%|▌         | 1/20 [00:13<03:47, 11.97s/iter, val_loss=0.00647, val_og_loss=0.0074]
+
 10%|█         | 2/20 [00:13<01:43,  5.75s/iter, val_loss=0.00647, val_og_loss=0.0074]
+
 10%|█         | 2/20 [00:14<01:43,  5.75s/iter, val_loss=0.00647, val_og_loss=0.0074]
+
 15%|█▌        | 3/20 [00:14<01:04,  3.77s/iter, val_loss=0.00647, val_og_loss=0.0074]
+
 15%|█▌        | 3/20 [00:16<01:04,  3.77s/iter, val_loss=0.00651, val_og_loss=0.0074]
+
 20%|██        | 4/20 [00:16<00:45,  2.83s/iter, val_loss=0.00651, val_og_loss=0.0074]
+
 20%|██        | 4/20 [00:17<00:45,  2.83s/iter, val_loss=0.00648, val_og_loss=0.0074]
+
 25%|██▌       | 5/20 [00:17<00:34,  2.31s/iter, val_loss=0.00648, val_og_loss=0.0074]
+
 25%|██▌       | 5/20 [00:18<00:34,  2.31s/iter, val_loss=0.00645, val_og_loss=0.0074]
+
 30%|███       | 6/20 [00:18<00:27,  2.00s/iter, val_loss=0.00645, val_og_loss=0.0074]
+
 30%|███       | 6/20 [00:20<00:27,  2.00s/iter, val_loss=0.00648, val_og_loss=0.0074]
+
 35%|███▌      | 7/20 [00:20<00:23,  1.80s/iter, val_loss=0.00648, val_og_loss=0.0074]
+
 35%|███▌      | 7/20 [00:21<00:23,  1.80s/iter, val_loss=0.00645, val_og_loss=0.0074]
+
 40%|████      | 8/20 [00:21<00:20,  1.68s/iter, val_loss=0.00645, val_og_loss=0.0074]
+
 40%|████      | 8/20 [00:23<00:20,  1.68s/iter, val_loss=0.00645, val_og_loss=0.0074]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.59s/iter, val_loss=0.00645, val_og_loss=0.0074]
+
 45%|████▌     | 9/20 [00:24<00:17,  1.59s/iter, val_loss=0.00639, val_og_loss=0.0074]
+
 50%|█████     | 10/20 [00:24<00:15,  1.52s/iter, val_loss=0.00639, val_og_loss=0.0074]
+
 50%|█████     | 10/20 [00:25<00:15,  1.52s/iter, val_loss=0.00641, val_og_loss=0.0074]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.48s/iter, val_loss=0.00641, val_og_loss=0.0074]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.48s/iter, val_loss=0.00638, val_og_loss=0.0074]
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.00638, val_og_loss=0.0074]
+
 60%|██████    | 12/20 [00:28<00:11,  1.45s/iter, val_loss=0.00639, val_og_loss=0.0074]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.44s/iter, val_loss=0.00639, val_og_loss=0.0074]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.44s/iter, val_loss=0.00636, val_og_loss=0.0074]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.00636, val_og_loss=0.0074]
+
 70%|███████   | 14/20 [00:31<00:08,  1.43s/iter, val_loss=0.00637, val_og_loss=0.0074]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.42s/iter, val_loss=0.00637, val_og_loss=0.0074]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.42s/iter, val_loss=0.00633, val_og_loss=0.0074]
+
 80%|████████  | 16/20 [00:32<00:05,  1.41s/iter, val_loss=0.00633, val_og_loss=0.0074]
+
 80%|████████  | 16/20 [00:34<00:05,  1.41s/iter, val_loss=0.00634, val_og_loss=0.0074]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00634, val_og_loss=0.0074]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.40s/iter, val_loss=0.00632, val_og_loss=0.0074]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00632, val_og_loss=0.0074]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.40s/iter, val_loss=0.0063, val_og_loss=0.0074] 
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.0063, val_og_loss=0.0074]
+
 95%|█████████▌| 19/20 [00:38<00:01,  1.40s/iter, val_loss=0.00631, val_og_loss=0.0074]
+
100%|██████████| 20/20 [00:38<00:00,  1.40s/iter, val_loss=0.00631, val_og_loss=0.0074]
100%|██████████| 20/20 [00:38<00:00,  1.92s/iter, val_loss=0.00631, val_og_loss=0.0074]
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:39:07<3:27:02, 887.34s/it]INFO:ParoQuant:Best val loss: 0.0063041094690561295, Original val loss: 0.0073998915031552315
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:39:10<3:27:02, 887.34s/it]2025-12-22 18:09:25,868 - INFO - Best val loss: 0.0063041094690561295, Original val loss: 0.0073998915031552315
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:12<?, ?iter/s, val_loss=0.000491, val_og_loss=0.000585]
+
  5%|▌         | 1/20 [00:12<03:48, 12.03s/iter, val_loss=0.000491, val_og_loss=0.000585]
+
  5%|▌         | 1/20 [00:13<03:48, 12.03s/iter, val_loss=0.000492, val_og_loss=0.000585]
+
 10%|█         | 2/20 [00:13<01:43,  5.77s/iter, val_loss=0.000492, val_og_loss=0.000585]
+
 10%|█         | 2/20 [00:14<01:43,  5.77s/iter, val_loss=0.000491, val_og_loss=0.000585]
+
 15%|█▌        | 3/20 [00:14<01:04,  3.78s/iter, val_loss=0.000491, val_og_loss=0.000585]
+
 15%|█▌        | 3/20 [00:16<01:04,  3.78s/iter, val_loss=0.000492, val_og_loss=0.000585]
+
 20%|██        | 4/20 [00:16<00:45,  2.84s/iter, val_loss=0.000492, val_og_loss=0.000585]
+
 20%|██        | 4/20 [00:17<00:45,  2.84s/iter, val_loss=0.000493, val_og_loss=0.000585]
+
 25%|██▌       | 5/20 [00:17<00:34,  2.32s/iter, val_loss=0.000493, val_og_loss=0.000585]
+
 25%|██▌       | 5/20 [00:19<00:34,  2.32s/iter, val_loss=0.000493, val_og_loss=0.000585]
+
 30%|███       | 6/20 [00:19<00:28,  2.00s/iter, val_loss=0.000493, val_og_loss=0.000585]
+
 30%|███       | 6/20 [00:20<00:28,  2.00s/iter, val_loss=0.000494, val_og_loss=0.000585]
+
 35%|███▌      | 7/20 [00:20<00:23,  1.81s/iter, val_loss=0.000494, val_og_loss=0.000585]
+
 35%|███▌      | 7/20 [00:21<00:23,  1.81s/iter, val_loss=0.000493, val_og_loss=0.000585]
+
 40%|████      | 8/20 [00:21<00:20,  1.68s/iter, val_loss=0.000493, val_og_loss=0.000585]
+
 40%|████      | 8/20 [00:23<00:20,  1.68s/iter, val_loss=0.000491, val_og_loss=0.000585]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.59s/iter, val_loss=0.000491, val_og_loss=0.000585]
+
 45%|████▌     | 9/20 [00:24<00:17,  1.59s/iter, val_loss=0.000493, val_og_loss=0.000585]
+
 50%|█████     | 10/20 [00:24<00:15,  1.53s/iter, val_loss=0.000493, val_og_loss=0.000585]
+
 50%|█████     | 10/20 [00:26<00:15,  1.53s/iter, val_loss=0.000492, val_og_loss=0.000585]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.49s/iter, val_loss=0.000492, val_og_loss=0.000585]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.49s/iter, val_loss=0.00049, val_og_loss=0.000585] 
+
 60%|██████    | 12/20 [00:27<00:11,  1.46s/iter, val_loss=0.00049, val_og_loss=0.000585]
+
 60%|██████    | 12/20 [00:28<00:11,  1.46s/iter, val_loss=0.000488, val_og_loss=0.000585]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.45s/iter, val_loss=0.000488, val_og_loss=0.000585]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.45s/iter, val_loss=0.000488, val_og_loss=0.000585]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.000488, val_og_loss=0.000585]
+
 70%|███████   | 14/20 [00:31<00:08,  1.43s/iter, val_loss=0.000488, val_og_loss=0.000585]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.42s/iter, val_loss=0.000488, val_og_loss=0.000585]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.42s/iter, val_loss=0.000486, val_og_loss=0.000585]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.000486, val_og_loss=0.000585]
+
 80%|████████  | 16/20 [00:34<00:05,  1.41s/iter, val_loss=0.000486, val_og_loss=0.000585]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.41s/iter, val_loss=0.000486, val_og_loss=0.000585]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.41s/iter, val_loss=0.000485, val_og_loss=0.000585]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.41s/iter, val_loss=0.000485, val_og_loss=0.000585]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.41s/iter, val_loss=0.000485, val_og_loss=0.000585]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.000485, val_og_loss=0.000585]
+
 95%|█████████▌| 19/20 [00:38<00:01,  1.40s/iter, val_loss=0.000484, val_og_loss=0.000585]
+
100%|██████████| 20/20 [00:38<00:00,  1.40s/iter, val_loss=0.000484, val_og_loss=0.000585]
100%|██████████| 20/20 [00:38<00:00,  1.93s/iter, val_loss=0.000484, val_og_loss=0.000585]
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:39:49<3:27:02, 887.34s/it]INFO:ParoQuant:Best val loss: 0.0004844138165935874, Original val loss: 0.0005853364709764719
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:39:51<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:39:51<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:40:22<3:27:02, 887.34s/it]2025-12-22 18:10:07,011 - INFO - Best val loss: 0.0004844138165935874, Original val loss: 0.0005853364709764719
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
  5%|▌         | 1/20 [00:15<05:03, 15.99s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
  5%|▌         | 1/20 [00:17<05:03, 15.99s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 10%|█         | 2/20 [00:17<02:16,  7.58s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 10%|█         | 2/20 [00:19<02:16,  7.58s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 15%|█▌        | 3/20 [00:19<01:23,  4.88s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 15%|█▌        | 3/20 [00:21<01:23,  4.88s/iter, val_loss=1.93e-5, val_og_loss=1.97e-5]
+
 20%|██        | 4/20 [00:21<00:58,  3.63s/iter, val_loss=1.93e-5, val_og_loss=1.97e-5]
+
 20%|██        | 4/20 [00:22<00:58,  3.63s/iter, val_loss=1.94e-5, val_og_loss=1.97e-5]
+
 25%|██▌       | 5/20 [00:22<00:43,  2.92s/iter, val_loss=1.94e-5, val_og_loss=1.97e-5]
+
 25%|██▌       | 5/20 [00:24<00:43,  2.92s/iter, val_loss=1.94e-5, val_og_loss=1.97e-5]
+
 30%|███       | 6/20 [00:24<00:34,  2.50s/iter, val_loss=1.94e-5, val_og_loss=1.97e-5]
+
 30%|███       | 6/20 [00:26<00:34,  2.50s/iter, val_loss=1.93e-5, val_og_loss=1.97e-5]
+
 35%|███▌      | 7/20 [00:26<00:29,  2.23s/iter, val_loss=1.93e-5, val_og_loss=1.97e-5]
+
 35%|███▌      | 7/20 [00:27<00:29,  2.23s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 40%|████      | 8/20 [00:27<00:24,  2.06s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 40%|████      | 8/20 [00:29<00:24,  2.06s/iter, val_loss=1.93e-5, val_og_loss=1.97e-5]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.94s/iter, val_loss=1.93e-5, val_og_loss=1.97e-5]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.94s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 50%|█████     | 10/20 [00:31<00:18,  1.86s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 50%|█████     | 10/20 [00:32<00:18,  1.86s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.81s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.81s/iter, val_loss=1.93e-5, val_og_loss=1.97e-5]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=1.93e-5, val_og_loss=1.97e-5]
+
 60%|██████    | 12/20 [00:36<00:14,  1.77s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.74s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.74s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 70%|███████   | 14/20 [00:39<00:10,  1.72s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.71s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.69s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.69s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.69s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.68s/iter, val_loss=1.92e-5, val_og_loss=1.97e-5]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.68s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
100%|██████████| 20/20 [00:47<00:00,  1.69s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
100%|██████████| 20/20 [00:47<00:00,  2.40s/iter, val_loss=1.91e-5, val_og_loss=1.97e-5]
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:41:10<3:27:02, 887.34s/it]INFO:ParoQuant:Best val loss: 1.9104074453935027e-05, Original val loss: 1.973151847778354e-05
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:41:10<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:41:10<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:41:55<3:27:02, 887.34s/it]2025-12-22 18:11:28,566 - INFO - Best val loss: 1.9104074453935027e-05, Original val loss: 1.973151847778354e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.00104, val_og_loss=0.00105]
+
  5%|▌         | 1/20 [00:30<09:43, 30.73s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
  5%|▌         | 1/20 [00:33<09:43, 30.73s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 10%|█         | 2/20 [00:33<04:16, 14.25s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 10%|█         | 2/20 [00:36<04:16, 14.25s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 15%|█▌        | 3/20 [00:36<02:32,  8.98s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 15%|█▌        | 3/20 [00:38<02:32,  8.98s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 20%|██        | 4/20 [00:38<01:44,  6.50s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 20%|██        | 4/20 [00:41<01:44,  6.50s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 25%|██▌       | 5/20 [00:41<01:17,  5.14s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.14s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 30%|███       | 6/20 [00:44<01:00,  4.32s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 30%|███       | 6/20 [00:47<01:00,  4.32s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.80s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.80s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 40%|████      | 8/20 [00:49<00:41,  3.45s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 40%|████      | 8/20 [00:52<00:41,  3.45s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.22s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.22s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 50%|█████     | 10/20 [00:55<00:30,  3.07s/iter, val_loss=0.00105, val_og_loss=0.00105]
+
 50%|█████     | 10/20 [00:57<00:30,  3.07s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.83s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.83s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.77s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.77s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 80%|████████  | 16/20 [01:11<00:11,  2.75s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 80%|████████  | 16/20 [01:14<00:11,  2.75s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.74s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.73s/iter, val_loss=0.00104, val_og_loss=0.00105]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.73s/iter, val_loss=0.00103, val_og_loss=0.00105]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.00103, val_og_loss=0.00105]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.00103, val_og_loss=0.00105]
+
100%|██████████| 20/20 [01:22<00:00,  2.72s/iter, val_loss=0.00103, val_og_loss=0.00105]
100%|██████████| 20/20 [01:22<00:00,  4.12s/iter, val_loss=0.00103, val_og_loss=0.00105]
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:43:18<3:27:02, 887.34s/it]INFO:ParoQuant:Best val loss: 0.0010345135815441608, Original val loss: 0.001054600696079433
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:43:22<3:27:02, 887.34s/it]2025-12-22 18:13:35,976 - INFO - Best val loss: 0.0010345135815441608, Original val loss: 0.001054600696079433
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:32<?, ?iter/s, val_loss=0.000559, val_og_loss=0.000571]
+
  5%|▌         | 1/20 [00:32<10:18, 32.56s/iter, val_loss=0.000559, val_og_loss=0.000571]
+
  5%|▌         | 1/20 [00:35<10:18, 32.56s/iter, val_loss=0.000562, val_og_loss=0.000571]
+
 10%|█         | 2/20 [00:35<04:30, 15.00s/iter, val_loss=0.000562, val_og_loss=0.000571]
+
 10%|█         | 2/20 [00:37<04:30, 15.00s/iter, val_loss=0.000566, val_og_loss=0.000571]
+
 15%|█▌        | 3/20 [00:37<02:39,  9.39s/iter, val_loss=0.000566, val_og_loss=0.000571]
+
 15%|█▌        | 3/20 [00:40<02:39,  9.39s/iter, val_loss=0.000568, val_og_loss=0.000571]
+
 20%|██        | 4/20 [00:40<01:48,  6.76s/iter, val_loss=0.000568, val_og_loss=0.000571]
+
 20%|██        | 4/20 [00:43<01:48,  6.76s/iter, val_loss=0.000568, val_og_loss=0.000571]
+
 25%|██▌       | 5/20 [00:43<01:19,  5.30s/iter, val_loss=0.000568, val_og_loss=0.000571]
+
 25%|██▌       | 5/20 [00:46<01:19,  5.30s/iter, val_loss=0.000569, val_og_loss=0.000571]
+
 30%|███       | 6/20 [00:46<01:01,  4.42s/iter, val_loss=0.000569, val_og_loss=0.000571]
+
 30%|███       | 6/20 [00:48<01:01,  4.42s/iter, val_loss=0.000568, val_og_loss=0.000571]
+
 35%|███▌      | 7/20 [00:48<00:50,  3.87s/iter, val_loss=0.000568, val_og_loss=0.000571]
+
 35%|███▌      | 7/20 [00:51<00:50,  3.87s/iter, val_loss=0.000569, val_og_loss=0.000571]
+
 40%|████      | 8/20 [00:51<00:41,  3.50s/iter, val_loss=0.000569, val_og_loss=0.000571]
+
 40%|████      | 8/20 [00:54<00:41,  3.50s/iter, val_loss=0.000567, val_og_loss=0.000571]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.26s/iter, val_loss=0.000567, val_og_loss=0.000571]
+
 45%|████▌     | 9/20 [00:57<00:35,  3.26s/iter, val_loss=0.000567, val_og_loss=0.000571]
+
 50%|█████     | 10/20 [00:57<00:30,  3.09s/iter, val_loss=0.000567, val_og_loss=0.000571]
+
 50%|█████     | 10/20 [00:59<00:30,  3.09s/iter, val_loss=0.000566, val_og_loss=0.000571]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.98s/iter, val_loss=0.000566, val_og_loss=0.000571]
+
 55%|█████▌    | 11/20 [01:02<00:26,  2.98s/iter, val_loss=0.000565, val_og_loss=0.000571]
+
 60%|██████    | 12/20 [01:02<00:23,  2.89s/iter, val_loss=0.000565, val_og_loss=0.000571]
+
 60%|██████    | 12/20 [01:05<00:23,  2.89s/iter, val_loss=0.000565, val_og_loss=0.000571]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.84s/iter, val_loss=0.000565, val_og_loss=0.000571]
+
 65%|██████▌   | 13/20 [01:07<00:19,  2.84s/iter, val_loss=0.000564, val_og_loss=0.000571]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.000564, val_og_loss=0.000571]
+
 70%|███████   | 14/20 [01:10<00:16,  2.80s/iter, val_loss=0.000564, val_og_loss=0.000571]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.000564, val_og_loss=0.000571]
+
 75%|███████▌  | 15/20 [01:13<00:13,  2.77s/iter, val_loss=0.000563, val_og_loss=0.000571]
+
 80%|████████  | 16/20 [01:13<00:11,  2.75s/iter, val_loss=0.000563, val_og_loss=0.000571]
+
 80%|████████  | 16/20 [01:15<00:11,  2.75s/iter, val_loss=0.000563, val_og_loss=0.000571]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.000563, val_og_loss=0.000571]
+
 85%|████████▌ | 17/20 [01:18<00:08,  2.74s/iter, val_loss=0.000562, val_og_loss=0.000571]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.000562, val_og_loss=0.000571]
+
 90%|█████████ | 18/20 [01:21<00:05,  2.73s/iter, val_loss=0.000562, val_og_loss=0.000571]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.72s/iter, val_loss=0.000562, val_og_loss=0.000571]
+
 95%|█████████▌| 19/20 [01:24<00:02,  2.72s/iter, val_loss=0.000561, val_og_loss=0.000571]
+
100%|██████████| 20/20 [01:24<00:00,  2.72s/iter, val_loss=0.000561, val_og_loss=0.000571]
100%|██████████| 20/20 [01:24<00:00,  4.21s/iter, val_loss=0.000561, val_og_loss=0.000571]
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:44:46<3:27:02, 887.34s/it]INFO:ParoQuant:Best val loss: 0.0005593994283117354, Original val loss: 0.000571220472920686
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:44:48<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:44:48<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:45:46<3:27:02, 887.34s/it]2025-12-22 18:15:04,769 - INFO - Best val loss: 0.0005593994283117354, Original val loss: 0.000571220472920686
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=9.42e-5, val_og_loss=9.56e-5]
+
  5%|▌         | 1/20 [00:33<10:37, 33.53s/iter, val_loss=9.42e-5, val_og_loss=9.56e-5]
+
  5%|▌         | 1/20 [00:38<10:37, 33.53s/iter, val_loss=9.46e-5, val_og_loss=9.56e-5]
+
 10%|█         | 2/20 [00:38<05:05, 17.00s/iter, val_loss=9.46e-5, val_og_loss=9.56e-5]
+
 10%|█         | 2/20 [00:44<05:05, 17.00s/iter, val_loss=9.5e-5, val_og_loss=9.56e-5] 
+
 15%|█▌        | 3/20 [00:44<03:19, 11.71s/iter, val_loss=9.5e-5, val_og_loss=9.56e-5]
+
 15%|█▌        | 3/20 [00:49<03:19, 11.71s/iter, val_loss=9.5e-5, val_og_loss=9.56e-5]
+
 20%|██        | 4/20 [00:49<02:27,  9.22s/iter, val_loss=9.5e-5, val_og_loss=9.56e-5]
+
 20%|██        | 4/20 [00:55<02:27,  9.22s/iter, val_loss=9.48e-5, val_og_loss=9.56e-5]
+
 25%|██▌       | 5/20 [00:55<01:57,  7.85s/iter, val_loss=9.48e-5, val_og_loss=9.56e-5]
+
 25%|██▌       | 5/20 [01:00<01:57,  7.85s/iter, val_loss=9.48e-5, val_og_loss=9.56e-5]
+
 30%|███       | 6/20 [01:00<01:38,  7.02s/iter, val_loss=9.48e-5, val_og_loss=9.56e-5]
+
 30%|███       | 6/20 [01:06<01:38,  7.02s/iter, val_loss=9.48e-5, val_og_loss=9.56e-5]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.50s/iter, val_loss=9.48e-5, val_og_loss=9.56e-5]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.50s/iter, val_loss=9.49e-5, val_og_loss=9.56e-5]
+
 40%|████      | 8/20 [01:11<01:13,  6.16s/iter, val_loss=9.49e-5, val_og_loss=9.56e-5]
+
 40%|████      | 8/20 [01:16<01:13,  6.16s/iter, val_loss=9.48e-5, val_og_loss=9.56e-5]
+
 45%|████▌     | 9/20 [01:16<01:05,  5.93s/iter, val_loss=9.48e-5, val_og_loss=9.56e-5]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.93s/iter, val_loss=9.47e-5, val_og_loss=9.56e-5]
+
 50%|█████     | 10/20 [01:22<00:57,  5.77s/iter, val_loss=9.47e-5, val_og_loss=9.56e-5]
+
 50%|█████     | 10/20 [01:27<00:57,  5.77s/iter, val_loss=9.47e-5, val_og_loss=9.56e-5]
+
 55%|█████▌    | 11/20 [01:27<00:50,  5.66s/iter, val_loss=9.47e-5, val_og_loss=9.56e-5]
+
 55%|█████▌    | 11/20 [01:33<00:50,  5.66s/iter, val_loss=9.46e-5, val_og_loss=9.56e-5]
+
 60%|██████    | 12/20 [01:33<00:44,  5.59s/iter, val_loss=9.46e-5, val_og_loss=9.56e-5]
+
 60%|██████    | 12/20 [01:38<00:44,  5.59s/iter, val_loss=9.46e-5, val_og_loss=9.56e-5]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.54s/iter, val_loss=9.46e-5, val_og_loss=9.56e-5]
+
 65%|██████▌   | 13/20 [01:43<00:38,  5.54s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 70%|███████   | 14/20 [01:43<00:33,  5.50s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 70%|███████   | 14/20 [01:49<00:33,  5.50s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.48s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 75%|███████▌  | 15/20 [01:54<00:27,  5.48s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 80%|████████  | 16/20 [01:54<00:21,  5.46s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 80%|████████  | 16/20 [02:00<00:21,  5.46s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.45s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.45s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.44s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=9.45e-5, val_og_loss=9.56e-5]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.44s/iter, val_loss=9.44e-5, val_og_loss=9.56e-5]
+
100%|██████████| 20/20 [02:16<00:00,  5.43s/iter, val_loss=9.44e-5, val_og_loss=9.56e-5]
100%|██████████| 20/20 [02:16<00:00,  6.83s/iter, val_loss=9.44e-5, val_og_loss=9.56e-5]
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:48:03<3:27:02, 887.34s/it]INFO:ParoQuant:Best val loss: 9.417571709491313e-05, Original val loss: 9.561668412061408e-05
+
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:48:04<3:27:02, 887.34s/it]
                                                                       

Processing Blocks:  56%|█████▋    | 18/32 [3:48:13<3:27:02, 887.34s/it]
Processing Blocks:  59%|█████▉    | 19/32 [3:48:52<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:48:52<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:49:12<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:49:12<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:51:29<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:51:29<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:52:25<3:11:53, 885.68s/it]2025-12-22 18:18:21,360 - INFO - Best val loss: 9.417571709491313e-05, Original val loss: 9.561668412061408e-05
+[INFO] Saving Layer 18 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 19...
+[INFO] Starting Sequential Training for Layer 19
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=0.00235, val_og_loss=0.00276]
+
  5%|▌         | 1/20 [00:13<04:19, 13.64s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
  5%|▌         | 1/20 [00:15<04:19, 13.64s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 10%|█         | 2/20 [00:15<01:58,  6.60s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 10%|█         | 2/20 [00:17<01:58,  6.60s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 15%|█▌        | 3/20 [00:17<01:14,  4.36s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 15%|█▌        | 3/20 [00:18<01:14,  4.36s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 20%|██        | 4/20 [00:18<00:52,  3.30s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 20%|██        | 4/20 [00:20<00:52,  3.30s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 25%|██▌       | 5/20 [00:20<00:40,  2.72s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 25%|██▌       | 5/20 [00:22<00:40,  2.72s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 30%|███       | 6/20 [00:22<00:33,  2.37s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 30%|███       | 6/20 [00:23<00:33,  2.37s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.14s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 35%|███▌      | 7/20 [00:25<00:27,  2.14s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 40%|████      | 8/20 [00:25<00:23,  1.99s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 40%|████      | 8/20 [00:27<00:23,  1.99s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.90s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.90s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 50%|█████     | 10/20 [00:28<00:18,  1.83s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 50%|█████     | 10/20 [00:30<00:18,  1.83s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.78s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.78s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
 60%|██████    | 12/20 [00:32<00:14,  1.75s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
 60%|██████    | 12/20 [00:33<00:14,  1.75s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.73s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=0.00236, val_og_loss=0.00276]
+
 70%|███████   | 14/20 [00:35<00:10,  1.72s/iter, val_loss=0.00236, val_og_loss=0.00276]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=0.00236, val_og_loss=0.00276]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.71s/iter, val_loss=0.00236, val_og_loss=0.00276]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
 80%|████████  | 16/20 [00:38<00:06,  1.70s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.69s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.70s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.70s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.69s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
100%|██████████| 20/20 [00:45<00:00,  1.69s/iter, val_loss=0.00235, val_og_loss=0.00276]
100%|██████████| 20/20 [00:45<00:00,  2.28s/iter, val_loss=0.00235, val_og_loss=0.00276]
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:53:11<3:11:53, 885.68s/it]INFO:ParoQuant:Best val loss: 0.002345314482226968, Original val loss: 0.002760852687060833
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:53:12<3:11:53, 885.68s/it]2025-12-22 18:23:29,015 - INFO - Best val loss: 0.002345314482226968, Original val loss: 0.002760852687060833
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.00588, val_og_loss=0.00671]
+
  5%|▌         | 1/20 [00:10<03:20, 10.54s/iter, val_loss=0.00588, val_og_loss=0.00671]
+
  5%|▌         | 1/20 [00:11<03:20, 10.54s/iter, val_loss=0.00592, val_og_loss=0.00671]
+
 10%|█         | 2/20 [00:11<01:32,  5.16s/iter, val_loss=0.00592, val_og_loss=0.00671]
+
 10%|█         | 2/20 [00:13<01:32,  5.16s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 15%|█▌        | 3/20 [00:13<00:58,  3.43s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 15%|█▌        | 3/20 [00:14<00:58,  3.43s/iter, val_loss=0.00593, val_og_loss=0.00671]
+
 20%|██        | 4/20 [00:14<00:42,  2.64s/iter, val_loss=0.00593, val_og_loss=0.00671]
+
 20%|██        | 4/20 [00:16<00:42,  2.64s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 25%|██▌       | 5/20 [00:16<00:32,  2.19s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 25%|██▌       | 5/20 [00:17<00:32,  2.19s/iter, val_loss=0.00591, val_og_loss=0.00671]
+
 30%|███       | 6/20 [00:17<00:26,  1.92s/iter, val_loss=0.00591, val_og_loss=0.00671]
+
 30%|███       | 6/20 [00:18<00:26,  1.92s/iter, val_loss=0.00595, val_og_loss=0.00671]
+
 35%|███▌      | 7/20 [00:18<00:22,  1.75s/iter, val_loss=0.00595, val_og_loss=0.00671]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.75s/iter, val_loss=0.00593, val_og_loss=0.00671]
+
 40%|████      | 8/20 [00:20<00:19,  1.63s/iter, val_loss=0.00593, val_og_loss=0.00671]
+
 40%|████      | 8/20 [00:21<00:19,  1.63s/iter, val_loss=0.00595, val_og_loss=0.00671]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.56s/iter, val_loss=0.00595, val_og_loss=0.00671]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.56s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 50%|█████     | 10/20 [00:23<00:15,  1.51s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 50%|█████     | 10/20 [00:24<00:15,  1.51s/iter, val_loss=0.0059, val_og_loss=0.00671] 
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.48s/iter, val_loss=0.0059, val_og_loss=0.00671]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.48s/iter, val_loss=0.0059, val_og_loss=0.00671]
+
 60%|██████    | 12/20 [00:25<00:11,  1.45s/iter, val_loss=0.0059, val_og_loss=0.00671]
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.00589, val_og_loss=0.00671]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.43s/iter, val_loss=0.00589, val_og_loss=0.00671]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.43s/iter, val_loss=0.00589, val_og_loss=0.00671]
+
 70%|███████   | 14/20 [00:28<00:08,  1.42s/iter, val_loss=0.00589, val_og_loss=0.00671]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.00587, val_og_loss=0.00671]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.00587, val_og_loss=0.00671]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.41s/iter, val_loss=0.00586, val_og_loss=0.00671]
+
 80%|████████  | 16/20 [00:31<00:05,  1.41s/iter, val_loss=0.00586, val_og_loss=0.00671]
+
 80%|████████  | 16/20 [00:32<00:05,  1.41s/iter, val_loss=0.00588, val_og_loss=0.00671]
+
 85%|████████▌ | 17/20 [00:32<00:04,  1.40s/iter, val_loss=0.00588, val_og_loss=0.00671]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00589, val_og_loss=0.00671]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.00589, val_og_loss=0.00671]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00588, val_og_loss=0.00671]
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.40s/iter, val_loss=0.00588, val_og_loss=0.00671]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00588, val_og_loss=0.00671]
+
100%|██████████| 20/20 [00:37<00:00,  1.39s/iter, val_loss=0.00588, val_og_loss=0.00671]
100%|██████████| 20/20 [00:37<00:00,  1.85s/iter, val_loss=0.00588, val_og_loss=0.00671]
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:53:49<3:11:53, 885.68s/it]INFO:ParoQuant:Best val loss: 0.005856872536242008, Original val loss: 0.006713856011629105
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:53:51<3:11:53, 885.68s/it]2025-12-22 18:24:07,899 - INFO - Best val loss: 0.005856872536242008, Original val loss: 0.006713856011629105
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.000565, val_og_loss=0.000663]
+
  5%|▌         | 1/20 [00:10<03:25, 10.80s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
  5%|▌         | 1/20 [00:12<03:25, 10.80s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 10%|█         | 2/20 [00:12<01:34,  5.26s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 10%|█         | 2/20 [00:13<01:34,  5.26s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 15%|█▌        | 3/20 [00:13<00:59,  3.49s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 15%|█▌        | 3/20 [00:14<00:59,  3.49s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 20%|██        | 4/20 [00:14<00:42,  2.66s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 20%|██        | 4/20 [00:16<00:42,  2.66s/iter, val_loss=0.000567, val_og_loss=0.000663]
+
 25%|██▌       | 5/20 [00:16<00:32,  2.20s/iter, val_loss=0.000567, val_og_loss=0.000663]
+
 25%|██▌       | 5/20 [00:17<00:32,  2.20s/iter, val_loss=0.000567, val_og_loss=0.000663]
+
 30%|███       | 6/20 [00:17<00:26,  1.93s/iter, val_loss=0.000567, val_og_loss=0.000663]
+
 30%|███       | 6/20 [00:19<00:26,  1.93s/iter, val_loss=0.000566, val_og_loss=0.000663]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.75s/iter, val_loss=0.000566, val_og_loss=0.000663]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.75s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 40%|████      | 8/20 [00:20<00:19,  1.63s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 40%|████      | 8/20 [00:21<00:19,  1.63s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.55s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.55s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 50%|█████     | 10/20 [00:23<00:15,  1.51s/iter, val_loss=0.000565, val_og_loss=0.000663]
+
 50%|█████     | 10/20 [00:24<00:15,  1.51s/iter, val_loss=0.000562, val_og_loss=0.000663]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.47s/iter, val_loss=0.000562, val_og_loss=0.000663]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.47s/iter, val_loss=0.000563, val_og_loss=0.000663]
+
 60%|██████    | 12/20 [00:26<00:11,  1.45s/iter, val_loss=0.000563, val_og_loss=0.000663]
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.000563, val_og_loss=0.000663]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.43s/iter, val_loss=0.000563, val_og_loss=0.000663]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.43s/iter, val_loss=0.000561, val_og_loss=0.000663]
+
 70%|███████   | 14/20 [00:28<00:08,  1.42s/iter, val_loss=0.000561, val_og_loss=0.000663]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.00056, val_og_loss=0.000663] 
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.00056, val_og_loss=0.000663]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.41s/iter, val_loss=0.00056, val_og_loss=0.000663]
+
 80%|████████  | 16/20 [00:31<00:05,  1.41s/iter, val_loss=0.00056, val_og_loss=0.000663]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.000557, val_og_loss=0.000663]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.40s/iter, val_loss=0.000557, val_og_loss=0.000663]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.000557, val_og_loss=0.000663]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.000557, val_og_loss=0.000663]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.000557, val_og_loss=0.000663]
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.39s/iter, val_loss=0.000557, val_og_loss=0.000663]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.39s/iter, val_loss=0.000557, val_og_loss=0.000663]
+
100%|██████████| 20/20 [00:37<00:00,  1.39s/iter, val_loss=0.000557, val_og_loss=0.000663]
100%|██████████| 20/20 [00:37<00:00,  1.86s/iter, val_loss=0.000557, val_og_loss=0.000663]
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:54:28<3:11:53, 885.68s/it]INFO:ParoQuant:Best val loss: 0.0005566952750086784, Original val loss: 0.0006630488205701113
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:54:31<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:54:31<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:55:01<3:11:53, 885.68s/it]2025-12-22 18:24:46,887 - INFO - Best val loss: 0.0005566952750086784, Original val loss: 0.0006630488205701113
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=1.73e-5, val_og_loss=1.77e-5]
+
  5%|▌         | 1/20 [00:13<04:25, 13.97s/iter, val_loss=1.73e-5, val_og_loss=1.77e-5]
+
  5%|▌         | 1/20 [00:15<04:25, 13.97s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 10%|█         | 2/20 [00:15<02:01,  6.74s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 10%|█         | 2/20 [00:17<02:01,  6.74s/iter, val_loss=1.77e-5, val_og_loss=1.77e-5]
+
 15%|█▌        | 3/20 [00:17<01:15,  4.43s/iter, val_loss=1.77e-5, val_og_loss=1.77e-5]
+
 15%|█▌        | 3/20 [00:19<01:15,  4.43s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 20%|██        | 4/20 [00:19<00:53,  3.35s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 20%|██        | 4/20 [00:20<00:53,  3.35s/iter, val_loss=1.74e-5, val_og_loss=1.77e-5]
+
 25%|██▌       | 5/20 [00:20<00:41,  2.75s/iter, val_loss=1.74e-5, val_og_loss=1.77e-5]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.75s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 30%|███       | 6/20 [00:22<00:33,  2.39s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 30%|███       | 6/20 [00:24<00:33,  2.39s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.16s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.16s/iter, val_loss=1.76e-5, val_og_loss=1.77e-5]
+
 40%|████      | 8/20 [00:25<00:24,  2.01s/iter, val_loss=1.76e-5, val_og_loss=1.77e-5]
+
 40%|████      | 8/20 [00:27<00:24,  2.01s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.91s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 45%|████▌     | 9/20 [00:29<00:20,  1.91s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 50%|█████     | 10/20 [00:29<00:18,  1.84s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.84s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.79s/iter, val_loss=1.75e-5, val_og_loss=1.77e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=1.74e-5, val_og_loss=1.77e-5]
+
 60%|██████    | 12/20 [00:32<00:14,  1.76s/iter, val_loss=1.74e-5, val_og_loss=1.77e-5]
+
 60%|██████    | 12/20 [00:34<00:14,  1.76s/iter, val_loss=1.74e-5, val_og_loss=1.77e-5]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.74s/iter, val_loss=1.74e-5, val_og_loss=1.77e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.74s/iter, val_loss=1.74e-5, val_og_loss=1.77e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.72s/iter, val_loss=1.74e-5, val_og_loss=1.77e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.70s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.70s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.69s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=1.73e-5, val_og_loss=1.77e-5]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.68s/iter, val_loss=1.73e-5, val_og_loss=1.77e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.68s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
100%|██████████| 20/20 [00:45<00:00,  1.69s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
100%|██████████| 20/20 [00:45<00:00,  2.30s/iter, val_loss=1.72e-5, val_og_loss=1.77e-5]
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:55:47<3:11:53, 885.68s/it]INFO:ParoQuant:Best val loss: 1.7168829799629748e-05, Original val loss: 1.7668517102720216e-05
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:55:48<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:55:48<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:56:32<3:11:53, 885.68s/it]2025-12-22 18:26:05,749 - INFO - Best val loss: 1.7168829799629748e-05, Original val loss: 1.7668517102720216e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00107, val_og_loss=0.00109]
+
  5%|▌         | 1/20 [00:29<09:26, 29.79s/iter, val_loss=0.00107, val_og_loss=0.00109]
+
  5%|▌         | 1/20 [00:32<09:26, 29.79s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 10%|█         | 2/20 [00:32<04:09, 13.87s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 10%|█         | 2/20 [00:35<04:09, 13.87s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 15%|█▌        | 3/20 [00:35<02:29,  8.78s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 15%|█▌        | 3/20 [00:37<02:29,  8.78s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 20%|██        | 4/20 [00:37<01:42,  6.39s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 20%|██        | 4/20 [00:40<01:42,  6.39s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 25%|██▌       | 5/20 [00:40<01:16,  5.07s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.07s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 30%|███       | 6/20 [00:43<00:59,  4.27s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 30%|███       | 6/20 [00:46<00:59,  4.27s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 35%|███▌      | 7/20 [00:46<00:48,  3.76s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.76s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 40%|████      | 8/20 [00:48<00:41,  3.43s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 40%|████      | 8/20 [00:51<00:41,  3.43s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.21s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.21s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.00109, val_og_loss=0.00109]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.96s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 60%|██████    | 12/20 [00:59<00:23,  2.89s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 60%|██████    | 12/20 [01:02<00:23,  2.89s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.83s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.83s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.78s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.78s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 80%|████████  | 16/20 [01:10<00:11,  2.76s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.75s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.75s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.74s/iter, val_loss=0.00108, val_og_loss=0.00109]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.00107, val_og_loss=0.00109]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.00107, val_og_loss=0.00109]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.00107, val_og_loss=0.00109]
+
100%|██████████| 20/20 [01:21<00:00,  2.73s/iter, val_loss=0.00107, val_og_loss=0.00109]
100%|██████████| 20/20 [01:21<00:00,  4.08s/iter, val_loss=0.00107, val_og_loss=0.00109]
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:57:53<3:11:53, 885.68s/it]INFO:ParoQuant:Best val loss: 0.0010734833776950836, Original val loss: 0.0010927070397883654
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:57:58<3:11:53, 885.68s/it]2025-12-22 18:28:11,563 - INFO - Best val loss: 0.0010734833776950836, Original val loss: 0.0010927070397883654
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:32<?, ?iter/s, val_loss=0.000571, val_og_loss=0.000581]
+
  5%|▌         | 1/20 [00:32<10:13, 32.29s/iter, val_loss=0.000571, val_og_loss=0.000581]
+
  5%|▌         | 1/20 [00:35<10:13, 32.29s/iter, val_loss=0.000574, val_og_loss=0.000581]
+
 10%|█         | 2/20 [00:35<04:28, 14.89s/iter, val_loss=0.000574, val_og_loss=0.000581]
+
 10%|█         | 2/20 [00:37<04:28, 14.89s/iter, val_loss=0.000578, val_og_loss=0.000581]
+
 15%|█▌        | 3/20 [00:37<02:38,  9.33s/iter, val_loss=0.000578, val_og_loss=0.000581]
+
 15%|█▌        | 3/20 [00:40<02:38,  9.33s/iter, val_loss=0.00058, val_og_loss=0.000581] 
+
 20%|██        | 4/20 [00:40<01:47,  6.72s/iter, val_loss=0.00058, val_og_loss=0.000581]
+
 20%|██        | 4/20 [00:43<01:47,  6.72s/iter, val_loss=0.000582, val_og_loss=0.000581]
+
 25%|██▌       | 5/20 [00:43<01:19,  5.27s/iter, val_loss=0.000582, val_og_loss=0.000581]
+
 25%|██▌       | 5/20 [00:45<01:19,  5.27s/iter, val_loss=0.000581, val_og_loss=0.000581]
+
 30%|███       | 6/20 [00:45<01:01,  4.41s/iter, val_loss=0.000581, val_og_loss=0.000581]
+
 30%|███       | 6/20 [00:48<01:01,  4.41s/iter, val_loss=0.000581, val_og_loss=0.000581]
+
 35%|███▌      | 7/20 [00:48<00:50,  3.85s/iter, val_loss=0.000581, val_og_loss=0.000581]
+
 35%|███▌      | 7/20 [00:51<00:50,  3.85s/iter, val_loss=0.000579, val_og_loss=0.000581]
+
 40%|████      | 8/20 [00:51<00:41,  3.49s/iter, val_loss=0.000579, val_og_loss=0.000581]
+
 40%|████      | 8/20 [00:53<00:41,  3.49s/iter, val_loss=0.000579, val_og_loss=0.000581]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.24s/iter, val_loss=0.000579, val_og_loss=0.000581]
+
 45%|████▌     | 9/20 [00:56<00:35,  3.24s/iter, val_loss=0.000578, val_og_loss=0.000581]
+
 50%|█████     | 10/20 [00:56<00:30,  3.08s/iter, val_loss=0.000578, val_og_loss=0.000581]
+
 50%|█████     | 10/20 [00:59<00:30,  3.08s/iter, val_loss=0.000578, val_og_loss=0.000581]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.97s/iter, val_loss=0.000578, val_og_loss=0.000581]
+
 55%|█████▌    | 11/20 [01:02<00:26,  2.97s/iter, val_loss=0.000577, val_og_loss=0.000581]
+
 60%|██████    | 12/20 [01:02<00:23,  2.90s/iter, val_loss=0.000577, val_og_loss=0.000581]
+
 60%|██████    | 12/20 [01:04<00:23,  2.90s/iter, val_loss=0.000577, val_og_loss=0.000581]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.000577, val_og_loss=0.000581]
+
 65%|██████▌   | 13/20 [01:07<00:19,  2.84s/iter, val_loss=0.000576, val_og_loss=0.000581]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.000576, val_og_loss=0.000581]
+
 70%|███████   | 14/20 [01:10<00:16,  2.80s/iter, val_loss=0.000576, val_og_loss=0.000581]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.78s/iter, val_loss=0.000576, val_og_loss=0.000581]
+
 75%|███████▌  | 15/20 [01:13<00:13,  2.78s/iter, val_loss=0.000575, val_og_loss=0.000581]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.000575, val_og_loss=0.000581]
+
 80%|████████  | 16/20 [01:15<00:11,  2.76s/iter, val_loss=0.000575, val_og_loss=0.000581]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.000575, val_og_loss=0.000581]
+
 85%|████████▌ | 17/20 [01:18<00:08,  2.74s/iter, val_loss=0.000575, val_og_loss=0.000581]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.000575, val_og_loss=0.000581]
+
 90%|█████████ | 18/20 [01:21<00:05,  2.73s/iter, val_loss=0.000574, val_og_loss=0.000581]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.000574, val_og_loss=0.000581]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.73s/iter, val_loss=0.000574, val_og_loss=0.000581]
+
100%|██████████| 20/20 [01:23<00:00,  2.72s/iter, val_loss=0.000574, val_og_loss=0.000581]
100%|██████████| 20/20 [01:23<00:00,  4.19s/iter, val_loss=0.000574, val_og_loss=0.000581]
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:59:22<3:11:53, 885.68s/it]INFO:ParoQuant:Best val loss: 0.0005705049261450768, Original val loss: 0.000581348140258342
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:59:23<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [3:59:23<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [4:00:24<3:11:53, 885.68s/it]2025-12-22 18:29:39,986 - INFO - Best val loss: 0.0005705049261450768, Original val loss: 0.000581348140258342
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:32<?, ?iter/s, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
  5%|▌         | 1/20 [00:32<10:19, 32.58s/iter, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
  5%|▌         | 1/20 [00:38<10:19, 32.58s/iter, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
 10%|█         | 2/20 [00:38<04:58, 16.61s/iter, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
 10%|█         | 2/20 [00:43<04:58, 16.61s/iter, val_loss=9.73e-5, val_og_loss=9.82e-5]
+
 15%|█▌        | 3/20 [00:43<03:15, 11.50s/iter, val_loss=9.73e-5, val_og_loss=9.82e-5]
+
 15%|█▌        | 3/20 [00:48<03:15, 11.50s/iter, val_loss=9.74e-5, val_og_loss=9.82e-5]
+
 20%|██        | 4/20 [00:48<02:25,  9.10s/iter, val_loss=9.74e-5, val_og_loss=9.82e-5]
+
 20%|██        | 4/20 [00:54<02:25,  9.10s/iter, val_loss=9.75e-5, val_og_loss=9.82e-5]
+
 25%|██▌       | 5/20 [00:54<01:56,  7.78s/iter, val_loss=9.75e-5, val_og_loss=9.82e-5]
+
 25%|██▌       | 5/20 [00:59<01:56,  7.78s/iter, val_loss=9.74e-5, val_og_loss=9.82e-5]
+
 30%|███       | 6/20 [00:59<01:37,  6.98s/iter, val_loss=9.74e-5, val_og_loss=9.82e-5]
+
 30%|███       | 6/20 [01:05<01:37,  6.98s/iter, val_loss=9.75e-5, val_og_loss=9.82e-5]
+
 35%|███▌      | 7/20 [01:05<01:24,  6.47s/iter, val_loss=9.75e-5, val_og_loss=9.82e-5]
+
 35%|███▌      | 7/20 [01:10<01:24,  6.47s/iter, val_loss=9.75e-5, val_og_loss=9.82e-5]
+
 40%|████      | 8/20 [01:10<01:13,  6.13s/iter, val_loss=9.75e-5, val_og_loss=9.82e-5]
+
 40%|████      | 8/20 [01:15<01:13,  6.13s/iter, val_loss=9.73e-5, val_og_loss=9.82e-5]
+
 45%|████▌     | 9/20 [01:15<01:04,  5.91s/iter, val_loss=9.73e-5, val_og_loss=9.82e-5]
+
 45%|████▌     | 9/20 [01:21<01:04,  5.91s/iter, val_loss=9.74e-5, val_og_loss=9.82e-5]
+
 50%|█████     | 10/20 [01:21<00:57,  5.76s/iter, val_loss=9.74e-5, val_og_loss=9.82e-5]
+
 50%|█████     | 10/20 [01:26<00:57,  5.76s/iter, val_loss=9.75e-5, val_og_loss=9.82e-5]
+
 55%|█████▌    | 11/20 [01:26<00:50,  5.66s/iter, val_loss=9.75e-5, val_og_loss=9.82e-5]
+
 55%|█████▌    | 11/20 [01:32<00:50,  5.66s/iter, val_loss=9.74e-5, val_og_loss=9.82e-5]
+
 60%|██████    | 12/20 [01:32<00:44,  5.58s/iter, val_loss=9.74e-5, val_og_loss=9.82e-5]
+
 60%|██████    | 12/20 [01:37<00:44,  5.58s/iter, val_loss=9.73e-5, val_og_loss=9.82e-5]
+
 65%|██████▌   | 13/20 [01:37<00:38,  5.53s/iter, val_loss=9.73e-5, val_og_loss=9.82e-5]
+
 65%|██████▌   | 13/20 [01:43<00:38,  5.53s/iter, val_loss=9.72e-5, val_og_loss=9.82e-5]
+
 70%|███████   | 14/20 [01:43<00:32,  5.50s/iter, val_loss=9.72e-5, val_og_loss=9.82e-5]
+
 70%|███████   | 14/20 [01:48<00:32,  5.50s/iter, val_loss=9.72e-5, val_og_loss=9.82e-5]
+
 75%|███████▌  | 15/20 [01:48<00:27,  5.47s/iter, val_loss=9.72e-5, val_og_loss=9.82e-5]
+
 75%|███████▌  | 15/20 [01:53<00:27,  5.47s/iter, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
 80%|████████  | 16/20 [01:53<00:21,  5.45s/iter, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
 80%|████████  | 16/20 [01:59<00:21,  5.45s/iter, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
 85%|████████▌ | 17/20 [01:59<00:16,  5.45s/iter, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
 85%|████████▌ | 17/20 [02:04<00:16,  5.45s/iter, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
 90%|█████████ | 18/20 [02:04<00:10,  5.44s/iter, val_loss=9.71e-5, val_og_loss=9.82e-5]
+
 90%|█████████ | 18/20 [02:10<00:10,  5.44s/iter, val_loss=9.7e-5, val_og_loss=9.82e-5] 
+
 95%|█████████▌| 19/20 [02:10<00:05,  5.44s/iter, val_loss=9.7e-5, val_og_loss=9.82e-5]
+
 95%|█████████▌| 19/20 [02:15<00:05,  5.44s/iter, val_loss=9.7e-5, val_og_loss=9.82e-5]
+
100%|██████████| 20/20 [02:15<00:00,  5.43s/iter, val_loss=9.7e-5, val_og_loss=9.82e-5]
100%|██████████| 20/20 [02:15<00:00,  6.78s/iter, val_loss=9.7e-5, val_og_loss=9.82e-5]
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [4:02:40<3:11:53, 885.68s/it]INFO:ParoQuant:Best val loss: 9.698369831312448e-05, Original val loss: 9.82189376372844e-05
+
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [4:02:41<3:11:53, 885.68s/it]
                                                                       

Processing Blocks:  59%|█████▉    | 19/32 [4:02:50<3:11:53, 885.68s/it]
Processing Blocks:  62%|██████▎   | 20/32 [4:03:28<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:03:28<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:03:45<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:03:45<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:06:02<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:06:02<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:07:03<2:56:32, 882.74s/it]2025-12-22 18:32:58,277 - INFO - Best val loss: 9.698369831312448e-05, Original val loss: 9.82189376372844e-05
+[INFO] Saving Layer 19 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 20...
+[INFO] Starting Sequential Training for Layer 20
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.00238, val_og_loss=0.00287]
+
  5%|▌         | 1/20 [00:14<04:35, 14.52s/iter, val_loss=0.00238, val_og_loss=0.00287]
+
  5%|▌         | 1/20 [00:16<04:35, 14.52s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 10%|█         | 2/20 [00:16<02:05,  6.97s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 10%|█         | 2/20 [00:17<02:05,  6.97s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 15%|█▌        | 3/20 [00:17<01:17,  4.57s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 15%|█▌        | 3/20 [00:19<01:17,  4.57s/iter, val_loss=0.0024, val_og_loss=0.00287] 
+
 20%|██        | 4/20 [00:19<00:54,  3.43s/iter, val_loss=0.0024, val_og_loss=0.00287]
+
 20%|██        | 4/20 [00:21<00:54,  3.43s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 25%|██▌       | 5/20 [00:21<00:42,  2.80s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 25%|██▌       | 5/20 [00:22<00:42,  2.80s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 30%|███       | 6/20 [00:22<00:33,  2.42s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 30%|███       | 6/20 [00:24<00:33,  2.42s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.19s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 35%|███▌      | 7/20 [00:26<00:28,  2.19s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 40%|████      | 8/20 [00:26<00:24,  2.03s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 40%|████      | 8/20 [00:28<00:24,  2.03s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 45%|████▌     | 9/20 [00:28<00:21,  1.92s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.92s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 50%|█████     | 10/20 [00:29<00:18,  1.84s/iter, val_loss=0.00239, val_og_loss=0.00287]
+
 50%|█████     | 10/20 [00:31<00:18,  1.84s/iter, val_loss=0.00238, val_og_loss=0.00287]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.80s/iter, val_loss=0.00238, val_og_loss=0.00287]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.80s/iter, val_loss=0.00237, val_og_loss=0.00287]
+
 60%|██████    | 12/20 [00:33<00:14,  1.76s/iter, val_loss=0.00237, val_og_loss=0.00287]
+
 60%|██████    | 12/20 [00:34<00:14,  1.76s/iter, val_loss=0.00238, val_og_loss=0.00287]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=0.00238, val_og_loss=0.00287]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.73s/iter, val_loss=0.00238, val_og_loss=0.00287]
+
 70%|███████   | 14/20 [00:36<00:10,  1.71s/iter, val_loss=0.00238, val_og_loss=0.00287]
+
 70%|███████   | 14/20 [00:38<00:10,  1.71s/iter, val_loss=0.00237, val_og_loss=0.00287]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=0.00237, val_og_loss=0.00287]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.00237, val_og_loss=0.00287]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=0.00237, val_og_loss=0.00287]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=0.00236, val_og_loss=0.00287]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.70s/iter, val_loss=0.00236, val_og_loss=0.00287]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.70s/iter, val_loss=0.00236, val_og_loss=0.00287]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=0.00236, val_og_loss=0.00287]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.00236, val_og_loss=0.00287]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=0.00236, val_og_loss=0.00287]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=0.00236, val_og_loss=0.00287]
+
100%|██████████| 20/20 [00:46<00:00,  1.69s/iter, val_loss=0.00236, val_og_loss=0.00287]
100%|██████████| 20/20 [00:46<00:00,  2.33s/iter, val_loss=0.00236, val_og_loss=0.00287]
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:07:50<2:56:32, 882.74s/it]INFO:ParoQuant:Best val loss: 0.002359530422836542, Original val loss: 0.0028723860159516335
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:07:51<2:56:32, 882.74s/it]2025-12-22 18:38:07,980 - INFO - Best val loss: 0.002359530422836542, Original val loss: 0.0028723860159516335
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=0.0061, val_og_loss=0.0071]
+
  5%|▌         | 1/20 [00:13<04:11, 13.23s/iter, val_loss=0.0061, val_og_loss=0.0071]
+
  5%|▌         | 1/20 [00:14<04:11, 13.23s/iter, val_loss=0.00627, val_og_loss=0.0071]
+
 10%|█         | 2/20 [00:14<01:52,  6.27s/iter, val_loss=0.00627, val_og_loss=0.0071]
+
 10%|█         | 2/20 [00:16<01:52,  6.27s/iter, val_loss=0.00623, val_og_loss=0.0071]
+
 15%|█▌        | 3/20 [00:16<01:08,  4.04s/iter, val_loss=0.00623, val_og_loss=0.0071]
+
 15%|█▌        | 3/20 [00:17<01:08,  4.04s/iter, val_loss=0.00624, val_og_loss=0.0071]
+
 20%|██        | 4/20 [00:17<00:47,  2.99s/iter, val_loss=0.00624, val_og_loss=0.0071]
+
 20%|██        | 4/20 [00:18<00:47,  2.99s/iter, val_loss=0.00624, val_og_loss=0.0071]
+
 25%|██▌       | 5/20 [00:18<00:36,  2.42s/iter, val_loss=0.00624, val_og_loss=0.0071]
+
 25%|██▌       | 5/20 [00:20<00:36,  2.42s/iter, val_loss=0.00623, val_og_loss=0.0071]
+
 30%|███       | 6/20 [00:20<00:28,  2.07s/iter, val_loss=0.00623, val_og_loss=0.0071]
+
 30%|███       | 6/20 [00:21<00:28,  2.07s/iter, val_loss=0.00621, val_og_loss=0.0071]
+
 35%|███▌      | 7/20 [00:21<00:23,  1.85s/iter, val_loss=0.00621, val_og_loss=0.0071]
+
 35%|███▌      | 7/20 [00:22<00:23,  1.85s/iter, val_loss=0.00624, val_og_loss=0.0071]
+
 40%|████      | 8/20 [00:22<00:20,  1.70s/iter, val_loss=0.00624, val_og_loss=0.0071]
+
 40%|████      | 8/20 [00:24<00:20,  1.70s/iter, val_loss=0.00622, val_og_loss=0.0071]
+
 45%|████▌     | 9/20 [00:24<00:17,  1.60s/iter, val_loss=0.00622, val_og_loss=0.0071]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.60s/iter, val_loss=0.0062, val_og_loss=0.0071] 
+
 50%|█████     | 10/20 [00:25<00:15,  1.55s/iter, val_loss=0.0062, val_og_loss=0.0071]
+
 50%|█████     | 10/20 [00:27<00:15,  1.55s/iter, val_loss=0.00622, val_og_loss=0.0071]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.50s/iter, val_loss=0.00622, val_og_loss=0.0071]
+
 55%|█████▌    | 11/20 [00:28<00:13,  1.50s/iter, val_loss=0.00619, val_og_loss=0.0071]
+
 60%|██████    | 12/20 [00:28<00:11,  1.46s/iter, val_loss=0.00619, val_og_loss=0.0071]
+
 60%|██████    | 12/20 [00:29<00:11,  1.46s/iter, val_loss=0.00616, val_og_loss=0.0071]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.44s/iter, val_loss=0.00616, val_og_loss=0.0071]
+
 65%|██████▌   | 13/20 [00:31<00:10,  1.44s/iter, val_loss=0.00619, val_og_loss=0.0071]
+
 70%|███████   | 14/20 [00:31<00:08,  1.42s/iter, val_loss=0.00619, val_og_loss=0.0071]
+
 70%|███████   | 14/20 [00:32<00:08,  1.42s/iter, val_loss=0.00619, val_og_loss=0.0071]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.42s/iter, val_loss=0.00619, val_og_loss=0.0071]
+
 75%|███████▌  | 15/20 [00:34<00:07,  1.42s/iter, val_loss=0.00616, val_og_loss=0.0071]
+
 80%|████████  | 16/20 [00:34<00:05,  1.41s/iter, val_loss=0.00616, val_og_loss=0.0071]
+
 80%|████████  | 16/20 [00:35<00:05,  1.41s/iter, val_loss=0.00614, val_og_loss=0.0071]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.40s/iter, val_loss=0.00614, val_og_loss=0.0071]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.40s/iter, val_loss=0.00613, val_og_loss=0.0071]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.00613, val_og_loss=0.0071]
+
 90%|█████████ | 18/20 [00:38<00:02,  1.40s/iter, val_loss=0.00614, val_og_loss=0.0071]
+
 95%|█████████▌| 19/20 [00:38<00:01,  1.40s/iter, val_loss=0.00614, val_og_loss=0.0071]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.40s/iter, val_loss=0.00611, val_og_loss=0.0071]
+
100%|██████████| 20/20 [00:39<00:00,  1.40s/iter, val_loss=0.00611, val_og_loss=0.0071]
100%|██████████| 20/20 [00:39<00:00,  1.98s/iter, val_loss=0.00611, val_og_loss=0.0071]
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:08:31<2:56:32, 882.74s/it]INFO:ParoQuant:Best val loss: 0.006101383361965418, Original val loss: 0.007098556961864233
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:08:34<2:56:32, 882.74s/it]2025-12-22 18:38:49,461 - INFO - Best val loss: 0.006101383361965418, Original val loss: 0.007098556961864233
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.000611, val_og_loss=0.000752]
+
  5%|▌         | 1/20 [00:14<04:37, 14.59s/iter, val_loss=0.000611, val_og_loss=0.000752]
+
  5%|▌         | 1/20 [00:15<04:37, 14.59s/iter, val_loss=0.000615, val_og_loss=0.000752]
+
 10%|█         | 2/20 [00:15<02:02,  6.82s/iter, val_loss=0.000615, val_og_loss=0.000752]
+
 10%|█         | 2/20 [00:17<02:02,  6.82s/iter, val_loss=0.000616, val_og_loss=0.000752]
+
 15%|█▌        | 3/20 [00:17<01:13,  4.35s/iter, val_loss=0.000616, val_og_loss=0.000752]
+
 15%|█▌        | 3/20 [00:18<01:13,  4.35s/iter, val_loss=0.000617, val_og_loss=0.000752]
+
 20%|██        | 4/20 [00:18<00:50,  3.18s/iter, val_loss=0.000617, val_og_loss=0.000752]
+
 20%|██        | 4/20 [00:20<00:50,  3.18s/iter, val_loss=0.000617, val_og_loss=0.000752]
+
 25%|██▌       | 5/20 [00:20<00:37,  2.53s/iter, val_loss=0.000617, val_og_loss=0.000752]
+
 25%|██▌       | 5/20 [00:21<00:37,  2.53s/iter, val_loss=0.000618, val_og_loss=0.000752]
+
 30%|███       | 6/20 [00:21<00:29,  2.14s/iter, val_loss=0.000618, val_og_loss=0.000752]
+
 30%|███       | 6/20 [00:22<00:29,  2.14s/iter, val_loss=0.000617, val_og_loss=0.000752]
+
 35%|███▌      | 7/20 [00:22<00:24,  1.90s/iter, val_loss=0.000617, val_og_loss=0.000752]
+
 35%|███▌      | 7/20 [00:24<00:24,  1.90s/iter, val_loss=0.000616, val_og_loss=0.000752]
+
 40%|████      | 8/20 [00:24<00:20,  1.74s/iter, val_loss=0.000616, val_og_loss=0.000752]
+
 40%|████      | 8/20 [00:25<00:20,  1.74s/iter, val_loss=0.000616, val_og_loss=0.000752]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.63s/iter, val_loss=0.000616, val_og_loss=0.000752]
+
 45%|████▌     | 9/20 [00:27<00:17,  1.63s/iter, val_loss=0.000615, val_og_loss=0.000752]
+
 50%|█████     | 10/20 [00:27<00:15,  1.55s/iter, val_loss=0.000615, val_og_loss=0.000752]
+
 50%|█████     | 10/20 [00:28<00:15,  1.55s/iter, val_loss=0.000615, val_og_loss=0.000752]
+
 55%|█████▌    | 11/20 [00:28<00:13,  1.50s/iter, val_loss=0.000615, val_og_loss=0.000752]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.50s/iter, val_loss=0.000613, val_og_loss=0.000752]
+
 60%|██████    | 12/20 [00:29<00:11,  1.47s/iter, val_loss=0.000613, val_og_loss=0.000752]
+
 60%|██████    | 12/20 [00:31<00:11,  1.47s/iter, val_loss=0.000611, val_og_loss=0.000752]
+
 65%|██████▌   | 13/20 [00:31<00:10,  1.45s/iter, val_loss=0.000611, val_og_loss=0.000752]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.45s/iter, val_loss=0.00061, val_og_loss=0.000752] 
+
 70%|███████   | 14/20 [00:32<00:08,  1.43s/iter, val_loss=0.00061, val_og_loss=0.000752]
+
 70%|███████   | 14/20 [00:34<00:08,  1.43s/iter, val_loss=0.000609, val_og_loss=0.000752]
+
 75%|███████▌  | 15/20 [00:34<00:07,  1.42s/iter, val_loss=0.000609, val_og_loss=0.000752]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.42s/iter, val_loss=0.00061, val_og_loss=0.000752] 
+
 80%|████████  | 16/20 [00:35<00:05,  1.41s/iter, val_loss=0.00061, val_og_loss=0.000752]
+
 80%|████████  | 16/20 [00:36<00:05,  1.41s/iter, val_loss=0.000609, val_og_loss=0.000752]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.40s/iter, val_loss=0.000609, val_og_loss=0.000752]
+
 85%|████████▌ | 17/20 [00:38<00:04,  1.40s/iter, val_loss=0.000607, val_og_loss=0.000752]
+
 90%|█████████ | 18/20 [00:38<00:02,  1.41s/iter, val_loss=0.000607, val_og_loss=0.000752]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.41s/iter, val_loss=0.000606, val_og_loss=0.000752]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.40s/iter, val_loss=0.000606, val_og_loss=0.000752]
+
 95%|█████████▌| 19/20 [00:41<00:01,  1.40s/iter, val_loss=0.000607, val_og_loss=0.000752]
+
100%|██████████| 20/20 [00:41<00:00,  1.40s/iter, val_loss=0.000607, val_og_loss=0.000752]
100%|██████████| 20/20 [00:41<00:00,  2.05s/iter, val_loss=0.000607, val_og_loss=0.000752]
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:09:15<2:56:32, 882.74s/it]INFO:ParoQuant:Best val loss: 0.0006061149761080742, Original val loss: 0.000751978310290724
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:09:17<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:09:17<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:09:48<2:56:32, 882.74s/it]2025-12-22 18:39:33,138 - INFO - Best val loss: 0.0006061149761080742, Original val loss: 0.000751978310290724
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
  5%|▌         | 1/20 [00:13<04:17, 13.57s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
  5%|▌         | 1/20 [00:15<04:17, 13.57s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 10%|█         | 2/20 [00:15<01:58,  6.58s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 10%|█         | 2/20 [00:16<01:58,  6.58s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
 15%|█▌        | 3/20 [00:16<01:13,  4.34s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
 15%|█▌        | 3/20 [00:18<01:13,  4.34s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 20%|██        | 4/20 [00:18<00:52,  3.30s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 20%|██        | 4/20 [00:20<00:52,  3.30s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
 25%|██▌       | 5/20 [00:20<00:40,  2.72s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
 25%|██▌       | 5/20 [00:22<00:40,  2.72s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 30%|███       | 6/20 [00:22<00:33,  2.37s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 30%|███       | 6/20 [00:23<00:33,  2.37s/iter, val_loss=2.13e-5, val_og_loss=2.18e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.15s/iter, val_loss=2.13e-5, val_og_loss=2.18e-5]
+
 35%|███▌      | 7/20 [00:25<00:27,  2.15s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 40%|████      | 8/20 [00:25<00:24,  2.00s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 40%|████      | 8/20 [00:27<00:24,  2.00s/iter, val_loss=2.17e-5, val_og_loss=2.18e-5]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.90s/iter, val_loss=2.17e-5, val_og_loss=2.18e-5]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.90s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 50%|█████     | 10/20 [00:28<00:18,  1.83s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.83s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.79s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 60%|██████    | 12/20 [00:32<00:14,  1.75s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 60%|██████    | 12/20 [00:33<00:14,  1.75s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.73s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.72s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.70s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.70s/iter, val_loss=2.13e-5, val_og_loss=2.18e-5]
+
 80%|████████  | 16/20 [00:38<00:06,  1.70s/iter, val_loss=2.13e-5, val_og_loss=2.18e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=2.13e-5, val_og_loss=2.18e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.70s/iter, val_loss=2.13e-5, val_og_loss=2.18e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=2.16e-5, val_og_loss=2.18e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=2.16e-5, val_og_loss=2.18e-5]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.68s/iter, val_loss=2.15e-5, val_og_loss=2.18e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.68s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
100%|██████████| 20/20 [00:45<00:00,  1.68s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
100%|██████████| 20/20 [00:45<00:00,  2.28s/iter, val_loss=2.14e-5, val_og_loss=2.18e-5]
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:10:34<2:56:32, 882.74s/it]INFO:ParoQuant:Best val loss: 2.1287552954163402e-05, Original val loss: 2.1794219719595276e-05
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:10:34<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:10:34<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:11:19<2:56:32, 882.74s/it]2025-12-22 18:40:52,482 - INFO - Best val loss: 2.1287552954163402e-05, Original val loss: 2.1794219719595276e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00114, val_og_loss=0.00116]
+
  5%|▌         | 1/20 [00:29<09:27, 29.88s/iter, val_loss=0.00114, val_og_loss=0.00116]
+
  5%|▌         | 1/20 [00:32<09:27, 29.88s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 10%|█         | 2/20 [00:32<04:10, 13.91s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 10%|█         | 2/20 [00:35<04:10, 13.91s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 15%|█▌        | 3/20 [00:35<02:29,  8.80s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 15%|█▌        | 3/20 [00:38<02:29,  8.80s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 20%|██        | 4/20 [00:38<01:42,  6.40s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 20%|██        | 4/20 [00:40<01:42,  6.40s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 25%|██▌       | 5/20 [00:40<01:16,  5.07s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.07s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 30%|███       | 6/20 [00:43<00:59,  4.28s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 30%|███       | 6/20 [00:46<00:59,  4.28s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 35%|███▌      | 7/20 [00:46<00:48,  3.76s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.76s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 40%|████      | 8/20 [00:48<00:41,  3.44s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 40%|████      | 8/20 [00:51<00:41,  3.44s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.21s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.21s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.95s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.95s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 60%|██████    | 12/20 [00:59<00:23,  2.88s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 60%|██████    | 12/20 [01:02<00:23,  2.88s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.84s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.84s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 80%|████████  | 16/20 [01:10<00:11,  2.75s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 80%|████████  | 16/20 [01:13<00:11,  2.75s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.73s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.00114, val_og_loss=0.00116]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.00114, val_og_loss=0.00116]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.00114, val_og_loss=0.00116]
+
100%|██████████| 20/20 [01:21<00:00,  2.72s/iter, val_loss=0.00114, val_og_loss=0.00116]
100%|██████████| 20/20 [01:21<00:00,  4.08s/iter, val_loss=0.00114, val_og_loss=0.00116]
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:12:40<2:56:32, 882.74s/it]INFO:ParoQuant:Best val loss: 0.0011437138309702277, Original val loss: 0.0011614924296736717
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:12:45<2:56:32, 882.74s/it]2025-12-22 18:42:58,805 - INFO - Best val loss: 0.0011437138309702277, Original val loss: 0.0011614924296736717
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.000613, val_og_loss=0.000625]
+
  5%|▌         | 1/20 [00:30<09:48, 30.98s/iter, val_loss=0.000613, val_og_loss=0.000625]
+
  5%|▌         | 1/20 [00:33<09:48, 30.98s/iter, val_loss=0.000616, val_og_loss=0.000625]
+
 10%|█         | 2/20 [00:33<04:18, 14.35s/iter, val_loss=0.000616, val_og_loss=0.000625]
+
 10%|█         | 2/20 [00:36<04:18, 14.35s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 15%|█▌        | 3/20 [00:36<02:33,  9.04s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 15%|█▌        | 3/20 [00:39<02:33,  9.04s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 20%|██        | 4/20 [00:39<01:44,  6.54s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 20%|██        | 4/20 [00:41<01:44,  6.54s/iter, val_loss=0.000623, val_og_loss=0.000625]
+
 25%|██▌       | 5/20 [00:41<01:17,  5.16s/iter, val_loss=0.000623, val_og_loss=0.000625]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.16s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 30%|███       | 6/20 [00:44<01:00,  4.33s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 30%|███       | 6/20 [00:47<01:00,  4.33s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.80s/iter, val_loss=0.000622, val_og_loss=0.000625]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.80s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 40%|████      | 8/20 [00:49<00:41,  3.45s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 40%|████      | 8/20 [00:52<00:41,  3.45s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.22s/iter, val_loss=0.000621, val_og_loss=0.000625]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.22s/iter, val_loss=0.00062, val_og_loss=0.000625] 
+
 50%|█████     | 10/20 [00:55<00:30,  3.08s/iter, val_loss=0.00062, val_og_loss=0.000625]
+
 50%|█████     | 10/20 [00:58<00:30,  3.08s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.97s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.97s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.84s/iter, val_loss=0.000619, val_og_loss=0.000625]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000618, val_og_loss=0.000625]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.000618, val_og_loss=0.000625]
+
 70%|███████   | 14/20 [01:09<00:16,  2.80s/iter, val_loss=0.000617, val_og_loss=0.000625]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.000617, val_og_loss=0.000625]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.000617, val_og_loss=0.000625]
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.000617, val_og_loss=0.000625]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.000616, val_og_loss=0.000625]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.75s/iter, val_loss=0.000616, val_og_loss=0.000625]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.75s/iter, val_loss=0.000616, val_og_loss=0.000625]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.000616, val_og_loss=0.000625]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.74s/iter, val_loss=0.000615, val_og_loss=0.000625]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.000615, val_og_loss=0.000625]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.000615, val_og_loss=0.000625]
+
100%|██████████| 20/20 [01:22<00:00,  2.72s/iter, val_loss=0.000615, val_og_loss=0.000625]
100%|██████████| 20/20 [01:22<00:00,  4.13s/iter, val_loss=0.000615, val_og_loss=0.000625]
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:14:08<2:56:32, 882.74s/it]INFO:ParoQuant:Best val loss: 0.0006134414579719305, Original val loss: 0.0006248772260732949
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:14:09<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:14:09<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:15:07<2:56:32, 882.74s/it]2025-12-22 18:44:26,115 - INFO - Best val loss: 0.0006134414579719305, Original val loss: 0.0006248772260732949
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:34<?, ?iter/s, val_loss=0.000105, val_og_loss=0.000107]
+
  5%|▌         | 1/20 [00:34<11:01, 34.80s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
  5%|▌         | 1/20 [00:40<11:01, 34.80s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 10%|█         | 2/20 [00:40<05:15, 17.51s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 10%|█         | 2/20 [00:45<05:15, 17.51s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 15%|█▌        | 3/20 [00:45<03:23, 11.99s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 15%|█▌        | 3/20 [00:51<03:23, 11.99s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 20%|██        | 4/20 [00:51<02:30,  9.40s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 20%|██        | 4/20 [00:56<02:30,  9.40s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 25%|██▌       | 5/20 [00:56<01:59,  7.97s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 25%|██▌       | 5/20 [01:01<01:59,  7.97s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 30%|███       | 6/20 [01:01<01:39,  7.10s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 30%|███       | 6/20 [01:07<01:39,  7.10s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 35%|███▌      | 7/20 [01:07<01:25,  6.55s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 35%|███▌      | 7/20 [01:12<01:25,  6.55s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 40%|████      | 8/20 [01:12<01:14,  6.19s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 40%|████      | 8/20 [01:18<01:14,  6.19s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 45%|████▌     | 9/20 [01:18<01:05,  5.95s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 45%|████▌     | 9/20 [01:23<01:05,  5.95s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 50%|█████     | 10/20 [01:23<00:57,  5.79s/iter, val_loss=0.000106, val_og_loss=0.000107]
+
 50%|█████     | 10/20 [01:29<00:57,  5.79s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 55%|█████▌    | 11/20 [01:29<00:51,  5.68s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 55%|█████▌    | 11/20 [01:34<00:51,  5.68s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 60%|██████    | 12/20 [01:34<00:44,  5.60s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 60%|██████    | 12/20 [01:39<00:44,  5.60s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 65%|██████▌   | 13/20 [01:39<00:38,  5.55s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 65%|██████▌   | 13/20 [01:45<00:38,  5.55s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 70%|███████   | 14/20 [01:45<00:33,  5.51s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 70%|███████   | 14/20 [01:50<00:33,  5.51s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 75%|███████▌  | 15/20 [01:50<00:27,  5.49s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 75%|███████▌  | 15/20 [01:56<00:27,  5.49s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 80%|████████  | 16/20 [01:56<00:21,  5.47s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 80%|████████  | 16/20 [02:01<00:21,  5.47s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 85%|████████▌ | 17/20 [02:01<00:16,  5.46s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 85%|████████▌ | 17/20 [02:07<00:16,  5.46s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 90%|█████████ | 18/20 [02:07<00:10,  5.45s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 90%|█████████ | 18/20 [02:12<00:10,  5.45s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 95%|█████████▌| 19/20 [02:12<00:05,  5.44s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
 95%|█████████▌| 19/20 [02:17<00:05,  5.44s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
100%|██████████| 20/20 [02:17<00:00,  5.43s/iter, val_loss=0.000105, val_og_loss=0.000107]
100%|██████████| 20/20 [02:17<00:00,  6.89s/iter, val_loss=0.000105, val_og_loss=0.000107]
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:17:25<2:56:32, 882.74s/it]INFO:ParoQuant:Best val loss: 0.00010502717486815527, Original val loss: 0.00010703172301873565
+
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:17:25<2:56:32, 882.74s/it]
                                                                       

Processing Blocks:  62%|██████▎   | 20/32 [4:17:36<2:56:32, 882.74s/it]
Processing Blocks:  66%|██████▌   | 21/32 [4:18:14<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:18:14<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:18:34<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:18:34<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:20:47<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:20:47<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:21:45<2:42:01, 883.81s/it]2025-12-22 18:47:43,200 - INFO - Best val loss: 0.00010502717486815527, Original val loss: 0.00010703172301873565
+[INFO] Saving Layer 20 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 21...
+[INFO] Starting Sequential Training for Layer 21
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=0.00229, val_og_loss=0.00272]
+
  5%|▌         | 1/20 [00:13<04:09, 13.15s/iter, val_loss=0.00229, val_og_loss=0.00272]
+
  5%|▌         | 1/20 [00:14<04:09, 13.15s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 10%|█         | 2/20 [00:14<01:55,  6.41s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 10%|█         | 2/20 [00:16<01:55,  6.41s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 15%|█▌        | 3/20 [00:16<01:12,  4.27s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 15%|█▌        | 3/20 [00:18<01:12,  4.27s/iter, val_loss=0.00232, val_og_loss=0.00272]
+
 20%|██        | 4/20 [00:18<00:52,  3.26s/iter, val_loss=0.00232, val_og_loss=0.00272]
+
 20%|██        | 4/20 [00:19<00:52,  3.26s/iter, val_loss=0.00232, val_og_loss=0.00272]
+
 25%|██▌       | 5/20 [00:19<00:40,  2.70s/iter, val_loss=0.00232, val_og_loss=0.00272]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.70s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 30%|███       | 6/20 [00:21<00:32,  2.36s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 30%|███       | 6/20 [00:23<00:32,  2.36s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.15s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 35%|███▌      | 7/20 [00:25<00:27,  2.15s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 40%|████      | 8/20 [00:25<00:24,  2.00s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 40%|████      | 8/20 [00:26<00:24,  2.00s/iter, val_loss=0.00232, val_og_loss=0.00272]
+
 45%|████▌     | 9/20 [00:26<00:20,  1.91s/iter, val_loss=0.00232, val_og_loss=0.00272]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.91s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 50%|█████     | 10/20 [00:28<00:18,  1.84s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 50%|█████     | 10/20 [00:30<00:18,  1.84s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.80s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.80s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 60%|██████    | 12/20 [00:31<00:14,  1.76s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 60%|██████    | 12/20 [00:33<00:14,  1.76s/iter, val_loss=0.0023, val_og_loss=0.00272] 
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.74s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.74s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 70%|███████   | 14/20 [00:35<00:10,  1.72s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 70%|███████   | 14/20 [00:36<00:10,  1.72s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 75%|███████▌  | 15/20 [00:36<00:08,  1.72s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.72s/iter, val_loss=0.00229, val_og_loss=0.00272]
+
 80%|████████  | 16/20 [00:38<00:06,  1.70s/iter, val_loss=0.00229, val_og_loss=0.00272]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=0.00229, val_og_loss=0.00272]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=0.00229, val_og_loss=0.00272]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.69s/iter, val_loss=0.00228, val_og_loss=0.00272]
+
 90%|█████████ | 18/20 [00:41<00:03,  1.69s/iter, val_loss=0.00228, val_og_loss=0.00272]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=0.00228, val_og_loss=0.00272]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.69s/iter, val_loss=0.00228, val_og_loss=0.00272]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=0.00227, val_og_loss=0.00272]
+
100%|██████████| 20/20 [00:45<00:00,  1.68s/iter, val_loss=0.00227, val_og_loss=0.00272]
100%|██████████| 20/20 [00:45<00:00,  2.27s/iter, val_loss=0.00227, val_og_loss=0.00272]
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:22:31<2:42:01, 883.81s/it]INFO:ParoQuant:Best val loss: 0.0022736878599971533, Original val loss: 0.0027178318705409765
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:22:33<2:42:01, 883.81s/it]2025-12-22 18:52:49,268 - INFO - Best val loss: 0.0022736878599971533, Original val loss: 0.0027178318705409765
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=0.00593, val_og_loss=0.00684]
+
  5%|▌         | 1/20 [00:13<04:21, 13.74s/iter, val_loss=0.00593, val_og_loss=0.00684]
+
  5%|▌         | 1/20 [00:15<04:21, 13.74s/iter, val_loss=0.00604, val_og_loss=0.00684]
+
 10%|█         | 2/20 [00:15<01:56,  6.48s/iter, val_loss=0.00604, val_og_loss=0.00684]
+
 10%|█         | 2/20 [00:16<01:56,  6.48s/iter, val_loss=0.00602, val_og_loss=0.00684]
+
 15%|█▌        | 3/20 [00:16<01:10,  4.15s/iter, val_loss=0.00602, val_og_loss=0.00684]
+
 15%|█▌        | 3/20 [00:17<01:10,  4.15s/iter, val_loss=0.00603, val_og_loss=0.00684]
+
 20%|██        | 4/20 [00:17<00:48,  3.06s/iter, val_loss=0.00603, val_og_loss=0.00684]
+
 20%|██        | 4/20 [00:19<00:48,  3.06s/iter, val_loss=0.00601, val_og_loss=0.00684]
+
 25%|██▌       | 5/20 [00:19<00:36,  2.46s/iter, val_loss=0.00601, val_og_loss=0.00684]
+
 25%|██▌       | 5/20 [00:20<00:36,  2.46s/iter, val_loss=0.00605, val_og_loss=0.00684]
+
 30%|███       | 6/20 [00:20<00:29,  2.10s/iter, val_loss=0.00605, val_og_loss=0.00684]
+
 30%|███       | 6/20 [00:22<00:29,  2.10s/iter, val_loss=0.00604, val_og_loss=0.00684]
+
 35%|███▌      | 7/20 [00:22<00:24,  1.86s/iter, val_loss=0.00604, val_og_loss=0.00684]
+
 35%|███▌      | 7/20 [00:23<00:24,  1.86s/iter, val_loss=0.006, val_og_loss=0.00684]  
+
 40%|████      | 8/20 [00:23<00:20,  1.71s/iter, val_loss=0.006, val_og_loss=0.00684]
+
 40%|████      | 8/20 [00:24<00:20,  1.71s/iter, val_loss=0.00601, val_og_loss=0.00684]
+
 45%|████▌     | 9/20 [00:24<00:17,  1.61s/iter, val_loss=0.00601, val_og_loss=0.00684]
+
 45%|████▌     | 9/20 [00:26<00:17,  1.61s/iter, val_loss=0.00603, val_og_loss=0.00684]
+
 50%|█████     | 10/20 [00:26<00:15,  1.55s/iter, val_loss=0.00603, val_og_loss=0.00684]
+
 50%|█████     | 10/20 [00:27<00:15,  1.55s/iter, val_loss=0.00599, val_og_loss=0.00684]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.50s/iter, val_loss=0.00599, val_og_loss=0.00684]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.50s/iter, val_loss=0.00599, val_og_loss=0.00684]
+
 60%|██████    | 12/20 [00:29<00:11,  1.47s/iter, val_loss=0.00599, val_og_loss=0.00684]
+
 60%|██████    | 12/20 [00:30<00:11,  1.47s/iter, val_loss=0.006, val_og_loss=0.00684]  
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.45s/iter, val_loss=0.006, val_og_loss=0.00684]
+
 65%|██████▌   | 13/20 [00:31<00:10,  1.45s/iter, val_loss=0.00594, val_og_loss=0.00684]
+
 70%|███████   | 14/20 [00:31<00:08,  1.43s/iter, val_loss=0.00594, val_og_loss=0.00684]
+
 70%|███████   | 14/20 [00:33<00:08,  1.43s/iter, val_loss=0.00597, val_og_loss=0.00684]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.42s/iter, val_loss=0.00597, val_og_loss=0.00684]
+
 75%|███████▌  | 15/20 [00:34<00:07,  1.42s/iter, val_loss=0.00597, val_og_loss=0.00684]
+
 80%|████████  | 16/20 [00:34<00:05,  1.41s/iter, val_loss=0.00597, val_og_loss=0.00684]
+
 80%|████████  | 16/20 [00:35<00:05,  1.41s/iter, val_loss=0.00596, val_og_loss=0.00684]
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.40s/iter, val_loss=0.00596, val_og_loss=0.00684]
+
 85%|████████▌ | 17/20 [00:37<00:04,  1.40s/iter, val_loss=0.00593, val_og_loss=0.00684]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.40s/iter, val_loss=0.00593, val_og_loss=0.00684]
+
 90%|█████████ | 18/20 [00:38<00:02,  1.40s/iter, val_loss=0.00595, val_og_loss=0.00684]
+
 95%|█████████▌| 19/20 [00:38<00:01,  1.40s/iter, val_loss=0.00595, val_og_loss=0.00684]
+
 95%|█████████▌| 19/20 [00:40<00:01,  1.40s/iter, val_loss=0.00594, val_og_loss=0.00684]
+
100%|██████████| 20/20 [00:40<00:00,  1.39s/iter, val_loss=0.00594, val_og_loss=0.00684]
100%|██████████| 20/20 [00:40<00:00,  2.01s/iter, val_loss=0.00594, val_og_loss=0.00684]
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:23:13<2:42:01, 883.81s/it]INFO:ParoQuant:Best val loss: 0.005928000435233116, Original val loss: 0.006837175227701664
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:23:15<2:42:01, 883.81s/it]2025-12-22 18:53:31,413 - INFO - Best val loss: 0.005928000435233116, Original val loss: 0.006837175227701664
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:12<?, ?iter/s, val_loss=0.000643, val_og_loss=0.000774]
+
  5%|▌         | 1/20 [00:12<04:03, 12.83s/iter, val_loss=0.000643, val_og_loss=0.000774]
+
  5%|▌         | 1/20 [00:14<04:03, 12.83s/iter, val_loss=0.000647, val_og_loss=0.000774]
+
 10%|█         | 2/20 [00:14<01:49,  6.10s/iter, val_loss=0.000647, val_og_loss=0.000774]
+
 10%|█         | 2/20 [00:15<01:49,  6.10s/iter, val_loss=0.000648, val_og_loss=0.000774]
+
 15%|█▌        | 3/20 [00:15<01:07,  3.95s/iter, val_loss=0.000648, val_og_loss=0.000774]
+
 15%|█▌        | 3/20 [00:17<01:07,  3.95s/iter, val_loss=0.000648, val_og_loss=0.000774]
+
 20%|██        | 4/20 [00:17<00:47,  2.95s/iter, val_loss=0.000648, val_og_loss=0.000774]
+
 20%|██        | 4/20 [00:18<00:47,  2.95s/iter, val_loss=0.00065, val_og_loss=0.000774] 
+
 25%|██▌       | 5/20 [00:18<00:35,  2.39s/iter, val_loss=0.00065, val_og_loss=0.000774]
+
 25%|██▌       | 5/20 [00:19<00:35,  2.39s/iter, val_loss=0.000651, val_og_loss=0.000774]
+
 30%|███       | 6/20 [00:19<00:28,  2.05s/iter, val_loss=0.000651, val_og_loss=0.000774]
+
 30%|███       | 6/20 [00:21<00:28,  2.05s/iter, val_loss=0.000649, val_og_loss=0.000774]
+
 35%|███▌      | 7/20 [00:21<00:23,  1.83s/iter, val_loss=0.000649, val_og_loss=0.000774]
+
 35%|███▌      | 7/20 [00:22<00:23,  1.83s/iter, val_loss=0.00065, val_og_loss=0.000774] 
+
 40%|████      | 8/20 [00:22<00:20,  1.69s/iter, val_loss=0.00065, val_og_loss=0.000774]
+
 40%|████      | 8/20 [00:23<00:20,  1.69s/iter, val_loss=0.00065, val_og_loss=0.000774]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.60s/iter, val_loss=0.00065, val_og_loss=0.000774]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.60s/iter, val_loss=0.000649, val_og_loss=0.000774]
+
 50%|█████     | 10/20 [00:25<00:15,  1.53s/iter, val_loss=0.000649, val_og_loss=0.000774]
+
 50%|█████     | 10/20 [00:26<00:15,  1.53s/iter, val_loss=0.000647, val_og_loss=0.000774]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.49s/iter, val_loss=0.000647, val_og_loss=0.000774]
+
 55%|█████▌    | 11/20 [00:28<00:13,  1.49s/iter, val_loss=0.000646, val_og_loss=0.000774]
+
 60%|██████    | 12/20 [00:28<00:11,  1.46s/iter, val_loss=0.000646, val_og_loss=0.000774]
+
 60%|██████    | 12/20 [00:29<00:11,  1.46s/iter, val_loss=0.000644, val_og_loss=0.000774]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.45s/iter, val_loss=0.000644, val_og_loss=0.000774]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.45s/iter, val_loss=0.000643, val_og_loss=0.000774]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.000643, val_og_loss=0.000774]
+
 70%|███████   | 14/20 [00:32<00:08,  1.43s/iter, val_loss=0.000643, val_og_loss=0.000774]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.42s/iter, val_loss=0.000643, val_og_loss=0.000774]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.42s/iter, val_loss=0.000641, val_og_loss=0.000774]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.000641, val_og_loss=0.000774]
+
 80%|████████  | 16/20 [00:35<00:05,  1.41s/iter, val_loss=0.00064, val_og_loss=0.000774] 
+
 85%|████████▌ | 17/20 [00:35<00:04,  1.40s/iter, val_loss=0.00064, val_og_loss=0.000774]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.40s/iter, val_loss=0.000639, val_og_loss=0.000774]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.000639, val_og_loss=0.000774]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.40s/iter, val_loss=0.000639, val_og_loss=0.000774]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.000639, val_og_loss=0.000774]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.40s/iter, val_loss=0.000638, val_og_loss=0.000774]
+
100%|██████████| 20/20 [00:39<00:00,  1.40s/iter, val_loss=0.000638, val_og_loss=0.000774]
100%|██████████| 20/20 [00:39<00:00,  1.97s/iter, val_loss=0.000638, val_og_loss=0.000774]
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:23:54<2:42:01, 883.81s/it]INFO:ParoQuant:Best val loss: 0.0006380589911714196, Original val loss: 0.0007735452381893992
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:23:57<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:23:57<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:24:28<2:42:01, 883.81s/it]2025-12-22 18:54:12,736 - INFO - Best val loss: 0.0006380589911714196, Original val loss: 0.0007735452381893992
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
  5%|▌         | 1/20 [00:13<04:18, 13.58s/iter, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
  5%|▌         | 1/20 [00:15<04:18, 13.58s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 10%|█         | 2/20 [00:15<01:58,  6.59s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 10%|█         | 2/20 [00:16<01:58,  6.59s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 15%|█▌        | 3/20 [00:16<01:13,  4.35s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 15%|█▌        | 3/20 [00:18<01:13,  4.35s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 20%|██        | 4/20 [00:18<00:52,  3.30s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 20%|██        | 4/20 [00:20<00:52,  3.30s/iter, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
 25%|██▌       | 5/20 [00:20<00:40,  2.72s/iter, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
 25%|██▌       | 5/20 [00:22<00:40,  2.72s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 30%|███       | 6/20 [00:22<00:33,  2.36s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 30%|███       | 6/20 [00:23<00:33,  2.36s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.14s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 35%|███▌      | 7/20 [00:25<00:27,  2.14s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 40%|████      | 8/20 [00:25<00:23,  1.99s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 40%|████      | 8/20 [00:27<00:23,  1.99s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.89s/iter, val_loss=2.86e-5, val_og_loss=2.96e-5]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.89s/iter, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
 50%|█████     | 10/20 [00:28<00:18,  1.83s/iter, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.83s/iter, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.79s/iter, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=2.84e-5, val_og_loss=2.96e-5]
+
 60%|██████    | 12/20 [00:32<00:14,  1.75s/iter, val_loss=2.84e-5, val_og_loss=2.96e-5]
+
 60%|██████    | 12/20 [00:33<00:14,  1.75s/iter, val_loss=2.84e-5, val_og_loss=2.96e-5]
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.73s/iter, val_loss=2.84e-5, val_og_loss=2.96e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=2.84e-5, val_og_loss=2.96e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.72s/iter, val_loss=2.84e-5, val_og_loss=2.96e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.71s/iter, val_loss=2.85e-5, val_og_loss=2.96e-5]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=2.83e-5, val_og_loss=2.96e-5]
+
 80%|████████  | 16/20 [00:38<00:06,  1.70s/iter, val_loss=2.83e-5, val_og_loss=2.96e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=2.83e-5, val_og_loss=2.96e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=2.83e-5, val_og_loss=2.96e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.69s/iter, val_loss=2.83e-5, val_og_loss=2.96e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.68s/iter, val_loss=2.83e-5, val_og_loss=2.96e-5]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.68s/iter, val_loss=2.83e-5, val_og_loss=2.96e-5]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.69s/iter, val_loss=2.83e-5, val_og_loss=2.96e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=2.82e-5, val_og_loss=2.96e-5]
+
100%|██████████| 20/20 [00:45<00:00,  1.68s/iter, val_loss=2.82e-5, val_og_loss=2.96e-5]
100%|██████████| 20/20 [00:45<00:00,  2.28s/iter, val_loss=2.82e-5, val_og_loss=2.96e-5]
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:25:13<2:42:01, 883.81s/it]INFO:ParoQuant:Best val loss: 2.8239099265192635e-05, Original val loss: 2.955353556899354e-05
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:25:14<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:25:14<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:25:58<2:42:01, 883.81s/it]2025-12-22 18:55:31,589 - INFO - Best val loss: 2.8239099265192635e-05, Original val loss: 2.955353556899354e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.00125, val_og_loss=0.00127]
+
  5%|▌         | 1/20 [00:31<09:55, 31.35s/iter, val_loss=0.00125, val_og_loss=0.00127]
+
  5%|▌         | 1/20 [00:34<09:55, 31.35s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
 10%|█         | 2/20 [00:34<04:21, 14.51s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
 10%|█         | 2/20 [00:36<04:21, 14.51s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 15%|█▌        | 3/20 [00:36<02:35,  9.12s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 15%|█▌        | 3/20 [00:39<02:35,  9.12s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 20%|██        | 4/20 [00:39<01:45,  6.60s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 20%|██        | 4/20 [00:42<01:45,  6.60s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 25%|██▌       | 5/20 [00:42<01:17,  5.19s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.19s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 30%|███       | 6/20 [00:44<01:01,  4.36s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 30%|███       | 6/20 [00:47<01:01,  4.36s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.82s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.82s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 40%|████      | 8/20 [00:50<00:41,  3.47s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 40%|████      | 8/20 [00:53<00:41,  3.47s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.23s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.23s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 50%|█████     | 10/20 [00:55<00:30,  3.07s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 50%|█████     | 10/20 [00:58<00:30,  3.07s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.97s/iter, val_loss=0.00128, val_og_loss=0.00127]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.97s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 60%|██████    | 12/20 [01:04<00:23,  2.89s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.85s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.85s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 70%|███████   | 14/20 [01:06<00:16,  2.81s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 70%|███████   | 14/20 [01:09<00:16,  2.81s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 75%|███████▌  | 15/20 [01:12<00:13,  2.78s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.00127, val_og_loss=0.00127]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.75s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.75s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.74s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.73s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
100%|██████████| 20/20 [01:23<00:00,  2.73s/iter, val_loss=0.00126, val_og_loss=0.00127]
100%|██████████| 20/20 [01:23<00:00,  4.15s/iter, val_loss=0.00126, val_og_loss=0.00127]
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:27:21<2:42:01, 883.81s/it]INFO:ParoQuant:Best val loss: 0.0012540678726509213, Original val loss: 0.0012739910744130611
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:27:25<2:42:01, 883.81s/it]2025-12-22 18:57:39,331 - INFO - Best val loss: 0.0012540678726509213, Original val loss: 0.0012739910744130611
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.000665, val_og_loss=0.000677]
+
  5%|▌         | 1/20 [00:30<09:48, 30.97s/iter, val_loss=0.000665, val_og_loss=0.000677]
+
  5%|▌         | 1/20 [00:33<09:48, 30.97s/iter, val_loss=0.000672, val_og_loss=0.000677]
+
 10%|█         | 2/20 [00:33<04:18, 14.35s/iter, val_loss=0.000672, val_og_loss=0.000677]
+
 10%|█         | 2/20 [00:36<04:18, 14.35s/iter, val_loss=0.000679, val_og_loss=0.000677]
+
 15%|█▌        | 3/20 [00:36<02:33,  9.05s/iter, val_loss=0.000679, val_og_loss=0.000677]
+
 15%|█▌        | 3/20 [00:39<02:33,  9.05s/iter, val_loss=0.000681, val_og_loss=0.000677]
+
 20%|██        | 4/20 [00:39<01:44,  6.55s/iter, val_loss=0.000681, val_og_loss=0.000677]
+
 20%|██        | 4/20 [00:41<01:44,  6.55s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 25%|██▌       | 5/20 [00:41<01:17,  5.17s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.17s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 30%|███       | 6/20 [00:44<01:00,  4.34s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 30%|███       | 6/20 [00:47<01:00,  4.34s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.80s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.80s/iter, val_loss=0.000689, val_og_loss=0.000677]
+
 40%|████      | 8/20 [00:50<00:41,  3.46s/iter, val_loss=0.000689, val_og_loss=0.000677]
+
 40%|████      | 8/20 [00:52<00:41,  3.46s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.23s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.23s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 50%|█████     | 10/20 [00:55<00:30,  3.07s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 50%|█████     | 10/20 [00:58<00:30,  3.07s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.96s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.84s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000688, val_og_loss=0.000677]
+
 70%|███████   | 14/20 [01:06<00:16,  2.81s/iter, val_loss=0.000688, val_og_loss=0.000677]
+
 70%|███████   | 14/20 [01:09<00:16,  2.81s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.75s/iter, val_loss=0.000686, val_og_loss=0.000677]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.75s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.74s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
100%|██████████| 20/20 [01:22<00:00,  2.72s/iter, val_loss=0.000687, val_og_loss=0.000677]
100%|██████████| 20/20 [01:22<00:00,  4.13s/iter, val_loss=0.000687, val_og_loss=0.000677]
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:28:48<2:42:01, 883.81s/it]INFO:ParoQuant:Best val loss: 0.000664760242216289, Original val loss: 0.0006769573083147407
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:28:50<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:28:50<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:29:50<2:42:01, 883.81s/it]2025-12-22 18:59:06,457 - INFO - Best val loss: 0.000664760242216289, Original val loss: 0.0006769573083147407
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.000125, val_og_loss=0.000128]
+
  5%|▌         | 1/20 [00:33<10:41, 33.75s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
  5%|▌         | 1/20 [00:39<10:41, 33.75s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 10%|█         | 2/20 [00:39<05:07, 17.08s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 10%|█         | 2/20 [00:44<05:07, 17.08s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 15%|█▌        | 3/20 [00:44<03:19, 11.76s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 15%|█▌        | 3/20 [00:50<03:19, 11.76s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 20%|██        | 4/20 [00:50<02:28,  9.25s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 20%|██        | 4/20 [00:55<02:28,  9.25s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 25%|██▌       | 5/20 [00:55<01:58,  7.87s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 25%|██▌       | 5/20 [01:00<01:58,  7.87s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 30%|███       | 6/20 [01:00<01:38,  7.04s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 30%|███       | 6/20 [01:06<01:38,  7.04s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.51s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.51s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 40%|████      | 8/20 [01:11<01:13,  6.16s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 40%|████      | 8/20 [01:17<01:13,  6.16s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 45%|████▌     | 9/20 [01:17<01:05,  5.93s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.93s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 50%|█████     | 10/20 [01:22<00:57,  5.77s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 50%|█████     | 10/20 [01:27<00:57,  5.77s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 55%|█████▌    | 11/20 [01:27<00:50,  5.66s/iter, val_loss=0.000126, val_og_loss=0.000128]
+
 55%|█████▌    | 11/20 [01:33<00:50,  5.66s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 60%|██████    | 12/20 [01:33<00:44,  5.59s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 60%|██████    | 12/20 [01:38<00:44,  5.59s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.54s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.54s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 70%|███████   | 14/20 [01:44<00:33,  5.50s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 70%|███████   | 14/20 [01:49<00:33,  5.50s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.48s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 75%|███████▌  | 15/20 [01:55<00:27,  5.48s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 80%|████████  | 16/20 [01:55<00:21,  5.46s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 80%|████████  | 16/20 [02:00<00:21,  5.46s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.45s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.45s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.44s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.44s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
100%|██████████| 20/20 [02:16<00:00,  5.43s/iter, val_loss=0.000125, val_og_loss=0.000128]
100%|██████████| 20/20 [02:16<00:00,  6.84s/iter, val_loss=0.000125, val_og_loss=0.000128]
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:32:07<2:42:01, 883.81s/it]INFO:ParoQuant:Best val loss: 0.0001248657499672845, Original val loss: 0.00012764464190695435
+
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:32:08<2:42:01, 883.81s/it]
                                                                       

Processing Blocks:  66%|██████▌   | 21/32 [4:32:19<2:42:01, 883.81s/it]
Processing Blocks:  69%|██████▉   | 22/32 [4:32:56<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:32:57<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:33:14<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:33:14<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:35:45<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:35:45<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:36:50<2:27:14, 883.46s/it]2025-12-22 19:02:25,472 - INFO - Best val loss: 0.0001248657499672845, Original val loss: 0.00012764464190695435
+[INFO] Saving Layer 21 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 22...
+[INFO] Starting Sequential Training for Layer 22
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:18<?, ?iter/s, val_loss=0.00229, val_og_loss=0.00281]
+
  5%|▌         | 1/20 [00:18<05:54, 18.64s/iter, val_loss=0.00229, val_og_loss=0.00281]
+
  5%|▌         | 1/20 [00:20<05:54, 18.64s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 10%|█         | 2/20 [00:20<02:36,  8.67s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 10%|█         | 2/20 [00:22<02:36,  8.67s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 15%|█▌        | 3/20 [00:22<01:33,  5.49s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 15%|█▌        | 3/20 [00:23<01:33,  5.49s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 20%|██        | 4/20 [00:23<01:03,  3.99s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 20%|██        | 4/20 [00:25<01:03,  3.99s/iter, val_loss=0.00235, val_og_loss=0.00281]
+
 25%|██▌       | 5/20 [00:25<00:47,  3.18s/iter, val_loss=0.00235, val_og_loss=0.00281]
+
 25%|██▌       | 5/20 [00:27<00:47,  3.18s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 30%|███       | 6/20 [00:27<00:37,  2.67s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 30%|███       | 6/20 [00:28<00:37,  2.67s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 35%|███▌      | 7/20 [00:28<00:30,  2.34s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 35%|███▌      | 7/20 [00:30<00:30,  2.34s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 40%|████      | 8/20 [00:30<00:25,  2.13s/iter, val_loss=0.00234, val_og_loss=0.00281]
+
 40%|████      | 8/20 [00:32<00:25,  2.13s/iter, val_loss=0.00233, val_og_loss=0.00281]
+
 45%|████▌     | 9/20 [00:32<00:21,  1.99s/iter, val_loss=0.00233, val_og_loss=0.00281]
+
 45%|████▌     | 9/20 [00:33<00:21,  1.99s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 50%|█████     | 10/20 [00:33<00:18,  1.90s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 50%|█████     | 10/20 [00:35<00:18,  1.90s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.83s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 55%|█████▌    | 11/20 [00:37<00:16,  1.83s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 60%|██████    | 12/20 [00:37<00:14,  1.78s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 60%|██████    | 12/20 [00:38<00:14,  1.78s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.75s/iter, val_loss=0.00232, val_og_loss=0.00281]
+
 65%|██████▌   | 13/20 [00:40<00:12,  1.75s/iter, val_loss=0.00231, val_og_loss=0.00281]
+
 70%|███████   | 14/20 [00:40<00:10,  1.73s/iter, val_loss=0.00231, val_og_loss=0.00281]
+
 70%|███████   | 14/20 [00:42<00:10,  1.73s/iter, val_loss=0.00231, val_og_loss=0.00281]
+
 75%|███████▌  | 15/20 [00:42<00:08,  1.71s/iter, val_loss=0.00231, val_og_loss=0.00281]
+
 75%|███████▌  | 15/20 [00:43<00:08,  1.71s/iter, val_loss=0.0023, val_og_loss=0.00281] 
+
 80%|████████  | 16/20 [00:43<00:06,  1.70s/iter, val_loss=0.0023, val_og_loss=0.00281]
+
 80%|████████  | 16/20 [00:45<00:06,  1.70s/iter, val_loss=0.0023, val_og_loss=0.00281]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.70s/iter, val_loss=0.0023, val_og_loss=0.00281]
+
 85%|████████▌ | 17/20 [00:47<00:05,  1.70s/iter, val_loss=0.00229, val_og_loss=0.00281]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.69s/iter, val_loss=0.00229, val_og_loss=0.00281]
+
 90%|█████████ | 18/20 [00:48<00:03,  1.69s/iter, val_loss=0.00229, val_og_loss=0.00281]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.69s/iter, val_loss=0.00229, val_og_loss=0.00281]
+
 95%|█████████▌| 19/20 [00:50<00:01,  1.69s/iter, val_loss=0.00229, val_og_loss=0.00281]
+
100%|██████████| 20/20 [00:50<00:00,  1.68s/iter, val_loss=0.00229, val_og_loss=0.00281]
100%|██████████| 20/20 [00:50<00:00,  2.53s/iter, val_loss=0.00229, val_og_loss=0.00281]
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:37:40<2:27:14, 883.46s/it]INFO:ParoQuant:Best val loss: 0.002288032555952668, Original val loss: 0.0028137839399278164
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:37:42<2:27:14, 883.46s/it]2025-12-22 19:07:58,838 - INFO - Best val loss: 0.002288032555952668, Original val loss: 0.0028137839399278164
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.00622, val_og_loss=0.00735]
+
  5%|▌         | 1/20 [00:16<05:22, 17.00s/iter, val_loss=0.00622, val_og_loss=0.00735]
+
  5%|▌         | 1/20 [00:18<05:22, 17.00s/iter, val_loss=0.0064, val_og_loss=0.00735] 
+
 10%|█         | 2/20 [00:18<02:20,  7.82s/iter, val_loss=0.0064, val_og_loss=0.00735]
+
 10%|█         | 2/20 [00:19<02:20,  7.82s/iter, val_loss=0.00638, val_og_loss=0.00735]
+
 15%|█▌        | 3/20 [00:19<01:23,  4.89s/iter, val_loss=0.00638, val_og_loss=0.00735]
+
 15%|█▌        | 3/20 [00:21<01:23,  4.89s/iter, val_loss=0.00635, val_og_loss=0.00735]
+
 20%|██        | 4/20 [00:21<00:56,  3.51s/iter, val_loss=0.00635, val_og_loss=0.00735]
+
 20%|██        | 4/20 [00:22<00:56,  3.51s/iter, val_loss=0.00635, val_og_loss=0.00735]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.74s/iter, val_loss=0.00635, val_og_loss=0.00735]
+
 25%|██▌       | 5/20 [00:23<00:41,  2.74s/iter, val_loss=0.00631, val_og_loss=0.00735]
+
 30%|███       | 6/20 [00:23<00:32,  2.29s/iter, val_loss=0.00631, val_og_loss=0.00735]
+
 30%|███       | 6/20 [00:25<00:32,  2.29s/iter, val_loss=0.00632, val_og_loss=0.00735]
+
 35%|███▌      | 7/20 [00:25<00:26,  2.00s/iter, val_loss=0.00632, val_og_loss=0.00735]
+
 35%|███▌      | 7/20 [00:26<00:26,  2.00s/iter, val_loss=0.00636, val_og_loss=0.00735]
+
 40%|████      | 8/20 [00:26<00:21,  1.81s/iter, val_loss=0.00636, val_og_loss=0.00735]
+
 40%|████      | 8/20 [00:28<00:21,  1.81s/iter, val_loss=0.0063, val_og_loss=0.00735] 
+
 45%|████▌     | 9/20 [00:28<00:18,  1.68s/iter, val_loss=0.0063, val_og_loss=0.00735]
+
 45%|████▌     | 9/20 [00:29<00:18,  1.68s/iter, val_loss=0.00632, val_og_loss=0.00735]
+
 50%|█████     | 10/20 [00:29<00:15,  1.60s/iter, val_loss=0.00632, val_og_loss=0.00735]
+
 50%|█████     | 10/20 [00:30<00:15,  1.60s/iter, val_loss=0.00629, val_og_loss=0.00735]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.53s/iter, val_loss=0.00629, val_og_loss=0.00735]
+
 55%|█████▌    | 11/20 [00:32<00:13,  1.53s/iter, val_loss=0.0063, val_og_loss=0.00735] 
+
 60%|██████    | 12/20 [00:32<00:11,  1.49s/iter, val_loss=0.0063, val_og_loss=0.00735]
+
 60%|██████    | 12/20 [00:33<00:11,  1.49s/iter, val_loss=0.00631, val_og_loss=0.00735]
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.46s/iter, val_loss=0.00631, val_og_loss=0.00735]
+
 65%|██████▌   | 13/20 [00:35<00:10,  1.46s/iter, val_loss=0.00627, val_og_loss=0.00735]
+
 70%|███████   | 14/20 [00:35<00:08,  1.44s/iter, val_loss=0.00627, val_og_loss=0.00735]
+
 70%|███████   | 14/20 [00:36<00:08,  1.44s/iter, val_loss=0.00622, val_og_loss=0.00735]
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.42s/iter, val_loss=0.00622, val_og_loss=0.00735]
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.42s/iter, val_loss=0.00625, val_og_loss=0.00735]
+
 80%|████████  | 16/20 [00:37<00:05,  1.42s/iter, val_loss=0.00625, val_og_loss=0.00735]
+
 80%|████████  | 16/20 [00:39<00:05,  1.42s/iter, val_loss=0.00625, val_og_loss=0.00735]
+
 85%|████████▌ | 17/20 [00:39<00:04,  1.42s/iter, val_loss=0.00625, val_og_loss=0.00735]
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.42s/iter, val_loss=0.00626, val_og_loss=0.00735]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.41s/iter, val_loss=0.00626, val_og_loss=0.00735]
+
 90%|█████████ | 18/20 [00:42<00:02,  1.41s/iter, val_loss=0.00624, val_og_loss=0.00735]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.41s/iter, val_loss=0.00624, val_og_loss=0.00735]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.41s/iter, val_loss=0.00622, val_og_loss=0.00735]
+
100%|██████████| 20/20 [00:43<00:00,  1.40s/iter, val_loss=0.00622, val_og_loss=0.00735]
100%|██████████| 20/20 [00:43<00:00,  2.18s/iter, val_loss=0.00622, val_og_loss=0.00735]
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:38:26<2:27:14, 883.46s/it]INFO:ParoQuant:Best val loss: 0.006217638961970806, Original val loss: 0.007346406579017639
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:38:29<2:27:14, 883.46s/it]2025-12-22 19:08:44,339 - INFO - Best val loss: 0.006217638961970806, Original val loss: 0.007346406579017639
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:19<?, ?iter/s, val_loss=0.000748, val_og_loss=0.000933]
+
  5%|▌         | 1/20 [00:19<06:11, 19.57s/iter, val_loss=0.000748, val_og_loss=0.000933]
+
  5%|▌         | 1/20 [00:20<06:11, 19.57s/iter, val_loss=0.000745, val_og_loss=0.000933]
+
 10%|█         | 2/20 [00:20<02:39,  8.88s/iter, val_loss=0.000745, val_og_loss=0.000933]
+
 10%|█         | 2/20 [00:22<02:39,  8.88s/iter, val_loss=0.000749, val_og_loss=0.000933]
+
 15%|█▌        | 3/20 [00:22<01:32,  5.46s/iter, val_loss=0.000749, val_og_loss=0.000933]
+
 15%|█▌        | 3/20 [00:23<01:32,  5.46s/iter, val_loss=0.000748, val_og_loss=0.000933]
+
 20%|██        | 4/20 [00:23<01:01,  3.85s/iter, val_loss=0.000748, val_og_loss=0.000933]
+
 20%|██        | 4/20 [00:25<01:01,  3.85s/iter, val_loss=0.000751, val_og_loss=0.000933]
+
 25%|██▌       | 5/20 [00:25<00:44,  2.98s/iter, val_loss=0.000751, val_og_loss=0.000933]
+
 25%|██▌       | 5/20 [00:26<00:44,  2.98s/iter, val_loss=0.000748, val_og_loss=0.000933]
+
 30%|███       | 6/20 [00:26<00:34,  2.44s/iter, val_loss=0.000748, val_og_loss=0.000933]
+
 30%|███       | 6/20 [00:27<00:34,  2.44s/iter, val_loss=0.000748, val_og_loss=0.000933]
+
 35%|███▌      | 7/20 [00:27<00:27,  2.09s/iter, val_loss=0.000748, val_og_loss=0.000933]
+
 35%|███▌      | 7/20 [00:29<00:27,  2.09s/iter, val_loss=0.000748, val_og_loss=0.000933]
+
 40%|████      | 8/20 [00:29<00:22,  1.87s/iter, val_loss=0.000748, val_og_loss=0.000933]
+
 40%|████      | 8/20 [00:30<00:22,  1.87s/iter, val_loss=0.00075, val_og_loss=0.000933] 
+
 45%|████▌     | 9/20 [00:30<00:18,  1.72s/iter, val_loss=0.00075, val_og_loss=0.000933]
+
 45%|████▌     | 9/20 [00:32<00:18,  1.72s/iter, val_loss=0.000747, val_og_loss=0.000933]
+
 50%|█████     | 10/20 [00:32<00:16,  1.62s/iter, val_loss=0.000747, val_og_loss=0.000933]
+
 50%|█████     | 10/20 [00:33<00:16,  1.62s/iter, val_loss=0.000745, val_og_loss=0.000933]
+
 55%|█████▌    | 11/20 [00:33<00:13,  1.55s/iter, val_loss=0.000745, val_og_loss=0.000933]
+
 55%|█████▌    | 11/20 [00:34<00:13,  1.55s/iter, val_loss=0.000742, val_og_loss=0.000933]
+
 60%|██████    | 12/20 [00:34<00:12,  1.50s/iter, val_loss=0.000742, val_og_loss=0.000933]
+
 60%|██████    | 12/20 [00:36<00:12,  1.50s/iter, val_loss=0.00074, val_og_loss=0.000933] 
+
 65%|██████▌   | 13/20 [00:36<00:10,  1.47s/iter, val_loss=0.00074, val_og_loss=0.000933]
+
 65%|██████▌   | 13/20 [00:37<00:10,  1.47s/iter, val_loss=0.000741, val_og_loss=0.000933]
+
 70%|███████   | 14/20 [00:37<00:08,  1.44s/iter, val_loss=0.000741, val_og_loss=0.000933]
+
 70%|███████   | 14/20 [00:39<00:08,  1.44s/iter, val_loss=0.000738, val_og_loss=0.000933]
+
 75%|███████▌  | 15/20 [00:39<00:07,  1.43s/iter, val_loss=0.000738, val_og_loss=0.000933]
+
 75%|███████▌  | 15/20 [00:40<00:07,  1.43s/iter, val_loss=0.000737, val_og_loss=0.000933]
+
 80%|████████  | 16/20 [00:40<00:05,  1.42s/iter, val_loss=0.000737, val_og_loss=0.000933]
+
 80%|████████  | 16/20 [00:41<00:05,  1.42s/iter, val_loss=0.000737, val_og_loss=0.000933]
+
 85%|████████▌ | 17/20 [00:41<00:04,  1.41s/iter, val_loss=0.000737, val_og_loss=0.000933]
+
 85%|████████▌ | 17/20 [00:43<00:04,  1.41s/iter, val_loss=0.000737, val_og_loss=0.000933]
+
 90%|█████████ | 18/20 [00:43<00:02,  1.41s/iter, val_loss=0.000737, val_og_loss=0.000933]
+
 90%|█████████ | 18/20 [00:44<00:02,  1.41s/iter, val_loss=0.000736, val_og_loss=0.000933]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.40s/iter, val_loss=0.000736, val_og_loss=0.000933]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.40s/iter, val_loss=0.000735, val_og_loss=0.000933]
+
100%|██████████| 20/20 [00:46<00:00,  1.41s/iter, val_loss=0.000735, val_og_loss=0.000933]
100%|██████████| 20/20 [00:46<00:00,  2.30s/iter, val_loss=0.000735, val_og_loss=0.000933]
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:39:15<2:27:14, 883.46s/it]INFO:ParoQuant:Best val loss: 0.0007351250387728214, Original val loss: 0.0009330702014267445
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:39:18<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:39:18<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:39:49<2:27:14, 883.46s/it]2025-12-22 19:09:33,301 - INFO - Best val loss: 0.0007351250387728214, Original val loss: 0.0009330702014267445
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=2.69e-5, val_og_loss=2.72e-5]
+
  5%|▌         | 1/20 [00:13<04:09, 13.14s/iter, val_loss=2.69e-5, val_og_loss=2.72e-5]
+
  5%|▌         | 1/20 [00:14<04:09, 13.14s/iter, val_loss=2.69e-5, val_og_loss=2.72e-5]
+
 10%|█         | 2/20 [00:14<01:55,  6.44s/iter, val_loss=2.69e-5, val_og_loss=2.72e-5]
+
 10%|█         | 2/20 [00:16<01:55,  6.44s/iter, val_loss=2.68e-5, val_og_loss=2.72e-5]
+
 15%|█▌        | 3/20 [00:16<01:12,  4.27s/iter, val_loss=2.68e-5, val_og_loss=2.72e-5]
+
 15%|█▌        | 3/20 [00:18<01:12,  4.27s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 20%|██        | 4/20 [00:18<00:52,  3.25s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 20%|██        | 4/20 [00:19<00:52,  3.25s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 25%|██▌       | 5/20 [00:19<00:40,  2.69s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.69s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 30%|███       | 6/20 [00:21<00:32,  2.34s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 30%|███       | 6/20 [00:23<00:32,  2.34s/iter, val_loss=2.68e-5, val_og_loss=2.72e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.12s/iter, val_loss=2.68e-5, val_og_loss=2.72e-5]
+
 35%|███▌      | 7/20 [00:24<00:27,  2.12s/iter, val_loss=2.69e-5, val_og_loss=2.72e-5]
+
 40%|████      | 8/20 [00:24<00:23,  1.98s/iter, val_loss=2.69e-5, val_og_loss=2.72e-5]
+
 40%|████      | 8/20 [00:26<00:23,  1.98s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 45%|████▌     | 9/20 [00:26<00:20,  1.89s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.89s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 50%|█████     | 10/20 [00:28<00:18,  1.82s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.82s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.78s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.78s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
 60%|██████    | 12/20 [00:31<00:13,  1.75s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
 60%|██████    | 12/20 [00:33<00:13,  1.75s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.73s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.71s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 70%|███████   | 14/20 [00:36<00:10,  1.71s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
 75%|███████▌  | 15/20 [00:36<00:08,  1.70s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.70s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 80%|████████  | 16/20 [00:38<00:06,  1.69s/iter, val_loss=2.67e-5, val_og_loss=2.72e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.69s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.68s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.68s/iter, val_loss=2.65e-5, val_og_loss=2.72e-5]
+
 90%|█████████ | 18/20 [00:41<00:03,  1.68s/iter, val_loss=2.65e-5, val_og_loss=2.72e-5]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.68s/iter, val_loss=2.65e-5, val_og_loss=2.72e-5]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.68s/iter, val_loss=2.65e-5, val_og_loss=2.72e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.68s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
100%|██████████| 20/20 [00:45<00:00,  1.68s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
100%|██████████| 20/20 [00:45<00:00,  2.25s/iter, val_loss=2.66e-5, val_og_loss=2.72e-5]
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:40:34<2:27:14, 883.46s/it]INFO:ParoQuant:Best val loss: 2.6540335966274142e-05, Original val loss: 2.7178502932656556e-05
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:40:35<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:40:35<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:41:21<2:27:14, 883.46s/it]2025-12-22 19:10:52,956 - INFO - Best val loss: 2.6540335966274142e-05, Original val loss: 2.7178502932656556e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:40<?, ?iter/s, val_loss=0.00128, val_og_loss=0.0013]
+
  5%|▌         | 1/20 [00:40<12:55, 40.81s/iter, val_loss=0.00128, val_og_loss=0.0013]
+
  5%|▌         | 1/20 [00:43<12:55, 40.81s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 10%|█         | 2/20 [00:43<05:31, 18.40s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 10%|█         | 2/20 [00:46<05:31, 18.40s/iter, val_loss=0.0013, val_og_loss=0.0013] 
+
 15%|█▌        | 3/20 [00:46<03:11, 11.25s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 15%|█▌        | 3/20 [00:48<03:11, 11.25s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 20%|██        | 4/20 [00:48<02:06,  7.88s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 20%|██        | 4/20 [00:51<02:06,  7.88s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 25%|██▌       | 5/20 [00:51<01:30,  6.02s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 25%|██▌       | 5/20 [00:54<01:30,  6.02s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 30%|███       | 6/20 [00:54<01:08,  4.89s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 30%|███       | 6/20 [00:57<01:08,  4.89s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 35%|███▌      | 7/20 [00:57<00:54,  4.18s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 35%|███▌      | 7/20 [00:59<00:54,  4.18s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 40%|████      | 8/20 [00:59<00:44,  3.71s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 40%|████      | 8/20 [01:02<00:44,  3.71s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 45%|████▌     | 9/20 [01:02<00:37,  3.40s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 45%|████▌     | 9/20 [01:05<00:37,  3.40s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 50%|█████     | 10/20 [01:05<00:31,  3.19s/iter, val_loss=0.0013, val_og_loss=0.0013]
+
 50%|█████     | 10/20 [01:07<00:31,  3.19s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 55%|█████▌    | 11/20 [01:07<00:27,  3.04s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 55%|█████▌    | 11/20 [01:10<00:27,  3.04s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 60%|██████    | 12/20 [01:10<00:23,  2.95s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 60%|██████    | 12/20 [01:13<00:23,  2.95s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 65%|██████▌   | 13/20 [01:13<00:20,  2.87s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 65%|██████▌   | 13/20 [01:16<00:20,  2.87s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 70%|███████   | 14/20 [01:16<00:16,  2.83s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 70%|███████   | 14/20 [01:18<00:16,  2.83s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 75%|███████▌  | 15/20 [01:18<00:13,  2.80s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 75%|███████▌  | 15/20 [01:21<00:13,  2.80s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 80%|████████  | 16/20 [01:21<00:11,  2.78s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 80%|████████  | 16/20 [01:24<00:11,  2.78s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 85%|████████▌ | 17/20 [01:24<00:08,  2.76s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 85%|████████▌ | 17/20 [01:27<00:08,  2.76s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 90%|█████████ | 18/20 [01:27<00:05,  2.74s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 90%|█████████ | 18/20 [01:29<00:05,  2.74s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 95%|█████████▌| 19/20 [01:29<00:02,  2.73s/iter, val_loss=0.00129, val_og_loss=0.0013]
+
 95%|█████████▌| 19/20 [01:32<00:02,  2.73s/iter, val_loss=0.00128, val_og_loss=0.0013]
+
100%|██████████| 20/20 [01:32<00:00,  2.73s/iter, val_loss=0.00128, val_og_loss=0.0013]
100%|██████████| 20/20 [01:32<00:00,  4.62s/iter, val_loss=0.00128, val_og_loss=0.0013]
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:42:53<2:27:14, 883.46s/it]INFO:ParoQuant:Best val loss: 0.0012822382850572467, Original val loss: 0.0013045994564890862
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:42:59<2:27:14, 883.46s/it]2025-12-22 19:13:11,712 - INFO - Best val loss: 0.0012822382850572467, Original val loss: 0.0013045994564890862
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:39<?, ?iter/s, val_loss=0.000686, val_og_loss=0.0007]
+
  5%|▌         | 1/20 [00:39<12:28, 39.40s/iter, val_loss=0.000686, val_og_loss=0.0007]
+
  5%|▌         | 1/20 [00:42<12:28, 39.40s/iter, val_loss=0.00069, val_og_loss=0.0007] 
+
 10%|█         | 2/20 [00:42<05:20, 17.82s/iter, val_loss=0.00069, val_og_loss=0.0007]
+
 10%|█         | 2/20 [00:44<05:20, 17.82s/iter, val_loss=0.000692, val_og_loss=0.0007]
+
 15%|█▌        | 3/20 [00:44<03:05, 10.93s/iter, val_loss=0.000692, val_og_loss=0.0007]
+
 15%|█▌        | 3/20 [00:47<03:05, 10.93s/iter, val_loss=0.000697, val_og_loss=0.0007]
+
 20%|██        | 4/20 [00:47<02:02,  7.69s/iter, val_loss=0.000697, val_og_loss=0.0007]
+
 20%|██        | 4/20 [00:50<02:02,  7.69s/iter, val_loss=0.000696, val_og_loss=0.0007]
+
 25%|██▌       | 5/20 [00:50<01:28,  5.89s/iter, val_loss=0.000696, val_og_loss=0.0007]
+
 25%|██▌       | 5/20 [00:53<01:28,  5.89s/iter, val_loss=0.000696, val_og_loss=0.0007]
+
 30%|███       | 6/20 [00:53<01:07,  4.82s/iter, val_loss=0.000696, val_og_loss=0.0007]
+
 30%|███       | 6/20 [00:55<01:07,  4.82s/iter, val_loss=0.000695, val_og_loss=0.0007]
+
 35%|███▌      | 7/20 [00:55<00:53,  4.13s/iter, val_loss=0.000695, val_og_loss=0.0007]
+
 35%|███▌      | 7/20 [00:58<00:53,  4.13s/iter, val_loss=0.000695, val_og_loss=0.0007]
+
 40%|████      | 8/20 [00:58<00:44,  3.68s/iter, val_loss=0.000695, val_og_loss=0.0007]
+
 40%|████      | 8/20 [01:01<00:44,  3.68s/iter, val_loss=0.000695, val_og_loss=0.0007]
+
 45%|████▌     | 9/20 [01:01<00:37,  3.38s/iter, val_loss=0.000695, val_og_loss=0.0007]
+
 45%|████▌     | 9/20 [01:03<00:37,  3.38s/iter, val_loss=0.000693, val_og_loss=0.0007]
+
 50%|█████     | 10/20 [01:03<00:31,  3.17s/iter, val_loss=0.000693, val_og_loss=0.0007]
+
 50%|█████     | 10/20 [01:06<00:31,  3.17s/iter, val_loss=0.000692, val_og_loss=0.0007]
+
 55%|█████▌    | 11/20 [01:06<00:27,  3.03s/iter, val_loss=0.000692, val_og_loss=0.0007]
+
 55%|█████▌    | 11/20 [01:09<00:27,  3.03s/iter, val_loss=0.000692, val_og_loss=0.0007]
+
 60%|██████    | 12/20 [01:09<00:23,  2.93s/iter, val_loss=0.000692, val_og_loss=0.0007]
+
 60%|██████    | 12/20 [01:11<00:23,  2.93s/iter, val_loss=0.000692, val_og_loss=0.0007]
+
 65%|██████▌   | 13/20 [01:11<00:20,  2.87s/iter, val_loss=0.000692, val_og_loss=0.0007]
+
 65%|██████▌   | 13/20 [01:14<00:20,  2.87s/iter, val_loss=0.000691, val_og_loss=0.0007]
+
 70%|███████   | 14/20 [01:14<00:16,  2.82s/iter, val_loss=0.000691, val_og_loss=0.0007]
+
 70%|███████   | 14/20 [01:17<00:16,  2.82s/iter, val_loss=0.000691, val_og_loss=0.0007]
+
 75%|███████▌  | 15/20 [01:17<00:13,  2.79s/iter, val_loss=0.000691, val_og_loss=0.0007]
+
 75%|███████▌  | 15/20 [01:20<00:13,  2.79s/iter, val_loss=0.00069, val_og_loss=0.0007] 
+
 80%|████████  | 16/20 [01:20<00:11,  2.77s/iter, val_loss=0.00069, val_og_loss=0.0007]
+
 80%|████████  | 16/20 [01:22<00:11,  2.77s/iter, val_loss=0.000689, val_og_loss=0.0007]
+
 85%|████████▌ | 17/20 [01:22<00:08,  2.75s/iter, val_loss=0.000689, val_og_loss=0.0007]
+
 85%|████████▌ | 17/20 [01:25<00:08,  2.75s/iter, val_loss=0.000689, val_og_loss=0.0007]
+
 90%|█████████ | 18/20 [01:25<00:05,  2.74s/iter, val_loss=0.000689, val_og_loss=0.0007]
+
 90%|█████████ | 18/20 [01:28<00:05,  2.74s/iter, val_loss=0.000688, val_og_loss=0.0007]
+
 95%|█████████▌| 19/20 [01:28<00:02,  2.73s/iter, val_loss=0.000688, val_og_loss=0.0007]
+
 95%|█████████▌| 19/20 [01:30<00:02,  2.73s/iter, val_loss=0.000688, val_og_loss=0.0007]
+
100%|██████████| 20/20 [01:30<00:00,  2.73s/iter, val_loss=0.000688, val_og_loss=0.0007]
100%|██████████| 20/20 [01:30<00:00,  4.55s/iter, val_loss=0.000688, val_og_loss=0.0007]
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:44:30<2:27:14, 883.46s/it]INFO:ParoQuant:Best val loss: 0.0006861991132609546, Original val loss: 0.0006996601587161422
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:44:32<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:44:32<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:45:33<2:27:14, 883.46s/it]2025-12-22 19:14:48,696 - INFO - Best val loss: 0.0006861991132609546, Original val loss: 0.0006996601587161422
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:39<?, ?iter/s, val_loss=0.00013, val_og_loss=0.000133]
+
  5%|▌         | 1/20 [00:39<12:30, 39.50s/iter, val_loss=0.00013, val_og_loss=0.000133]
+
  5%|▌         | 1/20 [00:44<12:30, 39.50s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 10%|█         | 2/20 [00:44<05:50, 19.45s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 10%|█         | 2/20 [00:50<05:50, 19.45s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 15%|█▌        | 3/20 [00:50<03:41, 13.04s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 15%|█▌        | 3/20 [00:55<03:41, 13.04s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 20%|██        | 4/20 [00:55<02:40, 10.04s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 20%|██        | 4/20 [01:01<02:40, 10.04s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 25%|██▌       | 5/20 [01:01<02:05,  8.37s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 25%|██▌       | 5/20 [01:06<02:05,  8.37s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 30%|███       | 6/20 [01:06<01:43,  7.38s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 30%|███       | 6/20 [01:12<01:43,  7.38s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 35%|███▌      | 7/20 [01:12<01:27,  6.74s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 35%|███▌      | 7/20 [01:17<01:27,  6.74s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 40%|████      | 8/20 [01:17<01:15,  6.32s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 40%|████      | 8/20 [01:22<01:15,  6.32s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 45%|████▌     | 9/20 [01:22<01:06,  6.04s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 45%|████▌     | 9/20 [01:28<01:06,  6.04s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 50%|█████     | 10/20 [01:28<00:58,  5.85s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 50%|█████     | 10/20 [01:33<00:58,  5.85s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 55%|█████▌    | 11/20 [01:33<00:51,  5.72s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 55%|█████▌    | 11/20 [01:39<00:51,  5.72s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 60%|██████    | 12/20 [01:39<00:45,  5.63s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 60%|██████    | 12/20 [01:44<00:45,  5.63s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.56s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 65%|██████▌   | 13/20 [01:50<00:38,  5.56s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 70%|███████   | 14/20 [01:50<00:33,  5.53s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 70%|███████   | 14/20 [01:55<00:33,  5.53s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 75%|███████▌  | 15/20 [01:55<00:27,  5.50s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 75%|███████▌  | 15/20 [02:00<00:27,  5.50s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 80%|████████  | 16/20 [02:00<00:21,  5.48s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 80%|████████  | 16/20 [02:06<00:21,  5.48s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 85%|████████▌ | 17/20 [02:06<00:16,  5.46s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 85%|████████▌ | 17/20 [02:11<00:16,  5.46s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.45s/iter, val_loss=0.000131, val_og_loss=0.000133]
+
 90%|█████████ | 18/20 [02:17<00:10,  5.45s/iter, val_loss=0.00013, val_og_loss=0.000133] 
+
 95%|█████████▌| 19/20 [02:17<00:05,  5.44s/iter, val_loss=0.00013, val_og_loss=0.000133]
+
 95%|█████████▌| 19/20 [02:22<00:05,  5.44s/iter, val_loss=0.00013, val_og_loss=0.000133]
+
100%|██████████| 20/20 [02:22<00:00,  5.44s/iter, val_loss=0.00013, val_og_loss=0.000133]
100%|██████████| 20/20 [02:22<00:00,  7.13s/iter, val_loss=0.00013, val_og_loss=0.000133]
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:47:55<2:27:14, 883.46s/it]INFO:ParoQuant:Best val loss: 0.00013009062968194485, Original val loss: 0.00013254091027192771
+
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:47:56<2:27:14, 883.46s/it]
                                                                       

Processing Blocks:  69%|██████▉   | 22/32 [4:48:07<2:27:14, 883.46s/it]
Processing Blocks:  72%|███████▏  | 23/32 [4:48:45<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:48:45<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:49:03<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:49:03<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:51:20<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:51:20<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:52:25<2:15:26, 902.95s/it]2025-12-22 19:18:13,886 - INFO - Best val loss: 0.00013009062968194485, Original val loss: 0.00013254091027192771
+[INFO] Saving Layer 22 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 23...
+[INFO] Starting Sequential Training for Layer 23
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.00241, val_og_loss=0.00279]
+
  5%|▌         | 1/20 [00:16<05:13, 16.48s/iter, val_loss=0.00241, val_og_loss=0.00279]
+
  5%|▌         | 1/20 [00:18<05:13, 16.48s/iter, val_loss=0.00245, val_og_loss=0.00279]
+
 10%|█         | 2/20 [00:18<02:20,  7.78s/iter, val_loss=0.00245, val_og_loss=0.00279]
+
 10%|█         | 2/20 [00:19<02:20,  7.78s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 15%|█▌        | 3/20 [00:19<01:25,  5.00s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 15%|█▌        | 3/20 [00:21<01:25,  5.00s/iter, val_loss=0.00247, val_og_loss=0.00279]
+
 20%|██        | 4/20 [00:21<00:59,  3.70s/iter, val_loss=0.00247, val_og_loss=0.00279]
+
 20%|██        | 4/20 [00:23<00:59,  3.70s/iter, val_loss=0.00247, val_og_loss=0.00279]
+
 25%|██▌       | 5/20 [00:23<00:44,  2.97s/iter, val_loss=0.00247, val_og_loss=0.00279]
+
 25%|██▌       | 5/20 [00:24<00:44,  2.97s/iter, val_loss=0.00247, val_og_loss=0.00279]
+
 30%|███       | 6/20 [00:24<00:35,  2.54s/iter, val_loss=0.00247, val_og_loss=0.00279]
+
 30%|███       | 6/20 [00:26<00:35,  2.54s/iter, val_loss=0.00247, val_og_loss=0.00279]
+
 35%|███▌      | 7/20 [00:26<00:29,  2.27s/iter, val_loss=0.00247, val_og_loss=0.00279]
+
 35%|███▌      | 7/20 [00:28<00:29,  2.27s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 40%|████      | 8/20 [00:28<00:24,  2.08s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 40%|████      | 8/20 [00:30<00:24,  2.08s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.96s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.96s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 50%|█████     | 10/20 [00:31<00:18,  1.87s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 50%|█████     | 10/20 [00:33<00:18,  1.87s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.82s/iter, val_loss=0.00246, val_og_loss=0.00279]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.82s/iter, val_loss=0.00245, val_og_loss=0.00279]
+
 60%|██████    | 12/20 [00:35<00:14,  1.77s/iter, val_loss=0.00245, val_og_loss=0.00279]
+
 60%|██████    | 12/20 [00:36<00:14,  1.77s/iter, val_loss=0.00244, val_og_loss=0.00279]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.74s/iter, val_loss=0.00244, val_og_loss=0.00279]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.74s/iter, val_loss=0.00244, val_og_loss=0.00279]
+
 70%|███████   | 14/20 [00:38<00:10,  1.72s/iter, val_loss=0.00244, val_og_loss=0.00279]
+
 70%|███████   | 14/20 [00:40<00:10,  1.72s/iter, val_loss=0.00245, val_og_loss=0.00279]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.71s/iter, val_loss=0.00245, val_og_loss=0.00279]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.71s/iter, val_loss=0.00245, val_og_loss=0.00279]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=0.00245, val_og_loss=0.00279]
+
 80%|████████  | 16/20 [00:43<00:06,  1.70s/iter, val_loss=0.00244, val_og_loss=0.00279]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.69s/iter, val_loss=0.00244, val_og_loss=0.00279]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.69s/iter, val_loss=0.00243, val_og_loss=0.00279]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.69s/iter, val_loss=0.00243, val_og_loss=0.00279]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.69s/iter, val_loss=0.00242, val_og_loss=0.00279]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.68s/iter, val_loss=0.00242, val_og_loss=0.00279]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.68s/iter, val_loss=0.00242, val_og_loss=0.00279]
+
100%|██████████| 20/20 [00:48<00:00,  1.69s/iter, val_loss=0.00242, val_og_loss=0.00279]
100%|██████████| 20/20 [00:48<00:00,  2.42s/iter, val_loss=0.00242, val_og_loss=0.00279]
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:53:13<2:15:26, 902.95s/it]INFO:ParoQuant:Best val loss: 0.0024137385189533234, Original val loss: 0.0027934787794947624
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:53:16<2:15:26, 902.95s/it]2025-12-22 19:23:31,718 - INFO - Best val loss: 0.0024137385189533234, Original val loss: 0.0027934787794947624
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.00619, val_og_loss=0.00696]
+
  5%|▌         | 1/20 [00:16<05:20, 16.85s/iter, val_loss=0.00619, val_og_loss=0.00696]
+
  5%|▌         | 1/20 [00:18<05:20, 16.85s/iter, val_loss=0.00631, val_og_loss=0.00696]
+
 10%|█         | 2/20 [00:18<02:19,  7.76s/iter, val_loss=0.00631, val_og_loss=0.00696]
+
 10%|█         | 2/20 [00:19<02:19,  7.76s/iter, val_loss=0.00633, val_og_loss=0.00696]
+
 15%|█▌        | 3/20 [00:19<01:22,  4.85s/iter, val_loss=0.00633, val_og_loss=0.00696]
+
 15%|█▌        | 3/20 [00:21<01:22,  4.85s/iter, val_loss=0.00632, val_og_loss=0.00696]
+
 20%|██        | 4/20 [00:21<00:55,  3.48s/iter, val_loss=0.00632, val_og_loss=0.00696]
+
 20%|██        | 4/20 [00:22<00:55,  3.48s/iter, val_loss=0.00634, val_og_loss=0.00696]
+
 25%|██▌       | 5/20 [00:22<00:40,  2.73s/iter, val_loss=0.00634, val_og_loss=0.00696]
+
 25%|██▌       | 5/20 [00:23<00:40,  2.73s/iter, val_loss=0.00631, val_og_loss=0.00696]
+
 30%|███       | 6/20 [00:23<00:31,  2.27s/iter, val_loss=0.00631, val_og_loss=0.00696]
+
 30%|███       | 6/20 [00:25<00:31,  2.27s/iter, val_loss=0.0063, val_og_loss=0.00696] 
+
 35%|███▌      | 7/20 [00:25<00:25,  1.99s/iter, val_loss=0.0063, val_og_loss=0.00696]
+
 35%|███▌      | 7/20 [00:26<00:25,  1.99s/iter, val_loss=0.00631, val_og_loss=0.00696]
+
 40%|████      | 8/20 [00:26<00:21,  1.81s/iter, val_loss=0.00631, val_og_loss=0.00696]
+
 40%|████      | 8/20 [00:28<00:21,  1.81s/iter, val_loss=0.00632, val_og_loss=0.00696]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.68s/iter, val_loss=0.00632, val_og_loss=0.00696]
+
 45%|████▌     | 9/20 [00:29<00:18,  1.68s/iter, val_loss=0.00631, val_og_loss=0.00696]
+
 50%|█████     | 10/20 [00:29<00:15,  1.59s/iter, val_loss=0.00631, val_og_loss=0.00696]
+
 50%|█████     | 10/20 [00:30<00:15,  1.59s/iter, val_loss=0.00633, val_og_loss=0.00696]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.53s/iter, val_loss=0.00633, val_og_loss=0.00696]
+
 55%|█████▌    | 11/20 [00:32<00:13,  1.53s/iter, val_loss=0.00626, val_og_loss=0.00696]
+
 60%|██████    | 12/20 [00:32<00:11,  1.49s/iter, val_loss=0.00626, val_og_loss=0.00696]
+
 60%|██████    | 12/20 [00:33<00:11,  1.49s/iter, val_loss=0.00624, val_og_loss=0.00696]
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.47s/iter, val_loss=0.00624, val_og_loss=0.00696]
+
 65%|██████▌   | 13/20 [00:35<00:10,  1.47s/iter, val_loss=0.00626, val_og_loss=0.00696]
+
 70%|███████   | 14/20 [00:35<00:08,  1.45s/iter, val_loss=0.00626, val_og_loss=0.00696]
+
 70%|███████   | 14/20 [00:36<00:08,  1.45s/iter, val_loss=0.00628, val_og_loss=0.00696]
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.43s/iter, val_loss=0.00628, val_og_loss=0.00696]
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.43s/iter, val_loss=0.00623, val_og_loss=0.00696]
+
 80%|████████  | 16/20 [00:37<00:05,  1.42s/iter, val_loss=0.00623, val_og_loss=0.00696]
+
 80%|████████  | 16/20 [00:39<00:05,  1.42s/iter, val_loss=0.00623, val_og_loss=0.00696]
+
 85%|████████▌ | 17/20 [00:39<00:04,  1.41s/iter, val_loss=0.00623, val_og_loss=0.00696]
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.41s/iter, val_loss=0.00622, val_og_loss=0.00696]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.41s/iter, val_loss=0.00622, val_og_loss=0.00696]
+
 90%|█████████ | 18/20 [00:41<00:02,  1.41s/iter, val_loss=0.00619, val_og_loss=0.00696]
+
 95%|█████████▌| 19/20 [00:41<00:01,  1.40s/iter, val_loss=0.00619, val_og_loss=0.00696]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.40s/iter, val_loss=0.00621, val_og_loss=0.00696]
+
100%|██████████| 20/20 [00:43<00:00,  1.40s/iter, val_loss=0.00621, val_og_loss=0.00696]
100%|██████████| 20/20 [00:43<00:00,  2.17s/iter, val_loss=0.00621, val_og_loss=0.00696]
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:54:00<2:15:26, 902.95s/it]INFO:ParoQuant:Best val loss: 0.006185045465826988, Original val loss: 0.006956971250474453
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:54:03<2:15:26, 902.95s/it]2025-12-22 19:24:18,278 - INFO - Best val loss: 0.006185045465826988, Original val loss: 0.006956971250474453
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:17<?, ?iter/s, val_loss=0.000849, val_og_loss=0.00102]
+
  5%|▌         | 1/20 [00:17<05:29, 17.33s/iter, val_loss=0.000849, val_og_loss=0.00102]
+
  5%|▌         | 1/20 [00:18<05:29, 17.33s/iter, val_loss=0.000851, val_og_loss=0.00102]
+
 10%|█         | 2/20 [00:18<02:23,  7.96s/iter, val_loss=0.000851, val_og_loss=0.00102]
+
 10%|█         | 2/20 [00:20<02:23,  7.96s/iter, val_loss=0.000852, val_og_loss=0.00102]
+
 15%|█▌        | 3/20 [00:20<01:24,  4.96s/iter, val_loss=0.000852, val_og_loss=0.00102]
+
 15%|█▌        | 3/20 [00:21<01:24,  4.96s/iter, val_loss=0.000853, val_og_loss=0.00102]
+
 20%|██        | 4/20 [00:21<00:56,  3.55s/iter, val_loss=0.000853, val_og_loss=0.00102]
+
 20%|██        | 4/20 [00:22<00:56,  3.55s/iter, val_loss=0.000852, val_og_loss=0.00102]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.77s/iter, val_loss=0.000852, val_og_loss=0.00102]
+
 25%|██▌       | 5/20 [00:24<00:41,  2.77s/iter, val_loss=0.000855, val_og_loss=0.00102]
+
 30%|███       | 6/20 [00:24<00:32,  2.31s/iter, val_loss=0.000855, val_og_loss=0.00102]
+
 30%|███       | 6/20 [00:25<00:32,  2.31s/iter, val_loss=0.000854, val_og_loss=0.00102]
+
 35%|███▌      | 7/20 [00:25<00:26,  2.01s/iter, val_loss=0.000854, val_og_loss=0.00102]
+
 35%|███▌      | 7/20 [00:27<00:26,  2.01s/iter, val_loss=0.000853, val_og_loss=0.00102]
+
 40%|████      | 8/20 [00:27<00:21,  1.81s/iter, val_loss=0.000853, val_og_loss=0.00102]
+
 40%|████      | 8/20 [00:28<00:21,  1.81s/iter, val_loss=0.000852, val_og_loss=0.00102]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.68s/iter, val_loss=0.000852, val_og_loss=0.00102]
+
 45%|████▌     | 9/20 [00:29<00:18,  1.68s/iter, val_loss=0.00085, val_og_loss=0.00102] 
+
 50%|█████     | 10/20 [00:29<00:15,  1.59s/iter, val_loss=0.00085, val_og_loss=0.00102]
+
 50%|█████     | 10/20 [00:31<00:15,  1.59s/iter, val_loss=0.000848, val_og_loss=0.00102]
+
 55%|█████▌    | 11/20 [00:31<00:13,  1.54s/iter, val_loss=0.000848, val_og_loss=0.00102]
+
 55%|█████▌    | 11/20 [00:32<00:13,  1.54s/iter, val_loss=0.000846, val_og_loss=0.00102]
+
 60%|██████    | 12/20 [00:32<00:11,  1.50s/iter, val_loss=0.000846, val_og_loss=0.00102]
+
 60%|██████    | 12/20 [00:34<00:11,  1.50s/iter, val_loss=0.000845, val_og_loss=0.00102]
+
 65%|██████▌   | 13/20 [00:34<00:10,  1.46s/iter, val_loss=0.000845, val_og_loss=0.00102]
+
 65%|██████▌   | 13/20 [00:35<00:10,  1.46s/iter, val_loss=0.000844, val_og_loss=0.00102]
+
 70%|███████   | 14/20 [00:35<00:08,  1.44s/iter, val_loss=0.000844, val_og_loss=0.00102]
+
 70%|███████   | 14/20 [00:36<00:08,  1.44s/iter, val_loss=0.000843, val_og_loss=0.00102]
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.42s/iter, val_loss=0.000843, val_og_loss=0.00102]
+
 75%|███████▌  | 15/20 [00:38<00:07,  1.42s/iter, val_loss=0.00084, val_og_loss=0.00102] 
+
 80%|████████  | 16/20 [00:38<00:05,  1.42s/iter, val_loss=0.00084, val_og_loss=0.00102]
+
 80%|████████  | 16/20 [00:39<00:05,  1.42s/iter, val_loss=0.000841, val_og_loss=0.00102]
+
 85%|████████▌ | 17/20 [00:39<00:04,  1.41s/iter, val_loss=0.000841, val_og_loss=0.00102]
+
 85%|████████▌ | 17/20 [00:41<00:04,  1.41s/iter, val_loss=0.000841, val_og_loss=0.00102]
+
 90%|█████████ | 18/20 [00:41<00:02,  1.40s/iter, val_loss=0.000841, val_og_loss=0.00102]
+
 90%|█████████ | 18/20 [00:42<00:02,  1.40s/iter, val_loss=0.00084, val_og_loss=0.00102] 
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.40s/iter, val_loss=0.00084, val_og_loss=0.00102]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.40s/iter, val_loss=0.000838, val_og_loss=0.00102]
+
100%|██████████| 20/20 [00:43<00:00,  1.39s/iter, val_loss=0.000838, val_og_loss=0.00102]
100%|██████████| 20/20 [00:43<00:00,  2.19s/iter, val_loss=0.000838, val_og_loss=0.00102]
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:54:46<2:15:26, 902.95s/it]INFO:ParoQuant:Best val loss: 0.000838392530567944, Original val loss: 0.0010171581525355577
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:54:49<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:54:49<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:55:22<2:15:26, 902.95s/it]2025-12-22 19:25:04,873 - INFO - Best val loss: 0.000838392530567944, Original val loss: 0.0010171581525355577
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=2.52e-5, val_og_loss=2.56e-5]
+
  5%|▌         | 1/20 [00:16<05:10, 16.36s/iter, val_loss=2.52e-5, val_og_loss=2.56e-5]
+
  5%|▌         | 1/20 [00:18<05:10, 16.36s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 10%|█         | 2/20 [00:18<02:19,  7.74s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 10%|█         | 2/20 [00:19<02:19,  7.74s/iter, val_loss=2.52e-5, val_og_loss=2.56e-5]
+
 15%|█▌        | 3/20 [00:19<01:24,  4.97s/iter, val_loss=2.52e-5, val_og_loss=2.56e-5]
+
 15%|█▌        | 3/20 [00:21<01:24,  4.97s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 20%|██        | 4/20 [00:21<00:58,  3.67s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 20%|██        | 4/20 [00:23<00:58,  3.67s/iter, val_loss=2.54e-5, val_og_loss=2.56e-5]
+
 25%|██▌       | 5/20 [00:23<00:44,  2.95s/iter, val_loss=2.54e-5, val_og_loss=2.56e-5]
+
 25%|██▌       | 5/20 [00:24<00:44,  2.95s/iter, val_loss=2.54e-5, val_og_loss=2.56e-5]
+
 30%|███       | 6/20 [00:24<00:35,  2.52s/iter, val_loss=2.54e-5, val_og_loss=2.56e-5]
+
 30%|███       | 6/20 [00:26<00:35,  2.52s/iter, val_loss=2.54e-5, val_og_loss=2.56e-5]
+
 35%|███▌      | 7/20 [00:26<00:29,  2.25s/iter, val_loss=2.54e-5, val_og_loss=2.56e-5]
+
 35%|███▌      | 7/20 [00:28<00:29,  2.25s/iter, val_loss=2.54e-5, val_og_loss=2.56e-5]
+
 40%|████      | 8/20 [00:28<00:24,  2.06s/iter, val_loss=2.54e-5, val_og_loss=2.56e-5]
+
 40%|████      | 8/20 [00:29<00:24,  2.06s/iter, val_loss=2.52e-5, val_og_loss=2.56e-5]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.95s/iter, val_loss=2.52e-5, val_og_loss=2.56e-5]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.95s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 50%|█████     | 10/20 [00:31<00:18,  1.86s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 50%|█████     | 10/20 [00:33<00:18,  1.86s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.81s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.81s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 60%|██████    | 12/20 [00:34<00:14,  1.77s/iter, val_loss=2.53e-5, val_og_loss=2.56e-5]
+
 60%|██████    | 12/20 [00:36<00:14,  1.77s/iter, val_loss=2.52e-5, val_og_loss=2.56e-5]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.75s/iter, val_loss=2.52e-5, val_og_loss=2.56e-5]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.75s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 70%|███████   | 14/20 [00:38<00:10,  1.73s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 70%|███████   | 14/20 [00:39<00:10,  1.73s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.71s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 80%|████████  | 16/20 [00:41<00:06,  1.71s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 80%|████████  | 16/20 [00:43<00:06,  1.71s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.71s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.71s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.70s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.70s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.69s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
100%|██████████| 20/20 [00:48<00:00,  1.68s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
100%|██████████| 20/20 [00:48<00:00,  2.42s/iter, val_loss=2.51e-5, val_og_loss=2.56e-5]
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:56:10<2:15:26, 902.95s/it]INFO:ParoQuant:Best val loss: 2.5069435650948435e-05, Original val loss: 2.560203938628547e-05
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:56:11<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:56:11<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:56:55<2:15:26, 902.95s/it]2025-12-22 19:26:28,899 - INFO - Best val loss: 2.5069435650948435e-05, Original val loss: 2.560203938628547e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00133, val_og_loss=0.00136]
+
  5%|▌         | 1/20 [00:29<09:13, 29.12s/iter, val_loss=0.00133, val_og_loss=0.00136]
+
  5%|▌         | 1/20 [00:31<09:13, 29.12s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 10%|█         | 2/20 [00:31<04:04, 13.58s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 10%|█         | 2/20 [00:34<04:04, 13.58s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 15%|█▌        | 3/20 [00:34<02:26,  8.62s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 15%|█▌        | 3/20 [00:37<02:26,  8.62s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 20%|██        | 4/20 [00:37<01:40,  6.29s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 20%|██        | 4/20 [00:39<01:40,  6.29s/iter, val_loss=0.00136, val_og_loss=0.00136]
+
 25%|██▌       | 5/20 [00:39<01:14,  5.00s/iter, val_loss=0.00136, val_og_loss=0.00136]
+
 25%|██▌       | 5/20 [00:42<01:14,  5.00s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 30%|███       | 6/20 [00:42<00:59,  4.22s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 30%|███       | 6/20 [00:45<00:59,  4.22s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 35%|███▌      | 7/20 [00:45<00:48,  3.73s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.73s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 40%|████      | 8/20 [00:48<00:40,  3.40s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 40%|████      | 8/20 [00:50<00:40,  3.40s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 45%|████▌     | 9/20 [00:50<00:35,  3.20s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.20s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 50%|█████     | 10/20 [00:53<00:30,  3.05s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 50%|█████     | 10/20 [00:56<00:30,  3.05s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.94s/iter, val_loss=0.00135, val_og_loss=0.00136]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.94s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 60%|██████    | 12/20 [00:58<00:22,  2.87s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 60%|██████    | 12/20 [01:01<00:22,  2.87s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 65%|██████▌   | 13/20 [01:01<00:19,  2.82s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.82s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 70%|███████   | 14/20 [01:04<00:16,  2.79s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 70%|███████   | 14/20 [01:07<00:16,  2.79s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.77s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 80%|████████  | 16/20 [01:09<00:11,  2.75s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 80%|████████  | 16/20 [01:12<00:11,  2.75s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 85%|████████▌ | 17/20 [01:12<00:08,  2.74s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.73s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.73s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 95%|█████████▌| 19/20 [01:17<00:02,  2.73s/iter, val_loss=0.00134, val_og_loss=0.00136]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.00133, val_og_loss=0.00136]
+
100%|██████████| 20/20 [01:20<00:00,  2.72s/iter, val_loss=0.00133, val_og_loss=0.00136]
100%|██████████| 20/20 [01:20<00:00,  4.03s/iter, val_loss=0.00133, val_og_loss=0.00136]
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:58:16<2:15:26, 902.95s/it]INFO:ParoQuant:Best val loss: 0.0013324746396392584, Original val loss: 0.0013596996432170272
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:58:22<2:15:26, 902.95s/it]2025-12-22 19:28:34,660 - INFO - Best val loss: 0.0013324746396392584, Original val loss: 0.0013596996432170272
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.000717, val_og_loss=0.000732]
+
  5%|▌         | 1/20 [00:31<10:00, 31.60s/iter, val_loss=0.000717, val_og_loss=0.000732]
+
  5%|▌         | 1/20 [00:34<10:00, 31.60s/iter, val_loss=0.000722, val_og_loss=0.000732]
+
 10%|█         | 2/20 [00:34<04:22, 14.61s/iter, val_loss=0.000722, val_og_loss=0.000732]
+
 10%|█         | 2/20 [00:37<04:22, 14.61s/iter, val_loss=0.000725, val_og_loss=0.000732]
+
 15%|█▌        | 3/20 [00:37<02:36,  9.18s/iter, val_loss=0.000725, val_og_loss=0.000732]
+
 15%|█▌        | 3/20 [00:39<02:36,  9.18s/iter, val_loss=0.00073, val_og_loss=0.000732] 
+
 20%|██        | 4/20 [00:39<01:46,  6.63s/iter, val_loss=0.00073, val_og_loss=0.000732]
+
 20%|██        | 4/20 [00:42<01:46,  6.63s/iter, val_loss=0.000733, val_og_loss=0.000732]
+
 25%|██▌       | 5/20 [00:42<01:18,  5.22s/iter, val_loss=0.000733, val_og_loss=0.000732]
+
 25%|██▌       | 5/20 [00:45<01:18,  5.22s/iter, val_loss=0.000734, val_og_loss=0.000732]
+
 30%|███       | 6/20 [00:45<01:01,  4.37s/iter, val_loss=0.000734, val_og_loss=0.000732]
+
 30%|███       | 6/20 [00:47<01:01,  4.37s/iter, val_loss=0.000733, val_og_loss=0.000732]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.83s/iter, val_loss=0.000733, val_og_loss=0.000732]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.83s/iter, val_loss=0.000734, val_og_loss=0.000732]
+
 40%|████      | 8/20 [00:50<00:41,  3.48s/iter, val_loss=0.000734, val_og_loss=0.000732]
+
 40%|████      | 8/20 [00:53<00:41,  3.48s/iter, val_loss=0.000731, val_og_loss=0.000732]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.24s/iter, val_loss=0.000731, val_og_loss=0.000732]
+
 45%|████▌     | 9/20 [00:56<00:35,  3.24s/iter, val_loss=0.00073, val_og_loss=0.000732] 
+
 50%|█████     | 10/20 [00:56<00:30,  3.08s/iter, val_loss=0.00073, val_og_loss=0.000732]
+
 50%|█████     | 10/20 [00:58<00:30,  3.08s/iter, val_loss=0.000729, val_og_loss=0.000732]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.96s/iter, val_loss=0.000729, val_og_loss=0.000732]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.96s/iter, val_loss=0.000728, val_og_loss=0.000732]
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.000728, val_og_loss=0.000732]
+
 60%|██████    | 12/20 [01:04<00:23,  2.89s/iter, val_loss=0.000728, val_og_loss=0.000732]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.000728, val_og_loss=0.000732]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000727, val_og_loss=0.000732]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.000727, val_og_loss=0.000732]
+
 70%|███████   | 14/20 [01:09<00:16,  2.80s/iter, val_loss=0.000726, val_og_loss=0.000732]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.77s/iter, val_loss=0.000726, val_og_loss=0.000732]
+
 75%|███████▌  | 15/20 [01:12<00:13,  2.77s/iter, val_loss=0.000725, val_og_loss=0.000732]
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.000725, val_og_loss=0.000732]
+
 80%|████████  | 16/20 [01:15<00:11,  2.76s/iter, val_loss=0.000724, val_og_loss=0.000732]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.000724, val_og_loss=0.000732]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.74s/iter, val_loss=0.000724, val_og_loss=0.000732]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.73s/iter, val_loss=0.000724, val_og_loss=0.000732]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.73s/iter, val_loss=0.000723, val_og_loss=0.000732]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.000723, val_og_loss=0.000732]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.73s/iter, val_loss=0.000723, val_og_loss=0.000732]
+
100%|██████████| 20/20 [01:23<00:00,  2.72s/iter, val_loss=0.000723, val_og_loss=0.000732]
100%|██████████| 20/20 [01:23<00:00,  4.16s/iter, val_loss=0.000723, val_og_loss=0.000732]
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:59:45<2:15:26, 902.95s/it]INFO:ParoQuant:Best val loss: 0.000717475893907249, Original val loss: 0.0007324873586185277
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:59:47<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [4:59:47<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [5:00:46<2:15:26, 902.95s/it]2025-12-22 19:30:03,571 - INFO - Best val loss: 0.000717475893907249, Original val loss: 0.0007324873586185277
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.000138, val_og_loss=0.000141]
+
  5%|▌         | 1/20 [00:33<10:37, 33.55s/iter, val_loss=0.000138, val_og_loss=0.000141]
+
  5%|▌         | 1/20 [00:38<10:37, 33.55s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 10%|█         | 2/20 [00:38<05:06, 17.00s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 10%|█         | 2/20 [00:44<05:06, 17.00s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 15%|█▌        | 3/20 [00:44<03:19, 11.72s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 15%|█▌        | 3/20 [00:49<03:19, 11.72s/iter, val_loss=0.00014, val_og_loss=0.000141] 
+
 20%|██        | 4/20 [00:49<02:27,  9.24s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 20%|██        | 4/20 [00:55<02:27,  9.24s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 25%|██▌       | 5/20 [00:55<01:57,  7.86s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 25%|██▌       | 5/20 [01:00<01:57,  7.86s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 30%|███       | 6/20 [01:00<01:38,  7.03s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 30%|███       | 6/20 [01:06<01:38,  7.03s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.50s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.50s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 40%|████      | 8/20 [01:11<01:13,  6.16s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 40%|████      | 8/20 [01:16<01:13,  6.16s/iter, val_loss=0.00014, val_og_loss=0.000141] 
+
 45%|████▌     | 9/20 [01:16<01:05,  5.93s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.93s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 50%|█████     | 10/20 [01:22<00:57,  5.77s/iter, val_loss=0.00014, val_og_loss=0.000141]
+
 50%|█████     | 10/20 [01:27<00:57,  5.77s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 55%|█████▌    | 11/20 [01:27<00:50,  5.67s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 55%|█████▌    | 11/20 [01:33<00:50,  5.67s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 60%|██████    | 12/20 [01:33<00:44,  5.59s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 60%|██████    | 12/20 [01:38<00:44,  5.59s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.54s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.54s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 70%|███████   | 14/20 [01:44<00:33,  5.51s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 70%|███████   | 14/20 [01:49<00:33,  5.51s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.49s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 75%|███████▌  | 15/20 [01:54<00:27,  5.49s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 80%|████████  | 16/20 [01:54<00:21,  5.46s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 80%|████████  | 16/20 [02:00<00:21,  5.46s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.45s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.45s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.44s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.44s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
100%|██████████| 20/20 [02:16<00:00,  5.43s/iter, val_loss=0.000139, val_og_loss=0.000141]
100%|██████████| 20/20 [02:16<00:00,  6.83s/iter, val_loss=0.000139, val_og_loss=0.000141]
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [5:03:03<2:15:26, 902.95s/it]INFO:ParoQuant:Best val loss: 0.0001381903130095452, Original val loss: 0.0001409030519425869
+
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [5:03:04<2:15:26, 902.95s/it]
                                                                       

Processing Blocks:  72%|███████▏  | 23/32 [5:03:15<2:15:26, 902.95s/it]
Processing Blocks:  75%|███████▌  | 24/32 [5:03:53<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:03:54<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:04:12<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:04:12<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:06:26<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:06:26<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:07:28<2:00:36, 904.50s/it]2025-12-22 19:33:21,515 - INFO - Best val loss: 0.0001381903130095452, Original val loss: 0.0001409030519425869
+[INFO] Saving Layer 23 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 24...
+[INFO] Starting Sequential Training for Layer 24
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:17<?, ?iter/s, val_loss=0.00236, val_og_loss=0.00276]
+
  5%|▌         | 1/20 [00:17<05:35, 17.68s/iter, val_loss=0.00236, val_og_loss=0.00276]
+
  5%|▌         | 1/20 [00:19<05:35, 17.68s/iter, val_loss=0.00242, val_og_loss=0.00276]
+
 10%|█         | 2/20 [00:19<02:29,  8.28s/iter, val_loss=0.00242, val_og_loss=0.00276]
+
 10%|█         | 2/20 [00:21<02:29,  8.28s/iter, val_loss=0.00242, val_og_loss=0.00276]
+
 15%|█▌        | 3/20 [00:21<01:29,  5.27s/iter, val_loss=0.00242, val_og_loss=0.00276]
+
 15%|█▌        | 3/20 [00:22<01:29,  5.27s/iter, val_loss=0.00242, val_og_loss=0.00276]
+
 20%|██        | 4/20 [00:22<01:01,  3.86s/iter, val_loss=0.00242, val_og_loss=0.00276]
+
 20%|██        | 4/20 [00:24<01:01,  3.86s/iter, val_loss=0.0024, val_og_loss=0.00276] 
+
 25%|██▌       | 5/20 [00:24<00:46,  3.07s/iter, val_loss=0.0024, val_og_loss=0.00276]
+
 25%|██▌       | 5/20 [00:26<00:46,  3.07s/iter, val_loss=0.00241, val_og_loss=0.00276]
+
 30%|███       | 6/20 [00:26<00:36,  2.60s/iter, val_loss=0.00241, val_og_loss=0.00276]
+
 30%|███       | 6/20 [00:27<00:36,  2.60s/iter, val_loss=0.00243, val_og_loss=0.00276]
+
 35%|███▌      | 7/20 [00:27<00:29,  2.30s/iter, val_loss=0.00243, val_og_loss=0.00276]
+
 35%|███▌      | 7/20 [00:29<00:29,  2.30s/iter, val_loss=0.00242, val_og_loss=0.00276]
+
 40%|████      | 8/20 [00:29<00:25,  2.10s/iter, val_loss=0.00242, val_og_loss=0.00276]
+
 40%|████      | 8/20 [00:31<00:25,  2.10s/iter, val_loss=0.00241, val_og_loss=0.00276]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.97s/iter, val_loss=0.00241, val_og_loss=0.00276]
+
 45%|████▌     | 9/20 [00:32<00:21,  1.97s/iter, val_loss=0.0024, val_og_loss=0.00276] 
+
 50%|█████     | 10/20 [00:32<00:18,  1.89s/iter, val_loss=0.0024, val_og_loss=0.00276]
+
 50%|█████     | 10/20 [00:34<00:18,  1.89s/iter, val_loss=0.0024, val_og_loss=0.00276]
+
 55%|█████▌    | 11/20 [00:34<00:16,  1.82s/iter, val_loss=0.0024, val_og_loss=0.00276]
+
 55%|█████▌    | 11/20 [00:36<00:16,  1.82s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 60%|██████    | 12/20 [00:36<00:14,  1.78s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 60%|██████    | 12/20 [00:37<00:14,  1.78s/iter, val_loss=0.00241, val_og_loss=0.00276]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.75s/iter, val_loss=0.00241, val_og_loss=0.00276]
+
 65%|██████▌   | 13/20 [00:39<00:12,  1.75s/iter, val_loss=0.0024, val_og_loss=0.00276] 
+
 70%|███████   | 14/20 [00:39<00:10,  1.73s/iter, val_loss=0.0024, val_og_loss=0.00276]
+
 70%|███████   | 14/20 [00:41<00:10,  1.73s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 75%|███████▌  | 15/20 [00:41<00:08,  1.71s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 75%|███████▌  | 15/20 [00:42<00:08,  1.71s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=0.00239, val_og_loss=0.00276]
+
 80%|████████  | 16/20 [00:44<00:06,  1.70s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
 85%|████████▌ | 17/20 [00:44<00:05,  1.70s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
 85%|████████▌ | 17/20 [00:46<00:05,  1.70s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 90%|█████████ | 18/20 [00:46<00:03,  1.69s/iter, val_loss=0.00238, val_og_loss=0.00276]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.69s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.68s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
 95%|█████████▌| 19/20 [00:49<00:01,  1.68s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
100%|██████████| 20/20 [00:49<00:00,  1.68s/iter, val_loss=0.00237, val_og_loss=0.00276]
100%|██████████| 20/20 [00:49<00:00,  2.48s/iter, val_loss=0.00237, val_og_loss=0.00276]
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:08:17<2:00:36, 904.50s/it]INFO:ParoQuant:Best val loss: 0.0023592421784996986, Original val loss: 0.0027630357071757317
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:08:20<2:00:36, 904.50s/it]2025-12-22 19:38:35,740 - INFO - Best val loss: 0.0023592421784996986, Original val loss: 0.0027630357071757317
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.0057, val_og_loss=0.00645]
+
  5%|▌         | 1/20 [00:15<04:57, 15.65s/iter, val_loss=0.0057, val_og_loss=0.00645]
+
  5%|▌         | 1/20 [00:17<04:57, 15.65s/iter, val_loss=0.00583, val_og_loss=0.00645]
+
 10%|█         | 2/20 [00:17<02:10,  7.26s/iter, val_loss=0.00583, val_og_loss=0.00645]
+
 10%|█         | 2/20 [00:18<02:10,  7.26s/iter, val_loss=0.00586, val_og_loss=0.00645]
+
 15%|█▌        | 3/20 [00:18<01:17,  4.58s/iter, val_loss=0.00586, val_og_loss=0.00645]
+
 15%|█▌        | 3/20 [00:19<01:17,  4.58s/iter, val_loss=0.00588, val_og_loss=0.00645]
+
 20%|██        | 4/20 [00:19<00:53,  3.32s/iter, val_loss=0.00588, val_og_loss=0.00645]
+
 20%|██        | 4/20 [00:21<00:53,  3.32s/iter, val_loss=0.00586, val_og_loss=0.00645]
+
 25%|██▌       | 5/20 [00:21<00:39,  2.62s/iter, val_loss=0.00586, val_og_loss=0.00645]
+
 25%|██▌       | 5/20 [00:22<00:39,  2.62s/iter, val_loss=0.00594, val_og_loss=0.00645]
+
 30%|███       | 6/20 [00:22<00:30,  2.20s/iter, val_loss=0.00594, val_og_loss=0.00645]
+
 30%|███       | 6/20 [00:23<00:30,  2.20s/iter, val_loss=0.00591, val_og_loss=0.00645]
+
 35%|███▌      | 7/20 [00:23<00:25,  1.94s/iter, val_loss=0.00591, val_og_loss=0.00645]
+
 35%|███▌      | 7/20 [00:25<00:25,  1.94s/iter, val_loss=0.00589, val_og_loss=0.00645]
+
 40%|████      | 8/20 [00:25<00:21,  1.77s/iter, val_loss=0.00589, val_og_loss=0.00645]
+
 40%|████      | 8/20 [00:26<00:21,  1.77s/iter, val_loss=0.00585, val_og_loss=0.00645]
+
 45%|████▌     | 9/20 [00:26<00:18,  1.65s/iter, val_loss=0.00585, val_og_loss=0.00645]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.65s/iter, val_loss=0.00584, val_og_loss=0.00645]
+
 50%|█████     | 10/20 [00:28<00:15,  1.57s/iter, val_loss=0.00584, val_og_loss=0.00645]
+
 50%|█████     | 10/20 [00:29<00:15,  1.57s/iter, val_loss=0.00589, val_og_loss=0.00645]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.51s/iter, val_loss=0.00589, val_og_loss=0.00645]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.51s/iter, val_loss=0.0058, val_og_loss=0.00645] 
+
 60%|██████    | 12/20 [00:30<00:11,  1.47s/iter, val_loss=0.0058, val_og_loss=0.00645]
+
 60%|██████    | 12/20 [00:32<00:11,  1.47s/iter, val_loss=0.00581, val_og_loss=0.00645]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.45s/iter, val_loss=0.00581, val_og_loss=0.00645]
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.45s/iter, val_loss=0.00581, val_og_loss=0.00645]
+
 70%|███████   | 14/20 [00:33<00:08,  1.43s/iter, val_loss=0.00581, val_og_loss=0.00645]
+
 70%|███████   | 14/20 [00:35<00:08,  1.43s/iter, val_loss=0.00582, val_og_loss=0.00645]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.42s/iter, val_loss=0.00582, val_og_loss=0.00645]
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.42s/iter, val_loss=0.0058, val_og_loss=0.00645] 
+
 80%|████████  | 16/20 [00:36<00:05,  1.41s/iter, val_loss=0.0058, val_og_loss=0.00645]
+
 80%|████████  | 16/20 [00:37<00:05,  1.41s/iter, val_loss=0.00578, val_og_loss=0.00645]
+
 85%|████████▌ | 17/20 [00:37<00:04,  1.40s/iter, val_loss=0.00578, val_og_loss=0.00645]
+
 85%|████████▌ | 17/20 [00:39<00:04,  1.40s/iter, val_loss=0.00576, val_og_loss=0.00645]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.40s/iter, val_loss=0.00576, val_og_loss=0.00645]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.40s/iter, val_loss=0.00576, val_og_loss=0.00645]
+
 95%|█████████▌| 19/20 [00:40<00:01,  1.40s/iter, val_loss=0.00576, val_og_loss=0.00645]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.40s/iter, val_loss=0.00578, val_og_loss=0.00645]
+
100%|██████████| 20/20 [00:42<00:00,  1.39s/iter, val_loss=0.00578, val_og_loss=0.00645]
100%|██████████| 20/20 [00:42<00:00,  2.10s/iter, val_loss=0.00578, val_og_loss=0.00645]
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:09:02<2:00:36, 904.50s/it]INFO:ParoQuant:Best val loss: 0.005696484353393316, Original val loss: 0.006450888235121965
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:09:04<2:00:36, 904.50s/it]2025-12-22 19:39:20,227 - INFO - Best val loss: 0.005696484353393316, Original val loss: 0.006450888235121965
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=0.00104, val_og_loss=0.00125]
+
  5%|▌         | 1/20 [00:16<05:09, 16.29s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
  5%|▌         | 1/20 [00:17<05:09, 16.29s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 10%|█         | 2/20 [00:17<02:15,  7.53s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 10%|█         | 2/20 [00:19<02:15,  7.53s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 15%|█▌        | 3/20 [00:19<01:20,  4.73s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 15%|█▌        | 3/20 [00:20<01:20,  4.73s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 20%|██        | 4/20 [00:20<00:54,  3.43s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 20%|██        | 4/20 [00:21<00:54,  3.43s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.69s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 25%|██▌       | 5/20 [00:23<00:40,  2.69s/iter, val_loss=0.00105, val_og_loss=0.00125]
+
 30%|███       | 6/20 [00:23<00:31,  2.25s/iter, val_loss=0.00105, val_og_loss=0.00125]
+
 30%|███       | 6/20 [00:24<00:31,  2.25s/iter, val_loss=0.00105, val_og_loss=0.00125]
+
 35%|███▌      | 7/20 [00:24<00:25,  1.97s/iter, val_loss=0.00105, val_og_loss=0.00125]
+
 35%|███▌      | 7/20 [00:26<00:25,  1.97s/iter, val_loss=0.00105, val_og_loss=0.00125]
+
 40%|████      | 8/20 [00:26<00:21,  1.79s/iter, val_loss=0.00105, val_og_loss=0.00125]
+
 40%|████      | 8/20 [00:27<00:21,  1.79s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 45%|████▌     | 9/20 [00:27<00:18,  1.67s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.67s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 50%|█████     | 10/20 [00:28<00:15,  1.59s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 50%|█████     | 10/20 [00:30<00:15,  1.59s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.53s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 55%|█████▌    | 11/20 [00:31<00:13,  1.53s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 60%|██████    | 12/20 [00:31<00:11,  1.49s/iter, val_loss=0.00104, val_og_loss=0.00125]
+
 60%|██████    | 12/20 [00:33<00:11,  1.49s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.46s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 65%|██████▌   | 13/20 [00:34<00:10,  1.46s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 70%|███████   | 14/20 [00:34<00:08,  1.44s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 70%|███████   | 14/20 [00:35<00:08,  1.44s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.43s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 75%|███████▌  | 15/20 [00:37<00:07,  1.43s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 80%|████████  | 16/20 [00:37<00:05,  1.42s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 80%|████████  | 16/20 [00:38<00:05,  1.42s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 85%|████████▌ | 17/20 [00:38<00:04,  1.41s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 85%|████████▌ | 17/20 [00:40<00:04,  1.41s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 90%|█████████ | 18/20 [00:41<00:02,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 95%|█████████▌| 19/20 [00:41<00:01,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
100%|██████████| 20/20 [00:42<00:00,  1.39s/iter, val_loss=0.00103, val_og_loss=0.00125]
100%|██████████| 20/20 [00:42<00:00,  2.14s/iter, val_loss=0.00103, val_og_loss=0.00125]
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:09:47<2:00:36, 904.50s/it]INFO:ParoQuant:Best val loss: 0.0010262615978717804, Original val loss: 0.0012464430183172226
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:09:50<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:09:50<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:10:21<2:00:36, 904.50s/it]2025-12-22 19:40:05,761 - INFO - Best val loss: 0.0010262615978717804, Original val loss: 0.0012464430183172226
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=2.9e-5, val_og_loss=2.97e-5]
+
  5%|▌         | 1/20 [00:13<04:23, 13.86s/iter, val_loss=2.9e-5, val_og_loss=2.97e-5]
+
  5%|▌         | 1/20 [00:15<04:23, 13.86s/iter, val_loss=2.91e-5, val_og_loss=2.97e-5]
+
 10%|█         | 2/20 [00:15<02:00,  6.70s/iter, val_loss=2.91e-5, val_og_loss=2.97e-5]
+
 10%|█         | 2/20 [00:17<02:00,  6.70s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 15%|█▌        | 3/20 [00:17<01:14,  4.41s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 15%|█▌        | 3/20 [00:18<01:14,  4.41s/iter, val_loss=2.92e-5, val_og_loss=2.97e-5]
+
 20%|██        | 4/20 [00:18<00:53,  3.33s/iter, val_loss=2.92e-5, val_og_loss=2.97e-5]
+
 20%|██        | 4/20 [00:20<00:53,  3.33s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 25%|██▌       | 5/20 [00:20<00:41,  2.74s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.74s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 30%|███       | 6/20 [00:22<00:33,  2.38s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 30%|███       | 6/20 [00:23<00:33,  2.38s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.15s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 35%|███▌      | 7/20 [00:25<00:27,  2.15s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 40%|████      | 8/20 [00:25<00:23,  2.00s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 40%|████      | 8/20 [00:27<00:23,  2.00s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.91s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 45%|████▌     | 9/20 [00:29<00:20,  1.91s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 50%|█████     | 10/20 [00:29<00:18,  1.84s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 50%|█████     | 10/20 [00:30<00:18,  1.84s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.79s/iter, val_loss=2.93e-5, val_og_loss=2.97e-5]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=2.92e-5, val_og_loss=2.97e-5]
+
 60%|██████    | 12/20 [00:32<00:14,  1.75s/iter, val_loss=2.92e-5, val_og_loss=2.97e-5]
+
 60%|██████    | 12/20 [00:34<00:14,  1.75s/iter, val_loss=2.91e-5, val_og_loss=2.97e-5]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=2.91e-5, val_og_loss=2.97e-5]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=2.91e-5, val_og_loss=2.97e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.71s/iter, val_loss=2.91e-5, val_og_loss=2.97e-5]
+
 70%|███████   | 14/20 [00:37<00:10,  1.71s/iter, val_loss=2.91e-5, val_og_loss=2.97e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.70s/iter, val_loss=2.91e-5, val_og_loss=2.97e-5]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.70s/iter, val_loss=2.9e-5, val_og_loss=2.97e-5] 
+
 80%|████████  | 16/20 [00:39<00:06,  1.69s/iter, val_loss=2.9e-5, val_og_loss=2.97e-5]
+
 80%|████████  | 16/20 [00:40<00:06,  1.69s/iter, val_loss=2.9e-5, val_og_loss=2.97e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=2.9e-5, val_og_loss=2.97e-5]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.69s/iter, val_loss=2.89e-5, val_og_loss=2.97e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.68s/iter, val_loss=2.89e-5, val_og_loss=2.97e-5]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.68s/iter, val_loss=2.89e-5, val_og_loss=2.97e-5]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.68s/iter, val_loss=2.89e-5, val_og_loss=2.97e-5]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.68s/iter, val_loss=2.89e-5, val_og_loss=2.97e-5]
+
100%|██████████| 20/20 [00:45<00:00,  1.67s/iter, val_loss=2.89e-5, val_og_loss=2.97e-5]
100%|██████████| 20/20 [00:45<00:00,  2.29s/iter, val_loss=2.89e-5, val_og_loss=2.97e-5]
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:11:07<2:00:36, 904.50s/it]INFO:ParoQuant:Best val loss: 2.891007716243621e-05, Original val loss: 2.973446862597484e-05
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:11:07<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:11:07<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:11:51<2:00:36, 904.50s/it]2025-12-22 19:41:25,206 - INFO - Best val loss: 2.891007716243621e-05, Original val loss: 2.973446862597484e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.00141, val_og_loss=0.00143]
+
  5%|▌         | 1/20 [00:30<09:33, 30.16s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
  5%|▌         | 1/20 [00:32<09:33, 30.16s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 10%|█         | 2/20 [00:32<04:12, 14.03s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 10%|█         | 2/20 [00:35<04:12, 14.03s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 15%|█▌        | 3/20 [00:35<02:30,  8.87s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 15%|█▌        | 3/20 [00:38<02:30,  8.87s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 20%|██        | 4/20 [00:38<01:42,  6.44s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 20%|██        | 4/20 [00:41<01:42,  6.44s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 25%|██▌       | 5/20 [00:41<01:16,  5.09s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.09s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 30%|███       | 6/20 [00:43<01:00,  4.29s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 30%|███       | 6/20 [00:46<01:00,  4.29s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 35%|███▌      | 7/20 [00:46<00:49,  3.78s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.78s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 40%|████      | 8/20 [00:49<00:41,  3.44s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 40%|████      | 8/20 [00:51<00:41,  3.44s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.21s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.21s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 60%|██████    | 12/20 [01:02<00:23,  2.89s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.83s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.83s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.00142, val_og_loss=0.00143]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.78s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.78s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 80%|████████  | 16/20 [01:10<00:11,  2.76s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.75s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.75s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.73s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.73s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
100%|██████████| 20/20 [01:21<00:00,  2.72s/iter, val_loss=0.00141, val_og_loss=0.00143]
100%|██████████| 20/20 [01:21<00:00,  4.09s/iter, val_loss=0.00141, val_og_loss=0.00143]
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:13:13<2:00:36, 904.50s/it]INFO:ParoQuant:Best val loss: 0.0014064835850149393, Original val loss: 0.0014300541952252388
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:13:18<2:00:36, 904.50s/it]2025-12-22 19:43:31,607 - INFO - Best val loss: 0.0014064835850149393, Original val loss: 0.0014300541952252388
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.000758, val_og_loss=0.000772]
+
  5%|▌         | 1/20 [00:31<09:55, 31.35s/iter, val_loss=0.000758, val_og_loss=0.000772]
+
  5%|▌         | 1/20 [00:34<09:55, 31.35s/iter, val_loss=0.000761, val_og_loss=0.000772]
+
 10%|█         | 2/20 [00:34<04:21, 14.50s/iter, val_loss=0.000761, val_og_loss=0.000772]
+
 10%|█         | 2/20 [00:36<04:21, 14.50s/iter, val_loss=0.000765, val_og_loss=0.000772]
+
 15%|█▌        | 3/20 [00:36<02:34,  9.11s/iter, val_loss=0.000765, val_og_loss=0.000772]
+
 15%|█▌        | 3/20 [00:39<02:34,  9.11s/iter, val_loss=0.000768, val_og_loss=0.000772]
+
 20%|██        | 4/20 [00:39<01:45,  6.59s/iter, val_loss=0.000768, val_og_loss=0.000772]
+
 20%|██        | 4/20 [00:42<01:45,  6.59s/iter, val_loss=0.000772, val_og_loss=0.000772]
+
 25%|██▌       | 5/20 [00:42<01:17,  5.19s/iter, val_loss=0.000772, val_og_loss=0.000772]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.19s/iter, val_loss=0.000772, val_og_loss=0.000772]
+
 30%|███       | 6/20 [00:44<01:00,  4.35s/iter, val_loss=0.000772, val_og_loss=0.000772]
+
 30%|███       | 6/20 [00:47<01:00,  4.35s/iter, val_loss=0.000772, val_og_loss=0.000772]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.81s/iter, val_loss=0.000772, val_og_loss=0.000772]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.81s/iter, val_loss=0.000771, val_og_loss=0.000772]
+
 40%|████      | 8/20 [00:50<00:41,  3.46s/iter, val_loss=0.000771, val_og_loss=0.000772]
+
 40%|████      | 8/20 [00:53<00:41,  3.46s/iter, val_loss=0.000771, val_og_loss=0.000772]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.23s/iter, val_loss=0.000771, val_og_loss=0.000772]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.23s/iter, val_loss=0.000769, val_og_loss=0.000772]
+
 50%|█████     | 10/20 [00:55<00:30,  3.07s/iter, val_loss=0.000769, val_og_loss=0.000772]
+
 50%|█████     | 10/20 [00:58<00:30,  3.07s/iter, val_loss=0.000767, val_og_loss=0.000772]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.97s/iter, val_loss=0.000767, val_og_loss=0.000772]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.97s/iter, val_loss=0.000767, val_og_loss=0.000772]
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.000767, val_og_loss=0.000772]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.000767, val_og_loss=0.000772]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.84s/iter, val_loss=0.000767, val_og_loss=0.000772]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000766, val_og_loss=0.000772]
+
 70%|███████   | 14/20 [01:06<00:16,  2.81s/iter, val_loss=0.000766, val_og_loss=0.000772]
+
 70%|███████   | 14/20 [01:09<00:16,  2.81s/iter, val_loss=0.000765, val_og_loss=0.000772]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.000765, val_og_loss=0.000772]
+
 75%|███████▌  | 15/20 [01:12<00:13,  2.78s/iter, val_loss=0.000763, val_og_loss=0.000772]
+
 80%|████████  | 16/20 [01:12<00:11,  2.75s/iter, val_loss=0.000763, val_og_loss=0.000772]
+
 80%|████████  | 16/20 [01:14<00:11,  2.75s/iter, val_loss=0.000763, val_og_loss=0.000772]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.74s/iter, val_loss=0.000763, val_og_loss=0.000772]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.74s/iter, val_loss=0.000763, val_og_loss=0.000772]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.73s/iter, val_loss=0.000763, val_og_loss=0.000772]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.73s/iter, val_loss=0.000762, val_og_loss=0.000772]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.000762, val_og_loss=0.000772]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.000762, val_og_loss=0.000772]
+
100%|██████████| 20/20 [01:22<00:00,  2.72s/iter, val_loss=0.000762, val_og_loss=0.000772]
100%|██████████| 20/20 [01:22<00:00,  4.15s/iter, val_loss=0.000762, val_og_loss=0.000772]
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:14:41<2:00:36, 904.50s/it]INFO:ParoQuant:Best val loss: 0.000757864210754633, Original val loss: 0.000772452331148088
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:14:42<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:14:42<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:15:41<2:00:36, 904.50s/it]2025-12-22 19:44:59,127 - INFO - Best val loss: 0.000757864210754633, Original val loss: 0.000772452331148088
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:34<?, ?iter/s, val_loss=0.000148, val_og_loss=0.000152]
+
  5%|▌         | 1/20 [00:34<10:47, 34.07s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
  5%|▌         | 1/20 [00:39<10:47, 34.07s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 10%|█         | 2/20 [00:39<05:09, 17.22s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 10%|█         | 2/20 [00:44<05:09, 17.22s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 15%|█▌        | 3/20 [00:44<03:21, 11.83s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 15%|█▌        | 3/20 [00:50<03:21, 11.83s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 20%|██        | 4/20 [00:50<02:28,  9.30s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 20%|██        | 4/20 [00:55<02:28,  9.30s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 25%|██▌       | 5/20 [00:55<01:58,  7.91s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 25%|██▌       | 5/20 [01:01<01:58,  7.91s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 30%|███       | 6/20 [01:01<01:38,  7.06s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 30%|███       | 6/20 [01:06<01:38,  7.06s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.53s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 35%|███▌      | 7/20 [01:12<01:24,  6.53s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 40%|████      | 8/20 [01:12<01:14,  6.18s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 40%|████      | 8/20 [01:17<01:14,  6.18s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 45%|████▌     | 9/20 [01:17<01:05,  5.94s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.94s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 50%|█████     | 10/20 [01:22<00:57,  5.78s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 50%|█████     | 10/20 [01:28<00:57,  5.78s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 55%|█████▌    | 11/20 [01:28<00:51,  5.67s/iter, val_loss=0.000149, val_og_loss=0.000152]
+
 55%|█████▌    | 11/20 [01:33<00:51,  5.67s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 60%|██████    | 12/20 [01:33<00:44,  5.59s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 60%|██████    | 12/20 [01:39<00:44,  5.59s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 65%|██████▌   | 13/20 [01:39<00:38,  5.54s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.54s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 70%|███████   | 14/20 [01:44<00:33,  5.50s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 70%|███████   | 14/20 [01:50<00:33,  5.50s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 75%|███████▌  | 15/20 [01:50<00:27,  5.48s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 75%|███████▌  | 15/20 [01:55<00:27,  5.48s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 80%|████████  | 16/20 [01:55<00:21,  5.46s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 80%|████████  | 16/20 [02:00<00:21,  5.46s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.45s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 85%|████████▌ | 17/20 [02:06<00:16,  5.45s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 90%|█████████ | 18/20 [02:06<00:10,  5.44s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
 95%|█████████▌| 19/20 [02:17<00:05,  5.44s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
100%|██████████| 20/20 [02:17<00:00,  5.43s/iter, val_loss=0.000148, val_og_loss=0.000152]
100%|██████████| 20/20 [02:17<00:00,  6.86s/iter, val_loss=0.000148, val_og_loss=0.000152]
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:17:58<2:00:36, 904.50s/it]INFO:ParoQuant:Best val loss: 0.00014794424350839108, Original val loss: 0.00015150032413657755
+
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:17:59<2:00:36, 904.50s/it]
                                                                       

Processing Blocks:  75%|███████▌  | 24/32 [5:18:09<2:00:36, 904.50s/it]
Processing Blocks:  78%|███████▊  | 25/32 [5:18:47<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:18:47<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:19:04<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:19:04<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:21:21<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:21:21<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:22:16<1:45:08, 901.27s/it]2025-12-22 19:48:16,623 - INFO - Best val loss: 0.00014794424350839108, Original val loss: 0.00015150032413657755
+[INFO] Saving Layer 24 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 25...
+[INFO] Starting Sequential Training for Layer 25
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:12<?, ?iter/s, val_loss=0.00238, val_og_loss=0.00272]
+
  5%|▌         | 1/20 [00:12<03:57, 12.49s/iter, val_loss=0.00238, val_og_loss=0.00272]
+
  5%|▌         | 1/20 [00:14<03:57, 12.49s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 10%|█         | 2/20 [00:14<01:50,  6.14s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 10%|█         | 2/20 [00:15<01:50,  6.14s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 15%|█▌        | 3/20 [00:15<01:09,  4.10s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 15%|█▌        | 3/20 [00:17<01:09,  4.10s/iter, val_loss=0.00244, val_og_loss=0.00272]
+
 20%|██        | 4/20 [00:17<00:50,  3.15s/iter, val_loss=0.00244, val_og_loss=0.00272]
+
 20%|██        | 4/20 [00:19<00:50,  3.15s/iter, val_loss=0.00244, val_og_loss=0.00272]
+
 25%|██▌       | 5/20 [00:19<00:39,  2.62s/iter, val_loss=0.00244, val_og_loss=0.00272]
+
 25%|██▌       | 5/20 [00:20<00:39,  2.62s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 30%|███       | 6/20 [00:20<00:32,  2.30s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 30%|███       | 6/20 [00:22<00:32,  2.30s/iter, val_loss=0.00242, val_og_loss=0.00272]
+
 35%|███▌      | 7/20 [00:22<00:27,  2.10s/iter, val_loss=0.00242, val_og_loss=0.00272]
+
 35%|███▌      | 7/20 [00:24<00:27,  2.10s/iter, val_loss=0.00244, val_og_loss=0.00272]
+
 40%|████      | 8/20 [00:24<00:23,  1.97s/iter, val_loss=0.00244, val_og_loss=0.00272]
+
 40%|████      | 8/20 [00:25<00:23,  1.97s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 45%|████▌     | 9/20 [00:25<00:20,  1.88s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.88s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 50%|█████     | 10/20 [00:27<00:18,  1.82s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 50%|█████     | 10/20 [00:29<00:18,  1.82s/iter, val_loss=0.00244, val_og_loss=0.00272]
+
 55%|█████▌    | 11/20 [00:29<00:15,  1.78s/iter, val_loss=0.00244, val_og_loss=0.00272]
+
 55%|█████▌    | 11/20 [00:31<00:15,  1.78s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 60%|██████    | 12/20 [00:31<00:13,  1.75s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 60%|██████    | 12/20 [00:32<00:13,  1.75s/iter, val_loss=0.00242, val_og_loss=0.00272]
+
 65%|██████▌   | 13/20 [00:32<00:12,  1.72s/iter, val_loss=0.00242, val_og_loss=0.00272]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.72s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 70%|███████   | 14/20 [00:34<00:10,  1.71s/iter, val_loss=0.00243, val_og_loss=0.00272]
+
 70%|███████   | 14/20 [00:36<00:10,  1.71s/iter, val_loss=0.00241, val_og_loss=0.00272]
+
 75%|███████▌  | 15/20 [00:36<00:08,  1.70s/iter, val_loss=0.00241, val_og_loss=0.00272]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.70s/iter, val_loss=0.0024, val_og_loss=0.00272] 
+
 80%|████████  | 16/20 [00:37<00:06,  1.69s/iter, val_loss=0.0024, val_og_loss=0.00272]
+
 80%|████████  | 16/20 [00:39<00:06,  1.69s/iter, val_loss=0.00239, val_og_loss=0.00272]
+
 85%|████████▌ | 17/20 [00:39<00:05,  1.69s/iter, val_loss=0.00239, val_og_loss=0.00272]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.69s/iter, val_loss=0.00239, val_og_loss=0.00272]
+
 90%|█████████ | 18/20 [00:41<00:03,  1.68s/iter, val_loss=0.00239, val_og_loss=0.00272]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.68s/iter, val_loss=0.00239, val_og_loss=0.00272]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.69s/iter, val_loss=0.00239, val_og_loss=0.00272]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=0.00239, val_og_loss=0.00272]
+
100%|██████████| 20/20 [00:44<00:00,  1.69s/iter, val_loss=0.00239, val_og_loss=0.00272]
100%|██████████| 20/20 [00:44<00:00,  2.22s/iter, val_loss=0.00239, val_og_loss=0.00272]
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:23:01<1:45:08, 901.27s/it]INFO:ParoQuant:Best val loss: 0.002383833285421133, Original val loss: 0.0027245290111750364
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:23:03<1:45:08, 901.27s/it]2025-12-22 19:53:19,285 - INFO - Best val loss: 0.002383833285421133, Original val loss: 0.0027245290111750364
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.00564, val_og_loss=0.00629]
+
  5%|▌         | 1/20 [00:10<03:21, 10.63s/iter, val_loss=0.00564, val_og_loss=0.00629]
+
  5%|▌         | 1/20 [00:12<03:21, 10.63s/iter, val_loss=0.00574, val_og_loss=0.00629]
+
 10%|█         | 2/20 [00:12<01:33,  5.19s/iter, val_loss=0.00574, val_og_loss=0.00629]
+
 10%|█         | 2/20 [00:13<01:33,  5.19s/iter, val_loss=0.00581, val_og_loss=0.00629]
+
 15%|█▌        | 3/20 [00:13<00:58,  3.46s/iter, val_loss=0.00581, val_og_loss=0.00629]
+
 15%|█▌        | 3/20 [00:14<00:58,  3.46s/iter, val_loss=0.00579, val_og_loss=0.00629]
+
 20%|██        | 4/20 [00:14<00:42,  2.64s/iter, val_loss=0.00579, val_og_loss=0.00629]
+
 20%|██        | 4/20 [00:16<00:42,  2.64s/iter, val_loss=0.00579, val_og_loss=0.00629]
+
 25%|██▌       | 5/20 [00:16<00:32,  2.20s/iter, val_loss=0.00579, val_og_loss=0.00629]
+
 25%|██▌       | 5/20 [00:17<00:32,  2.20s/iter, val_loss=0.00575, val_og_loss=0.00629]
+
 30%|███       | 6/20 [00:17<00:26,  1.92s/iter, val_loss=0.00575, val_og_loss=0.00629]
+
 30%|███       | 6/20 [00:18<00:26,  1.92s/iter, val_loss=0.00584, val_og_loss=0.00629]
+
 35%|███▌      | 7/20 [00:18<00:22,  1.75s/iter, val_loss=0.00584, val_og_loss=0.00629]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.75s/iter, val_loss=0.00578, val_og_loss=0.00629]
+
 40%|████      | 8/20 [00:20<00:19,  1.63s/iter, val_loss=0.00578, val_og_loss=0.00629]
+
 40%|████      | 8/20 [00:21<00:19,  1.63s/iter, val_loss=0.00576, val_og_loss=0.00629]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.56s/iter, val_loss=0.00576, val_og_loss=0.00629]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.56s/iter, val_loss=0.0058, val_og_loss=0.00629] 
+
 50%|█████     | 10/20 [00:23<00:15,  1.50s/iter, val_loss=0.0058, val_og_loss=0.00629]
+
 50%|█████     | 10/20 [00:24<00:15,  1.50s/iter, val_loss=0.00577, val_og_loss=0.00629]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.48s/iter, val_loss=0.00577, val_og_loss=0.00629]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.48s/iter, val_loss=0.00574, val_og_loss=0.00629]
+
 60%|██████    | 12/20 [00:25<00:11,  1.45s/iter, val_loss=0.00574, val_og_loss=0.00629]
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.00576, val_og_loss=0.00629]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.43s/iter, val_loss=0.00576, val_og_loss=0.00629]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.43s/iter, val_loss=0.00574, val_og_loss=0.00629]
+
 70%|███████   | 14/20 [00:28<00:08,  1.42s/iter, val_loss=0.00574, val_og_loss=0.00629]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.00573, val_og_loss=0.00629]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.00573, val_og_loss=0.00629]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.41s/iter, val_loss=0.00567, val_og_loss=0.00629]
+
 80%|████████  | 16/20 [00:31<00:05,  1.41s/iter, val_loss=0.00567, val_og_loss=0.00629]
+
 80%|████████  | 16/20 [00:32<00:05,  1.41s/iter, val_loss=0.0057, val_og_loss=0.00629] 
+
 85%|████████▌ | 17/20 [00:32<00:04,  1.41s/iter, val_loss=0.0057, val_og_loss=0.00629]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.41s/iter, val_loss=0.0057, val_og_loss=0.00629]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.0057, val_og_loss=0.00629]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00569, val_og_loss=0.00629]
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.40s/iter, val_loss=0.00569, val_og_loss=0.00629]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00568, val_og_loss=0.00629]
+
100%|██████████| 20/20 [00:37<00:00,  1.40s/iter, val_loss=0.00568, val_og_loss=0.00629]
100%|██████████| 20/20 [00:37<00:00,  1.86s/iter, val_loss=0.00568, val_og_loss=0.00629]
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:23:40<1:45:08, 901.27s/it]INFO:ParoQuant:Best val loss: 0.005635721143335104, Original val loss: 0.006290259305387735
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:23:43<1:45:08, 901.27s/it]2025-12-22 19:53:58,093 - INFO - Best val loss: 0.005635721143335104, Original val loss: 0.006290259305387735
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:12<?, ?iter/s, val_loss=0.0011, val_og_loss=0.00131]
+
  5%|▌         | 1/20 [00:12<04:00, 12.64s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
  5%|▌         | 1/20 [00:14<04:00, 12.64s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 10%|█         | 2/20 [00:14<01:48,  6.02s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 10%|█         | 2/20 [00:15<01:48,  6.02s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 15%|█▌        | 3/20 [00:15<01:06,  3.91s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 15%|█▌        | 3/20 [00:16<01:06,  3.91s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 20%|██        | 4/20 [00:16<00:46,  2.91s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 20%|██        | 4/20 [00:18<00:46,  2.91s/iter, val_loss=0.00111, val_og_loss=0.00131]
+
 25%|██▌       | 5/20 [00:18<00:35,  2.37s/iter, val_loss=0.00111, val_og_loss=0.00131]
+
 25%|██▌       | 5/20 [00:19<00:35,  2.37s/iter, val_loss=0.0011, val_og_loss=0.00131] 
+
 30%|███       | 6/20 [00:19<00:28,  2.04s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 30%|███       | 6/20 [00:20<00:28,  2.04s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 35%|███▌      | 7/20 [00:20<00:23,  1.83s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 35%|███▌      | 7/20 [00:22<00:23,  1.83s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 40%|████      | 8/20 [00:22<00:20,  1.69s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 40%|████      | 8/20 [00:23<00:20,  1.69s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.59s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.59s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 50%|█████     | 10/20 [00:25<00:15,  1.53s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 50%|█████     | 10/20 [00:26<00:15,  1.53s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.49s/iter, val_loss=0.0011, val_og_loss=0.00131]
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.49s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 60%|██████    | 12/20 [00:27<00:11,  1.46s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 60%|██████    | 12/20 [00:29<00:11,  1.46s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.44s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.44s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 70%|███████   | 14/20 [00:32<00:08,  1.43s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.41s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.41s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.00109, val_og_loss=0.00131]
+
 80%|████████  | 16/20 [00:34<00:05,  1.41s/iter, val_loss=0.00108, val_og_loss=0.00131]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00108, val_og_loss=0.00131]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.40s/iter, val_loss=0.00108, val_og_loss=0.00131]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.41s/iter, val_loss=0.00108, val_og_loss=0.00131]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.41s/iter, val_loss=0.00108, val_og_loss=0.00131]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00108, val_og_loss=0.00131]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.40s/iter, val_loss=0.00108, val_og_loss=0.00131]
+
100%|██████████| 20/20 [00:39<00:00,  1.40s/iter, val_loss=0.00108, val_og_loss=0.00131]
100%|██████████| 20/20 [00:39<00:00,  1.96s/iter, val_loss=0.00108, val_og_loss=0.00131]
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:24:22<1:45:08, 901.27s/it]INFO:ParoQuant:Best val loss: 0.0010808230144903064, Original val loss: 0.0013050460256636143
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:24:24<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:24:24<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:24:55<1:45:08, 901.27s/it]2025-12-22 19:54:40,222 - INFO - Best val loss: 0.0010808230144903064, Original val loss: 0.0013050460256636143
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
  5%|▌         | 1/20 [00:16<05:18, 16.77s/iter, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
  5%|▌         | 1/20 [00:18<05:18, 16.77s/iter, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
 10%|█         | 2/20 [00:18<02:22,  7.91s/iter, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
 10%|█         | 2/20 [00:20<02:22,  7.91s/iter, val_loss=3.31e-5, val_og_loss=3.34e-5]
+
 15%|█▌        | 3/20 [00:20<01:26,  5.07s/iter, val_loss=3.31e-5, val_og_loss=3.34e-5]
+
 15%|█▌        | 3/20 [00:21<01:26,  5.07s/iter, val_loss=3.32e-5, val_og_loss=3.34e-5]
+
 20%|██        | 4/20 [00:21<00:59,  3.73s/iter, val_loss=3.32e-5, val_og_loss=3.34e-5]
+
 20%|██        | 4/20 [00:23<00:59,  3.73s/iter, val_loss=3.31e-5, val_og_loss=3.34e-5]
+
 25%|██▌       | 5/20 [00:23<00:44,  2.99s/iter, val_loss=3.31e-5, val_og_loss=3.34e-5]
+
 25%|██▌       | 5/20 [00:25<00:44,  2.99s/iter, val_loss=3.33e-5, val_og_loss=3.34e-5]
+
 30%|███       | 6/20 [00:25<00:35,  2.55s/iter, val_loss=3.33e-5, val_og_loss=3.34e-5]
+
 30%|███       | 6/20 [00:26<00:35,  2.55s/iter, val_loss=3.34e-5, val_og_loss=3.34e-5]
+
 35%|███▌      | 7/20 [00:26<00:29,  2.26s/iter, val_loss=3.34e-5, val_og_loss=3.34e-5]
+
 35%|███▌      | 7/20 [00:28<00:29,  2.26s/iter, val_loss=3.31e-5, val_og_loss=3.34e-5]
+
 40%|████      | 8/20 [00:28<00:24,  2.08s/iter, val_loss=3.31e-5, val_og_loss=3.34e-5]
+
 40%|████      | 8/20 [00:30<00:24,  2.08s/iter, val_loss=3.32e-5, val_og_loss=3.34e-5]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.95s/iter, val_loss=3.32e-5, val_og_loss=3.34e-5]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.95s/iter, val_loss=3.31e-5, val_og_loss=3.34e-5]
+
 50%|█████     | 10/20 [00:31<00:18,  1.86s/iter, val_loss=3.31e-5, val_og_loss=3.34e-5]
+
 50%|█████     | 10/20 [00:33<00:18,  1.86s/iter, val_loss=3.3e-5, val_og_loss=3.34e-5] 
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.81s/iter, val_loss=3.3e-5, val_og_loss=3.34e-5]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.81s/iter, val_loss=3.3e-5, val_og_loss=3.34e-5]
+
 60%|██████    | 12/20 [00:35<00:14,  1.77s/iter, val_loss=3.3e-5, val_og_loss=3.34e-5]
+
 60%|██████    | 12/20 [00:36<00:14,  1.77s/iter, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.74s/iter, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.74s/iter, val_loss=3.3e-5, val_og_loss=3.34e-5] 
+
 70%|███████   | 14/20 [00:38<00:10,  1.72s/iter, val_loss=3.3e-5, val_og_loss=3.34e-5]
+
 70%|███████   | 14/20 [00:40<00:10,  1.72s/iter, val_loss=3.28e-5, val_og_loss=3.34e-5]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.71s/iter, val_loss=3.28e-5, val_og_loss=3.34e-5]
+
 75%|███████▌  | 15/20 [00:42<00:08,  1.71s/iter, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
 80%|████████  | 16/20 [00:43<00:06,  1.70s/iter, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.69s/iter, val_loss=3.29e-5, val_og_loss=3.34e-5]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.69s/iter, val_loss=3.28e-5, val_og_loss=3.34e-5]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.69s/iter, val_loss=3.28e-5, val_og_loss=3.34e-5]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.69s/iter, val_loss=3.27e-5, val_og_loss=3.34e-5]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=3.27e-5, val_og_loss=3.34e-5]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.69s/iter, val_loss=3.27e-5, val_og_loss=3.34e-5]
+
100%|██████████| 20/20 [00:48<00:00,  1.68s/iter, val_loss=3.27e-5, val_og_loss=3.34e-5]
100%|██████████| 20/20 [00:48<00:00,  2.44s/iter, val_loss=3.27e-5, val_og_loss=3.34e-5]
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:25:44<1:45:08, 901.27s/it]INFO:ParoQuant:Best val loss: 3.26699300785549e-05, Original val loss: 3.338920942042023e-05
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:25:44<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:25:44<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:26:28<1:45:08, 901.27s/it]2025-12-22 19:56:01,984 - INFO - Best val loss: 3.26699300785549e-05, Original val loss: 3.338920942042023e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.00148, val_og_loss=0.00152]
+
  5%|▌         | 1/20 [00:30<09:34, 30.24s/iter, val_loss=0.00148, val_og_loss=0.00152]
+
  5%|▌         | 1/20 [00:32<09:34, 30.24s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 10%|█         | 2/20 [00:32<04:12, 14.05s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 10%|█         | 2/20 [00:35<04:12, 14.05s/iter, val_loss=0.0015, val_og_loss=0.00152] 
+
 15%|█▌        | 3/20 [00:35<02:30,  8.87s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 15%|█▌        | 3/20 [00:38<02:30,  8.87s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 20%|██        | 4/20 [00:38<01:43,  6.44s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 20%|██        | 4/20 [00:41<01:43,  6.44s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 25%|██▌       | 5/20 [00:41<01:16,  5.10s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.10s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 30%|███       | 6/20 [00:43<01:00,  4.29s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 30%|███       | 6/20 [00:46<01:00,  4.29s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 35%|███▌      | 7/20 [00:46<00:49,  3.78s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.78s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 40%|████      | 8/20 [00:49<00:41,  3.45s/iter, val_loss=0.00151, val_og_loss=0.00152]
+
 40%|████      | 8/20 [00:51<00:41,  3.45s/iter, val_loss=0.0015, val_og_loss=0.00152] 
+
 45%|████▌     | 9/20 [00:51<00:35,  3.22s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.22s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 50%|█████     | 10/20 [00:54<00:30,  3.07s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 50%|█████     | 10/20 [00:57<00:30,  3.07s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 60%|██████    | 12/20 [01:02<00:23,  2.89s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.84s/iter, val_loss=0.0015, val_og_loss=0.00152]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.84s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.77s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.77s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.73s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.73s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
100%|██████████| 20/20 [01:21<00:00,  2.72s/iter, val_loss=0.00149, val_og_loss=0.00152]
100%|██████████| 20/20 [01:21<00:00,  4.09s/iter, val_loss=0.00149, val_og_loss=0.00152]
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:27:50<1:45:08, 901.27s/it]INFO:ParoQuant:Best val loss: 0.0014843815006315708, Original val loss: 0.001515964511781931
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:27:54<1:45:08, 901.27s/it]2025-12-22 19:58:08,750 - INFO - Best val loss: 0.0014843815006315708, Original val loss: 0.001515964511781931
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.000804, val_og_loss=0.000819]
+
  5%|▌         | 1/20 [00:30<09:43, 30.68s/iter, val_loss=0.000804, val_og_loss=0.000819]
+
  5%|▌         | 1/20 [00:33<09:43, 30.68s/iter, val_loss=0.000808, val_og_loss=0.000819]
+
 10%|█         | 2/20 [00:33<04:16, 14.24s/iter, val_loss=0.000808, val_og_loss=0.000819]
+
 10%|█         | 2/20 [00:36<04:16, 14.24s/iter, val_loss=0.000812, val_og_loss=0.000819]
+
 15%|█▌        | 3/20 [00:36<02:32,  8.98s/iter, val_loss=0.000812, val_og_loss=0.000819]
+
 15%|█▌        | 3/20 [00:38<02:32,  8.98s/iter, val_loss=0.000816, val_og_loss=0.000819]
+
 20%|██        | 4/20 [00:38<01:44,  6.51s/iter, val_loss=0.000816, val_og_loss=0.000819]
+
 20%|██        | 4/20 [00:41<01:44,  6.51s/iter, val_loss=0.000817, val_og_loss=0.000819]
+
 25%|██▌       | 5/20 [00:41<01:17,  5.14s/iter, val_loss=0.000817, val_og_loss=0.000819]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.14s/iter, val_loss=0.000817, val_og_loss=0.000819]
+
 30%|███       | 6/20 [00:44<01:00,  4.32s/iter, val_loss=0.000817, val_og_loss=0.000819]
+
 30%|███       | 6/20 [00:46<01:00,  4.32s/iter, val_loss=0.000818, val_og_loss=0.000819]
+
 35%|███▌      | 7/20 [00:46<00:49,  3.79s/iter, val_loss=0.000818, val_og_loss=0.000819]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.79s/iter, val_loss=0.000816, val_og_loss=0.000819]
+
 40%|████      | 8/20 [00:49<00:41,  3.45s/iter, val_loss=0.000816, val_og_loss=0.000819]
+
 40%|████      | 8/20 [00:52<00:41,  3.45s/iter, val_loss=0.000816, val_og_loss=0.000819]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.22s/iter, val_loss=0.000816, val_og_loss=0.000819]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.22s/iter, val_loss=0.000817, val_og_loss=0.000819]
+
 50%|█████     | 10/20 [00:55<00:30,  3.07s/iter, val_loss=0.000817, val_og_loss=0.000819]
+
 50%|█████     | 10/20 [00:57<00:30,  3.07s/iter, val_loss=0.000816, val_og_loss=0.000819]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.000816, val_og_loss=0.000819]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.000814, val_og_loss=0.000819]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.000814, val_og_loss=0.000819]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.000813, val_og_loss=0.000819]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.84s/iter, val_loss=0.000813, val_og_loss=0.000819]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000812, val_og_loss=0.000819]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.000812, val_og_loss=0.000819]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.000811, val_og_loss=0.000819]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.78s/iter, val_loss=0.000811, val_og_loss=0.000819]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.00081, val_og_loss=0.000819] 
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.00081, val_og_loss=0.000819]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.000809, val_og_loss=0.000819]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.75s/iter, val_loss=0.000809, val_og_loss=0.000819]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.75s/iter, val_loss=0.000809, val_og_loss=0.000819]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.74s/iter, val_loss=0.000809, val_og_loss=0.000819]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.74s/iter, val_loss=0.000808, val_og_loss=0.000819]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.000808, val_og_loss=0.000819]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.000807, val_og_loss=0.000819]
+
100%|██████████| 20/20 [01:22<00:00,  2.74s/iter, val_loss=0.000807, val_og_loss=0.000819]
100%|██████████| 20/20 [01:22<00:00,  4.12s/iter, val_loss=0.000807, val_og_loss=0.000819]
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:29:17<1:45:08, 901.27s/it]INFO:ParoQuant:Best val loss: 0.0008039674139581621, Original val loss: 0.000819383014459163
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:29:18<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:29:18<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:30:18<1:45:08, 901.27s/it]2025-12-22 19:59:35,318 - INFO - Best val loss: 0.0008039674139581621, Original val loss: 0.000819383014459163
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.000165, val_og_loss=0.000168]
+
  5%|▌         | 1/20 [00:33<10:40, 33.69s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
  5%|▌         | 1/20 [00:39<10:40, 33.69s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 10%|█         | 2/20 [00:39<05:07, 17.06s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 10%|█         | 2/20 [00:44<05:07, 17.06s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 15%|█▌        | 3/20 [00:44<03:19, 11.74s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 15%|█▌        | 3/20 [00:49<03:19, 11.74s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 20%|██        | 4/20 [00:49<02:27,  9.24s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 20%|██        | 4/20 [00:55<02:27,  9.24s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 25%|██▌       | 5/20 [00:55<01:57,  7.86s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 25%|██▌       | 5/20 [01:00<01:57,  7.86s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 30%|███       | 6/20 [01:00<01:38,  7.03s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 30%|███       | 6/20 [01:06<01:38,  7.03s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.51s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.51s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 40%|████      | 8/20 [01:11<01:13,  6.16s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 40%|████      | 8/20 [01:17<01:13,  6.16s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 45%|████▌     | 9/20 [01:17<01:05,  5.93s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.93s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 50%|█████     | 10/20 [01:22<00:57,  5.77s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 50%|█████     | 10/20 [01:27<00:57,  5.77s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 55%|█████▌    | 11/20 [01:27<00:51,  5.67s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 55%|█████▌    | 11/20 [01:33<00:51,  5.67s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 60%|██████    | 12/20 [01:33<00:44,  5.59s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 60%|██████    | 12/20 [01:38<00:44,  5.59s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.54s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.54s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 70%|███████   | 14/20 [01:44<00:33,  5.51s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 70%|███████   | 14/20 [01:49<00:33,  5.51s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.48s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 75%|███████▌  | 15/20 [01:55<00:27,  5.48s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 80%|████████  | 16/20 [01:55<00:21,  5.46s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
 80%|████████  | 16/20 [02:00<00:21,  5.46s/iter, val_loss=0.000164, val_og_loss=0.000168]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.45s/iter, val_loss=0.000164, val_og_loss=0.000168]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.45s/iter, val_loss=0.000164, val_og_loss=0.000168]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.44s/iter, val_loss=0.000164, val_og_loss=0.000168]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=0.000164, val_og_loss=0.000168]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=0.000164, val_og_loss=0.000168]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.44s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
100%|██████████| 20/20 [02:16<00:00,  5.43s/iter, val_loss=0.000165, val_og_loss=0.000168]
100%|██████████| 20/20 [02:16<00:00,  6.83s/iter, val_loss=0.000165, val_og_loss=0.000168]
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:32:35<1:45:08, 901.27s/it]INFO:ParoQuant:Best val loss: 0.00016441376646980643, Original val loss: 0.00016819470329210162
+
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:32:35<1:45:08, 901.27s/it]
                                                                       

Processing Blocks:  78%|███████▊  | 25/32 [5:32:46<1:45:08, 901.27s/it]
Processing Blocks:  81%|████████▏ | 26/32 [5:33:24<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:33:24<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:33:42<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:33:42<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:35:58<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:35:58<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:36:53<1:29:24, 894.05s/it]2025-12-22 20:02:53,041 - INFO - Best val loss: 0.00016441376646980643, Original val loss: 0.00016819470329210162
+[INFO] Saving Layer 25 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 26...
+[INFO] Starting Sequential Training for Layer 26
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=0.00225, val_og_loss=0.00266]
+
  5%|▌         | 1/20 [00:13<04:13, 13.34s/iter, val_loss=0.00225, val_og_loss=0.00266]
+
  5%|▌         | 1/20 [00:15<04:13, 13.34s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 10%|█         | 2/20 [00:15<01:56,  6.50s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 10%|█         | 2/20 [00:16<01:56,  6.50s/iter, val_loss=0.0023, val_og_loss=0.00266] 
+
 15%|█▌        | 3/20 [00:16<01:13,  4.30s/iter, val_loss=0.0023, val_og_loss=0.00266]
+
 15%|█▌        | 3/20 [00:18<01:13,  4.30s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 20%|██        | 4/20 [00:18<00:52,  3.27s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 20%|██        | 4/20 [00:20<00:52,  3.27s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 25%|██▌       | 5/20 [00:20<00:40,  2.69s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.69s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 30%|███       | 6/20 [00:21<00:32,  2.35s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 30%|███       | 6/20 [00:23<00:32,  2.35s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.13s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 35%|███▌      | 7/20 [00:25<00:27,  2.13s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 40%|████      | 8/20 [00:25<00:23,  1.98s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 40%|████      | 8/20 [00:26<00:23,  1.98s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 45%|████▌     | 9/20 [00:26<00:20,  1.89s/iter, val_loss=0.00229, val_og_loss=0.00266]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.89s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 50%|█████     | 10/20 [00:28<00:18,  1.82s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 50%|█████     | 10/20 [00:30<00:18,  1.82s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.78s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.78s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 60%|██████    | 12/20 [00:31<00:14,  1.75s/iter, val_loss=0.00228, val_og_loss=0.00266]
+
 60%|██████    | 12/20 [00:33<00:14,  1.75s/iter, val_loss=0.00227, val_og_loss=0.00266]
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.73s/iter, val_loss=0.00227, val_og_loss=0.00266]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=0.00227, val_og_loss=0.00266]
+
 70%|███████   | 14/20 [00:35<00:10,  1.72s/iter, val_loss=0.00227, val_og_loss=0.00266]
+
 70%|███████   | 14/20 [00:36<00:10,  1.72s/iter, val_loss=0.00227, val_og_loss=0.00266]
+
 75%|███████▌  | 15/20 [00:36<00:08,  1.72s/iter, val_loss=0.00227, val_og_loss=0.00266]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.72s/iter, val_loss=0.00227, val_og_loss=0.00266]
+
 80%|████████  | 16/20 [00:38<00:06,  1.71s/iter, val_loss=0.00227, val_og_loss=0.00266]
+
 80%|████████  | 16/20 [00:40<00:06,  1.71s/iter, val_loss=0.00226, val_og_loss=0.00266]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.70s/iter, val_loss=0.00226, val_og_loss=0.00266]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.70s/iter, val_loss=0.00226, val_og_loss=0.00266]
+
 90%|█████████ | 18/20 [00:41<00:03,  1.69s/iter, val_loss=0.00226, val_og_loss=0.00266]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=0.00226, val_og_loss=0.00266]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.69s/iter, val_loss=0.00226, val_og_loss=0.00266]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=0.00225, val_og_loss=0.00266]
+
100%|██████████| 20/20 [00:45<00:00,  1.69s/iter, val_loss=0.00225, val_og_loss=0.00266]
100%|██████████| 20/20 [00:45<00:00,  2.27s/iter, val_loss=0.00225, val_og_loss=0.00266]
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:37:38<1:29:24, 894.05s/it]INFO:ParoQuant:Best val loss: 0.002251715399324894, Original val loss: 0.002664221916347742
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:37:40<1:29:24, 894.05s/it]2025-12-22 20:07:56,790 - INFO - Best val loss: 0.002251715399324894, Original val loss: 0.002664221916347742
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.00583, val_og_loss=0.00671]
+
  5%|▌         | 1/20 [00:10<03:24, 10.76s/iter, val_loss=0.00583, val_og_loss=0.00671]
+
  5%|▌         | 1/20 [00:12<03:24, 10.76s/iter, val_loss=0.00597, val_og_loss=0.00671]
+
 10%|█         | 2/20 [00:12<01:34,  5.25s/iter, val_loss=0.00597, val_og_loss=0.00671]
+
 10%|█         | 2/20 [00:13<01:34,  5.25s/iter, val_loss=0.00607, val_og_loss=0.00671]
+
 15%|█▌        | 3/20 [00:13<00:59,  3.48s/iter, val_loss=0.00607, val_og_loss=0.00671]
+
 15%|█▌        | 3/20 [00:14<00:59,  3.48s/iter, val_loss=0.00599, val_og_loss=0.00671]
+
 20%|██        | 4/20 [00:14<00:42,  2.65s/iter, val_loss=0.00599, val_og_loss=0.00671]
+
 20%|██        | 4/20 [00:16<00:42,  2.65s/iter, val_loss=0.00602, val_og_loss=0.00671]
+
 25%|██▌       | 5/20 [00:16<00:32,  2.19s/iter, val_loss=0.00602, val_og_loss=0.00671]
+
 25%|██▌       | 5/20 [00:17<00:32,  2.19s/iter, val_loss=0.00603, val_og_loss=0.00671]
+
 30%|███       | 6/20 [00:17<00:26,  1.91s/iter, val_loss=0.00603, val_og_loss=0.00671]
+
 30%|███       | 6/20 [00:19<00:26,  1.91s/iter, val_loss=0.00595, val_og_loss=0.00671]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.74s/iter, val_loss=0.00595, val_og_loss=0.00671]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.74s/iter, val_loss=0.00598, val_og_loss=0.00671]
+
 40%|████      | 8/20 [00:20<00:19,  1.64s/iter, val_loss=0.00598, val_og_loss=0.00671]
+
 40%|████      | 8/20 [00:21<00:19,  1.64s/iter, val_loss=0.00595, val_og_loss=0.00671]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.56s/iter, val_loss=0.00595, val_og_loss=0.00671]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.56s/iter, val_loss=0.006, val_og_loss=0.00671]  
+
 50%|█████     | 10/20 [00:23<00:15,  1.51s/iter, val_loss=0.006, val_og_loss=0.00671]
+
 50%|█████     | 10/20 [00:24<00:15,  1.51s/iter, val_loss=0.00593, val_og_loss=0.00671]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.47s/iter, val_loss=0.00593, val_og_loss=0.00671]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.47s/iter, val_loss=0.006, val_og_loss=0.00671]  
+
 60%|██████    | 12/20 [00:26<00:11,  1.45s/iter, val_loss=0.006, val_og_loss=0.00671]
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.44s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.44s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 70%|███████   | 14/20 [00:28<00:08,  1.42s/iter, val_loss=0.00594, val_og_loss=0.00671]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.0059, val_og_loss=0.00671] 
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.0059, val_og_loss=0.00671]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.41s/iter, val_loss=0.00593, val_og_loss=0.00671]
+
 80%|████████  | 16/20 [00:31<00:05,  1.41s/iter, val_loss=0.00593, val_og_loss=0.00671]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.00591, val_og_loss=0.00671]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.40s/iter, val_loss=0.00591, val_og_loss=0.00671]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00591, val_og_loss=0.00671]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.00591, val_og_loss=0.00671]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00592, val_og_loss=0.00671]
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.40s/iter, val_loss=0.00592, val_og_loss=0.00671]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00591, val_og_loss=0.00671]
+
100%|██████████| 20/20 [00:37<00:00,  1.40s/iter, val_loss=0.00591, val_og_loss=0.00671]
100%|██████████| 20/20 [00:37<00:00,  1.86s/iter, val_loss=0.00591, val_og_loss=0.00671]
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:38:17<1:29:24, 894.05s/it]INFO:ParoQuant:Best val loss: 0.005834463983774185, Original val loss: 0.006710827816277742
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:38:19<1:29:24, 894.05s/it]2025-12-22 20:08:35,794 - INFO - Best val loss: 0.005834463983774185, Original val loss: 0.006710827816277742
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=0.00104, val_og_loss=0.00129]
+
  5%|▌         | 1/20 [00:11<03:30, 11.06s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
  5%|▌         | 1/20 [00:12<03:30, 11.06s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 10%|█         | 2/20 [00:12<01:36,  5.37s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 10%|█         | 2/20 [00:13<01:36,  5.37s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 15%|█▌        | 3/20 [00:13<01:00,  3.55s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 15%|█▌        | 3/20 [00:15<01:00,  3.55s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 20%|██        | 4/20 [00:15<00:43,  2.70s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 20%|██        | 4/20 [00:16<00:43,  2.70s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 25%|██▌       | 5/20 [00:16<00:33,  2.23s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 25%|██▌       | 5/20 [00:18<00:33,  2.23s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 30%|███       | 6/20 [00:18<00:27,  1.94s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 30%|███       | 6/20 [00:19<00:27,  1.94s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.76s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.76s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 40%|████      | 8/20 [00:20<00:19,  1.65s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 40%|████      | 8/20 [00:22<00:19,  1.65s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.57s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.57s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 50%|█████     | 10/20 [00:23<00:15,  1.51s/iter, val_loss=0.00105, val_og_loss=0.00129]
+
 50%|█████     | 10/20 [00:24<00:15,  1.51s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.48s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.48s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 60%|██████    | 12/20 [00:26<00:11,  1.46s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 60%|██████    | 12/20 [00:27<00:11,  1.46s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.44s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.44s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 70%|███████   | 14/20 [00:29<00:08,  1.42s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.00104, val_og_loss=0.00129]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.41s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
 80%|████████  | 16/20 [00:31<00:05,  1.41s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
100%|██████████| 20/20 [00:37<00:00,  1.40s/iter, val_loss=0.00103, val_og_loss=0.00129]
100%|██████████| 20/20 [00:37<00:00,  1.88s/iter, val_loss=0.00103, val_og_loss=0.00129]
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:38:57<1:29:24, 894.05s/it]INFO:ParoQuant:Best val loss: 0.0010292230872437358, Original val loss: 0.0012871318031102419
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:38:59<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:38:59<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:39:30<1:29:24, 894.05s/it]2025-12-22 20:09:15,248 - INFO - Best val loss: 0.0010292230872437358, Original val loss: 0.0012871318031102419
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:16<?, ?iter/s, val_loss=4.79e-5, val_og_loss=4.91e-5]
+
  5%|▌         | 1/20 [00:16<05:20, 16.88s/iter, val_loss=4.79e-5, val_og_loss=4.91e-5]
+
  5%|▌         | 1/20 [00:18<05:20, 16.88s/iter, val_loss=4.78e-5, val_og_loss=4.91e-5]
+
 10%|█         | 2/20 [00:18<02:23,  7.94s/iter, val_loss=4.78e-5, val_og_loss=4.91e-5]
+
 10%|█         | 2/20 [00:20<02:23,  7.94s/iter, val_loss=4.82e-5, val_og_loss=4.91e-5]
+
 15%|█▌        | 3/20 [00:20<01:26,  5.09s/iter, val_loss=4.82e-5, val_og_loss=4.91e-5]
+
 15%|█▌        | 3/20 [00:21<01:26,  5.09s/iter, val_loss=4.84e-5, val_og_loss=4.91e-5]
+
 20%|██        | 4/20 [00:21<00:59,  3.74s/iter, val_loss=4.84e-5, val_og_loss=4.91e-5]
+
 20%|██        | 4/20 [00:23<00:59,  3.74s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 25%|██▌       | 5/20 [00:23<00:44,  3.00s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 25%|██▌       | 5/20 [00:25<00:44,  3.00s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 30%|███       | 6/20 [00:25<00:35,  2.55s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 30%|███       | 6/20 [00:26<00:35,  2.55s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 35%|███▌      | 7/20 [00:26<00:29,  2.27s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 35%|███▌      | 7/20 [00:28<00:29,  2.27s/iter, val_loss=4.84e-5, val_og_loss=4.91e-5]
+
 40%|████      | 8/20 [00:28<00:24,  2.08s/iter, val_loss=4.84e-5, val_og_loss=4.91e-5]
+
 40%|████      | 8/20 [00:30<00:24,  2.08s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.95s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 45%|████▌     | 9/20 [00:31<00:21,  1.95s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 50%|█████     | 10/20 [00:31<00:18,  1.86s/iter, val_loss=4.85e-5, val_og_loss=4.91e-5]
+
 50%|█████     | 10/20 [00:33<00:18,  1.86s/iter, val_loss=4.81e-5, val_og_loss=4.91e-5]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.81s/iter, val_loss=4.81e-5, val_og_loss=4.91e-5]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.81s/iter, val_loss=4.81e-5, val_og_loss=4.91e-5]
+
 60%|██████    | 12/20 [00:35<00:14,  1.77s/iter, val_loss=4.81e-5, val_og_loss=4.91e-5]
+
 60%|██████    | 12/20 [00:37<00:14,  1.77s/iter, val_loss=4.81e-5, val_og_loss=4.91e-5]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.74s/iter, val_loss=4.81e-5, val_og_loss=4.91e-5]
+
 65%|██████▌   | 13/20 [00:38<00:12,  1.74s/iter, val_loss=4.79e-5, val_og_loss=4.91e-5]
+
 70%|███████   | 14/20 [00:38<00:10,  1.72s/iter, val_loss=4.79e-5, val_og_loss=4.91e-5]
+
 70%|███████   | 14/20 [00:40<00:10,  1.72s/iter, val_loss=4.78e-5, val_og_loss=4.91e-5]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.71s/iter, val_loss=4.78e-5, val_og_loss=4.91e-5]
+
 75%|███████▌  | 15/20 [00:42<00:08,  1.71s/iter, val_loss=4.77e-5, val_og_loss=4.91e-5]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=4.77e-5, val_og_loss=4.91e-5]
+
 80%|████████  | 16/20 [00:43<00:06,  1.70s/iter, val_loss=4.77e-5, val_og_loss=4.91e-5]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.70s/iter, val_loss=4.77e-5, val_og_loss=4.91e-5]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.70s/iter, val_loss=4.77e-5, val_og_loss=4.91e-5]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.69s/iter, val_loss=4.77e-5, val_og_loss=4.91e-5]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.69s/iter, val_loss=4.78e-5, val_og_loss=4.91e-5]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.69s/iter, val_loss=4.78e-5, val_og_loss=4.91e-5]
+
 95%|█████████▌| 19/20 [00:48<00:01,  1.69s/iter, val_loss=4.76e-5, val_og_loss=4.91e-5]
+
100%|██████████| 20/20 [00:48<00:00,  1.68s/iter, val_loss=4.76e-5, val_og_loss=4.91e-5]
100%|██████████| 20/20 [00:48<00:00,  2.44s/iter, val_loss=4.76e-5, val_og_loss=4.91e-5]
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:40:18<1:29:24, 894.05s/it]INFO:ParoQuant:Best val loss: 4.7553829062962905e-05, Original val loss: 4.9099191528512165e-05
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:40:19<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:40:19<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:41:02<1:29:24, 894.05s/it]2025-12-22 20:10:36,852 - INFO - Best val loss: 4.7553829062962905e-05, Original val loss: 4.9099191528512165e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00161, val_og_loss=0.00164]
+
  5%|▌         | 1/20 [00:29<09:21, 29.57s/iter, val_loss=0.00161, val_og_loss=0.00164]
+
  5%|▌         | 1/20 [00:32<09:21, 29.57s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 10%|█         | 2/20 [00:32<04:07, 13.77s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 10%|█         | 2/20 [00:34<04:07, 13.77s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 15%|█▌        | 3/20 [00:34<02:28,  8.72s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 15%|█▌        | 3/20 [00:37<02:28,  8.72s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 20%|██        | 4/20 [00:37<01:41,  6.35s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 20%|██        | 4/20 [00:40<01:41,  6.35s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 25%|██▌       | 5/20 [00:40<01:15,  5.04s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 25%|██▌       | 5/20 [00:43<01:15,  5.04s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 30%|███       | 6/20 [00:43<00:59,  4.26s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 30%|███       | 6/20 [00:45<00:59,  4.26s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 35%|███▌      | 7/20 [00:45<00:48,  3.75s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.75s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 40%|████      | 8/20 [00:48<00:41,  3.42s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 40%|████      | 8/20 [00:51<00:41,  3.42s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.21s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.21s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.00163, val_og_loss=0.00164]
+
 50%|█████     | 10/20 [00:56<00:30,  3.06s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.95s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.95s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 60%|██████    | 12/20 [00:59<00:23,  2.88s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 60%|██████    | 12/20 [01:02<00:23,  2.88s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.83s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.83s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 70%|███████   | 14/20 [01:04<00:16,  2.79s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 70%|███████   | 14/20 [01:07<00:16,  2.79s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 80%|████████  | 16/20 [01:10<00:11,  2.75s/iter, val_loss=0.00162, val_og_loss=0.00164]
+
 80%|████████  | 16/20 [01:13<00:11,  2.75s/iter, val_loss=0.00161, val_og_loss=0.00164]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.00161, val_og_loss=0.00164]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.00161, val_og_loss=0.00164]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.73s/iter, val_loss=0.00161, val_og_loss=0.00164]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.00161, val_og_loss=0.00164]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.00161, val_og_loss=0.00164]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.00161, val_og_loss=0.00164]
+
100%|██████████| 20/20 [01:21<00:00,  2.72s/iter, val_loss=0.00161, val_og_loss=0.00164]
100%|██████████| 20/20 [01:21<00:00,  4.06s/iter, val_loss=0.00161, val_og_loss=0.00164]
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:42:23<1:29:24, 894.05s/it]INFO:ParoQuant:Best val loss: 0.0016080747591331601, Original val loss: 0.0016417662845924497
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:42:28<1:29:24, 894.05s/it]2025-12-22 20:12:41,596 - INFO - Best val loss: 0.0016080747591331601, Original val loss: 0.0016417662845924497
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:31<?, ?iter/s, val_loss=0.00087, val_og_loss=0.000891]
+
  5%|▌         | 1/20 [00:31<09:58, 31.51s/iter, val_loss=0.00087, val_og_loss=0.000891]
+
  5%|▌         | 1/20 [00:34<09:58, 31.51s/iter, val_loss=0.000872, val_og_loss=0.000891]
+
 10%|█         | 2/20 [00:34<04:22, 14.56s/iter, val_loss=0.000872, val_og_loss=0.000891]
+
 10%|█         | 2/20 [00:36<04:22, 14.56s/iter, val_loss=0.000878, val_og_loss=0.000891]
+
 15%|█▌        | 3/20 [00:36<02:35,  9.15s/iter, val_loss=0.000878, val_og_loss=0.000891]
+
 15%|█▌        | 3/20 [00:39<02:35,  9.15s/iter, val_loss=0.000879, val_og_loss=0.000891]
+
 20%|██        | 4/20 [00:39<01:45,  6.62s/iter, val_loss=0.000879, val_og_loss=0.000891]
+
 20%|██        | 4/20 [00:42<01:45,  6.62s/iter, val_loss=0.000882, val_og_loss=0.000891]
+
 25%|██▌       | 5/20 [00:42<01:18,  5.21s/iter, val_loss=0.000882, val_og_loss=0.000891]
+
 25%|██▌       | 5/20 [00:45<01:18,  5.21s/iter, val_loss=0.000882, val_og_loss=0.000891]
+
 30%|███       | 6/20 [00:45<01:01,  4.36s/iter, val_loss=0.000882, val_og_loss=0.000891]
+
 30%|███       | 6/20 [00:47<01:01,  4.36s/iter, val_loss=0.000882, val_og_loss=0.000891]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.82s/iter, val_loss=0.000882, val_og_loss=0.000891]
+
 35%|███▌      | 7/20 [00:50<00:49,  3.82s/iter, val_loss=0.000879, val_og_loss=0.000891]
+
 40%|████      | 8/20 [00:50<00:41,  3.47s/iter, val_loss=0.000879, val_og_loss=0.000891]
+
 40%|████      | 8/20 [00:53<00:41,  3.47s/iter, val_loss=0.00088, val_og_loss=0.000891] 
+
 45%|████▌     | 9/20 [00:53<00:35,  3.24s/iter, val_loss=0.00088, val_og_loss=0.000891]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.24s/iter, val_loss=0.000878, val_og_loss=0.000891]
+
 50%|█████     | 10/20 [00:55<00:30,  3.08s/iter, val_loss=0.000878, val_og_loss=0.000891]
+
 50%|█████     | 10/20 [00:58<00:30,  3.08s/iter, val_loss=0.000877, val_og_loss=0.000891]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.97s/iter, val_loss=0.000877, val_og_loss=0.000891]
+
 55%|█████▌    | 11/20 [01:01<00:26,  2.97s/iter, val_loss=0.000876, val_og_loss=0.000891]
+
 60%|██████    | 12/20 [01:01<00:23,  2.89s/iter, val_loss=0.000876, val_og_loss=0.000891]
+
 60%|██████    | 12/20 [01:04<00:23,  2.89s/iter, val_loss=0.000874, val_og_loss=0.000891]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.000874, val_og_loss=0.000891]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.84s/iter, val_loss=0.000873, val_og_loss=0.000891]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.000873, val_og_loss=0.000891]
+
 70%|███████   | 14/20 [01:09<00:16,  2.80s/iter, val_loss=0.000871, val_og_loss=0.000891]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.77s/iter, val_loss=0.000871, val_og_loss=0.000891]
+
 75%|███████▌  | 15/20 [01:12<00:13,  2.77s/iter, val_loss=0.000871, val_og_loss=0.000891]
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.000871, val_og_loss=0.000891]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.00087, val_og_loss=0.000891] 
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.74s/iter, val_loss=0.00087, val_og_loss=0.000891]
+
 85%|████████▌ | 17/20 [01:17<00:08,  2.74s/iter, val_loss=0.00087, val_og_loss=0.000891]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.73s/iter, val_loss=0.00087, val_og_loss=0.000891]
+
 90%|█████████ | 18/20 [01:20<00:05,  2.73s/iter, val_loss=0.00087, val_og_loss=0.000891]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.00087, val_og_loss=0.000891]
+
 95%|█████████▌| 19/20 [01:23<00:02,  2.73s/iter, val_loss=0.000869, val_og_loss=0.000891]
+
100%|██████████| 20/20 [01:23<00:00,  2.72s/iter, val_loss=0.000869, val_og_loss=0.000891]
100%|██████████| 20/20 [01:23<00:00,  4.16s/iter, val_loss=0.000869, val_og_loss=0.000891]
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:43:51<1:29:24, 894.05s/it]INFO:ParoQuant:Best val loss: 0.0008691346738487482, Original val loss: 0.0008908282034099102
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:43:52<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:43:52<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:44:50<1:29:24, 894.05s/it]2025-12-22 20:14:09,371 - INFO - Best val loss: 0.0008691346738487482, Original val loss: 0.0008908282034099102
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.000189, val_og_loss=0.000196]
+
  5%|▌         | 1/20 [00:33<10:32, 33.31s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
  5%|▌         | 1/20 [00:38<10:32, 33.31s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 10%|█         | 2/20 [00:38<05:04, 16.90s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 10%|█         | 2/20 [00:44<05:04, 16.90s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 15%|█▌        | 3/20 [00:44<03:18, 11.66s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 15%|█▌        | 3/20 [00:49<03:18, 11.66s/iter, val_loss=0.00019, val_og_loss=0.000196] 
+
 20%|██        | 4/20 [00:49<02:27,  9.19s/iter, val_loss=0.00019, val_og_loss=0.000196]
+
 20%|██        | 4/20 [00:54<02:27,  9.19s/iter, val_loss=0.00019, val_og_loss=0.000196]
+
 25%|██▌       | 5/20 [00:54<01:57,  7.83s/iter, val_loss=0.00019, val_og_loss=0.000196]
+
 25%|██▌       | 5/20 [01:00<01:57,  7.83s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 30%|███       | 6/20 [01:00<01:38,  7.01s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 30%|███       | 6/20 [01:05<01:38,  7.01s/iter, val_loss=0.00019, val_og_loss=0.000196] 
+
 35%|███▌      | 7/20 [01:05<01:24,  6.49s/iter, val_loss=0.00019, val_og_loss=0.000196]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.49s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 40%|████      | 8/20 [01:11<01:13,  6.15s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 40%|████      | 8/20 [01:16<01:13,  6.15s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 45%|████▌     | 9/20 [01:16<01:05,  5.92s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.92s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 50%|█████     | 10/20 [01:22<00:57,  5.77s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 50%|█████     | 10/20 [01:27<00:57,  5.77s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 55%|█████▌    | 11/20 [01:27<00:50,  5.66s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 55%|█████▌    | 11/20 [01:32<00:50,  5.66s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 60%|██████    | 12/20 [01:32<00:44,  5.59s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 60%|██████    | 12/20 [01:38<00:44,  5.59s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.54s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 65%|██████▌   | 13/20 [01:43<00:38,  5.54s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 70%|███████   | 14/20 [01:43<00:33,  5.50s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 70%|███████   | 14/20 [01:49<00:33,  5.50s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.48s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 75%|███████▌  | 15/20 [01:54<00:27,  5.48s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 80%|████████  | 16/20 [01:54<00:21,  5.47s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 80%|████████  | 16/20 [02:00<00:21,  5.47s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.46s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.46s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.45s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 90%|█████████ | 18/20 [02:10<00:10,  5.45s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 95%|█████████▌| 19/20 [02:10<00:05,  5.44s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.44s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
100%|██████████| 20/20 [02:16<00:00,  5.44s/iter, val_loss=0.000189, val_og_loss=0.000196]
100%|██████████| 20/20 [02:16<00:00,  6.82s/iter, val_loss=0.000189, val_og_loss=0.000196]
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:47:06<1:29:24, 894.05s/it]INFO:ParoQuant:Best val loss: 0.00018856232054531574, Original val loss: 0.00019554731261450797
+
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:47:07<1:29:24, 894.05s/it]
                                                                       

Processing Blocks:  81%|████████▏ | 26/32 [5:47:18<1:29:24, 894.05s/it]
Processing Blocks:  84%|████████▍ | 27/32 [5:47:56<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:47:57<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:48:15<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:48:15<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:50:28<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:50:28<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:51:22<1:13:57, 887.45s/it]2025-12-22 20:17:24,473 - INFO - Best val loss: 0.00018856232054531574, Original val loss: 0.00019554731261450797
+[INFO] Saving Layer 26 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 27...
+[INFO] Starting Sequential Training for Layer 27
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=0.00236, val_og_loss=0.00268]
+
  5%|▌         | 1/20 [00:13<04:09, 13.11s/iter, val_loss=0.00236, val_og_loss=0.00268]
+
  5%|▌         | 1/20 [00:14<04:09, 13.11s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 10%|█         | 2/20 [00:14<01:55,  6.39s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 10%|█         | 2/20 [00:16<01:55,  6.39s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 15%|█▌        | 3/20 [00:16<01:12,  4.25s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 15%|█▌        | 3/20 [00:18<01:12,  4.25s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 20%|██        | 4/20 [00:18<00:51,  3.24s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 20%|██        | 4/20 [00:19<00:51,  3.24s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 25%|██▌       | 5/20 [00:19<00:40,  2.67s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 25%|██▌       | 5/20 [00:21<00:40,  2.67s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 30%|███       | 6/20 [00:21<00:32,  2.34s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 30%|███       | 6/20 [00:23<00:32,  2.34s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.13s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 35%|███▌      | 7/20 [00:24<00:27,  2.13s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 40%|████      | 8/20 [00:24<00:23,  1.98s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 40%|████      | 8/20 [00:26<00:23,  1.98s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 45%|████▌     | 9/20 [00:26<00:20,  1.89s/iter, val_loss=0.00242, val_og_loss=0.00268]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.89s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 50%|█████     | 10/20 [00:28<00:18,  1.82s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 50%|█████     | 10/20 [00:29<00:18,  1.82s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 55%|█████▌    | 11/20 [00:29<00:16,  1.79s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.79s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 60%|██████    | 12/20 [00:31<00:14,  1.75s/iter, val_loss=0.00241, val_og_loss=0.00268]
+
 60%|██████    | 12/20 [00:33<00:14,  1.75s/iter, val_loss=0.0024, val_og_loss=0.00268] 
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.73s/iter, val_loss=0.0024, val_og_loss=0.00268]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=0.00239, val_og_loss=0.00268]
+
 70%|███████   | 14/20 [00:34<00:10,  1.71s/iter, val_loss=0.00239, val_og_loss=0.00268]
+
 70%|███████   | 14/20 [00:36<00:10,  1.71s/iter, val_loss=0.00239, val_og_loss=0.00268]
+
 75%|███████▌  | 15/20 [00:36<00:08,  1.71s/iter, val_loss=0.00239, val_og_loss=0.00268]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=0.00238, val_og_loss=0.00268]
+
 80%|████████  | 16/20 [00:38<00:06,  1.70s/iter, val_loss=0.00238, val_og_loss=0.00268]
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=0.00237, val_og_loss=0.00268]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=0.00237, val_og_loss=0.00268]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.69s/iter, val_loss=0.00238, val_og_loss=0.00268]
+
 90%|█████████ | 18/20 [00:41<00:03,  1.69s/iter, val_loss=0.00238, val_og_loss=0.00268]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=0.00237, val_og_loss=0.00268]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.70s/iter, val_loss=0.00237, val_og_loss=0.00268]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.70s/iter, val_loss=0.00236, val_og_loss=0.00268]
+
100%|██████████| 20/20 [00:45<00:00,  1.69s/iter, val_loss=0.00236, val_og_loss=0.00268]
100%|██████████| 20/20 [00:45<00:00,  2.26s/iter, val_loss=0.00236, val_og_loss=0.00268]
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:52:07<1:13:57, 887.45s/it]INFO:ParoQuant:Best val loss: 0.00235559931024909, Original val loss: 0.0026816329918801785
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:52:09<1:13:57, 887.45s/it]2025-12-22 20:22:25,819 - INFO - Best val loss: 0.00235559931024909, Original val loss: 0.0026816329918801785
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.00615, val_og_loss=0.00681]
+
  5%|▌         | 1/20 [00:10<03:22, 10.65s/iter, val_loss=0.00615, val_og_loss=0.00681]
+
  5%|▌         | 1/20 [00:12<03:22, 10.65s/iter, val_loss=0.00629, val_og_loss=0.00681]
+
 10%|█         | 2/20 [00:12<01:33,  5.21s/iter, val_loss=0.00629, val_og_loss=0.00681]
+
 10%|█         | 2/20 [00:13<01:33,  5.21s/iter, val_loss=0.0063, val_og_loss=0.00681] 
+
 15%|█▌        | 3/20 [00:13<00:58,  3.47s/iter, val_loss=0.0063, val_og_loss=0.00681]
+
 15%|█▌        | 3/20 [00:14<00:58,  3.47s/iter, val_loss=0.00632, val_og_loss=0.00681]
+
 20%|██        | 4/20 [00:14<00:42,  2.65s/iter, val_loss=0.00632, val_og_loss=0.00681]
+
 20%|██        | 4/20 [00:16<00:42,  2.65s/iter, val_loss=0.00632, val_og_loss=0.00681]
+
 25%|██▌       | 5/20 [00:16<00:32,  2.19s/iter, val_loss=0.00632, val_og_loss=0.00681]
+
 25%|██▌       | 5/20 [00:17<00:32,  2.19s/iter, val_loss=0.00624, val_og_loss=0.00681]
+
 30%|███       | 6/20 [00:17<00:26,  1.92s/iter, val_loss=0.00624, val_og_loss=0.00681]
+
 30%|███       | 6/20 [00:19<00:26,  1.92s/iter, val_loss=0.00632, val_og_loss=0.00681]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.75s/iter, val_loss=0.00632, val_og_loss=0.00681]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.75s/iter, val_loss=0.00629, val_og_loss=0.00681]
+
 40%|████      | 8/20 [00:20<00:19,  1.64s/iter, val_loss=0.00629, val_og_loss=0.00681]
+
 40%|████      | 8/20 [00:21<00:19,  1.64s/iter, val_loss=0.00626, val_og_loss=0.00681]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.56s/iter, val_loss=0.00626, val_og_loss=0.00681]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.56s/iter, val_loss=0.0063, val_og_loss=0.00681] 
+
 50%|█████     | 10/20 [00:23<00:15,  1.51s/iter, val_loss=0.0063, val_og_loss=0.00681]
+
 50%|█████     | 10/20 [00:24<00:15,  1.51s/iter, val_loss=0.00628, val_og_loss=0.00681]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.48s/iter, val_loss=0.00628, val_og_loss=0.00681]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.48s/iter, val_loss=0.00629, val_og_loss=0.00681]
+
 60%|██████    | 12/20 [00:25<00:11,  1.45s/iter, val_loss=0.00629, val_og_loss=0.00681]
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.00624, val_og_loss=0.00681]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.43s/iter, val_loss=0.00624, val_og_loss=0.00681]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.43s/iter, val_loss=0.00628, val_og_loss=0.00681]
+
 70%|███████   | 14/20 [00:28<00:08,  1.42s/iter, val_loss=0.00628, val_og_loss=0.00681]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.00628, val_og_loss=0.00681]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.00628, val_og_loss=0.00681]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.41s/iter, val_loss=0.00626, val_og_loss=0.00681]
+
 80%|████████  | 16/20 [00:31<00:05,  1.40s/iter, val_loss=0.00626, val_og_loss=0.00681]
+
 80%|████████  | 16/20 [00:32<00:05,  1.40s/iter, val_loss=0.00622, val_og_loss=0.00681]
+
 85%|████████▌ | 17/20 [00:32<00:04,  1.40s/iter, val_loss=0.00622, val_og_loss=0.00681]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00621, val_og_loss=0.00681]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.00621, val_og_loss=0.00681]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.0062, val_og_loss=0.00681] 
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.39s/iter, val_loss=0.0062, val_og_loss=0.00681]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.39s/iter, val_loss=0.00621, val_og_loss=0.00681]
+
100%|██████████| 20/20 [00:37<00:00,  1.39s/iter, val_loss=0.00621, val_og_loss=0.00681]
100%|██████████| 20/20 [00:37<00:00,  1.85s/iter, val_loss=0.00621, val_og_loss=0.00681]
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:52:46<1:13:57, 887.45s/it]INFO:ParoQuant:Best val loss: 0.0061473846435546875, Original val loss: 0.006814828608185053
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:52:48<1:13:57, 887.45s/it]2025-12-22 20:23:04,757 - INFO - Best val loss: 0.0061473846435546875, Original val loss: 0.006814828608185053
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=0.00145, val_og_loss=0.00168]
+
  5%|▌         | 1/20 [00:11<03:30, 11.09s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
  5%|▌         | 1/20 [00:12<03:30, 11.09s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 10%|█         | 2/20 [00:12<01:36,  5.39s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 10%|█         | 2/20 [00:13<01:36,  5.39s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 15%|█▌        | 3/20 [00:13<01:00,  3.57s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 15%|█▌        | 3/20 [00:15<01:00,  3.57s/iter, val_loss=0.00146, val_og_loss=0.00168]
+
 20%|██        | 4/20 [00:15<00:43,  2.71s/iter, val_loss=0.00146, val_og_loss=0.00168]
+
 20%|██        | 4/20 [00:16<00:43,  2.71s/iter, val_loss=0.00146, val_og_loss=0.00168]
+
 25%|██▌       | 5/20 [00:16<00:33,  2.24s/iter, val_loss=0.00146, val_og_loss=0.00168]
+
 25%|██▌       | 5/20 [00:18<00:33,  2.24s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 30%|███       | 6/20 [00:18<00:27,  1.95s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 30%|███       | 6/20 [00:19<00:27,  1.95s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.77s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.77s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 40%|████      | 8/20 [00:20<00:19,  1.65s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 40%|████      | 8/20 [00:22<00:19,  1.65s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.58s/iter, val_loss=0.00145, val_og_loss=0.00168]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.58s/iter, val_loss=0.00144, val_og_loss=0.00168]
+
 50%|█████     | 10/20 [00:23<00:15,  1.52s/iter, val_loss=0.00144, val_og_loss=0.00168]
+
 50%|█████     | 10/20 [00:25<00:15,  1.52s/iter, val_loss=0.00144, val_og_loss=0.00168]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.48s/iter, val_loss=0.00144, val_og_loss=0.00168]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.48s/iter, val_loss=0.00144, val_og_loss=0.00168]
+
 60%|██████    | 12/20 [00:26<00:11,  1.45s/iter, val_loss=0.00144, val_og_loss=0.00168]
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.43s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 65%|██████▌   | 13/20 [00:29<00:10,  1.43s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 70%|███████   | 14/20 [00:29<00:08,  1.43s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 70%|███████   | 14/20 [00:30<00:08,  1.43s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 75%|███████▌  | 15/20 [00:32<00:07,  1.41s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 80%|████████  | 16/20 [00:32<00:05,  1.41s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.40s/iter, val_loss=0.00143, val_og_loss=0.00168]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
 90%|█████████ | 18/20 [00:36<00:02,  1.40s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
100%|██████████| 20/20 [00:37<00:00,  1.40s/iter, val_loss=0.00142, val_og_loss=0.00168]
100%|██████████| 20/20 [00:37<00:00,  1.88s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:53:26<1:13:57, 887.45s/it]INFO:ParoQuant:Best val loss: 0.0014204855542629957, Original val loss: 0.0016800264129415154
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:53:28<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:53:28<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:53:59<1:13:57, 887.45s/it]2025-12-22 20:23:44,209 - INFO - Best val loss: 0.0014204855542629957, Original val loss: 0.0016800264129415154
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:12<?, ?iter/s, val_loss=6.21e-5, val_og_loss=6.36e-5]
+
  5%|▌         | 1/20 [00:12<04:05, 12.92s/iter, val_loss=6.21e-5, val_og_loss=6.36e-5]
+
  5%|▌         | 1/20 [00:14<04:05, 12.92s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 10%|█         | 2/20 [00:14<01:53,  6.31s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 10%|█         | 2/20 [00:16<01:53,  6.31s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 15%|█▌        | 3/20 [00:16<01:11,  4.20s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 15%|█▌        | 3/20 [00:17<01:11,  4.20s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 20%|██        | 4/20 [00:17<00:51,  3.21s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 20%|██        | 4/20 [00:19<00:51,  3.21s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 25%|██▌       | 5/20 [00:19<00:39,  2.66s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 25%|██▌       | 5/20 [00:21<00:39,  2.66s/iter, val_loss=6.24e-5, val_og_loss=6.36e-5]
+
 30%|███       | 6/20 [00:21<00:32,  2.32s/iter, val_loss=6.24e-5, val_og_loss=6.36e-5]
+
 30%|███       | 6/20 [00:23<00:32,  2.32s/iter, val_loss=6.24e-5, val_og_loss=6.36e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.11s/iter, val_loss=6.24e-5, val_og_loss=6.36e-5]
+
 35%|███▌      | 7/20 [00:24<00:27,  2.11s/iter, val_loss=6.21e-5, val_og_loss=6.36e-5]
+
 40%|████      | 8/20 [00:24<00:23,  1.97s/iter, val_loss=6.21e-5, val_og_loss=6.36e-5]
+
 40%|████      | 8/20 [00:26<00:23,  1.97s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 45%|████▌     | 9/20 [00:26<00:20,  1.88s/iter, val_loss=6.22e-5, val_og_loss=6.36e-5]
+
 45%|████▌     | 9/20 [00:28<00:20,  1.88s/iter, val_loss=6.21e-5, val_og_loss=6.36e-5]
+
 50%|█████     | 10/20 [00:28<00:18,  1.82s/iter, val_loss=6.21e-5, val_og_loss=6.36e-5]
+
 50%|█████     | 10/20 [00:29<00:18,  1.82s/iter, val_loss=6.23e-5, val_og_loss=6.36e-5]
+
 55%|█████▌    | 11/20 [00:29<00:15,  1.77s/iter, val_loss=6.23e-5, val_og_loss=6.36e-5]
+
 55%|█████▌    | 11/20 [00:31<00:15,  1.77s/iter, val_loss=6.2e-5, val_og_loss=6.36e-5] 
+
 60%|██████    | 12/20 [00:31<00:13,  1.74s/iter, val_loss=6.2e-5, val_og_loss=6.36e-5]
+
 60%|██████    | 12/20 [00:33<00:13,  1.74s/iter, val_loss=6.2e-5, val_og_loss=6.36e-5]
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.72s/iter, val_loss=6.2e-5, val_og_loss=6.36e-5]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.72s/iter, val_loss=6.17e-5, val_og_loss=6.36e-5]
+
 70%|███████   | 14/20 [00:34<00:10,  1.71s/iter, val_loss=6.17e-5, val_og_loss=6.36e-5]
+
 70%|███████   | 14/20 [00:36<00:10,  1.71s/iter, val_loss=6.19e-5, val_og_loss=6.36e-5]
+
 75%|███████▌  | 15/20 [00:36<00:08,  1.70s/iter, val_loss=6.19e-5, val_og_loss=6.36e-5]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.70s/iter, val_loss=6.17e-5, val_og_loss=6.36e-5]
+
 80%|████████  | 16/20 [00:38<00:06,  1.69s/iter, val_loss=6.17e-5, val_og_loss=6.36e-5]
+
 80%|████████  | 16/20 [00:39<00:06,  1.69s/iter, val_loss=6.19e-5, val_og_loss=6.36e-5]
+
 85%|████████▌ | 17/20 [00:39<00:05,  1.69s/iter, val_loss=6.19e-5, val_og_loss=6.36e-5]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.69s/iter, val_loss=6.2e-5, val_og_loss=6.36e-5] 
+
 90%|█████████ | 18/20 [00:41<00:03,  1.68s/iter, val_loss=6.2e-5, val_og_loss=6.36e-5]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.68s/iter, val_loss=6.19e-5, val_og_loss=6.36e-5]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.69s/iter, val_loss=6.19e-5, val_og_loss=6.36e-5]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=6.21e-5, val_og_loss=6.36e-5]
+
100%|██████████| 20/20 [00:44<00:00,  1.68s/iter, val_loss=6.21e-5, val_og_loss=6.36e-5]
100%|██████████| 20/20 [00:44<00:00,  2.24s/iter, val_loss=6.21e-5, val_og_loss=6.36e-5]
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:54:44<1:13:57, 887.45s/it]INFO:ParoQuant:Best val loss: 6.172335997689515e-05, Original val loss: 6.357505480991676e-05
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:54:44<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:54:44<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:55:29<1:13:57, 887.45s/it]2025-12-22 20:25:02,367 - INFO - Best val loss: 6.172335997689515e-05, Original val loss: 6.357505480991676e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00178, val_og_loss=0.0018]
+
  5%|▌         | 1/20 [00:29<09:19, 29.42s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
  5%|▌         | 1/20 [00:32<09:19, 29.42s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 10%|█         | 2/20 [00:32<04:06, 13.71s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 10%|█         | 2/20 [00:34<04:06, 13.71s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 15%|█▌        | 3/20 [00:34<02:27,  8.69s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 15%|█▌        | 3/20 [00:37<02:27,  8.69s/iter, val_loss=0.0018, val_og_loss=0.0018] 
+
 20%|██        | 4/20 [00:37<01:41,  6.34s/iter, val_loss=0.0018, val_og_loss=0.0018]
+
 20%|██        | 4/20 [00:40<01:41,  6.34s/iter, val_loss=0.0018, val_og_loss=0.0018]
+
 25%|██▌       | 5/20 [00:40<01:15,  5.03s/iter, val_loss=0.0018, val_og_loss=0.0018]
+
 25%|██▌       | 5/20 [00:43<01:15,  5.03s/iter, val_loss=0.0018, val_og_loss=0.0018]
+
 30%|███       | 6/20 [00:43<00:59,  4.24s/iter, val_loss=0.0018, val_og_loss=0.0018]
+
 30%|███       | 6/20 [00:45<00:59,  4.24s/iter, val_loss=0.0018, val_og_loss=0.0018]
+
 35%|███▌      | 7/20 [00:45<00:48,  3.74s/iter, val_loss=0.0018, val_og_loss=0.0018]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.74s/iter, val_loss=0.0018, val_og_loss=0.0018]
+
 40%|████      | 8/20 [00:48<00:40,  3.42s/iter, val_loss=0.0018, val_og_loss=0.0018]
+
 40%|████      | 8/20 [00:51<00:40,  3.42s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.19s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.19s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 50%|█████     | 10/20 [00:53<00:30,  3.05s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 50%|█████     | 10/20 [00:56<00:30,  3.05s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.94s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.94s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 60%|██████    | 12/20 [00:59<00:22,  2.87s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 60%|██████    | 12/20 [01:01<00:22,  2.87s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.83s/iter, val_loss=0.00179, val_og_loss=0.0018]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.83s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 70%|███████   | 14/20 [01:04<00:16,  2.79s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 70%|███████   | 14/20 [01:07<00:16,  2.79s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 80%|████████  | 16/20 [01:10<00:11,  2.75s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 80%|████████  | 16/20 [01:12<00:11,  2.75s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 85%|████████▌ | 17/20 [01:12<00:08,  2.74s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.73s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.00178, val_og_loss=0.0018]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.00177, val_og_loss=0.0018]
+
100%|██████████| 20/20 [01:20<00:00,  2.72s/iter, val_loss=0.00177, val_og_loss=0.0018]
100%|██████████| 20/20 [01:20<00:00,  4.05s/iter, val_loss=0.00177, val_og_loss=0.0018]
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:56:50<1:13:57, 887.45s/it]INFO:ParoQuant:Best val loss: 0.0017745860386639833, Original val loss: 0.0018041864968836308
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:56:54<1:13:57, 887.45s/it]2025-12-22 20:27:08,348 - INFO - Best val loss: 0.0017745860386639833, Original val loss: 0.0018041864968836308
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00097, val_og_loss=0.000988]
+
  5%|▌         | 1/20 [00:29<09:29, 29.96s/iter, val_loss=0.00097, val_og_loss=0.000988]
+
  5%|▌         | 1/20 [00:32<09:29, 29.96s/iter, val_loss=0.000974, val_og_loss=0.000988]
+
 10%|█         | 2/20 [00:32<04:10, 13.93s/iter, val_loss=0.000974, val_og_loss=0.000988]
+
 10%|█         | 2/20 [00:35<04:10, 13.93s/iter, val_loss=0.000977, val_og_loss=0.000988]
+
 15%|█▌        | 3/20 [00:35<02:29,  8.81s/iter, val_loss=0.000977, val_og_loss=0.000988]
+
 15%|█▌        | 3/20 [00:38<02:29,  8.81s/iter, val_loss=0.000983, val_og_loss=0.000988]
+
 20%|██        | 4/20 [00:38<01:42,  6.40s/iter, val_loss=0.000983, val_og_loss=0.000988]
+
 20%|██        | 4/20 [00:40<01:42,  6.40s/iter, val_loss=0.000985, val_og_loss=0.000988]
+
 25%|██▌       | 5/20 [00:40<01:16,  5.08s/iter, val_loss=0.000985, val_og_loss=0.000988]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.08s/iter, val_loss=0.000983, val_og_loss=0.000988]
+
 30%|███       | 6/20 [00:43<00:59,  4.28s/iter, val_loss=0.000983, val_og_loss=0.000988]
+
 30%|███       | 6/20 [00:46<00:59,  4.28s/iter, val_loss=0.000983, val_og_loss=0.000988]
+
 35%|███▌      | 7/20 [00:46<00:49,  3.77s/iter, val_loss=0.000983, val_og_loss=0.000988]
+
 35%|███▌      | 7/20 [00:48<00:49,  3.77s/iter, val_loss=0.000981, val_og_loss=0.000988]
+
 40%|████      | 8/20 [00:48<00:41,  3.44s/iter, val_loss=0.000981, val_og_loss=0.000988]
+
 40%|████      | 8/20 [00:51<00:41,  3.44s/iter, val_loss=0.000983, val_og_loss=0.000988]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.21s/iter, val_loss=0.000983, val_og_loss=0.000988]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.21s/iter, val_loss=0.000982, val_og_loss=0.000988]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.000982, val_og_loss=0.000988]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.00098, val_og_loss=0.000988] 
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.00098, val_og_loss=0.000988]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.96s/iter, val_loss=0.000979, val_og_loss=0.000988]
+
 60%|██████    | 12/20 [00:59<00:23,  2.89s/iter, val_loss=0.000979, val_og_loss=0.000988]
+
 60%|██████    | 12/20 [01:02<00:23,  2.89s/iter, val_loss=0.000978, val_og_loss=0.000988]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.84s/iter, val_loss=0.000978, val_og_loss=0.000988]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.84s/iter, val_loss=0.000976, val_og_loss=0.000988]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.000976, val_og_loss=0.000988]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.000975, val_og_loss=0.000988]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.77s/iter, val_loss=0.000975, val_og_loss=0.000988]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.000974, val_og_loss=0.000988]
+
 80%|████████  | 16/20 [01:10<00:11,  2.75s/iter, val_loss=0.000974, val_og_loss=0.000988]
+
 80%|████████  | 16/20 [01:13<00:11,  2.75s/iter, val_loss=0.000973, val_og_loss=0.000988]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.000973, val_og_loss=0.000988]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.000973, val_og_loss=0.000988]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.73s/iter, val_loss=0.000973, val_og_loss=0.000988]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.73s/iter, val_loss=0.000971, val_og_loss=0.000988]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.000971, val_og_loss=0.000988]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.000971, val_og_loss=0.000988]
+
100%|██████████| 20/20 [01:21<00:00,  2.73s/iter, val_loss=0.000971, val_og_loss=0.000988]
100%|██████████| 20/20 [01:21<00:00,  4.08s/iter, val_loss=0.000971, val_og_loss=0.000988]
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:58:16<1:13:57, 887.45s/it]INFO:ParoQuant:Best val loss: 0.0009701523813419044, Original val loss: 0.000988280982710421
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:58:18<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:58:18<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [5:59:17<1:13:57, 887.45s/it]2025-12-22 20:28:34,269 - INFO - Best val loss: 0.0009701523813419044, Original val loss: 0.000988280982710421
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:34<?, ?iter/s, val_loss=0.000234, val_og_loss=0.00025]
+
  5%|▌         | 1/20 [00:34<11:01, 34.81s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
  5%|▌         | 1/20 [00:40<11:01, 34.81s/iter, val_loss=0.000236, val_og_loss=0.00025]
+
 10%|█         | 2/20 [00:40<05:15, 17.53s/iter, val_loss=0.000236, val_og_loss=0.00025]
+
 10%|█         | 2/20 [00:45<05:15, 17.53s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 15%|█▌        | 3/20 [00:45<03:23, 12.00s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 15%|█▌        | 3/20 [00:51<03:23, 12.00s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 20%|██        | 4/20 [00:51<02:30,  9.40s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 20%|██        | 4/20 [00:56<02:30,  9.40s/iter, val_loss=0.000236, val_og_loss=0.00025]
+
 25%|██▌       | 5/20 [00:56<01:59,  7.96s/iter, val_loss=0.000236, val_og_loss=0.00025]
+
 25%|██▌       | 5/20 [01:01<01:59,  7.96s/iter, val_loss=0.000236, val_og_loss=0.00025]
+
 30%|███       | 6/20 [01:01<01:39,  7.10s/iter, val_loss=0.000236, val_og_loss=0.00025]
+
 30%|███       | 6/20 [01:07<01:39,  7.10s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 35%|███▌      | 7/20 [01:07<01:25,  6.55s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 35%|███▌      | 7/20 [01:12<01:25,  6.55s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 40%|████      | 8/20 [01:12<01:14,  6.19s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 40%|████      | 8/20 [01:18<01:14,  6.19s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 45%|████▌     | 9/20 [01:18<01:05,  5.95s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 45%|████▌     | 9/20 [01:23<01:05,  5.95s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 50%|█████     | 10/20 [01:23<00:57,  5.78s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 50%|█████     | 10/20 [01:28<00:57,  5.78s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 55%|█████▌    | 11/20 [01:28<00:51,  5.67s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 55%|█████▌    | 11/20 [01:34<00:51,  5.67s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 60%|██████    | 12/20 [01:34<00:44,  5.60s/iter, val_loss=0.000235, val_og_loss=0.00025]
+
 60%|██████    | 12/20 [01:39<00:44,  5.60s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 65%|██████▌   | 13/20 [01:39<00:38,  5.55s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 65%|██████▌   | 13/20 [01:45<00:38,  5.55s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 70%|███████   | 14/20 [01:45<00:33,  5.51s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 70%|███████   | 14/20 [01:50<00:33,  5.51s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 75%|███████▌  | 15/20 [01:50<00:27,  5.49s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 75%|███████▌  | 15/20 [01:56<00:27,  5.49s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 80%|████████  | 16/20 [01:56<00:21,  5.47s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 80%|████████  | 16/20 [02:01<00:21,  5.47s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 85%|████████▌ | 17/20 [02:01<00:16,  5.45s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 85%|████████▌ | 17/20 [02:06<00:16,  5.45s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 90%|█████████ | 18/20 [02:06<00:10,  5.44s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 90%|█████████ | 18/20 [02:12<00:10,  5.44s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 95%|█████████▌| 19/20 [02:12<00:05,  5.44s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
 95%|█████████▌| 19/20 [02:17<00:05,  5.44s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
100%|██████████| 20/20 [02:17<00:00,  5.43s/iter, val_loss=0.000234, val_og_loss=0.00025]
100%|██████████| 20/20 [02:17<00:00,  6.89s/iter, val_loss=0.000234, val_og_loss=0.00025]
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [6:01:35<1:13:57, 887.45s/it]INFO:ParoQuant:Best val loss: 0.00023367530957330018, Original val loss: 0.00024966703495010734
+
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [6:01:36<1:13:57, 887.45s/it]
                                                                       

Processing Blocks:  84%|████████▍ | 27/32 [6:01:46<1:13:57, 887.45s/it]
Processing Blocks:  88%|████████▊ | 28/32 [6:02:25<58:47, 881.80s/it]  
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:02:25<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:02:42<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:02:42<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:04:57<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:04:57<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:05:59<58:47, 881.80s/it]2025-12-22 20:31:53,671 - INFO - Best val loss: 0.00023367530957330018, Original val loss: 0.00024966703495010734
+[INFO] Saving Layer 27 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 28...
+[INFO] Starting Sequential Training for Layer 28
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00231, val_og_loss=0.00267]
+
  5%|▌         | 1/20 [00:15<04:53, 15.44s/iter, val_loss=0.00231, val_og_loss=0.00267]
+
  5%|▌         | 1/20 [00:17<04:53, 15.44s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 10%|█         | 2/20 [00:17<02:12,  7.35s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 10%|█         | 2/20 [00:18<02:12,  7.35s/iter, val_loss=0.00235, val_og_loss=0.00267]
+
 15%|█▌        | 3/20 [00:18<01:20,  4.76s/iter, val_loss=0.00235, val_og_loss=0.00267]
+
 15%|█▌        | 3/20 [00:20<01:20,  4.76s/iter, val_loss=0.00234, val_og_loss=0.00267]
+
 20%|██        | 4/20 [00:20<00:56,  3.54s/iter, val_loss=0.00234, val_og_loss=0.00267]
+
 20%|██        | 4/20 [00:22<00:56,  3.54s/iter, val_loss=0.00235, val_og_loss=0.00267]
+
 25%|██▌       | 5/20 [00:22<00:43,  2.87s/iter, val_loss=0.00235, val_og_loss=0.00267]
+
 25%|██▌       | 5/20 [00:23<00:43,  2.87s/iter, val_loss=0.00235, val_og_loss=0.00267]
+
 30%|███       | 6/20 [00:23<00:34,  2.46s/iter, val_loss=0.00235, val_og_loss=0.00267]
+
 30%|███       | 6/20 [00:25<00:34,  2.46s/iter, val_loss=0.00234, val_og_loss=0.00267]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.21s/iter, val_loss=0.00234, val_og_loss=0.00267]
+
 35%|███▌      | 7/20 [00:27<00:28,  2.21s/iter, val_loss=0.00234, val_og_loss=0.00267]
+
 40%|████      | 8/20 [00:27<00:24,  2.04s/iter, val_loss=0.00234, val_og_loss=0.00267]
+
 40%|████      | 8/20 [00:28<00:24,  2.04s/iter, val_loss=0.00234, val_og_loss=0.00267]
+
 45%|████▌     | 9/20 [00:28<00:21,  1.93s/iter, val_loss=0.00234, val_og_loss=0.00267]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.93s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 50%|█████     | 10/20 [00:30<00:18,  1.85s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 50%|█████     | 10/20 [00:32<00:18,  1.85s/iter, val_loss=0.00235, val_og_loss=0.00267]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.80s/iter, val_loss=0.00235, val_og_loss=0.00267]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.80s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 60%|██████    | 12/20 [00:33<00:14,  1.76s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 60%|██████    | 12/20 [00:35<00:14,  1.76s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.73s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.73s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=0.00233, val_og_loss=0.00267]
+
 70%|███████   | 14/20 [00:38<00:10,  1.72s/iter, val_loss=0.00231, val_og_loss=0.00267]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=0.00231, val_og_loss=0.00267]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.71s/iter, val_loss=0.0023, val_og_loss=0.00267] 
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=0.0023, val_og_loss=0.00267]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=0.00231, val_og_loss=0.00267]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.69s/iter, val_loss=0.00231, val_og_loss=0.00267]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.69s/iter, val_loss=0.0023, val_og_loss=0.00267] 
+
 90%|█████████ | 18/20 [00:43<00:03,  1.68s/iter, val_loss=0.0023, val_og_loss=0.00267]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.68s/iter, val_loss=0.00229, val_og_loss=0.00267]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.68s/iter, val_loss=0.00229, val_og_loss=0.00267]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.68s/iter, val_loss=0.0023, val_og_loss=0.00267] 
+
100%|██████████| 20/20 [00:47<00:00,  1.68s/iter, val_loss=0.0023, val_og_loss=0.00267]
100%|██████████| 20/20 [00:47<00:00,  2.37s/iter, val_loss=0.0023, val_og_loss=0.00267]
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:06:46<58:47, 881.80s/it]INFO:ParoQuant:Best val loss: 0.0022940426133573055, Original val loss: 0.002665424020960927
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:06:47<58:47, 881.80s/it]2025-12-22 20:37:04,468 - INFO - Best val loss: 0.0022940426133573055, Original val loss: 0.002665424020960927
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.0054, val_og_loss=0.00603]
+
  5%|▌         | 1/20 [00:10<03:18, 10.45s/iter, val_loss=0.0054, val_og_loss=0.00603]
+
  5%|▌         | 1/20 [00:11<03:18, 10.45s/iter, val_loss=0.00553, val_og_loss=0.00603]
+
 10%|█         | 2/20 [00:11<01:32,  5.12s/iter, val_loss=0.00553, val_og_loss=0.00603]
+
 10%|█         | 2/20 [00:13<01:32,  5.12s/iter, val_loss=0.00554, val_og_loss=0.00603]
+
 15%|█▌        | 3/20 [00:13<00:58,  3.42s/iter, val_loss=0.00554, val_og_loss=0.00603]
+
 15%|█▌        | 3/20 [00:14<00:58,  3.42s/iter, val_loss=0.00553, val_og_loss=0.00603]
+
 20%|██        | 4/20 [00:14<00:41,  2.62s/iter, val_loss=0.00553, val_og_loss=0.00603]
+
 20%|██        | 4/20 [00:16<00:41,  2.62s/iter, val_loss=0.00556, val_og_loss=0.00603]
+
 25%|██▌       | 5/20 [00:16<00:32,  2.17s/iter, val_loss=0.00556, val_og_loss=0.00603]
+
 25%|██▌       | 5/20 [00:17<00:32,  2.17s/iter, val_loss=0.00551, val_og_loss=0.00603]
+
 30%|███       | 6/20 [00:17<00:26,  1.92s/iter, val_loss=0.00551, val_og_loss=0.00603]
+
 30%|███       | 6/20 [00:18<00:26,  1.92s/iter, val_loss=0.00556, val_og_loss=0.00603]
+
 35%|███▌      | 7/20 [00:18<00:22,  1.74s/iter, val_loss=0.00556, val_og_loss=0.00603]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.74s/iter, val_loss=0.00557, val_og_loss=0.00603]
+
 40%|████      | 8/20 [00:20<00:19,  1.63s/iter, val_loss=0.00557, val_og_loss=0.00603]
+
 40%|████      | 8/20 [00:21<00:19,  1.63s/iter, val_loss=0.00553, val_og_loss=0.00603]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.55s/iter, val_loss=0.00553, val_og_loss=0.00603]
+
 45%|████▌     | 9/20 [00:22<00:17,  1.55s/iter, val_loss=0.00554, val_og_loss=0.00603]
+
 50%|█████     | 10/20 [00:22<00:15,  1.51s/iter, val_loss=0.00554, val_og_loss=0.00603]
+
 50%|█████     | 10/20 [00:24<00:15,  1.51s/iter, val_loss=0.00553, val_og_loss=0.00603]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.47s/iter, val_loss=0.00553, val_og_loss=0.00603]
+
 55%|█████▌    | 11/20 [00:25<00:13,  1.47s/iter, val_loss=0.00551, val_og_loss=0.00603]
+
 60%|██████    | 12/20 [00:25<00:11,  1.44s/iter, val_loss=0.00551, val_og_loss=0.00603]
+
 60%|██████    | 12/20 [00:27<00:11,  1.44s/iter, val_loss=0.00544, val_og_loss=0.00603]
+
 65%|██████▌   | 13/20 [00:27<00:09,  1.42s/iter, val_loss=0.00544, val_og_loss=0.00603]
+
 65%|██████▌   | 13/20 [00:28<00:09,  1.42s/iter, val_loss=0.00548, val_og_loss=0.00603]
+
 70%|███████   | 14/20 [00:28<00:08,  1.41s/iter, val_loss=0.00548, val_og_loss=0.00603]
+
 70%|███████   | 14/20 [00:29<00:08,  1.41s/iter, val_loss=0.00544, val_og_loss=0.00603]
+
 75%|███████▌  | 15/20 [00:29<00:07,  1.40s/iter, val_loss=0.00544, val_og_loss=0.00603]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.40s/iter, val_loss=0.00545, val_og_loss=0.00603]
+
 80%|████████  | 16/20 [00:31<00:05,  1.40s/iter, val_loss=0.00545, val_og_loss=0.00603]
+
 80%|████████  | 16/20 [00:32<00:05,  1.40s/iter, val_loss=0.00544, val_og_loss=0.00603]
+
 85%|████████▌ | 17/20 [00:32<00:04,  1.40s/iter, val_loss=0.00544, val_og_loss=0.00603]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.40s/iter, val_loss=0.00541, val_og_loss=0.00603]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.00541, val_og_loss=0.00603]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00544, val_og_loss=0.00603]
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.40s/iter, val_loss=0.00544, val_og_loss=0.00603]
+
 95%|█████████▌| 19/20 [00:36<00:01,  1.40s/iter, val_loss=0.00543, val_og_loss=0.00603]
+
100%|██████████| 20/20 [00:36<00:00,  1.40s/iter, val_loss=0.00543, val_og_loss=0.00603]
100%|██████████| 20/20 [00:36<00:00,  1.84s/iter, val_loss=0.00543, val_og_loss=0.00603]
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:07:24<58:47, 881.80s/it]INFO:ParoQuant:Best val loss: 0.005397523753345013, Original val loss: 0.006029196083545685
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:07:26<58:47, 881.80s/it]2025-12-22 20:37:42,550 - INFO - Best val loss: 0.005397523753345013, Original val loss: 0.006029196083545685
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:10<?, ?iter/s, val_loss=0.0014, val_og_loss=0.00168]
+
  5%|▌         | 1/20 [00:10<03:22, 10.68s/iter, val_loss=0.0014, val_og_loss=0.00168]
+
  5%|▌         | 1/20 [00:12<03:22, 10.68s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 10%|█         | 2/20 [00:12<01:34,  5.24s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 10%|█         | 2/20 [00:13<01:34,  5.24s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 15%|█▌        | 3/20 [00:13<00:59,  3.48s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 15%|█▌        | 3/20 [00:14<00:59,  3.48s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 20%|██        | 4/20 [00:14<00:42,  2.65s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 20%|██        | 4/20 [00:16<00:42,  2.65s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
 25%|██▌       | 5/20 [00:16<00:32,  2.20s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
 25%|██▌       | 5/20 [00:17<00:32,  2.20s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 30%|███       | 6/20 [00:17<00:26,  1.92s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 30%|███       | 6/20 [00:19<00:26,  1.92s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
 35%|███▌      | 7/20 [00:19<00:22,  1.76s/iter, val_loss=0.00142, val_og_loss=0.00168]
+
 35%|███▌      | 7/20 [00:20<00:22,  1.76s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 40%|████      | 8/20 [00:20<00:19,  1.64s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 40%|████      | 8/20 [00:21<00:19,  1.64s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 45%|████▌     | 9/20 [00:21<00:17,  1.56s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 45%|████▌     | 9/20 [00:23<00:17,  1.56s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 50%|█████     | 10/20 [00:23<00:15,  1.51s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 50%|█████     | 10/20 [00:24<00:15,  1.51s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 55%|█████▌    | 11/20 [00:24<00:13,  1.48s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 55%|█████▌    | 11/20 [00:26<00:13,  1.48s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 60%|██████    | 12/20 [00:26<00:11,  1.45s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 60%|██████    | 12/20 [00:27<00:11,  1.45s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 65%|██████▌   | 13/20 [00:27<00:10,  1.43s/iter, val_loss=0.00141, val_og_loss=0.00168]
+
 65%|██████▌   | 13/20 [00:28<00:10,  1.43s/iter, val_loss=0.0014, val_og_loss=0.00168] 
+
 70%|███████   | 14/20 [00:28<00:08,  1.42s/iter, val_loss=0.0014, val_og_loss=0.00168]
+
 70%|███████   | 14/20 [00:30<00:08,  1.42s/iter, val_loss=0.0014, val_og_loss=0.00168]
+
 75%|███████▌  | 15/20 [00:30<00:07,  1.41s/iter, val_loss=0.0014, val_og_loss=0.00168]
+
 75%|███████▌  | 15/20 [00:31<00:07,  1.41s/iter, val_loss=0.0014, val_og_loss=0.00168]
+
 80%|████████  | 16/20 [00:31<00:05,  1.41s/iter, val_loss=0.0014, val_og_loss=0.00168]
+
 80%|████████  | 16/20 [00:33<00:05,  1.41s/iter, val_loss=0.00139, val_og_loss=0.00168]
+
 85%|████████▌ | 17/20 [00:33<00:04,  1.41s/iter, val_loss=0.00139, val_og_loss=0.00168]
+
 85%|████████▌ | 17/20 [00:34<00:04,  1.41s/iter, val_loss=0.00139, val_og_loss=0.00168]
+
 90%|█████████ | 18/20 [00:34<00:02,  1.40s/iter, val_loss=0.00139, val_og_loss=0.00168]
+
 90%|█████████ | 18/20 [00:35<00:02,  1.40s/iter, val_loss=0.00139, val_og_loss=0.00168]
+
 95%|█████████▌| 19/20 [00:35<00:01,  1.40s/iter, val_loss=0.00139, val_og_loss=0.00168]
+
 95%|█████████▌| 19/20 [00:37<00:01,  1.40s/iter, val_loss=0.00139, val_og_loss=0.00168]
+
100%|██████████| 20/20 [00:37<00:00,  1.40s/iter, val_loss=0.00139, val_og_loss=0.00168]
100%|██████████| 20/20 [00:37<00:00,  1.86s/iter, val_loss=0.00139, val_og_loss=0.00168]
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:08:03<58:47, 881.80s/it]INFO:ParoQuant:Best val loss: 0.0013875557342544198, Original val loss: 0.0016815138515084982
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:08:06<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:08:06<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:08:38<58:47, 881.80s/it]2025-12-22 20:38:21,553 - INFO - Best val loss: 0.0013875557342544198, Original val loss: 0.0016815138515084982
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:11<?, ?iter/s, val_loss=8.1e-5, val_og_loss=8.36e-5]
+
  5%|▌         | 1/20 [00:11<03:46, 11.91s/iter, val_loss=8.1e-5, val_og_loss=8.36e-5]
+
  5%|▌         | 1/20 [00:13<03:46, 11.91s/iter, val_loss=8.11e-5, val_og_loss=8.36e-5]
+
 10%|█         | 2/20 [00:13<01:46,  5.90s/iter, val_loss=8.11e-5, val_og_loss=8.36e-5]
+
 10%|█         | 2/20 [00:15<01:46,  5.90s/iter, val_loss=8.15e-5, val_og_loss=8.36e-5]
+
 15%|█▌        | 3/20 [00:15<01:07,  3.97s/iter, val_loss=8.15e-5, val_og_loss=8.36e-5]
+
 15%|█▌        | 3/20 [00:16<01:07,  3.97s/iter, val_loss=8.16e-5, val_og_loss=8.36e-5]
+
 20%|██        | 4/20 [00:16<00:49,  3.07s/iter, val_loss=8.16e-5, val_og_loss=8.36e-5]
+
 20%|██        | 4/20 [00:18<00:49,  3.07s/iter, val_loss=8.17e-5, val_og_loss=8.36e-5]
+
 25%|██▌       | 5/20 [00:18<00:38,  2.57s/iter, val_loss=8.17e-5, val_og_loss=8.36e-5]
+
 25%|██▌       | 5/20 [00:20<00:38,  2.57s/iter, val_loss=8.17e-5, val_og_loss=8.36e-5]
+
 30%|███       | 6/20 [00:20<00:31,  2.28s/iter, val_loss=8.17e-5, val_og_loss=8.36e-5]
+
 30%|███       | 6/20 [00:22<00:31,  2.28s/iter, val_loss=8.15e-5, val_og_loss=8.36e-5]
+
 35%|███▌      | 7/20 [00:22<00:27,  2.08s/iter, val_loss=8.15e-5, val_og_loss=8.36e-5]
+
 35%|███▌      | 7/20 [00:23<00:27,  2.08s/iter, val_loss=8.15e-5, val_og_loss=8.36e-5]
+
 40%|████      | 8/20 [00:23<00:23,  1.95s/iter, val_loss=8.15e-5, val_og_loss=8.36e-5]
+
 40%|████      | 8/20 [00:25<00:23,  1.95s/iter, val_loss=8.12e-5, val_og_loss=8.36e-5]
+
 45%|████▌     | 9/20 [00:25<00:20,  1.86s/iter, val_loss=8.12e-5, val_og_loss=8.36e-5]
+
 45%|████▌     | 9/20 [00:27<00:20,  1.86s/iter, val_loss=8.12e-5, val_og_loss=8.36e-5]
+
 50%|█████     | 10/20 [00:27<00:18,  1.81s/iter, val_loss=8.12e-5, val_og_loss=8.36e-5]
+
 50%|█████     | 10/20 [00:28<00:18,  1.81s/iter, val_loss=8.14e-5, val_og_loss=8.36e-5]
+
 55%|█████▌    | 11/20 [00:28<00:15,  1.77s/iter, val_loss=8.14e-5, val_og_loss=8.36e-5]
+
 55%|█████▌    | 11/20 [00:30<00:15,  1.77s/iter, val_loss=8.08e-5, val_og_loss=8.36e-5]
+
 60%|██████    | 12/20 [00:30<00:13,  1.74s/iter, val_loss=8.08e-5, val_og_loss=8.36e-5]
+
 60%|██████    | 12/20 [00:32<00:13,  1.74s/iter, val_loss=8.14e-5, val_og_loss=8.36e-5]
+
 65%|██████▌   | 13/20 [00:32<00:12,  1.72s/iter, val_loss=8.14e-5, val_og_loss=8.36e-5]
+
 65%|██████▌   | 13/20 [00:33<00:12,  1.72s/iter, val_loss=8.09e-5, val_og_loss=8.36e-5]
+
 70%|███████   | 14/20 [00:33<00:10,  1.70s/iter, val_loss=8.09e-5, val_og_loss=8.36e-5]
+
 70%|███████   | 14/20 [00:35<00:10,  1.70s/iter, val_loss=8.07e-5, val_og_loss=8.36e-5]
+
 75%|███████▌  | 15/20 [00:35<00:08,  1.70s/iter, val_loss=8.07e-5, val_og_loss=8.36e-5]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.70s/iter, val_loss=8.06e-5, val_og_loss=8.36e-5]
+
 80%|████████  | 16/20 [00:37<00:06,  1.69s/iter, val_loss=8.06e-5, val_og_loss=8.36e-5]
+
 80%|████████  | 16/20 [00:38<00:06,  1.69s/iter, val_loss=8.03e-5, val_og_loss=8.36e-5]
+
 85%|████████▌ | 17/20 [00:38<00:05,  1.69s/iter, val_loss=8.03e-5, val_og_loss=8.36e-5]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.69s/iter, val_loss=8.09e-5, val_og_loss=8.36e-5]
+
 90%|█████████ | 18/20 [00:40<00:03,  1.68s/iter, val_loss=8.09e-5, val_og_loss=8.36e-5]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.68s/iter, val_loss=8.05e-5, val_og_loss=8.36e-5]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.69s/iter, val_loss=8.05e-5, val_og_loss=8.36e-5]
+
 95%|█████████▌| 19/20 [00:43<00:01,  1.69s/iter, val_loss=8.08e-5, val_og_loss=8.36e-5]
+
100%|██████████| 20/20 [00:43<00:00,  1.68s/iter, val_loss=8.08e-5, val_og_loss=8.36e-5]
100%|██████████| 20/20 [00:43<00:00,  2.19s/iter, val_loss=8.08e-5, val_og_loss=8.36e-5]
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:09:22<58:47, 881.80s/it]INFO:ParoQuant:Best val loss: 8.029799937503412e-05, Original val loss: 8.356954640476033e-05
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:09:22<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:09:22<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:10:06<58:47, 881.80s/it]2025-12-22 20:39:40,063 - INFO - Best val loss: 8.029799937503412e-05, Original val loss: 8.356954640476033e-05
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:28<?, ?iter/s, val_loss=0.00197, val_og_loss=0.00201]
+
  5%|▌         | 1/20 [00:29<09:11, 29.00s/iter, val_loss=0.00197, val_og_loss=0.00201]
+
  5%|▌         | 1/20 [00:31<09:11, 29.00s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 10%|█         | 2/20 [00:31<04:03, 13.54s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 10%|█         | 2/20 [00:34<04:03, 13.54s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 15%|█▌        | 3/20 [00:34<02:26,  8.59s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 15%|█▌        | 3/20 [00:37<02:26,  8.59s/iter, val_loss=0.002, val_og_loss=0.00201]  
+
 20%|██        | 4/20 [00:37<01:40,  6.28s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 20%|██        | 4/20 [00:39<01:40,  6.28s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 25%|██▌       | 5/20 [00:39<01:14,  4.99s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 25%|██▌       | 5/20 [00:42<01:14,  4.99s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 30%|███       | 6/20 [00:42<00:59,  4.22s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 30%|███       | 6/20 [00:45<00:59,  4.22s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 35%|███▌      | 7/20 [00:45<00:48,  3.73s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.73s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 40%|████      | 8/20 [00:48<00:40,  3.40s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 40%|████      | 8/20 [00:50<00:40,  3.40s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 45%|████▌     | 9/20 [00:50<00:35,  3.19s/iter, val_loss=0.002, val_og_loss=0.00201]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.19s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 50%|█████     | 10/20 [00:53<00:30,  3.04s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 50%|█████     | 10/20 [00:56<00:30,  3.04s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.94s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.94s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 60%|██████    | 12/20 [00:58<00:22,  2.87s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 60%|██████    | 12/20 [01:01<00:22,  2.87s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 65%|██████▌   | 13/20 [01:01<00:19,  2.82s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.82s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 70%|███████   | 14/20 [01:04<00:16,  2.79s/iter, val_loss=0.00199, val_og_loss=0.00201]
+
 70%|███████   | 14/20 [01:07<00:16,  2.79s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.77s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 80%|████████  | 16/20 [01:09<00:11,  2.75s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 80%|████████  | 16/20 [01:12<00:11,  2.75s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 85%|████████▌ | 17/20 [01:12<00:08,  2.74s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.74s/iter, val_loss=0.00198, val_og_loss=0.00201]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.00197, val_og_loss=0.00201]
+
 95%|█████████▌| 19/20 [01:17<00:02,  2.73s/iter, val_loss=0.00197, val_og_loss=0.00201]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.00197, val_og_loss=0.00201]
+
100%|██████████| 20/20 [01:20<00:00,  2.73s/iter, val_loss=0.00197, val_og_loss=0.00201]
100%|██████████| 20/20 [01:20<00:00,  4.03s/iter, val_loss=0.00197, val_og_loss=0.00201]
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:11:27<58:47, 881.80s/it]INFO:ParoQuant:Best val loss: 0.001965629169717431, Original val loss: 0.0020098527893424034
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:11:32<58:47, 881.80s/it]2025-12-22 20:41:45,165 - INFO - Best val loss: 0.001965629169717431, Original val loss: 0.0020098527893424034
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.00114, val_og_loss=0.00116]
+
  5%|▌         | 1/20 [00:30<09:32, 30.12s/iter, val_loss=0.00114, val_og_loss=0.00116]
+
  5%|▌         | 1/20 [00:32<09:32, 30.12s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 10%|█         | 2/20 [00:32<04:12, 14.00s/iter, val_loss=0.00115, val_og_loss=0.00116]
+
 10%|█         | 2/20 [00:35<04:12, 14.00s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 15%|█▌        | 3/20 [00:35<02:30,  8.85s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 15%|█▌        | 3/20 [00:38<02:30,  8.85s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 20%|██        | 4/20 [00:38<01:42,  6.43s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 20%|██        | 4/20 [00:40<01:42,  6.43s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 25%|██▌       | 5/20 [00:40<01:16,  5.09s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.09s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 30%|███       | 6/20 [00:43<00:59,  4.28s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 30%|███       | 6/20 [00:46<00:59,  4.28s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 35%|███▌      | 7/20 [00:46<00:48,  3.76s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 35%|███▌      | 7/20 [00:49<00:48,  3.76s/iter, val_loss=0.00118, val_og_loss=0.00116]
+
 40%|████      | 8/20 [00:49<00:41,  3.43s/iter, val_loss=0.00118, val_og_loss=0.00116]
+
 40%|████      | 8/20 [00:51<00:41,  3.43s/iter, val_loss=0.00118, val_og_loss=0.00116]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.21s/iter, val_loss=0.00118, val_og_loss=0.00116]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.21s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.96s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 60%|██████    | 12/20 [00:59<00:23,  2.88s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 60%|██████    | 12/20 [01:02<00:23,  2.88s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.83s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.83s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 70%|███████   | 14/20 [01:05<00:16,  2.79s/iter, val_loss=0.00117, val_og_loss=0.00116]
+
 70%|███████   | 14/20 [01:08<00:16,  2.79s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.77s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 80%|████████  | 16/20 [01:10<00:11,  2.76s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.74s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
100%|██████████| 20/20 [01:21<00:00,  2.72s/iter, val_loss=0.00116, val_og_loss=0.00116]
100%|██████████| 20/20 [01:21<00:00,  4.08s/iter, val_loss=0.00116, val_og_loss=0.00116]
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:12:54<58:47, 881.80s/it]INFO:ParoQuant:Best val loss: 0.0011374240275472403, Original val loss: 0.0011574503732845187
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:12:55<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:12:55<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:13:53<58:47, 881.80s/it]2025-12-22 20:43:12,231 - INFO - Best val loss: 0.0011374240275472403, Original val loss: 0.0011574503732845187
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.000322, val_og_loss=0.000343]
+
  5%|▌         | 1/20 [00:33<10:34, 33.41s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
  5%|▌         | 1/20 [00:38<10:34, 33.41s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 10%|█         | 2/20 [00:38<05:05, 16.95s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 10%|█         | 2/20 [00:44<05:05, 16.95s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 15%|█▌        | 3/20 [00:44<03:18, 11.69s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 15%|█▌        | 3/20 [00:49<03:18, 11.69s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 20%|██        | 4/20 [00:49<02:27,  9.21s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 20%|██        | 4/20 [00:55<02:27,  9.21s/iter, val_loss=0.000323, val_og_loss=0.000343]
+
 25%|██▌       | 5/20 [00:55<01:57,  7.85s/iter, val_loss=0.000323, val_og_loss=0.000343]
+
 25%|██▌       | 5/20 [01:00<01:57,  7.85s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 30%|███       | 6/20 [01:00<01:38,  7.02s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 30%|███       | 6/20 [01:05<01:38,  7.02s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 35%|███▌      | 7/20 [01:05<01:24,  6.50s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.50s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 40%|████      | 8/20 [01:11<01:13,  6.15s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 40%|████      | 8/20 [01:16<01:13,  6.15s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 45%|████▌     | 9/20 [01:16<01:05,  5.93s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.93s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 50%|█████     | 10/20 [01:22<00:57,  5.77s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 50%|█████     | 10/20 [01:27<00:57,  5.77s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 55%|█████▌    | 11/20 [01:27<00:50,  5.67s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 55%|█████▌    | 11/20 [01:33<00:50,  5.67s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 60%|██████    | 12/20 [01:33<00:44,  5.59s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 60%|██████    | 12/20 [01:38<00:44,  5.59s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.54s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 65%|██████▌   | 13/20 [01:43<00:38,  5.54s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 70%|███████   | 14/20 [01:43<00:33,  5.51s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 70%|███████   | 14/20 [01:49<00:33,  5.51s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.48s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 75%|███████▌  | 15/20 [01:54<00:27,  5.48s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 80%|████████  | 16/20 [01:54<00:21,  5.46s/iter, val_loss=0.000322, val_og_loss=0.000343]
+
 80%|████████  | 16/20 [02:00<00:21,  5.46s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.45s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.45s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.45s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.45s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.44s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
100%|██████████| 20/20 [02:16<00:00,  5.44s/iter, val_loss=0.000321, val_og_loss=0.000343]
100%|██████████| 20/20 [02:16<00:00,  6.82s/iter, val_loss=0.000321, val_og_loss=0.000343]
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:16:10<58:47, 881.80s/it]INFO:ParoQuant:Best val loss: 0.00032092080800794065, Original val loss: 0.0003429403295740485
+
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:16:10<58:47, 881.80s/it]
                                                                     

Processing Blocks:  88%|████████▊ | 28/32 [6:16:21<58:47, 881.80s/it]
Processing Blocks:  91%|█████████ | 29/32 [6:16:58<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:16:59<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:17:16<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:17:16<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:19:31<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:19:31<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:20:34<43:58, 879.39s/it]2025-12-22 20:46:28,157 - INFO - Best val loss: 0.00032092080800794065, Original val loss: 0.0003429403295740485
+[INFO] Saving Layer 28 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 29...
+[INFO] Starting Sequential Training for Layer 29
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00251, val_og_loss=0.00271]
+
  5%|▌         | 1/20 [00:15<04:49, 15.21s/iter, val_loss=0.00251, val_og_loss=0.00271]
+
  5%|▌         | 1/20 [00:16<04:49, 15.21s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 10%|█         | 2/20 [00:16<02:10,  7.28s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 10%|█         | 2/20 [00:18<02:10,  7.28s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 15%|█▌        | 3/20 [00:18<01:20,  4.72s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 15%|█▌        | 3/20 [00:20<01:20,  4.72s/iter, val_loss=0.00254, val_og_loss=0.00271]
+
 20%|██        | 4/20 [00:20<00:56,  3.53s/iter, val_loss=0.00254, val_og_loss=0.00271]
+
 20%|██        | 4/20 [00:21<00:56,  3.53s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 25%|██▌       | 5/20 [00:21<00:42,  2.86s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 25%|██▌       | 5/20 [00:23<00:42,  2.86s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 30%|███       | 6/20 [00:23<00:34,  2.46s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 30%|███       | 6/20 [00:25<00:34,  2.46s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.21s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 35%|███▌      | 7/20 [00:27<00:28,  2.21s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 40%|████      | 8/20 [00:27<00:24,  2.04s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 40%|████      | 8/20 [00:28<00:24,  2.04s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 45%|████▌     | 9/20 [00:28<00:21,  1.93s/iter, val_loss=0.00253, val_og_loss=0.00271]
+
 45%|████▌     | 9/20 [00:30<00:21,  1.93s/iter, val_loss=0.00252, val_og_loss=0.00271]
+
 50%|█████     | 10/20 [00:30<00:18,  1.85s/iter, val_loss=0.00252, val_og_loss=0.00271]
+
 50%|█████     | 10/20 [00:32<00:18,  1.85s/iter, val_loss=0.00251, val_og_loss=0.00271]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.80s/iter, val_loss=0.00251, val_og_loss=0.00271]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.80s/iter, val_loss=0.00252, val_og_loss=0.00271]
+
 60%|██████    | 12/20 [00:33<00:14,  1.76s/iter, val_loss=0.00252, val_og_loss=0.00271]
+
 60%|██████    | 12/20 [00:35<00:14,  1.76s/iter, val_loss=0.0025, val_og_loss=0.00271] 
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.74s/iter, val_loss=0.0025, val_og_loss=0.00271]
+
 65%|██████▌   | 13/20 [00:37<00:12,  1.74s/iter, val_loss=0.0025, val_og_loss=0.00271]
+
 70%|███████   | 14/20 [00:37<00:10,  1.73s/iter, val_loss=0.0025, val_og_loss=0.00271]
+
 70%|███████   | 14/20 [00:38<00:10,  1.73s/iter, val_loss=0.00251, val_og_loss=0.00271]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=0.00251, val_og_loss=0.00271]
+
 75%|███████▌  | 15/20 [00:40<00:08,  1.71s/iter, val_loss=0.0025, val_og_loss=0.00271] 
+
 80%|████████  | 16/20 [00:40<00:06,  1.70s/iter, val_loss=0.0025, val_og_loss=0.00271]
+
 80%|████████  | 16/20 [00:42<00:06,  1.70s/iter, val_loss=0.0025, val_og_loss=0.00271]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.70s/iter, val_loss=0.0025, val_og_loss=0.00271]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.70s/iter, val_loss=0.00249, val_og_loss=0.00271]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.70s/iter, val_loss=0.00249, val_og_loss=0.00271]
+
 90%|█████████ | 18/20 [00:45<00:03,  1.70s/iter, val_loss=0.00249, val_og_loss=0.00271]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.70s/iter, val_loss=0.00249, val_og_loss=0.00271]
+
 95%|█████████▌| 19/20 [00:47<00:01,  1.70s/iter, val_loss=0.00249, val_og_loss=0.00271]
+
100%|██████████| 20/20 [00:47<00:00,  1.69s/iter, val_loss=0.00249, val_og_loss=0.00271]
100%|██████████| 20/20 [00:47<00:00,  2.36s/iter, val_loss=0.00249, val_og_loss=0.00271]
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:21:21<43:58, 879.39s/it]INFO:ParoQuant:Best val loss: 0.002485012635588646, Original val loss: 0.0027076767291873693
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:21:25<43:58, 879.39s/it]2025-12-22 20:51:39,567 - INFO - Best val loss: 0.002485012635588646, Original val loss: 0.0027076767291873693
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.0062, val_og_loss=0.00663]
+
  5%|▌         | 1/20 [00:14<04:37, 14.60s/iter, val_loss=0.0062, val_og_loss=0.00663]
+
  5%|▌         | 1/20 [00:15<04:37, 14.60s/iter, val_loss=0.00633, val_og_loss=0.00663]
+
 10%|█         | 2/20 [00:15<02:02,  6.83s/iter, val_loss=0.00633, val_og_loss=0.00663]
+
 10%|█         | 2/20 [00:17<02:02,  6.83s/iter, val_loss=0.00633, val_og_loss=0.00663]
+
 15%|█▌        | 3/20 [00:17<01:13,  4.34s/iter, val_loss=0.00633, val_og_loss=0.00663]
+
 15%|█▌        | 3/20 [00:18<01:13,  4.34s/iter, val_loss=0.00631, val_og_loss=0.00663]
+
 20%|██        | 4/20 [00:18<00:50,  3.18s/iter, val_loss=0.00631, val_og_loss=0.00663]
+
 20%|██        | 4/20 [00:20<00:50,  3.18s/iter, val_loss=0.00634, val_og_loss=0.00663]
+
 25%|██▌       | 5/20 [00:20<00:37,  2.53s/iter, val_loss=0.00634, val_og_loss=0.00663]
+
 25%|██▌       | 5/20 [00:21<00:37,  2.53s/iter, val_loss=0.00631, val_og_loss=0.00663]
+
 30%|███       | 6/20 [00:21<00:30,  2.15s/iter, val_loss=0.00631, val_og_loss=0.00663]
+
 30%|███       | 6/20 [00:22<00:30,  2.15s/iter, val_loss=0.00629, val_og_loss=0.00663]
+
 35%|███▌      | 7/20 [00:22<00:24,  1.90s/iter, val_loss=0.00629, val_og_loss=0.00663]
+
 35%|███▌      | 7/20 [00:24<00:24,  1.90s/iter, val_loss=0.00627, val_og_loss=0.00663]
+
 40%|████      | 8/20 [00:24<00:20,  1.74s/iter, val_loss=0.00627, val_og_loss=0.00663]
+
 40%|████      | 8/20 [00:25<00:20,  1.74s/iter, val_loss=0.00627, val_og_loss=0.00663]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.63s/iter, val_loss=0.00627, val_og_loss=0.00663]
+
 45%|████▌     | 9/20 [00:27<00:17,  1.63s/iter, val_loss=0.00632, val_og_loss=0.00663]
+
 50%|█████     | 10/20 [00:27<00:15,  1.56s/iter, val_loss=0.00632, val_og_loss=0.00663]
+
 50%|█████     | 10/20 [00:28<00:15,  1.56s/iter, val_loss=0.00628, val_og_loss=0.00663]
+
 55%|█████▌    | 11/20 [00:28<00:13,  1.51s/iter, val_loss=0.00628, val_og_loss=0.00663]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.51s/iter, val_loss=0.00629, val_og_loss=0.00663]
+
 60%|██████    | 12/20 [00:29<00:11,  1.47s/iter, val_loss=0.00629, val_og_loss=0.00663]
+
 60%|██████    | 12/20 [00:31<00:11,  1.47s/iter, val_loss=0.00625, val_og_loss=0.00663]
+
 65%|██████▌   | 13/20 [00:31<00:10,  1.45s/iter, val_loss=0.00625, val_og_loss=0.00663]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.45s/iter, val_loss=0.00615, val_og_loss=0.00663]
+
 70%|███████   | 14/20 [00:32<00:08,  1.43s/iter, val_loss=0.00615, val_og_loss=0.00663]
+
 70%|███████   | 14/20 [00:34<00:08,  1.43s/iter, val_loss=0.00623, val_og_loss=0.00663]
+
 75%|███████▌  | 15/20 [00:34<00:07,  1.42s/iter, val_loss=0.00623, val_og_loss=0.00663]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.42s/iter, val_loss=0.00621, val_og_loss=0.00663]
+
 80%|████████  | 16/20 [00:35<00:05,  1.41s/iter, val_loss=0.00621, val_og_loss=0.00663]
+
 80%|████████  | 16/20 [00:36<00:05,  1.41s/iter, val_loss=0.00617, val_og_loss=0.00663]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.41s/iter, val_loss=0.00617, val_og_loss=0.00663]
+
 85%|████████▌ | 17/20 [00:38<00:04,  1.41s/iter, val_loss=0.00614, val_og_loss=0.00663]
+
 90%|█████████ | 18/20 [00:38<00:02,  1.40s/iter, val_loss=0.00614, val_og_loss=0.00663]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.40s/iter, val_loss=0.00618, val_og_loss=0.00663]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.40s/iter, val_loss=0.00618, val_og_loss=0.00663]
+
 95%|█████████▌| 19/20 [00:41<00:01,  1.40s/iter, val_loss=0.00616, val_og_loss=0.00663]
+
100%|██████████| 20/20 [00:41<00:00,  1.39s/iter, val_loss=0.00616, val_og_loss=0.00663]
100%|██████████| 20/20 [00:41<00:00,  2.05s/iter, val_loss=0.00616, val_og_loss=0.00663]
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:22:06<43:58, 879.39s/it]INFO:ParoQuant:Best val loss: 0.006136463023722172, Original val loss: 0.006633281242102385
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:22:08<43:58, 879.39s/it]2025-12-22 20:52:24,018 - INFO - Best val loss: 0.006136463023722172, Original val loss: 0.006633281242102385
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.0018, val_og_loss=0.00198]
+
  5%|▌         | 1/20 [00:15<04:55, 15.54s/iter, val_loss=0.0018, val_og_loss=0.00198]
+
  5%|▌         | 1/20 [00:16<04:55, 15.54s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 10%|█         | 2/20 [00:16<02:09,  7.22s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 10%|█         | 2/20 [00:18<02:09,  7.22s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 15%|█▌        | 3/20 [00:18<01:17,  4.57s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 15%|█▌        | 3/20 [00:19<01:17,  4.57s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 20%|██        | 4/20 [00:19<00:53,  3.32s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 20%|██        | 4/20 [00:21<00:53,  3.32s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 25%|██▌       | 5/20 [00:21<00:39,  2.62s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 25%|██▌       | 5/20 [00:22<00:39,  2.62s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 30%|███       | 6/20 [00:22<00:30,  2.20s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 30%|███       | 6/20 [00:23<00:30,  2.20s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 35%|███▌      | 7/20 [00:23<00:25,  1.94s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 35%|███▌      | 7/20 [00:25<00:25,  1.94s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 40%|████      | 8/20 [00:25<00:21,  1.78s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 40%|████      | 8/20 [00:26<00:21,  1.78s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 45%|████▌     | 9/20 [00:26<00:18,  1.66s/iter, val_loss=0.00182, val_og_loss=0.00198]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.66s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 50%|█████     | 10/20 [00:28<00:15,  1.58s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 50%|█████     | 10/20 [00:29<00:15,  1.58s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.52s/iter, val_loss=0.00181, val_og_loss=0.00198]
+
 55%|█████▌    | 11/20 [00:30<00:13,  1.52s/iter, val_loss=0.0018, val_og_loss=0.00198] 
+
 60%|██████    | 12/20 [00:30<00:11,  1.49s/iter, val_loss=0.0018, val_og_loss=0.00198]
+
 60%|██████    | 12/20 [00:32<00:11,  1.49s/iter, val_loss=0.0018, val_og_loss=0.00198]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.46s/iter, val_loss=0.0018, val_og_loss=0.00198]
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.46s/iter, val_loss=0.0018, val_og_loss=0.00198]
+
 70%|███████   | 14/20 [00:33<00:08,  1.45s/iter, val_loss=0.0018, val_og_loss=0.00198]
+
 70%|███████   | 14/20 [00:35<00:08,  1.45s/iter, val_loss=0.00179, val_og_loss=0.00198]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.43s/iter, val_loss=0.00179, val_og_loss=0.00198]
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.43s/iter, val_loss=0.0018, val_og_loss=0.00198] 
+
 80%|████████  | 16/20 [00:36<00:05,  1.43s/iter, val_loss=0.0018, val_og_loss=0.00198]
+
 80%|████████  | 16/20 [00:37<00:05,  1.43s/iter, val_loss=0.00178, val_og_loss=0.00198]
+
 85%|████████▌ | 17/20 [00:37<00:04,  1.41s/iter, val_loss=0.00178, val_og_loss=0.00198]
+
 85%|████████▌ | 17/20 [00:39<00:04,  1.41s/iter, val_loss=0.00178, val_og_loss=0.00198]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.41s/iter, val_loss=0.00178, val_og_loss=0.00198]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.41s/iter, val_loss=0.00178, val_og_loss=0.00198]
+
 95%|█████████▌| 19/20 [00:40<00:01,  1.41s/iter, val_loss=0.00178, val_og_loss=0.00198]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.41s/iter, val_loss=0.00178, val_og_loss=0.00198]
+
100%|██████████| 20/20 [00:42<00:00,  1.40s/iter, val_loss=0.00178, val_og_loss=0.00198]
100%|██████████| 20/20 [00:42<00:00,  2.11s/iter, val_loss=0.00178, val_og_loss=0.00198]
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:22:50<43:58, 879.39s/it]INFO:ParoQuant:Best val loss: 0.0017779826885089278, Original val loss: 0.0019846854265779257
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:22:53<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:22:53<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:23:24<43:58, 879.39s/it]2025-12-22 20:53:08,753 - INFO - Best val loss: 0.0017779826885089278, Original val loss: 0.0019846854265779257
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.00011, val_og_loss=0.000116]
+
  5%|▌         | 1/20 [00:14<04:35, 14.48s/iter, val_loss=0.00011, val_og_loss=0.000116]
+
  5%|▌         | 1/20 [00:16<04:35, 14.48s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 10%|█         | 2/20 [00:16<02:05,  6.97s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 10%|█         | 2/20 [00:17<02:05,  6.97s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 15%|█▌        | 3/20 [00:17<01:17,  4.55s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 15%|█▌        | 3/20 [00:19<01:17,  4.55s/iter, val_loss=0.000112, val_og_loss=0.000116]
+
 20%|██        | 4/20 [00:19<00:54,  3.42s/iter, val_loss=0.000112, val_og_loss=0.000116]
+
 20%|██        | 4/20 [00:21<00:54,  3.42s/iter, val_loss=0.00011, val_og_loss=0.000116] 
+
 25%|██▌       | 5/20 [00:21<00:41,  2.79s/iter, val_loss=0.00011, val_og_loss=0.000116]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.79s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 30%|███       | 6/20 [00:22<00:33,  2.42s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 30%|███       | 6/20 [00:24<00:33,  2.42s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.18s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 35%|███▌      | 7/20 [00:26<00:28,  2.18s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 40%|████      | 8/20 [00:26<00:24,  2.02s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 40%|████      | 8/20 [00:27<00:24,  2.02s/iter, val_loss=0.00011, val_og_loss=0.000116] 
+
 45%|████▌     | 9/20 [00:27<00:20,  1.91s/iter, val_loss=0.00011, val_og_loss=0.000116]
+
 45%|████▌     | 9/20 [00:29<00:20,  1.91s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 50%|█████     | 10/20 [00:29<00:18,  1.84s/iter, val_loss=0.000111, val_og_loss=0.000116]
+
 50%|█████     | 10/20 [00:31<00:18,  1.84s/iter, val_loss=0.00011, val_og_loss=0.000116] 
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.79s/iter, val_loss=0.00011, val_og_loss=0.000116]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=0.00011, val_og_loss=0.000116]
+
 60%|██████    | 12/20 [00:32<00:14,  1.76s/iter, val_loss=0.00011, val_og_loss=0.000116]
+
 60%|██████    | 12/20 [00:34<00:14,  1.76s/iter, val_loss=0.00011, val_og_loss=0.000116]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=0.00011, val_og_loss=0.000116]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.73s/iter, val_loss=0.000108, val_og_loss=0.000116]
+
 70%|███████   | 14/20 [00:36<00:10,  1.72s/iter, val_loss=0.000108, val_og_loss=0.000116]
+
 70%|███████   | 14/20 [00:38<00:10,  1.72s/iter, val_loss=0.000109, val_og_loss=0.000116]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=0.000109, val_og_loss=0.000116]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.000108, val_og_loss=0.000116]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=0.000108, val_og_loss=0.000116]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=0.000108, val_og_loss=0.000116]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.69s/iter, val_loss=0.000108, val_og_loss=0.000116]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.69s/iter, val_loss=0.000107, val_og_loss=0.000116]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.69s/iter, val_loss=0.000107, val_og_loss=0.000116]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.000107, val_og_loss=0.000116]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=0.000107, val_og_loss=0.000116]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=0.000108, val_og_loss=0.000116]
+
100%|██████████| 20/20 [00:46<00:00,  1.68s/iter, val_loss=0.000108, val_og_loss=0.000116]
100%|██████████| 20/20 [00:46<00:00,  2.32s/iter, val_loss=0.000108, val_og_loss=0.000116]
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:24:11<43:58, 879.39s/it]INFO:ParoQuant:Best val loss: 0.00010710969218052924, Original val loss: 0.0001163055349024944
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:24:11<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:24:11<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:24:55<43:58, 879.39s/it]2025-12-22 20:54:29,066 - INFO - Best val loss: 0.00010710969218052924, Original val loss: 0.0001163055349024944
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00205, val_og_loss=0.0021]
+
  5%|▌         | 1/20 [00:29<09:22, 29.59s/iter, val_loss=0.00205, val_og_loss=0.0021]
+
  5%|▌         | 1/20 [00:32<09:22, 29.59s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 10%|█         | 2/20 [00:32<04:08, 13.79s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 10%|█         | 2/20 [00:35<04:08, 13.79s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 15%|█▌        | 3/20 [00:35<02:28,  8.74s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 15%|█▌        | 3/20 [00:37<02:28,  8.74s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 20%|██        | 4/20 [00:37<01:41,  6.36s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 20%|██        | 4/20 [00:40<01:41,  6.36s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 25%|██▌       | 5/20 [00:40<01:15,  5.06s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 25%|██▌       | 5/20 [00:43<01:15,  5.06s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 30%|███       | 6/20 [00:43<00:59,  4.26s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 30%|███       | 6/20 [00:45<00:59,  4.26s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 35%|███▌      | 7/20 [00:45<00:48,  3.75s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.75s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 40%|████      | 8/20 [00:48<00:41,  3.42s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 40%|████      | 8/20 [00:51<00:41,  3.42s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 45%|████▌     | 9/20 [00:51<00:35,  3.20s/iter, val_loss=0.00208, val_og_loss=0.0021]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.20s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 50%|█████     | 10/20 [00:54<00:30,  3.05s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 50%|█████     | 10/20 [00:56<00:30,  3.05s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.95s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.95s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 60%|██████    | 12/20 [00:59<00:23,  2.88s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 60%|██████    | 12/20 [01:02<00:23,  2.88s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.83s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.83s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 70%|███████   | 14/20 [01:04<00:16,  2.80s/iter, val_loss=0.00207, val_og_loss=0.0021]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.77s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 75%|███████▌  | 15/20 [01:10<00:13,  2.77s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 80%|████████  | 16/20 [01:10<00:11,  2.75s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 80%|████████  | 16/20 [01:13<00:11,  2.75s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.74s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.74s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.74s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.00206, val_og_loss=0.0021]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.00205, val_og_loss=0.0021]
+
100%|██████████| 20/20 [01:21<00:00,  2.73s/iter, val_loss=0.00205, val_og_loss=0.0021]
100%|██████████| 20/20 [01:21<00:00,  4.06s/iter, val_loss=0.00205, val_og_loss=0.0021]
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:26:16<43:58, 879.39s/it]INFO:ParoQuant:Best val loss: 0.0020545872393995523, Original val loss: 0.0020999389234930277
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:26:21<43:58, 879.39s/it]2025-12-22 20:56:34,703 - INFO - Best val loss: 0.0020545872393995523, Original val loss: 0.0020999389234930277
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.00126, val_og_loss=0.00129]
+
  5%|▌         | 1/20 [00:30<09:39, 30.52s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
  5%|▌         | 1/20 [00:33<09:39, 30.52s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 10%|█         | 2/20 [00:33<04:14, 14.16s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 10%|█         | 2/20 [00:35<04:14, 14.16s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 15%|█▌        | 3/20 [00:35<02:31,  8.94s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 15%|█▌        | 3/20 [00:38<02:31,  8.94s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 20%|██        | 4/20 [00:38<01:43,  6.49s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 20%|██        | 4/20 [00:41<01:43,  6.49s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 25%|██▌       | 5/20 [00:41<01:16,  5.13s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 25%|██▌       | 5/20 [00:44<01:16,  5.13s/iter, val_loss=0.00128, val_og_loss=0.00129]
+
 30%|███       | 6/20 [00:44<01:00,  4.31s/iter, val_loss=0.00128, val_og_loss=0.00129]
+
 30%|███       | 6/20 [00:46<01:00,  4.31s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 35%|███▌      | 7/20 [00:46<00:49,  3.79s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.79s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 40%|████      | 8/20 [00:49<00:41,  3.45s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 40%|████      | 8/20 [00:52<00:41,  3.45s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.22s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.22s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 50%|█████     | 10/20 [00:54<00:30,  3.06s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 60%|██████    | 12/20 [01:03<00:23,  2.89s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.84s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.84s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.00127, val_og_loss=0.00129]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.77s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.77s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.74s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.74s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.74s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
100%|██████████| 20/20 [01:22<00:00,  2.72s/iter, val_loss=0.00126, val_og_loss=0.00129]
100%|██████████| 20/20 [01:22<00:00,  4.11s/iter, val_loss=0.00126, val_og_loss=0.00129]
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:27:43<43:58, 879.39s/it]INFO:ParoQuant:Best val loss: 0.0012600459158420563, Original val loss: 0.0012909644283354282
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:27:44<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:27:44<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:28:42<43:58, 879.39s/it]2025-12-22 20:58:01,352 - INFO - Best val loss: 0.0012600459158420563, Original val loss: 0.0012909644283354282
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.000448, val_og_loss=0.000495]
+
  5%|▌         | 1/20 [00:33<10:43, 33.87s/iter, val_loss=0.000448, val_og_loss=0.000495]
+
  5%|▌         | 1/20 [00:39<10:43, 33.87s/iter, val_loss=0.000449, val_og_loss=0.000495]
+
 10%|█         | 2/20 [00:39<05:08, 17.15s/iter, val_loss=0.000449, val_og_loss=0.000495]
+
 10%|█         | 2/20 [00:44<05:08, 17.15s/iter, val_loss=0.00045, val_og_loss=0.000495] 
+
 15%|█▌        | 3/20 [00:44<03:20, 11.80s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 15%|█▌        | 3/20 [00:50<03:20, 11.80s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 20%|██        | 4/20 [00:50<02:28,  9.29s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 20%|██        | 4/20 [00:55<02:28,  9.29s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 25%|██▌       | 5/20 [00:55<01:58,  7.89s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 25%|██▌       | 5/20 [01:01<01:58,  7.89s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 30%|███       | 6/20 [01:01<01:38,  7.06s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 30%|███       | 6/20 [01:06<01:38,  7.06s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 35%|███▌      | 7/20 [01:06<01:24,  6.52s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 35%|███▌      | 7/20 [01:11<01:24,  6.52s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 40%|████      | 8/20 [01:11<01:14,  6.17s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 40%|████      | 8/20 [01:17<01:14,  6.17s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 45%|████▌     | 9/20 [01:17<01:05,  5.93s/iter, val_loss=0.00045, val_og_loss=0.000495]
+
 45%|████▌     | 9/20 [01:22<01:05,  5.93s/iter, val_loss=0.000449, val_og_loss=0.000495]
+
 50%|█████     | 10/20 [01:22<00:57,  5.78s/iter, val_loss=0.000449, val_og_loss=0.000495]
+
 50%|█████     | 10/20 [01:28<00:57,  5.78s/iter, val_loss=0.000448, val_og_loss=0.000495]
+
 55%|█████▌    | 11/20 [01:28<00:51,  5.67s/iter, val_loss=0.000448, val_og_loss=0.000495]
+
 55%|█████▌    | 11/20 [01:33<00:51,  5.67s/iter, val_loss=0.000449, val_og_loss=0.000495]
+
 60%|██████    | 12/20 [01:33<00:44,  5.60s/iter, val_loss=0.000449, val_og_loss=0.000495]
+
 60%|██████    | 12/20 [01:38<00:44,  5.60s/iter, val_loss=0.000448, val_og_loss=0.000495]
+
 65%|██████▌   | 13/20 [01:39<00:38,  5.55s/iter, val_loss=0.000448, val_og_loss=0.000495]
+
 65%|██████▌   | 13/20 [01:44<00:38,  5.55s/iter, val_loss=0.000447, val_og_loss=0.000495]
+
 70%|███████   | 14/20 [01:44<00:33,  5.51s/iter, val_loss=0.000447, val_og_loss=0.000495]
+
 70%|███████   | 14/20 [01:49<00:33,  5.51s/iter, val_loss=0.000447, val_og_loss=0.000495]
+
 75%|███████▌  | 15/20 [01:49<00:27,  5.48s/iter, val_loss=0.000447, val_og_loss=0.000495]
+
 75%|███████▌  | 15/20 [01:55<00:27,  5.48s/iter, val_loss=0.000446, val_og_loss=0.000495]
+
 80%|████████  | 16/20 [01:55<00:21,  5.47s/iter, val_loss=0.000446, val_og_loss=0.000495]
+
 80%|████████  | 16/20 [02:00<00:21,  5.47s/iter, val_loss=0.000446, val_og_loss=0.000495]
+
 85%|████████▌ | 17/20 [02:00<00:16,  5.45s/iter, val_loss=0.000446, val_og_loss=0.000495]
+
 85%|████████▌ | 17/20 [02:06<00:16,  5.45s/iter, val_loss=0.000447, val_og_loss=0.000495]
+
 90%|█████████ | 18/20 [02:06<00:10,  5.44s/iter, val_loss=0.000447, val_og_loss=0.000495]
+
 90%|█████████ | 18/20 [02:11<00:10,  5.44s/iter, val_loss=0.000446, val_og_loss=0.000495]
+
 95%|█████████▌| 19/20 [02:11<00:05,  5.44s/iter, val_loss=0.000446, val_og_loss=0.000495]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.44s/iter, val_loss=0.000446, val_og_loss=0.000495]
+
100%|██████████| 20/20 [02:16<00:00,  5.43s/iter, val_loss=0.000446, val_og_loss=0.000495]
100%|██████████| 20/20 [02:16<00:00,  6.85s/iter, val_loss=0.000446, val_og_loss=0.000495]
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:30:59<43:58, 879.39s/it]INFO:ParoQuant:Best val loss: 0.0004463549121282995, Original val loss: 0.0004952335148118436
+
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:31:00<43:58, 879.39s/it]
                                                                     

Processing Blocks:  91%|█████████ | 29/32 [6:31:10<43:58, 879.39s/it]
Processing Blocks:  94%|█████████▍| 30/32 [6:31:48<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:31:49<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:32:08<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:32:08<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:34:22<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:34:22<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:35:26<29:25, 882.56s/it]2025-12-22 21:01:17,699 - INFO - Best val loss: 0.0004463549121282995, Original val loss: 0.0004952335148118436
+[INFO] Saving Layer 29 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 30...
+[INFO] Starting Sequential Training for Layer 30
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.00223, val_og_loss=0.00252]
+
  5%|▌         | 1/20 [00:14<04:29, 14.17s/iter, val_loss=0.00223, val_og_loss=0.00252]
+
  5%|▌         | 1/20 [00:15<04:29, 14.17s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 10%|█         | 2/20 [00:15<02:03,  6.84s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 10%|█         | 2/20 [00:17<02:03,  6.84s/iter, val_loss=0.00226, val_og_loss=0.00252]
+
 15%|█▌        | 3/20 [00:17<01:16,  4.50s/iter, val_loss=0.00226, val_og_loss=0.00252]
+
 15%|█▌        | 3/20 [00:19<01:16,  4.50s/iter, val_loss=0.00225, val_og_loss=0.00252]
+
 20%|██        | 4/20 [00:19<00:54,  3.40s/iter, val_loss=0.00225, val_og_loss=0.00252]
+
 20%|██        | 4/20 [00:21<00:54,  3.40s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 25%|██▌       | 5/20 [00:21<00:41,  2.79s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.79s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 30%|███       | 6/20 [00:22<00:33,  2.41s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 30%|███       | 6/20 [00:24<00:33,  2.41s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.18s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 35%|███▌      | 7/20 [00:26<00:28,  2.18s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 40%|████      | 8/20 [00:26<00:24,  2.02s/iter, val_loss=0.00227, val_og_loss=0.00252]
+
 40%|████      | 8/20 [00:27<00:24,  2.02s/iter, val_loss=0.00226, val_og_loss=0.00252]
+
 45%|████▌     | 9/20 [00:27<00:21,  1.92s/iter, val_loss=0.00226, val_og_loss=0.00252]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.92s/iter, val_loss=0.00225, val_og_loss=0.00252]
+
 50%|█████     | 10/20 [00:29<00:18,  1.84s/iter, val_loss=0.00225, val_og_loss=0.00252]
+
 50%|█████     | 10/20 [00:31<00:18,  1.84s/iter, val_loss=0.00225, val_og_loss=0.00252]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.80s/iter, val_loss=0.00225, val_og_loss=0.00252]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.80s/iter, val_loss=0.00224, val_og_loss=0.00252]
+
 60%|██████    | 12/20 [00:32<00:14,  1.76s/iter, val_loss=0.00224, val_og_loss=0.00252]
+
 60%|██████    | 12/20 [00:34<00:14,  1.76s/iter, val_loss=0.00224, val_og_loss=0.00252]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=0.00224, val_og_loss=0.00252]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.73s/iter, val_loss=0.00223, val_og_loss=0.00252]
+
 70%|███████   | 14/20 [00:36<00:10,  1.72s/iter, val_loss=0.00223, val_og_loss=0.00252]
+
 70%|███████   | 14/20 [00:37<00:10,  1.72s/iter, val_loss=0.00224, val_og_loss=0.00252]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.71s/iter, val_loss=0.00224, val_og_loss=0.00252]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.00223, val_og_loss=0.00252]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=0.00223, val_og_loss=0.00252]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=0.00222, val_og_loss=0.00252]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.69s/iter, val_loss=0.00222, val_og_loss=0.00252]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.69s/iter, val_loss=0.00222, val_og_loss=0.00252]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.69s/iter, val_loss=0.00222, val_og_loss=0.00252]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.69s/iter, val_loss=0.00223, val_og_loss=0.00252]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=0.00223, val_og_loss=0.00252]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.69s/iter, val_loss=0.00222, val_og_loss=0.00252]
+
100%|██████████| 20/20 [00:46<00:00,  1.69s/iter, val_loss=0.00222, val_og_loss=0.00252]
100%|██████████| 20/20 [00:46<00:00,  2.31s/iter, val_loss=0.00222, val_og_loss=0.00252]
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:36:12<29:25, 882.56s/it]INFO:ParoQuant:Best val loss: 0.002216869266703725, Original val loss: 0.002516241278499365
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:36:15<29:25, 882.56s/it]2025-12-22 21:06:30,479 - INFO - Best val loss: 0.002216869266703725, Original val loss: 0.002516241278499365
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:15<?, ?iter/s, val_loss=0.00516, val_og_loss=0.00556]
+
  5%|▌         | 1/20 [00:15<05:01, 15.86s/iter, val_loss=0.00516, val_og_loss=0.00556]
+
  5%|▌         | 1/20 [00:17<05:01, 15.86s/iter, val_loss=0.00528, val_og_loss=0.00556]
+
 10%|█         | 2/20 [00:17<02:12,  7.35s/iter, val_loss=0.00528, val_og_loss=0.00556]
+
 10%|█         | 2/20 [00:18<02:12,  7.35s/iter, val_loss=0.00527, val_og_loss=0.00556]
+
 15%|█▌        | 3/20 [00:18<01:18,  4.63s/iter, val_loss=0.00527, val_og_loss=0.00556]
+
 15%|█▌        | 3/20 [00:20<01:18,  4.63s/iter, val_loss=0.00531, val_og_loss=0.00556]
+
 20%|██        | 4/20 [00:20<00:53,  3.36s/iter, val_loss=0.00531, val_og_loss=0.00556]
+
 20%|██        | 4/20 [00:21<00:53,  3.36s/iter, val_loss=0.00527, val_og_loss=0.00556]
+
 25%|██▌       | 5/20 [00:21<00:39,  2.65s/iter, val_loss=0.00527, val_og_loss=0.00556]
+
 25%|██▌       | 5/20 [00:22<00:39,  2.65s/iter, val_loss=0.00526, val_og_loss=0.00556]
+
 30%|███       | 6/20 [00:22<00:31,  2.22s/iter, val_loss=0.00526, val_og_loss=0.00556]
+
 30%|███       | 6/20 [00:24<00:31,  2.22s/iter, val_loss=0.0053, val_og_loss=0.00556] 
+
 35%|███▌      | 7/20 [00:24<00:25,  1.95s/iter, val_loss=0.0053, val_og_loss=0.00556]
+
 35%|███▌      | 7/20 [00:25<00:25,  1.95s/iter, val_loss=0.00527, val_og_loss=0.00556]
+
 40%|████      | 8/20 [00:25<00:21,  1.77s/iter, val_loss=0.00527, val_og_loss=0.00556]
+
 40%|████      | 8/20 [00:27<00:21,  1.77s/iter, val_loss=0.00523, val_og_loss=0.00556]
+
 45%|████▌     | 9/20 [00:27<00:18,  1.65s/iter, val_loss=0.00523, val_og_loss=0.00556]
+
 45%|████▌     | 9/20 [00:28<00:18,  1.65s/iter, val_loss=0.00526, val_og_loss=0.00556]
+
 50%|█████     | 10/20 [00:28<00:15,  1.57s/iter, val_loss=0.00526, val_og_loss=0.00556]
+
 50%|█████     | 10/20 [00:29<00:15,  1.57s/iter, val_loss=0.00524, val_og_loss=0.00556]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.52s/iter, val_loss=0.00524, val_og_loss=0.00556]
+
 55%|█████▌    | 11/20 [00:31<00:13,  1.52s/iter, val_loss=0.00524, val_og_loss=0.00556]
+
 60%|██████    | 12/20 [00:31<00:11,  1.48s/iter, val_loss=0.00524, val_og_loss=0.00556]
+
 60%|██████    | 12/20 [00:32<00:11,  1.48s/iter, val_loss=0.00526, val_og_loss=0.00556]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.46s/iter, val_loss=0.00526, val_og_loss=0.00556]
+
 65%|██████▌   | 13/20 [00:33<00:10,  1.46s/iter, val_loss=0.00525, val_og_loss=0.00556]
+
 70%|███████   | 14/20 [00:33<00:08,  1.44s/iter, val_loss=0.00525, val_og_loss=0.00556]
+
 70%|███████   | 14/20 [00:35<00:08,  1.44s/iter, val_loss=0.00522, val_og_loss=0.00556]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.42s/iter, val_loss=0.00522, val_og_loss=0.00556]
+
 75%|███████▌  | 15/20 [00:36<00:07,  1.42s/iter, val_loss=0.00523, val_og_loss=0.00556]
+
 80%|████████  | 16/20 [00:36<00:05,  1.41s/iter, val_loss=0.00523, val_og_loss=0.00556]
+
 80%|████████  | 16/20 [00:38<00:05,  1.41s/iter, val_loss=0.00518, val_og_loss=0.00556]
+
 85%|████████▌ | 17/20 [00:38<00:04,  1.41s/iter, val_loss=0.00518, val_og_loss=0.00556]
+
 85%|████████▌ | 17/20 [00:39<00:04,  1.41s/iter, val_loss=0.00514, val_og_loss=0.00556]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.40s/iter, val_loss=0.00514, val_og_loss=0.00556]
+
 90%|█████████ | 18/20 [00:40<00:02,  1.40s/iter, val_loss=0.00519, val_og_loss=0.00556]
+
 95%|█████████▌| 19/20 [00:40<00:01,  1.40s/iter, val_loss=0.00519, val_og_loss=0.00556]
+
 95%|█████████▌| 19/20 [00:42<00:01,  1.40s/iter, val_loss=0.00515, val_og_loss=0.00556]
+
100%|██████████| 20/20 [00:42<00:00,  1.40s/iter, val_loss=0.00515, val_og_loss=0.00556]
100%|██████████| 20/20 [00:42<00:00,  2.12s/iter, val_loss=0.00515, val_og_loss=0.00556]
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:36:58<29:25, 882.56s/it]INFO:ParoQuant:Best val loss: 0.0051377397030591965, Original val loss: 0.005555348936468363
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:37:01<29:25, 882.56s/it]2025-12-22 21:07:16,053 - INFO - Best val loss: 0.0051377397030591965, Original val loss: 0.005555348936468363
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.00234, val_og_loss=0.00272]
+
  5%|▌         | 1/20 [00:14<04:37, 14.62s/iter, val_loss=0.00234, val_og_loss=0.00272]
+
  5%|▌         | 1/20 [00:16<04:37, 14.62s/iter, val_loss=0.00237, val_og_loss=0.00272]
+
 10%|█         | 2/20 [00:16<02:03,  6.84s/iter, val_loss=0.00237, val_og_loss=0.00272]
+
 10%|█         | 2/20 [00:17<02:03,  6.84s/iter, val_loss=0.00236, val_og_loss=0.00272]
+
 15%|█▌        | 3/20 [00:17<01:14,  4.36s/iter, val_loss=0.00236, val_og_loss=0.00272]
+
 15%|█▌        | 3/20 [00:18<01:14,  4.36s/iter, val_loss=0.00234, val_og_loss=0.00272]
+
 20%|██        | 4/20 [00:18<00:51,  3.19s/iter, val_loss=0.00234, val_og_loss=0.00272]
+
 20%|██        | 4/20 [00:20<00:51,  3.19s/iter, val_loss=0.00235, val_og_loss=0.00272]
+
 25%|██▌       | 5/20 [00:20<00:38,  2.54s/iter, val_loss=0.00235, val_og_loss=0.00272]
+
 25%|██▌       | 5/20 [00:21<00:38,  2.54s/iter, val_loss=0.00236, val_og_loss=0.00272]
+
 30%|███       | 6/20 [00:21<00:30,  2.15s/iter, val_loss=0.00236, val_og_loss=0.00272]
+
 30%|███       | 6/20 [00:22<00:30,  2.15s/iter, val_loss=0.00236, val_og_loss=0.00272]
+
 35%|███▌      | 7/20 [00:22<00:24,  1.90s/iter, val_loss=0.00236, val_og_loss=0.00272]
+
 35%|███▌      | 7/20 [00:24<00:24,  1.90s/iter, val_loss=0.00235, val_og_loss=0.00272]
+
 40%|████      | 8/20 [00:24<00:20,  1.74s/iter, val_loss=0.00235, val_og_loss=0.00272]
+
 40%|████      | 8/20 [00:25<00:20,  1.74s/iter, val_loss=0.00235, val_og_loss=0.00272]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.63s/iter, val_loss=0.00235, val_og_loss=0.00272]
+
 45%|████▌     | 9/20 [00:27<00:17,  1.63s/iter, val_loss=0.00234, val_og_loss=0.00272]
+
 50%|█████     | 10/20 [00:27<00:15,  1.55s/iter, val_loss=0.00234, val_og_loss=0.00272]
+
 50%|█████     | 10/20 [00:28<00:15,  1.55s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 55%|█████▌    | 11/20 [00:28<00:13,  1.50s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.50s/iter, val_loss=0.00234, val_og_loss=0.00272]
+
 60%|██████    | 12/20 [00:29<00:11,  1.47s/iter, val_loss=0.00234, val_og_loss=0.00272]
+
 60%|██████    | 12/20 [00:31<00:11,  1.47s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 65%|██████▌   | 13/20 [00:31<00:10,  1.45s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.45s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 70%|███████   | 14/20 [00:32<00:08,  1.43s/iter, val_loss=0.00233, val_og_loss=0.00272]
+
 70%|███████   | 14/20 [00:34<00:08,  1.43s/iter, val_loss=0.00232, val_og_loss=0.00272]
+
 75%|███████▌  | 15/20 [00:34<00:07,  1.42s/iter, val_loss=0.00232, val_og_loss=0.00272]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.42s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 80%|████████  | 16/20 [00:35<00:05,  1.41s/iter, val_loss=0.00231, val_og_loss=0.00272]
+
 80%|████████  | 16/20 [00:36<00:05,  1.41s/iter, val_loss=0.0023, val_og_loss=0.00272] 
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.42s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 85%|████████▌ | 17/20 [00:38<00:04,  1.42s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 90%|█████████ | 18/20 [00:38<00:02,  1.41s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.41s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.41s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
 95%|█████████▌| 19/20 [00:41<00:01,  1.41s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
100%|██████████| 20/20 [00:41<00:00,  1.40s/iter, val_loss=0.0023, val_og_loss=0.00272]
100%|██████████| 20/20 [00:41<00:00,  2.06s/iter, val_loss=0.0023, val_og_loss=0.00272]
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:37:42<29:25, 882.56s/it]INFO:ParoQuant:Best val loss: 0.0022965141106396914, Original val loss: 0.002718192059546709
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:37:45<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:37:45<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:38:16<29:25, 882.56s/it]2025-12-22 21:08:00,546 - INFO - Best val loss: 0.0022965141106396914, Original val loss: 0.002718192059546709
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:18<?, ?iter/s, val_loss=0.000196, val_og_loss=0.000199]
+
  5%|▌         | 1/20 [00:18<05:58, 18.86s/iter, val_loss=0.000196, val_og_loss=0.000199]
+
  5%|▌         | 1/20 [00:20<05:58, 18.86s/iter, val_loss=0.000196, val_og_loss=0.000199]
+
 10%|█         | 2/20 [00:20<02:37,  8.75s/iter, val_loss=0.000196, val_og_loss=0.000199]
+
 10%|█         | 2/20 [00:22<02:37,  8.75s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 15%|█▌        | 3/20 [00:22<01:33,  5.53s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 15%|█▌        | 3/20 [00:23<01:33,  5.53s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 20%|██        | 4/20 [00:23<01:04,  4.01s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 20%|██        | 4/20 [00:25<01:04,  4.01s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 25%|██▌       | 5/20 [00:25<00:47,  3.17s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 25%|██▌       | 5/20 [00:27<00:47,  3.17s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 30%|███       | 6/20 [00:27<00:37,  2.66s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 30%|███       | 6/20 [00:28<00:37,  2.66s/iter, val_loss=0.000196, val_og_loss=0.000199]
+
 35%|███▌      | 7/20 [00:28<00:30,  2.35s/iter, val_loss=0.000196, val_og_loss=0.000199]
+
 35%|███▌      | 7/20 [00:30<00:30,  2.35s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 40%|████      | 8/20 [00:30<00:25,  2.14s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 40%|████      | 8/20 [00:32<00:25,  2.14s/iter, val_loss=0.000196, val_og_loss=0.000199]
+
 45%|████▌     | 9/20 [00:32<00:21,  1.99s/iter, val_loss=0.000196, val_og_loss=0.000199]
+
 45%|████▌     | 9/20 [00:34<00:21,  1.99s/iter, val_loss=0.000196, val_og_loss=0.000199]
+
 50%|█████     | 10/20 [00:34<00:18,  1.90s/iter, val_loss=0.000196, val_og_loss=0.000199]
+
 50%|█████     | 10/20 [00:35<00:18,  1.90s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 55%|█████▌    | 11/20 [00:35<00:16,  1.83s/iter, val_loss=0.000197, val_og_loss=0.000199]
+
 55%|█████▌    | 11/20 [00:37<00:16,  1.83s/iter, val_loss=0.000195, val_og_loss=0.000199]
+
 60%|██████    | 12/20 [00:37<00:14,  1.78s/iter, val_loss=0.000195, val_og_loss=0.000199]
+
 60%|██████    | 12/20 [00:39<00:14,  1.78s/iter, val_loss=0.000195, val_og_loss=0.000199]
+
 65%|██████▌   | 13/20 [00:39<00:12,  1.75s/iter, val_loss=0.000195, val_og_loss=0.000199]
+
 65%|██████▌   | 13/20 [00:40<00:12,  1.75s/iter, val_loss=0.000194, val_og_loss=0.000199]
+
 70%|███████   | 14/20 [00:40<00:10,  1.73s/iter, val_loss=0.000194, val_og_loss=0.000199]
+
 70%|███████   | 14/20 [00:42<00:10,  1.73s/iter, val_loss=0.000195, val_og_loss=0.000199]
+
 75%|███████▌  | 15/20 [00:42<00:08,  1.72s/iter, val_loss=0.000195, val_og_loss=0.000199]
+
 75%|███████▌  | 15/20 [00:44<00:08,  1.72s/iter, val_loss=0.000194, val_og_loss=0.000199]
+
 80%|████████  | 16/20 [00:44<00:06,  1.71s/iter, val_loss=0.000194, val_og_loss=0.000199]
+
 80%|████████  | 16/20 [00:45<00:06,  1.71s/iter, val_loss=0.000193, val_og_loss=0.000199]
+
 85%|████████▌ | 17/20 [00:45<00:05,  1.70s/iter, val_loss=0.000193, val_og_loss=0.000199]
+
 85%|████████▌ | 17/20 [00:47<00:05,  1.70s/iter, val_loss=0.000192, val_og_loss=0.000199]
+
 90%|█████████ | 18/20 [00:47<00:03,  1.69s/iter, val_loss=0.000192, val_og_loss=0.000199]
+
 90%|█████████ | 18/20 [00:49<00:03,  1.69s/iter, val_loss=0.000193, val_og_loss=0.000199]
+
 95%|█████████▌| 19/20 [00:49<00:01,  1.69s/iter, val_loss=0.000193, val_og_loss=0.000199]
+
 95%|█████████▌| 19/20 [00:50<00:01,  1.69s/iter, val_loss=0.000192, val_og_loss=0.000199]
+
100%|██████████| 20/20 [00:50<00:00,  1.69s/iter, val_loss=0.000192, val_og_loss=0.000199]
100%|██████████| 20/20 [00:50<00:00,  2.54s/iter, val_loss=0.000192, val_og_loss=0.000199]
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:39:07<29:25, 882.56s/it]INFO:ParoQuant:Best val loss: 0.00019205459102522582, Original val loss: 0.0001987398136407137
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:39:07<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:39:07<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:39:51<29:25, 882.56s/it]2025-12-22 21:09:25,172 - INFO - Best val loss: 0.00019205459102522582, Original val loss: 0.0001987398136407137
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:28<?, ?iter/s, val_loss=0.00228, val_og_loss=0.0023]
+
  5%|▌         | 1/20 [00:28<09:07, 28.84s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
  5%|▌         | 1/20 [00:31<09:07, 28.84s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 10%|█         | 2/20 [00:31<04:02, 13.47s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 10%|█         | 2/20 [00:34<04:02, 13.47s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 15%|█▌        | 3/20 [00:34<02:25,  8.55s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 15%|█▌        | 3/20 [00:36<02:25,  8.55s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 20%|██        | 4/20 [00:36<01:39,  6.25s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 20%|██        | 4/20 [00:39<01:39,  6.25s/iter, val_loss=0.0023, val_og_loss=0.0023] 
+
 25%|██▌       | 5/20 [00:39<01:14,  4.98s/iter, val_loss=0.0023, val_og_loss=0.0023]
+
 25%|██▌       | 5/20 [00:42<01:14,  4.98s/iter, val_loss=0.0023, val_og_loss=0.0023]
+
 30%|███       | 6/20 [00:42<00:58,  4.21s/iter, val_loss=0.0023, val_og_loss=0.0023]
+
 30%|███       | 6/20 [00:45<00:58,  4.21s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 35%|███▌      | 7/20 [00:45<00:48,  3.72s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 35%|███▌      | 7/20 [00:47<00:48,  3.72s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 40%|████      | 8/20 [00:47<00:40,  3.40s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 40%|████      | 8/20 [00:50<00:40,  3.40s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 45%|████▌     | 9/20 [00:50<00:35,  3.19s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.19s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 50%|█████     | 10/20 [00:53<00:30,  3.05s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 50%|█████     | 10/20 [00:56<00:30,  3.05s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.95s/iter, val_loss=0.00229, val_og_loss=0.0023]
+
 55%|█████▌    | 11/20 [00:58<00:26,  2.95s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 60%|██████    | 12/20 [00:58<00:23,  2.88s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 60%|██████    | 12/20 [01:01<00:23,  2.88s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 65%|██████▌   | 13/20 [01:01<00:19,  2.84s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 70%|███████   | 14/20 [01:04<00:16,  2.80s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 75%|███████▌  | 15/20 [01:06<00:13,  2.78s/iter, val_loss=0.00228, val_og_loss=0.0023]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
 80%|████████  | 16/20 [01:09<00:11,  2.76s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
 85%|████████▌ | 17/20 [01:12<00:08,  2.75s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.75s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.74s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
 90%|█████████ | 18/20 [01:17<00:05,  2.74s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
 95%|█████████▌| 19/20 [01:17<00:02,  2.73s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
100%|██████████| 20/20 [01:20<00:00,  2.73s/iter, val_loss=0.00227, val_og_loss=0.0023]
100%|██████████| 20/20 [01:20<00:00,  4.02s/iter, val_loss=0.00227, val_og_loss=0.0023]
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:41:12<29:25, 882.56s/it]INFO:ParoQuant:Best val loss: 0.0022660624235868454, Original val loss: 0.0023007476702332497
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:41:17<29:25, 882.56s/it]2025-12-22 21:11:30,204 - INFO - Best val loss: 0.0022660624235868454, Original val loss: 0.0023007476702332497
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.00137, val_og_loss=0.00139]
+
  5%|▌         | 1/20 [00:30<09:43, 30.71s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
  5%|▌         | 1/20 [00:33<09:43, 30.71s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 10%|█         | 2/20 [00:33<04:16, 14.25s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 10%|█         | 2/20 [00:36<04:16, 14.25s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 15%|█▌        | 3/20 [00:36<02:32,  8.99s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 15%|█▌        | 3/20 [00:38<02:32,  8.99s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 20%|██        | 4/20 [00:38<01:44,  6.51s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 20%|██        | 4/20 [00:41<01:44,  6.51s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 25%|██▌       | 5/20 [00:41<01:17,  5.15s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 25%|██▌       | 5/20 [00:44<01:17,  5.15s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 30%|███       | 6/20 [00:44<01:00,  4.32s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 30%|███       | 6/20 [00:47<01:00,  4.32s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 35%|███▌      | 7/20 [00:47<00:49,  3.79s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.79s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 40%|████      | 8/20 [00:49<00:41,  3.46s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 40%|████      | 8/20 [00:52<00:41,  3.46s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.22s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 45%|████▌     | 9/20 [00:55<00:35,  3.22s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 50%|█████     | 10/20 [00:55<00:30,  3.06s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 50%|█████     | 10/20 [00:57<00:30,  3.06s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.00138, val_og_loss=0.00139]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 60%|██████    | 12/20 [01:00<00:23,  2.88s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 60%|██████    | 12/20 [01:03<00:23,  2.88s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 65%|██████▌   | 13/20 [01:03<00:19,  2.83s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 65%|██████▌   | 13/20 [01:06<00:19,  2.83s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 70%|███████   | 14/20 [01:06<00:16,  2.80s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.77s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.77s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 80%|████████  | 16/20 [01:14<00:11,  2.76s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 85%|████████▌ | 17/20 [01:14<00:08,  2.74s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.74s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.74s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.74s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
 95%|█████████▌| 19/20 [01:22<00:02,  2.73s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
100%|██████████| 20/20 [01:22<00:00,  2.72s/iter, val_loss=0.00137, val_og_loss=0.00139]
100%|██████████| 20/20 [01:22<00:00,  4.12s/iter, val_loss=0.00137, val_og_loss=0.00139]
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:42:39<29:25, 882.56s/it]INFO:ParoQuant:Best val loss: 0.001365162432193756, Original val loss: 0.0013886085944250226
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:42:40<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:42:40<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:43:39<29:25, 882.56s/it]2025-12-22 21:12:57,326 - INFO - Best val loss: 0.001365162432193756, Original val loss: 0.0013886085944250226
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:32<?, ?iter/s, val_loss=0.000761, val_og_loss=0.00092]
+
  5%|▌         | 1/20 [00:32<10:26, 32.97s/iter, val_loss=0.000761, val_og_loss=0.00092]
+
  5%|▌         | 1/20 [00:38<10:26, 32.97s/iter, val_loss=0.000755, val_og_loss=0.00092]
+
 10%|█         | 2/20 [00:38<05:01, 16.77s/iter, val_loss=0.000755, val_og_loss=0.00092]
+
 10%|█         | 2/20 [00:43<05:01, 16.77s/iter, val_loss=0.000757, val_og_loss=0.00092]
+
 15%|█▌        | 3/20 [00:43<03:16, 11.59s/iter, val_loss=0.000757, val_og_loss=0.00092]
+
 15%|█▌        | 3/20 [00:49<03:16, 11.59s/iter, val_loss=0.000755, val_og_loss=0.00092]
+
 20%|██        | 4/20 [00:49<02:26,  9.15s/iter, val_loss=0.000755, val_og_loss=0.00092]
+
 20%|██        | 4/20 [00:54<02:26,  9.15s/iter, val_loss=0.000755, val_og_loss=0.00092]
+
 25%|██▌       | 5/20 [00:54<01:57,  7.81s/iter, val_loss=0.000755, val_og_loss=0.00092]
+
 25%|██▌       | 5/20 [01:00<01:57,  7.81s/iter, val_loss=0.000753, val_og_loss=0.00092]
+
 30%|███       | 6/20 [01:00<01:38,  7.00s/iter, val_loss=0.000753, val_og_loss=0.00092]
+
 30%|███       | 6/20 [01:05<01:38,  7.00s/iter, val_loss=0.000756, val_og_loss=0.00092]
+
 35%|███▌      | 7/20 [01:05<01:24,  6.49s/iter, val_loss=0.000756, val_og_loss=0.00092]
+
 35%|███▌      | 7/20 [01:10<01:24,  6.49s/iter, val_loss=0.000752, val_og_loss=0.00092]
+
 40%|████      | 8/20 [01:10<01:13,  6.15s/iter, val_loss=0.000752, val_og_loss=0.00092]
+
 40%|████      | 8/20 [01:16<01:13,  6.15s/iter, val_loss=0.00075, val_og_loss=0.00092] 
+
 45%|████▌     | 9/20 [01:16<01:05,  5.93s/iter, val_loss=0.00075, val_og_loss=0.00092]
+
 45%|████▌     | 9/20 [01:21<01:05,  5.93s/iter, val_loss=0.000753, val_og_loss=0.00092]
+
 50%|█████     | 10/20 [01:21<00:57,  5.77s/iter, val_loss=0.000753, val_og_loss=0.00092]
+
 50%|█████     | 10/20 [01:27<00:57,  5.77s/iter, val_loss=0.000752, val_og_loss=0.00092]
+
 55%|█████▌    | 11/20 [01:27<00:51,  5.67s/iter, val_loss=0.000752, val_og_loss=0.00092]
+
 55%|█████▌    | 11/20 [01:32<00:51,  5.67s/iter, val_loss=0.000753, val_og_loss=0.00092]
+
 60%|██████    | 12/20 [01:32<00:44,  5.59s/iter, val_loss=0.000753, val_og_loss=0.00092]
+
 60%|██████    | 12/20 [01:38<00:44,  5.59s/iter, val_loss=0.000752, val_og_loss=0.00092]
+
 65%|██████▌   | 13/20 [01:38<00:38,  5.55s/iter, val_loss=0.000752, val_og_loss=0.00092]
+
 65%|██████▌   | 13/20 [01:43<00:38,  5.55s/iter, val_loss=0.000748, val_og_loss=0.00092]
+
 70%|███████   | 14/20 [01:43<00:33,  5.51s/iter, val_loss=0.000748, val_og_loss=0.00092]
+
 70%|███████   | 14/20 [01:48<00:33,  5.51s/iter, val_loss=0.000749, val_og_loss=0.00092]
+
 75%|███████▌  | 15/20 [01:48<00:27,  5.49s/iter, val_loss=0.000749, val_og_loss=0.00092]
+
 75%|███████▌  | 15/20 [01:54<00:27,  5.49s/iter, val_loss=0.000749, val_og_loss=0.00092]
+
 80%|████████  | 16/20 [01:54<00:21,  5.47s/iter, val_loss=0.000749, val_og_loss=0.00092]
+
 80%|████████  | 16/20 [01:59<00:21,  5.47s/iter, val_loss=0.000747, val_og_loss=0.00092]
+
 85%|████████▌ | 17/20 [01:59<00:16,  5.45s/iter, val_loss=0.000747, val_og_loss=0.00092]
+
 85%|████████▌ | 17/20 [02:05<00:16,  5.45s/iter, val_loss=0.000745, val_og_loss=0.00092]
+
 90%|█████████ | 18/20 [02:05<00:10,  5.45s/iter, val_loss=0.000745, val_og_loss=0.00092]
+
 90%|█████████ | 18/20 [02:10<00:10,  5.45s/iter, val_loss=0.000746, val_og_loss=0.00092]
+
 95%|█████████▌| 19/20 [02:10<00:05,  5.44s/iter, val_loss=0.000746, val_og_loss=0.00092]
+
 95%|█████████▌| 19/20 [02:16<00:05,  5.44s/iter, val_loss=0.000746, val_og_loss=0.00092]
+
100%|██████████| 20/20 [02:16<00:00,  5.44s/iter, val_loss=0.000746, val_og_loss=0.00092]
100%|██████████| 20/20 [02:16<00:00,  6.81s/iter, val_loss=0.000746, val_og_loss=0.00092]
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:45:55<29:25, 882.56s/it]INFO:ParoQuant:Best val loss: 0.0007454190053977072, Original val loss: 0.0009195362217724323
+
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:45:55<29:25, 882.56s/it]
                                                                     

Processing Blocks:  94%|█████████▍| 30/32 [6:46:06<29:25, 882.56s/it]
Processing Blocks:  97%|█████████▋| 31/32 [6:46:43<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:46:44<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:47:02<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:47:02<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:49:16<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:49:16<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:50:18<14:46, 886.16s/it]2025-12-22 21:16:13,338 - INFO - Best val loss: 0.0007454190053977072, Original val loss: 0.0009195362217724323
+[INFO] Saving Layer 30 results...
+[INFO] Generating output for next block...
+[INFO] Initializing quantization parameters for Layer 31...
+[INFO] Starting Sequential Training for Layer 31
+[INFO] Pre-capturing ALL clean targets for 7 modules...
+[INFO]   > Group 0: ['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.q_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:13<?, ?iter/s, val_loss=0.002, val_og_loss=0.00215]
+
  5%|▌         | 1/20 [00:13<04:19, 13.64s/iter, val_loss=0.002, val_og_loss=0.00215]
+
  5%|▌         | 1/20 [00:15<04:19, 13.64s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 10%|█         | 2/20 [00:15<01:59,  6.64s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 10%|█         | 2/20 [00:17<01:59,  6.64s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 15%|█▌        | 3/20 [00:17<01:14,  4.38s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 15%|█▌        | 3/20 [00:18<01:14,  4.38s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 20%|██        | 4/20 [00:18<00:53,  3.32s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 20%|██        | 4/20 [00:20<00:53,  3.32s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 25%|██▌       | 5/20 [00:20<00:41,  2.74s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.74s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 30%|███       | 6/20 [00:22<00:33,  2.38s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 30%|███       | 6/20 [00:23<00:33,  2.38s/iter, val_loss=0.00203, val_og_loss=0.00215]
+
 35%|███▌      | 7/20 [00:23<00:28,  2.16s/iter, val_loss=0.00203, val_og_loss=0.00215]
+
 35%|███▌      | 7/20 [00:25<00:28,  2.16s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 40%|████      | 8/20 [00:25<00:24,  2.01s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 40%|████      | 8/20 [00:27<00:24,  2.01s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 45%|████▌     | 9/20 [00:27<00:21,  1.91s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 45%|████▌     | 9/20 [00:28<00:21,  1.91s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 50%|█████     | 10/20 [00:28<00:18,  1.84s/iter, val_loss=0.00201, val_og_loss=0.00215]
+
 50%|█████     | 10/20 [00:30<00:18,  1.84s/iter, val_loss=0.002, val_og_loss=0.00215]  
+
 55%|█████▌    | 11/20 [00:30<00:16,  1.79s/iter, val_loss=0.002, val_og_loss=0.00215]
+
 55%|█████▌    | 11/20 [00:32<00:16,  1.79s/iter, val_loss=0.002, val_og_loss=0.00215]
+
 60%|██████    | 12/20 [00:32<00:14,  1.76s/iter, val_loss=0.002, val_og_loss=0.00215]
+
 60%|██████    | 12/20 [00:34<00:14,  1.76s/iter, val_loss=0.002, val_og_loss=0.00215]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.74s/iter, val_loss=0.002, val_og_loss=0.00215]
+
 65%|██████▌   | 13/20 [00:35<00:12,  1.74s/iter, val_loss=0.00198, val_og_loss=0.00215]
+
 70%|███████   | 14/20 [00:35<00:10,  1.73s/iter, val_loss=0.00198, val_og_loss=0.00215]
+
 70%|███████   | 14/20 [00:37<00:10,  1.73s/iter, val_loss=0.00198, val_og_loss=0.00215]
+
 75%|███████▌  | 15/20 [00:37<00:08,  1.72s/iter, val_loss=0.00198, val_og_loss=0.00215]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.72s/iter, val_loss=0.00198, val_og_loss=0.00215]
+
 80%|████████  | 16/20 [00:39<00:06,  1.71s/iter, val_loss=0.00198, val_og_loss=0.00215]
+
 80%|████████  | 16/20 [00:40<00:06,  1.71s/iter, val_loss=0.00197, val_og_loss=0.00215]
+
 85%|████████▌ | 17/20 [00:40<00:05,  1.71s/iter, val_loss=0.00197, val_og_loss=0.00215]
+
 85%|████████▌ | 17/20 [00:42<00:05,  1.71s/iter, val_loss=0.00198, val_og_loss=0.00215]
+
 90%|█████████ | 18/20 [00:42<00:03,  1.70s/iter, val_loss=0.00198, val_og_loss=0.00215]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.70s/iter, val_loss=0.00197, val_og_loss=0.00215]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.69s/iter, val_loss=0.00197, val_og_loss=0.00215]
+
 95%|█████████▌| 19/20 [00:45<00:01,  1.69s/iter, val_loss=0.00196, val_og_loss=0.00215]
+
100%|██████████| 20/20 [00:45<00:00,  1.69s/iter, val_loss=0.00196, val_og_loss=0.00215]
100%|██████████| 20/20 [00:45<00:00,  2.29s/iter, val_loss=0.00196, val_og_loss=0.00215]
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:51:04<14:46, 886.16s/it]INFO:ParoQuant:Best val loss: 0.001962420530617237, Original val loss: 0.0021503418684005737
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:51:08<14:46, 886.16s/it]2025-12-22 21:21:22,207 - INFO - Best val loss: 0.001962420530617237, Original val loss: 0.0021503418684005737
+[INFO]     Optimizing self_attn.k_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.00418, val_og_loss=0.00443]
+
  5%|▌         | 1/20 [00:14<04:27, 14.06s/iter, val_loss=0.00418, val_og_loss=0.00443]
+
  5%|▌         | 1/20 [00:15<04:27, 14.06s/iter, val_loss=0.00423, val_og_loss=0.00443]
+
 10%|█         | 2/20 [00:15<01:59,  6.61s/iter, val_loss=0.00423, val_og_loss=0.00443]
+
 10%|█         | 2/20 [00:16<01:59,  6.61s/iter, val_loss=0.00425, val_og_loss=0.00443]
+
 15%|█▌        | 3/20 [00:16<01:11,  4.22s/iter, val_loss=0.00425, val_og_loss=0.00443]
+
 15%|█▌        | 3/20 [00:18<01:11,  4.22s/iter, val_loss=0.00424, val_og_loss=0.00443]
+
 20%|██        | 4/20 [00:18<00:49,  3.10s/iter, val_loss=0.00424, val_og_loss=0.00443]
+
 20%|██        | 4/20 [00:19<00:49,  3.10s/iter, val_loss=0.00426, val_og_loss=0.00443]
+
 25%|██▌       | 5/20 [00:19<00:37,  2.48s/iter, val_loss=0.00426, val_og_loss=0.00443]
+
 25%|██▌       | 5/20 [00:20<00:37,  2.48s/iter, val_loss=0.00425, val_og_loss=0.00443]
+
 30%|███       | 6/20 [00:20<00:29,  2.11s/iter, val_loss=0.00425, val_og_loss=0.00443]
+
 30%|███       | 6/20 [00:22<00:29,  2.11s/iter, val_loss=0.00422, val_og_loss=0.00443]
+
 35%|███▌      | 7/20 [00:22<00:24,  1.87s/iter, val_loss=0.00422, val_og_loss=0.00443]
+
 35%|███▌      | 7/20 [00:23<00:24,  1.87s/iter, val_loss=0.00422, val_og_loss=0.00443]
+
 40%|████      | 8/20 [00:23<00:20,  1.72s/iter, val_loss=0.00422, val_og_loss=0.00443]
+
 40%|████      | 8/20 [00:25<00:20,  1.72s/iter, val_loss=0.00421, val_og_loss=0.00443]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.62s/iter, val_loss=0.00421, val_og_loss=0.00443]
+
 45%|████▌     | 9/20 [00:26<00:17,  1.62s/iter, val_loss=0.00421, val_og_loss=0.00443]
+
 50%|█████     | 10/20 [00:26<00:15,  1.56s/iter, val_loss=0.00421, val_og_loss=0.00443]
+
 50%|█████     | 10/20 [00:27<00:15,  1.56s/iter, val_loss=0.0042, val_og_loss=0.00443] 
+
 55%|█████▌    | 11/20 [00:27<00:13,  1.51s/iter, val_loss=0.0042, val_og_loss=0.00443]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.51s/iter, val_loss=0.00421, val_og_loss=0.00443]
+
 60%|██████    | 12/20 [00:29<00:11,  1.47s/iter, val_loss=0.00421, val_og_loss=0.00443]
+
 60%|██████    | 12/20 [00:30<00:11,  1.47s/iter, val_loss=0.00421, val_og_loss=0.00443]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.45s/iter, val_loss=0.00421, val_og_loss=0.00443]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.45s/iter, val_loss=0.00419, val_og_loss=0.00443]
+
 70%|███████   | 14/20 [00:32<00:08,  1.43s/iter, val_loss=0.00419, val_og_loss=0.00443]
+
 70%|███████   | 14/20 [00:33<00:08,  1.43s/iter, val_loss=0.00416, val_og_loss=0.00443]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.43s/iter, val_loss=0.00416, val_og_loss=0.00443]
+
 75%|███████▌  | 15/20 [00:34<00:07,  1.43s/iter, val_loss=0.00416, val_og_loss=0.00443]
+
 80%|████████  | 16/20 [00:34<00:05,  1.42s/iter, val_loss=0.00416, val_og_loss=0.00443]
+
 80%|████████  | 16/20 [00:36<00:05,  1.42s/iter, val_loss=0.00413, val_og_loss=0.00443]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.41s/iter, val_loss=0.00413, val_og_loss=0.00443]
+
 85%|████████▌ | 17/20 [00:37<00:04,  1.41s/iter, val_loss=0.00411, val_og_loss=0.00443]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.41s/iter, val_loss=0.00411, val_og_loss=0.00443]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.41s/iter, val_loss=0.00415, val_og_loss=0.00443]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.40s/iter, val_loss=0.00415, val_og_loss=0.00443]
+
 95%|█████████▌| 19/20 [00:40<00:01,  1.40s/iter, val_loss=0.00416, val_og_loss=0.00443]
+
100%|██████████| 20/20 [00:40<00:00,  1.41s/iter, val_loss=0.00416, val_og_loss=0.00443]
100%|██████████| 20/20 [00:40<00:00,  2.03s/iter, val_loss=0.00416, val_og_loss=0.00443]
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:51:48<14:46, 886.16s/it]INFO:ParoQuant:Best val loss: 0.0041051520965993404, Original val loss: 0.004428991116583347
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:51:52<14:46, 886.16s/it]2025-12-22 21:22:06,587 - INFO - Best val loss: 0.0041051520965993404, Original val loss: 0.004428991116583347
+[INFO]     Optimizing self_attn.v_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.00157, val_og_loss=0.00167]
+
  5%|▌         | 1/20 [00:14<04:28, 14.11s/iter, val_loss=0.00157, val_og_loss=0.00167]
+
  5%|▌         | 1/20 [00:15<04:28, 14.11s/iter, val_loss=0.00159, val_og_loss=0.00167]
+
 10%|█         | 2/20 [00:15<01:59,  6.63s/iter, val_loss=0.00159, val_og_loss=0.00167]
+
 10%|█         | 2/20 [00:16<01:59,  6.63s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 15%|█▌        | 3/20 [00:16<01:11,  4.23s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 15%|█▌        | 3/20 [00:18<01:11,  4.23s/iter, val_loss=0.00159, val_og_loss=0.00167]
+
 20%|██        | 4/20 [00:18<00:49,  3.11s/iter, val_loss=0.00159, val_og_loss=0.00167]
+
 20%|██        | 4/20 [00:19<00:49,  3.11s/iter, val_loss=0.00159, val_og_loss=0.00167]
+
 25%|██▌       | 5/20 [00:19<00:37,  2.50s/iter, val_loss=0.00159, val_og_loss=0.00167]
+
 25%|██▌       | 5/20 [00:21<00:37,  2.50s/iter, val_loss=0.0016, val_og_loss=0.00167] 
+
 30%|███       | 6/20 [00:21<00:29,  2.12s/iter, val_loss=0.0016, val_og_loss=0.00167]
+
 30%|███       | 6/20 [00:22<00:29,  2.12s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 35%|███▌      | 7/20 [00:22<00:24,  1.89s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 35%|███▌      | 7/20 [00:23<00:24,  1.89s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 40%|████      | 8/20 [00:23<00:20,  1.73s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 40%|████      | 8/20 [00:25<00:20,  1.73s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 45%|████▌     | 9/20 [00:25<00:17,  1.62s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 45%|████▌     | 9/20 [00:26<00:17,  1.62s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 50%|█████     | 10/20 [00:26<00:15,  1.55s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 50%|█████     | 10/20 [00:28<00:15,  1.55s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 55%|█████▌    | 11/20 [00:28<00:13,  1.51s/iter, val_loss=0.00158, val_og_loss=0.00167]
+
 55%|█████▌    | 11/20 [00:29<00:13,  1.51s/iter, val_loss=0.00157, val_og_loss=0.00167]
+
 60%|██████    | 12/20 [00:29<00:11,  1.47s/iter, val_loss=0.00157, val_og_loss=0.00167]
+
 60%|██████    | 12/20 [00:30<00:11,  1.47s/iter, val_loss=0.00157, val_og_loss=0.00167]
+
 65%|██████▌   | 13/20 [00:30<00:10,  1.45s/iter, val_loss=0.00157, val_og_loss=0.00167]
+
 65%|██████▌   | 13/20 [00:32<00:10,  1.45s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 70%|███████   | 14/20 [00:32<00:08,  1.43s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 70%|███████   | 14/20 [00:33<00:08,  1.43s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 75%|███████▌  | 15/20 [00:33<00:07,  1.42s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 75%|███████▌  | 15/20 [00:35<00:07,  1.42s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 80%|████████  | 16/20 [00:35<00:05,  1.42s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 80%|████████  | 16/20 [00:36<00:05,  1.42s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 85%|████████▌ | 17/20 [00:36<00:04,  1.41s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 85%|████████▌ | 17/20 [00:37<00:04,  1.41s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 90%|█████████ | 18/20 [00:37<00:02,  1.41s/iter, val_loss=0.00156, val_og_loss=0.00167]
+
 90%|█████████ | 18/20 [00:39<00:02,  1.41s/iter, val_loss=0.00155, val_og_loss=0.00167]
+
 95%|█████████▌| 19/20 [00:39<00:01,  1.40s/iter, val_loss=0.00155, val_og_loss=0.00167]
+
 95%|█████████▌| 19/20 [00:40<00:01,  1.40s/iter, val_loss=0.00155, val_og_loss=0.00167]
+
100%|██████████| 20/20 [00:40<00:00,  1.41s/iter, val_loss=0.00155, val_og_loss=0.00167]
100%|██████████| 20/20 [00:40<00:00,  2.03s/iter, val_loss=0.00155, val_og_loss=0.00167]
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:52:33<14:46, 886.16s/it]INFO:ParoQuant:Best val loss: 0.0015535852871835232, Original val loss: 0.00166537263430655
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:52:35<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:52:35<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:53:07<14:46, 886.16s/it]2025-12-22 21:22:51,059 - INFO - Best val loss: 0.0015535852871835232, Original val loss: 0.00166537263430655
+[INFO]   > Group 1: ['self_attn.o_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing self_attn.o_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:14<?, ?iter/s, val_loss=0.000371, val_og_loss=0.000426]
+
  5%|▌         | 1/20 [00:14<04:36, 14.53s/iter, val_loss=0.000371, val_og_loss=0.000426]
+
  5%|▌         | 1/20 [00:16<04:36, 14.53s/iter, val_loss=0.000367, val_og_loss=0.000426]
+
 10%|█         | 2/20 [00:16<02:05,  6.98s/iter, val_loss=0.000367, val_og_loss=0.000426]
+
 10%|█         | 2/20 [00:17<02:05,  6.98s/iter, val_loss=0.000371, val_og_loss=0.000426]
+
 15%|█▌        | 3/20 [00:17<01:17,  4.57s/iter, val_loss=0.000371, val_og_loss=0.000426]
+
 15%|█▌        | 3/20 [00:19<01:17,  4.57s/iter, val_loss=0.000372, val_og_loss=0.000426]
+
 20%|██        | 4/20 [00:19<00:54,  3.43s/iter, val_loss=0.000372, val_og_loss=0.000426]
+
 20%|██        | 4/20 [00:21<00:54,  3.43s/iter, val_loss=0.000377, val_og_loss=0.000426]
+
 25%|██▌       | 5/20 [00:21<00:41,  2.80s/iter, val_loss=0.000377, val_og_loss=0.000426]
+
 25%|██▌       | 5/20 [00:22<00:41,  2.80s/iter, val_loss=0.000373, val_og_loss=0.000426]
+
 30%|███       | 6/20 [00:22<00:33,  2.42s/iter, val_loss=0.000373, val_og_loss=0.000426]
+
 30%|███       | 6/20 [00:24<00:33,  2.42s/iter, val_loss=0.000368, val_og_loss=0.000426]
+
 35%|███▌      | 7/20 [00:24<00:28,  2.19s/iter, val_loss=0.000368, val_og_loss=0.000426]
+
 35%|███▌      | 7/20 [00:26<00:28,  2.19s/iter, val_loss=0.000367, val_og_loss=0.000426]
+
 40%|████      | 8/20 [00:26<00:24,  2.02s/iter, val_loss=0.000367, val_og_loss=0.000426]
+
 40%|████      | 8/20 [00:28<00:24,  2.02s/iter, val_loss=0.000373, val_og_loss=0.000426]
+
 45%|████▌     | 9/20 [00:28<00:21,  1.91s/iter, val_loss=0.000373, val_og_loss=0.000426]
+
 45%|████▌     | 9/20 [00:29<00:21,  1.91s/iter, val_loss=0.000372, val_og_loss=0.000426]
+
 50%|█████     | 10/20 [00:29<00:18,  1.85s/iter, val_loss=0.000372, val_og_loss=0.000426]
+
 50%|█████     | 10/20 [00:31<00:18,  1.85s/iter, val_loss=0.000366, val_og_loss=0.000426]
+
 55%|█████▌    | 11/20 [00:31<00:16,  1.80s/iter, val_loss=0.000366, val_og_loss=0.000426]
+
 55%|█████▌    | 11/20 [00:33<00:16,  1.80s/iter, val_loss=0.000372, val_og_loss=0.000426]
+
 60%|██████    | 12/20 [00:33<00:14,  1.76s/iter, val_loss=0.000372, val_og_loss=0.000426]
+
 60%|██████    | 12/20 [00:34<00:14,  1.76s/iter, val_loss=0.000368, val_og_loss=0.000426]
+
 65%|██████▌   | 13/20 [00:34<00:12,  1.73s/iter, val_loss=0.000368, val_og_loss=0.000426]
+
 65%|██████▌   | 13/20 [00:36<00:12,  1.73s/iter, val_loss=0.000367, val_og_loss=0.000426]
+
 70%|███████   | 14/20 [00:36<00:10,  1.73s/iter, val_loss=0.000367, val_og_loss=0.000426]
+
 70%|███████   | 14/20 [00:38<00:10,  1.73s/iter, val_loss=0.000367, val_og_loss=0.000426]
+
 75%|███████▌  | 15/20 [00:38<00:08,  1.71s/iter, val_loss=0.000367, val_og_loss=0.000426]
+
 75%|███████▌  | 15/20 [00:39<00:08,  1.71s/iter, val_loss=0.000365, val_og_loss=0.000426]
+
 80%|████████  | 16/20 [00:39<00:06,  1.70s/iter, val_loss=0.000365, val_og_loss=0.000426]
+
 80%|████████  | 16/20 [00:41<00:06,  1.70s/iter, val_loss=0.000363, val_og_loss=0.000426]
+
 85%|████████▌ | 17/20 [00:41<00:05,  1.69s/iter, val_loss=0.000363, val_og_loss=0.000426]
+
 85%|████████▌ | 17/20 [00:43<00:05,  1.69s/iter, val_loss=0.000364, val_og_loss=0.000426]
+
 90%|█████████ | 18/20 [00:43<00:03,  1.70s/iter, val_loss=0.000364, val_og_loss=0.000426]
+
 90%|█████████ | 18/20 [00:44<00:03,  1.70s/iter, val_loss=0.000366, val_og_loss=0.000426]
+
 95%|█████████▌| 19/20 [00:44<00:01,  1.70s/iter, val_loss=0.000366, val_og_loss=0.000426]
+
 95%|█████████▌| 19/20 [00:46<00:01,  1.70s/iter, val_loss=0.000366, val_og_loss=0.000426]
+
100%|██████████| 20/20 [00:46<00:00,  1.69s/iter, val_loss=0.000366, val_og_loss=0.000426]
100%|██████████| 20/20 [00:46<00:00,  2.33s/iter, val_loss=0.000366, val_og_loss=0.000426]
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:53:54<14:46, 886.16s/it]INFO:ParoQuant:Best val loss: 0.0003625757235568017, Original val loss: 0.0004264394810888916
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:53:54<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:53:54<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:54:39<14:46, 886.16s/it]2025-12-22 21:24:12,453 - INFO - Best val loss: 0.0003625757235568017, Original val loss: 0.0004264394810888916
+[INFO]   > Group 2: ['mlp.gate_proj', 'mlp.up_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.gate_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:29<?, ?iter/s, val_loss=0.00233, val_og_loss=0.00234]
+
  5%|▌         | 1/20 [00:29<09:13, 29.15s/iter, val_loss=0.00233, val_og_loss=0.00234]
+
  5%|▌         | 1/20 [00:31<09:13, 29.15s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 10%|█         | 2/20 [00:31<04:04, 13.60s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 10%|█         | 2/20 [00:34<04:04, 13.60s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 15%|█▌        | 3/20 [00:34<02:26,  8.63s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 15%|█▌        | 3/20 [00:37<02:26,  8.63s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 20%|██        | 4/20 [00:37<01:40,  6.30s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 20%|██        | 4/20 [00:40<01:40,  6.30s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 25%|██▌       | 5/20 [00:40<01:15,  5.01s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 25%|██▌       | 5/20 [00:42<01:15,  5.01s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 30%|███       | 6/20 [00:42<00:59,  4.23s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 30%|███       | 6/20 [00:45<00:59,  4.23s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 35%|███▌      | 7/20 [00:45<00:48,  3.75s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 35%|███▌      | 7/20 [00:48<00:48,  3.75s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 40%|████      | 8/20 [00:48<00:40,  3.42s/iter, val_loss=0.00235, val_og_loss=0.00234]
+
 40%|████      | 8/20 [00:50<00:40,  3.42s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 45%|████▌     | 9/20 [00:50<00:35,  3.20s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 45%|████▌     | 9/20 [00:53<00:35,  3.20s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 50%|█████     | 10/20 [00:53<00:30,  3.05s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 50%|█████     | 10/20 [00:56<00:30,  3.05s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 55%|█████▌    | 11/20 [00:56<00:26,  2.95s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 55%|█████▌    | 11/20 [00:59<00:26,  2.95s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 60%|██████    | 12/20 [00:59<00:23,  2.88s/iter, val_loss=0.00234, val_og_loss=0.00234]
+
 60%|██████    | 12/20 [01:01<00:23,  2.88s/iter, val_loss=0.00233, val_og_loss=0.00234]
+
 65%|██████▌   | 13/20 [01:01<00:19,  2.84s/iter, val_loss=0.00233, val_og_loss=0.00234]
+
 65%|██████▌   | 13/20 [01:04<00:19,  2.84s/iter, val_loss=0.00233, val_og_loss=0.00234]
+
 70%|███████   | 14/20 [01:04<00:16,  2.80s/iter, val_loss=0.00233, val_og_loss=0.00234]
+
 70%|███████   | 14/20 [01:07<00:16,  2.80s/iter, val_loss=0.00233, val_og_loss=0.00234]
+
 75%|███████▌  | 15/20 [01:07<00:13,  2.78s/iter, val_loss=0.00233, val_og_loss=0.00234]
+
 75%|███████▌  | 15/20 [01:09<00:13,  2.78s/iter, val_loss=0.00233, val_og_loss=0.00234]
+
 80%|████████  | 16/20 [01:09<00:11,  2.76s/iter, val_loss=0.00233, val_og_loss=0.00234]
+
 80%|████████  | 16/20 [01:12<00:11,  2.76s/iter, val_loss=0.00232, val_og_loss=0.00234]
+
 85%|████████▌ | 17/20 [01:12<00:08,  2.75s/iter, val_loss=0.00232, val_og_loss=0.00234]
+
 85%|████████▌ | 17/20 [01:15<00:08,  2.75s/iter, val_loss=0.00232, val_og_loss=0.00234]
+
 90%|█████████ | 18/20 [01:15<00:05,  2.74s/iter, val_loss=0.00232, val_og_loss=0.00234]
+
 90%|█████████ | 18/20 [01:18<00:05,  2.74s/iter, val_loss=0.00232, val_og_loss=0.00234]
+
 95%|█████████▌| 19/20 [01:18<00:02,  2.73s/iter, val_loss=0.00232, val_og_loss=0.00234]
+
 95%|█████████▌| 19/20 [01:20<00:02,  2.73s/iter, val_loss=0.00232, val_og_loss=0.00234]
+
100%|██████████| 20/20 [01:20<00:00,  2.73s/iter, val_loss=0.00232, val_og_loss=0.00234]
100%|██████████| 20/20 [01:20<00:00,  4.04s/iter, val_loss=0.00232, val_og_loss=0.00234]
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:56:00<14:46, 886.16s/it]INFO:ParoQuant:Best val loss: 0.0023188393097370863, Original val loss: 0.002342933090403676
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:56:05<14:46, 886.16s/it]2025-12-22 21:26:18,612 - INFO - Best val loss: 0.0023188393097370863, Original val loss: 0.002342933090403676
+[INFO]     Optimizing mlp.up_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:30<?, ?iter/s, val_loss=0.00143, val_og_loss=0.00143]
+
  5%|▌         | 1/20 [00:30<09:34, 30.21s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
  5%|▌         | 1/20 [00:32<09:34, 30.21s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 10%|█         | 2/20 [00:32<04:12, 14.04s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 10%|█         | 2/20 [00:35<04:12, 14.04s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 15%|█▌        | 3/20 [00:35<02:30,  8.88s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 15%|█▌        | 3/20 [00:38<02:30,  8.88s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 20%|██        | 4/20 [00:38<01:43,  6.44s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 20%|██        | 4/20 [00:41<01:43,  6.44s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 25%|██▌       | 5/20 [00:41<01:16,  5.10s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 25%|██▌       | 5/20 [00:43<01:16,  5.10s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 30%|███       | 6/20 [00:43<01:00,  4.29s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 30%|███       | 6/20 [00:46<01:00,  4.29s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 35%|███▌      | 7/20 [00:46<00:49,  3.78s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 35%|███▌      | 7/20 [00:49<00:49,  3.78s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 40%|████      | 8/20 [00:49<00:41,  3.45s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 40%|████      | 8/20 [00:52<00:41,  3.45s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 45%|████▌     | 9/20 [00:52<00:35,  3.22s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 45%|████▌     | 9/20 [00:54<00:35,  3.22s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 50%|█████     | 10/20 [00:54<00:30,  3.07s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 50%|█████     | 10/20 [00:57<00:30,  3.07s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 55%|█████▌    | 11/20 [00:57<00:26,  2.96s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 55%|█████▌    | 11/20 [01:00<00:26,  2.96s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 60%|██████    | 12/20 [01:00<00:23,  2.89s/iter, val_loss=0.00144, val_og_loss=0.00143]
+
 60%|██████    | 12/20 [01:02<00:23,  2.89s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 65%|██████▌   | 13/20 [01:02<00:19,  2.84s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 65%|██████▌   | 13/20 [01:05<00:19,  2.84s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 70%|███████   | 14/20 [01:05<00:16,  2.80s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 70%|███████   | 14/20 [01:08<00:16,  2.80s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 75%|███████▌  | 15/20 [01:08<00:13,  2.78s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 75%|███████▌  | 15/20 [01:11<00:13,  2.78s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 80%|████████  | 16/20 [01:11<00:11,  2.76s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 80%|████████  | 16/20 [01:13<00:11,  2.76s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 85%|████████▌ | 17/20 [01:13<00:08,  2.75s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 85%|████████▌ | 17/20 [01:16<00:08,  2.75s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 90%|█████████ | 18/20 [01:16<00:05,  2.74s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 90%|█████████ | 18/20 [01:19<00:05,  2.74s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 95%|█████████▌| 19/20 [01:19<00:02,  2.73s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
 95%|█████████▌| 19/20 [01:21<00:02,  2.73s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
100%|██████████| 20/20 [01:21<00:00,  2.73s/iter, val_loss=0.00143, val_og_loss=0.00143]
100%|██████████| 20/20 [01:21<00:00,  4.10s/iter, val_loss=0.00143, val_og_loss=0.00143]
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:57:27<14:46, 886.16s/it]INFO:ParoQuant:Best val loss: 0.0014255967689678073, Original val loss: 0.001434665871784091
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:57:28<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:57:28<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [6:58:28<14:46, 886.16s/it]2025-12-22 21:27:45,185 - INFO - Best val loss: 0.0014255967689678073, Original val loss: 0.001434665871784091
+[INFO]   > Group 3: ['mlp.down_proj']
+[INFO]     Capturing Dirty Inputs (Student Forcing)...
+[INFO]     Optimizing mlp.down_proj | Step 1: ['channel_scales', 'angles']
+
+
  0%|          | 0/20 [00:00<?, ?iter/s]
+
  0%|          | 0/20 [00:33<?, ?iter/s, val_loss=0.298, val_og_loss=0.00602]
+
  5%|▌         | 1/20 [00:33<10:45, 33.96s/iter, val_loss=0.298, val_og_loss=0.00602]
+
  5%|▌         | 1/20 [00:39<10:45, 33.96s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 10%|█         | 2/20 [00:39<05:08, 17.12s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 10%|█         | 2/20 [00:44<05:08, 17.12s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 15%|█▌        | 3/20 [00:44<03:19, 11.73s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 15%|█▌        | 3/20 [00:49<03:19, 11.73s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 20%|██        | 4/20 [00:49<02:27,  9.20s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 20%|██        | 4/20 [00:55<02:27,  9.20s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 25%|██▌       | 5/20 [00:55<01:57,  7.81s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 25%|██▌       | 5/20 [01:00<01:57,  7.81s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 30%|███       | 6/20 [01:00<01:37,  6.97s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 30%|███       | 6/20 [01:05<01:37,  6.97s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 35%|███▌      | 7/20 [01:05<01:23,  6.43s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 35%|███▌      | 7/20 [01:11<01:23,  6.43s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 40%|████      | 8/20 [01:11<01:12,  6.08s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 40%|████      | 8/20 [01:16<01:12,  6.08s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 45%|████▌     | 9/20 [01:16<01:04,  5.84s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 45%|████▌     | 9/20 [01:21<01:04,  5.84s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 50%|█████     | 10/20 [01:21<00:56,  5.69s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 50%|█████     | 10/20 [01:27<00:56,  5.69s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 55%|█████▌    | 11/20 [01:27<00:50,  5.58s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 55%|█████▌    | 11/20 [01:32<00:50,  5.58s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 60%|██████    | 12/20 [01:32<00:44,  5.51s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 60%|██████    | 12/20 [01:37<00:44,  5.51s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 65%|██████▌   | 13/20 [01:37<00:38,  5.45s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 65%|██████▌   | 13/20 [01:43<00:38,  5.45s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 70%|███████   | 14/20 [01:43<00:32,  5.42s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 70%|███████   | 14/20 [01:48<00:32,  5.42s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 75%|███████▌  | 15/20 [01:48<00:26,  5.39s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 75%|███████▌  | 15/20 [01:53<00:26,  5.39s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 80%|████████  | 16/20 [01:53<00:21,  5.37s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 80%|████████  | 16/20 [01:59<00:21,  5.37s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 85%|████████▌ | 17/20 [01:59<00:16,  5.36s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 85%|████████▌ | 17/20 [02:04<00:16,  5.36s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 90%|█████████ | 18/20 [02:04<00:10,  5.35s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 90%|█████████ | 18/20 [02:09<00:10,  5.35s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 95%|█████████▌| 19/20 [02:09<00:05,  5.35s/iter, val_loss=0.298, val_og_loss=0.00602]
+
 95%|█████████▌| 19/20 [02:15<00:05,  5.35s/iter, val_loss=0.298, val_og_loss=0.00602]
+
100%|██████████| 20/20 [02:15<00:00,  5.34s/iter, val_loss=0.298, val_og_loss=0.00602]
100%|██████████| 20/20 [02:15<00:00,  6.76s/iter, val_loss=0.298, val_og_loss=0.00602]
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [7:00:43<14:46, 886.16s/it]INFO:ParoQuant:Best val loss: 0.00601727981120348, Original val loss: 0.00601727981120348
+
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [7:00:44<14:46, 886.16s/it]
                                                                     

Processing Blocks:  97%|█████████▋| 31/32 [7:00:54<14:46, 886.16s/it]
Processing Blocks: 100%|██████████| 32/32 [7:01:31<00:00, 886.81s/it]
Processing Blocks: 100%|██████████| 32/32 [7:01:31<00:00, 790.37s/it]
+2025-12-22 21:31:01,579 - INFO - Best val loss: 0.00601727981120348, Original val loss: 0.00601727981120348
+[INFO] Saving Layer 31 results...
+[INFO] Generating output for next block...
+[INFO] Optimization Completed.
+INFO 12-22 21:32:02 [__init__.py:241] Automatically detected platform cuda.
+`torch_dtype` is deprecated! Use `dtype` instead!
+Loading model and tokenizer...
+
Loading checkpoint shards:   0%|          | 0/4 [00:00<?, ?it/s]
Loading checkpoint shards:  25%|██▌       | 1/4 [00:02<00:07,  2.42s/it]
Loading checkpoint shards:  50%|█████     | 2/4 [00:05<00:05,  2.76s/it]
Loading checkpoint shards:  75%|███████▌  | 3/4 [00:08<00:02,  2.72s/it]
Loading checkpoint shards: 100%|██████████| 4/4 [00:08<00:00,  1.81s/it]
Loading checkpoint shards: 100%|██████████| 4/4 [00:08<00:00,  2.12s/it]
+Replacing weights with pseudo-dequantized versions...
+
Layers:   0%|          | 0/32 [00:00<?, ?it/s]
Layers:   3%|▎         | 1/32 [00:03<01:46,  3.44s/it]
Layers:   6%|▋         | 2/32 [00:04<01:08,  2.27s/it]
Layers:   9%|▉         | 3/32 [00:06<00:54,  1.89s/it]
Layers:  12%|█▎        | 4/32 [00:07<00:47,  1.69s/it]
Layers:  16%|█▌        | 5/32 [00:08<00:41,  1.54s/it]
Layers:  19%|█▉        | 6/32 [00:10<00:37,  1.44s/it]
Layers:  22%|██▏       | 7/32 [00:11<00:33,  1.34s/it]
Layers:  25%|██▌       | 8/32 [00:12<00:29,  1.25s/it]
Layers:  28%|██▊       | 9/32 [00:13<00:27,  1.19s/it]
Layers:  31%|███▏      | 10/32 [00:14<00:26,  1.19s/it]
Layers:  34%|███▍      | 11/32 [00:15<00:24,  1.19s/it]
Layers:  38%|███▊      | 12/32 [00:17<00:23,  1.19s/it]
Layers:  41%|████      | 13/32 [00:18<00:23,  1.22s/it]
Layers:  44%|████▍     | 14/32 [00:19<00:21,  1.21s/it]
Layers:  47%|████▋     | 15/32 [00:20<00:20,  1.21s/it]
Layers:  50%|█████     | 16/32 [00:21<00:19,  1.21s/it]
Layers:  53%|█████▎    | 17/32 [00:23<00:18,  1.23s/it]
Layers:  56%|█████▋    | 18/32 [00:24<00:17,  1.25s/it]
Layers:  59%|█████▉    | 19/32 [00:25<00:16,  1.23s/it]
Layers:  62%|██████▎   | 20/32 [00:26<00:14,  1.23s/it]
Layers:  66%|██████▌   | 21/32 [00:28<00:13,  1.22s/it]
Layers:  69%|██████▉   | 22/32 [00:29<00:12,  1.22s/it]
Layers:  72%|███████▏  | 23/32 [00:30<00:11,  1.24s/it]
Layers:  75%|███████▌  | 24/32 [00:31<00:09,  1.24s/it]
Layers:  78%|███████▊  | 25/32 [00:33<00:08,  1.23s/it]
Layers:  81%|████████▏ | 26/32 [00:34<00:07,  1.22s/it]
Layers:  84%|████████▍ | 27/32 [00:35<00:06,  1.22s/it]
Layers:  88%|████████▊ | 28/32 [00:36<00:04,  1.23s/it]
Layers:  91%|█████████ | 29/32 [00:38<00:03,  1.24s/it]
Layers:  94%|█████████▍| 30/32 [00:39<00:02,  1.24s/it]
Layers:  97%|█████████▋| 31/32 [00:40<00:01,  1.31s/it]
Layers: 100%|██████████| 32/32 [00:41<00:00,  1.25s/it]
Layers: 100%|██████████| 32/32 [00:41<00:00,  1.31s/it]
+
Evaluating c4:   0%|          | 0/256 [00:00<?, ?it/s]
Evaluating c4:   0%|          | 0/256 [00:01<?, ?it/s, ppl=5.40]
Evaluating c4:   0%|          | 1/256 [00:01<05:16,  1.24s/it, ppl=5.40]
Evaluating c4:   0%|          | 1/256 [00:02<05:16,  1.24s/it, ppl=6.55]
Evaluating c4:   1%|          | 2/256 [00:02<04:19,  1.02s/it, ppl=6.55]
Evaluating c4:   1%|          | 2/256 [00:02<04:19,  1.02s/it, ppl=8.48]
Evaluating c4:   1%|          | 3/256 [00:02<04:00,  1.05it/s, ppl=8.48]
Evaluating c4:   1%|          | 3/256 [00:03<04:00,  1.05it/s, ppl=9.54]
Evaluating c4:   2%|▏         | 4/256 [00:03<03:51,  1.09it/s, ppl=9.54]
Evaluating c4:   2%|▏         | 4/256 [00:04<03:51,  1.09it/s, ppl=7.43]
Evaluating c4:   2%|▏         | 5/256 [00:04<03:46,  1.11it/s, ppl=7.43]
Evaluating c4:   2%|▏         | 5/256 [00:05<03:46,  1.11it/s, ppl=5.95]
Evaluating c4:   2%|▏         | 6/256 [00:05<03:43,  1.12it/s, ppl=5.95]
Evaluating c4:   2%|▏         | 6/256 [00:06<03:43,  1.12it/s, ppl=5.97]
Evaluating c4:   3%|▎         | 7/256 [00:06<03:40,  1.13it/s, ppl=5.97]
Evaluating c4:   3%|▎         | 7/256 [00:07<03:40,  1.13it/s, ppl=6.20]
Evaluating c4:   3%|▎         | 8/256 [00:07<03:41,  1.12it/s, ppl=6.20]
Evaluating c4:   3%|▎         | 8/256 [00:08<03:41,  1.12it/s, ppl=6.99]
Evaluating c4:   4%|▎         | 9/256 [00:08<03:39,  1.13it/s, ppl=6.99]
Evaluating c4:   4%|▎         | 9/256 [00:09<03:39,  1.13it/s, ppl=7.29]
Evaluating c4:   4%|▍         | 10/256 [00:09<03:36,  1.14it/s, ppl=7.29]
Evaluating c4:   4%|▍         | 10/256 [00:09<03:36,  1.14it/s, ppl=6.74]
Evaluating c4:   4%|▍         | 11/256 [00:09<03:34,  1.14it/s, ppl=6.74]
Evaluating c4:   4%|▍         | 11/256 [00:10<03:34,  1.14it/s, ppl=6.70]
Evaluating c4:   5%|▍         | 12/256 [00:10<03:32,  1.15it/s, ppl=6.70]
Evaluating c4:   5%|▍         | 12/256 [00:11<03:32,  1.15it/s, ppl=6.72]
Evaluating c4:   5%|▌         | 13/256 [00:11<03:31,  1.15it/s, ppl=6.72]
Evaluating c4:   5%|▌         | 13/256 [00:12<03:31,  1.15it/s, ppl=6.59]
Evaluating c4:   5%|▌         | 14/256 [00:12<03:30,  1.15it/s, ppl=6.59]
Evaluating c4:   5%|▌         | 14/256 [00:13<03:30,  1.15it/s, ppl=7.05]
Evaluating c4:   6%|▌         | 15/256 [00:13<03:29,  1.15it/s, ppl=7.05]
Evaluating c4:   6%|▌         | 15/256 [00:14<03:29,  1.15it/s, ppl=7.27]
Evaluating c4:   6%|▋         | 16/256 [00:14<03:30,  1.14it/s, ppl=7.27]
Evaluating c4:   6%|▋         | 16/256 [00:15<03:30,  1.14it/s, ppl=7.22]
Evaluating c4:   7%|▋         | 17/256 [00:15<03:28,  1.14it/s, ppl=7.22]
Evaluating c4:   7%|▋         | 17/256 [00:16<03:28,  1.14it/s, ppl=7.19]
Evaluating c4:   7%|▋         | 18/256 [00:16<03:27,  1.15it/s, ppl=7.19]
Evaluating c4:   7%|▋         | 18/256 [00:16<03:27,  1.15it/s, ppl=7.42]
Evaluating c4:   7%|▋         | 19/256 [00:16<03:25,  1.15it/s, ppl=7.42]
Evaluating c4:   7%|▋         | 19/256 [00:17<03:25,  1.15it/s, ppl=7.80]
Evaluating c4:   8%|▊         | 20/256 [00:17<03:24,  1.15it/s, ppl=7.80]
Evaluating c4:   8%|▊         | 20/256 [00:18<03:24,  1.15it/s, ppl=8.05]
Evaluating c4:   8%|▊         | 21/256 [00:18<03:23,  1.15it/s, ppl=8.05]
Evaluating c4:   8%|▊         | 21/256 [00:19<03:23,  1.15it/s, ppl=8.07]
Evaluating c4:   9%|▊         | 22/256 [00:19<03:22,  1.15it/s, ppl=8.07]
Evaluating c4:   9%|▊         | 22/256 [00:20<03:22,  1.15it/s, ppl=8.18]
Evaluating c4:   9%|▉         | 23/256 [00:20<03:21,  1.15it/s, ppl=8.18]
Evaluating c4:   9%|▉         | 23/256 [00:21<03:21,  1.15it/s, ppl=8.18]
Evaluating c4:   9%|▉         | 24/256 [00:21<03:22,  1.14it/s, ppl=8.18]
Evaluating c4:   9%|▉         | 24/256 [00:22<03:22,  1.14it/s, ppl=8.28]
Evaluating c4:  10%|▉         | 25/256 [00:22<03:21,  1.15it/s, ppl=8.28]
Evaluating c4:  10%|▉         | 25/256 [00:23<03:21,  1.15it/s, ppl=8.46]
Evaluating c4:  10%|█         | 26/256 [00:23<03:20,  1.15it/s, ppl=8.46]
Evaluating c4:  10%|█         | 26/256 [00:23<03:20,  1.15it/s, ppl=8.33]
Evaluating c4:  11%|█         | 27/256 [00:23<03:18,  1.15it/s, ppl=8.33]
Evaluating c4:  11%|█         | 27/256 [00:24<03:18,  1.15it/s, ppl=8.38]
Evaluating c4:  11%|█         | 28/256 [00:24<03:17,  1.15it/s, ppl=8.38]
Evaluating c4:  11%|█         | 28/256 [00:25<03:17,  1.15it/s, ppl=8.48]
Evaluating c4:  11%|█▏        | 29/256 [00:25<03:16,  1.15it/s, ppl=8.48]
Evaluating c4:  11%|█▏        | 29/256 [00:26<03:16,  1.15it/s, ppl=8.73]
Evaluating c4:  12%|█▏        | 30/256 [00:26<03:15,  1.15it/s, ppl=8.73]
Evaluating c4:  12%|█▏        | 30/256 [00:27<03:15,  1.15it/s, ppl=8.78]
Evaluating c4:  12%|█▏        | 31/256 [00:27<03:14,  1.15it/s, ppl=8.78]
Evaluating c4:  12%|█▏        | 31/256 [00:28<03:14,  1.15it/s, ppl=8.67]
Evaluating c4:  12%|█▎        | 32/256 [00:28<03:16,  1.14it/s, ppl=8.67]
Evaluating c4:  12%|█▎        | 32/256 [00:29<03:16,  1.14it/s, ppl=8.57]
Evaluating c4:  13%|█▎        | 33/256 [00:29<03:14,  1.15it/s, ppl=8.57]
Evaluating c4:  13%|█▎        | 33/256 [00:29<03:14,  1.15it/s, ppl=8.37]
Evaluating c4:  13%|█▎        | 34/256 [00:29<03:13,  1.15it/s, ppl=8.37]
Evaluating c4:  13%|█▎        | 34/256 [00:30<03:13,  1.15it/s, ppl=8.38]
Evaluating c4:  14%|█▎        | 35/256 [00:30<03:11,  1.15it/s, ppl=8.38]
Evaluating c4:  14%|█▎        | 35/256 [00:31<03:11,  1.15it/s, ppl=8.32]
Evaluating c4:  14%|█▍        | 36/256 [00:31<03:10,  1.15it/s, ppl=8.32]
Evaluating c4:  14%|█▍        | 36/256 [00:32<03:10,  1.15it/s, ppl=8.36]
Evaluating c4:  14%|█▍        | 37/256 [00:32<03:09,  1.15it/s, ppl=8.36]
Evaluating c4:  14%|█▍        | 37/256 [00:33<03:09,  1.15it/s, ppl=8.29]
Evaluating c4:  15%|█▍        | 38/256 [00:33<03:08,  1.15it/s, ppl=8.29]
Evaluating c4:  15%|█▍        | 38/256 [00:34<03:08,  1.15it/s, ppl=8.46]
Evaluating c4:  15%|█▌        | 39/256 [00:34<03:07,  1.16it/s, ppl=8.46]
Evaluating c4:  15%|█▌        | 39/256 [00:35<03:07,  1.16it/s, ppl=8.56]
Evaluating c4:  16%|█▌        | 40/256 [00:35<03:06,  1.16it/s, ppl=8.56]
Evaluating c4:  16%|█▌        | 40/256 [00:36<03:06,  1.16it/s, ppl=8.39]
Evaluating c4:  16%|█▌        | 41/256 [00:36<03:08,  1.14it/s, ppl=8.39]
Evaluating c4:  16%|█▌        | 41/256 [00:36<03:08,  1.14it/s, ppl=8.41]
Evaluating c4:  16%|█▋        | 42/256 [00:36<03:06,  1.15it/s, ppl=8.41]
Evaluating c4:  16%|█▋        | 42/256 [00:37<03:06,  1.15it/s, ppl=8.46]
Evaluating c4:  17%|█▋        | 43/256 [00:37<03:05,  1.15it/s, ppl=8.46]
Evaluating c4:  17%|█▋        | 43/256 [00:38<03:05,  1.15it/s, ppl=8.10]
Evaluating c4:  17%|█▋        | 44/256 [00:38<03:04,  1.15it/s, ppl=8.10]
Evaluating c4:  17%|█▋        | 44/256 [00:39<03:04,  1.15it/s, ppl=8.19]
Evaluating c4:  18%|█▊        | 45/256 [00:39<03:03,  1.15it/s, ppl=8.19]
Evaluating c4:  18%|█▊        | 45/256 [00:40<03:03,  1.15it/s, ppl=8.22]
Evaluating c4:  18%|█▊        | 46/256 [00:40<03:02,  1.15it/s, ppl=8.22]
Evaluating c4:  18%|█▊        | 46/256 [00:41<03:02,  1.15it/s, ppl=8.19]
Evaluating c4:  18%|█▊        | 47/256 [00:41<03:01,  1.15it/s, ppl=8.19]
Evaluating c4:  18%|█▊        | 47/256 [00:42<03:01,  1.15it/s, ppl=8.31]
Evaluating c4:  19%|█▉        | 48/256 [00:42<03:00,  1.16it/s, ppl=8.31]
Evaluating c4:  19%|█▉        | 48/256 [00:43<03:00,  1.16it/s, ppl=8.13]
Evaluating c4:  19%|█▉        | 49/256 [00:43<03:01,  1.14it/s, ppl=8.13]
Evaluating c4:  19%|█▉        | 49/256 [00:43<03:01,  1.14it/s, ppl=8.15]
Evaluating c4:  20%|█▉        | 50/256 [00:43<02:59,  1.14it/s, ppl=8.15]
Evaluating c4:  20%|█▉        | 50/256 [00:44<02:59,  1.14it/s, ppl=8.22]
Evaluating c4:  20%|█▉        | 51/256 [00:44<02:58,  1.15it/s, ppl=8.22]
Evaluating c4:  20%|█▉        | 51/256 [00:45<02:58,  1.15it/s, ppl=8.24]
Evaluating c4:  20%|██        | 52/256 [00:45<02:57,  1.15it/s, ppl=8.24]
Evaluating c4:  20%|██        | 52/256 [00:46<02:57,  1.15it/s, ppl=8.35]
Evaluating c4:  21%|██        | 53/256 [00:46<02:56,  1.15it/s, ppl=8.35]
Evaluating c4:  21%|██        | 53/256 [00:47<02:56,  1.15it/s, ppl=8.48]
Evaluating c4:  21%|██        | 54/256 [00:47<02:55,  1.15it/s, ppl=8.48]
Evaluating c4:  21%|██        | 54/256 [00:48<02:55,  1.15it/s, ppl=8.50]
Evaluating c4:  21%|██▏       | 55/256 [00:48<02:54,  1.15it/s, ppl=8.50]
Evaluating c4:  21%|██▏       | 55/256 [00:49<02:54,  1.15it/s, ppl=8.47]
Evaluating c4:  22%|██▏       | 56/256 [00:49<02:53,  1.15it/s, ppl=8.47]
Evaluating c4:  22%|██▏       | 56/256 [00:49<02:53,  1.15it/s, ppl=8.49]
Evaluating c4:  22%|██▏       | 57/256 [00:49<02:54,  1.14it/s, ppl=8.49]
Evaluating c4:  22%|██▏       | 57/256 [00:50<02:54,  1.14it/s, ppl=8.61]
Evaluating c4:  23%|██▎       | 58/256 [00:50<02:52,  1.15it/s, ppl=8.61]
Evaluating c4:  23%|██▎       | 58/256 [00:51<02:52,  1.15it/s, ppl=8.54]
Evaluating c4:  23%|██▎       | 59/256 [00:51<02:51,  1.15it/s, ppl=8.54]
Evaluating c4:  23%|██▎       | 59/256 [00:52<02:51,  1.15it/s, ppl=8.59]
Evaluating c4:  23%|██▎       | 60/256 [00:52<02:50,  1.15it/s, ppl=8.59]
Evaluating c4:  23%|██▎       | 60/256 [00:53<02:50,  1.15it/s, ppl=8.38]
Evaluating c4:  24%|██▍       | 61/256 [00:53<02:49,  1.15it/s, ppl=8.38]
Evaluating c4:  24%|██▍       | 61/256 [00:54<02:49,  1.15it/s, ppl=8.38]
Evaluating c4:  24%|██▍       | 62/256 [00:54<02:48,  1.15it/s, ppl=8.38]
Evaluating c4:  24%|██▍       | 62/256 [00:55<02:48,  1.15it/s, ppl=8.29]
Evaluating c4:  25%|██▍       | 63/256 [00:55<02:47,  1.15it/s, ppl=8.29]
Evaluating c4:  25%|██▍       | 63/256 [00:56<02:47,  1.15it/s, ppl=8.38]
Evaluating c4:  25%|██▌       | 64/256 [00:56<02:49,  1.13it/s, ppl=8.38]
Evaluating c4:  25%|██▌       | 64/256 [00:56<02:49,  1.13it/s, ppl=8.36]
Evaluating c4:  25%|██▌       | 65/256 [00:56<02:48,  1.14it/s, ppl=8.36]
Evaluating c4:  25%|██▌       | 65/256 [00:57<02:48,  1.14it/s, ppl=8.23]
Evaluating c4:  26%|██▌       | 66/256 [00:57<02:46,  1.14it/s, ppl=8.23]
Evaluating c4:  26%|██▌       | 66/256 [00:58<02:46,  1.14it/s, ppl=8.27]
Evaluating c4:  26%|██▌       | 67/256 [00:58<02:45,  1.14it/s, ppl=8.27]
Evaluating c4:  26%|██▌       | 67/256 [00:59<02:45,  1.14it/s, ppl=8.52]
Evaluating c4:  27%|██▋       | 68/256 [00:59<02:44,  1.15it/s, ppl=8.52]
Evaluating c4:  27%|██▋       | 68/256 [01:00<02:44,  1.15it/s, ppl=8.50]
Evaluating c4:  27%|██▋       | 69/256 [01:00<02:42,  1.15it/s, ppl=8.50]
Evaluating c4:  27%|██▋       | 69/256 [01:01<02:42,  1.15it/s, ppl=8.46]
Evaluating c4:  27%|██▋       | 70/256 [01:01<02:41,  1.15it/s, ppl=8.46]
Evaluating c4:  27%|██▋       | 70/256 [01:02<02:41,  1.15it/s, ppl=8.50]
Evaluating c4:  28%|██▊       | 71/256 [01:02<02:40,  1.15it/s, ppl=8.50]
Evaluating c4:  28%|██▊       | 71/256 [01:03<02:40,  1.15it/s, ppl=8.55]
Evaluating c4:  28%|██▊       | 72/256 [01:03<02:41,  1.14it/s, ppl=8.55]
Evaluating c4:  28%|██▊       | 72/256 [01:03<02:41,  1.14it/s, ppl=8.60]
Evaluating c4:  29%|██▊       | 73/256 [01:03<02:39,  1.15it/s, ppl=8.60]
Evaluating c4:  29%|██▊       | 73/256 [01:04<02:39,  1.15it/s, ppl=8.62]
Evaluating c4:  29%|██▉       | 74/256 [01:04<02:38,  1.15it/s, ppl=8.62]
Evaluating c4:  29%|██▉       | 74/256 [01:05<02:38,  1.15it/s, ppl=8.64]
Evaluating c4:  29%|██▉       | 75/256 [01:05<02:37,  1.15it/s, ppl=8.64]
Evaluating c4:  29%|██▉       | 75/256 [01:06<02:37,  1.15it/s, ppl=8.67]
Evaluating c4:  30%|██▉       | 76/256 [01:06<02:36,  1.15it/s, ppl=8.67]
Evaluating c4:  30%|██▉       | 76/256 [01:07<02:36,  1.15it/s, ppl=8.70]
Evaluating c4:  30%|███       | 77/256 [01:07<02:35,  1.15it/s, ppl=8.70]
Evaluating c4:  30%|███       | 77/256 [01:08<02:35,  1.15it/s, ppl=8.69]
Evaluating c4:  30%|███       | 78/256 [01:08<02:34,  1.15it/s, ppl=8.69]
Evaluating c4:  30%|███       | 78/256 [01:09<02:34,  1.15it/s, ppl=8.71]
Evaluating c4:  31%|███       | 79/256 [01:09<02:35,  1.14it/s, ppl=8.71]
Evaluating c4:  31%|███       | 79/256 [01:10<02:35,  1.14it/s, ppl=8.74]
Evaluating c4:  31%|███▏      | 80/256 [01:10<02:33,  1.15it/s, ppl=8.74]
Evaluating c4:  31%|███▏      | 80/256 [01:10<02:33,  1.15it/s, ppl=8.76]
Evaluating c4:  32%|███▏      | 81/256 [01:10<02:32,  1.15it/s, ppl=8.76]
Evaluating c4:  32%|███▏      | 81/256 [01:11<02:32,  1.15it/s, ppl=8.78]
Evaluating c4:  32%|███▏      | 82/256 [01:11<02:31,  1.15it/s, ppl=8.78]
Evaluating c4:  32%|███▏      | 82/256 [01:12<02:31,  1.15it/s, ppl=8.77]
Evaluating c4:  32%|███▏      | 83/256 [01:12<02:30,  1.15it/s, ppl=8.77]
Evaluating c4:  32%|███▏      | 83/256 [01:13<02:30,  1.15it/s, ppl=8.86]
Evaluating c4:  33%|███▎      | 84/256 [01:13<02:30,  1.15it/s, ppl=8.86]
Evaluating c4:  33%|███▎      | 84/256 [01:14<02:30,  1.15it/s, ppl=8.80]
Evaluating c4:  33%|███▎      | 85/256 [01:14<02:29,  1.15it/s, ppl=8.80]
Evaluating c4:  33%|███▎      | 85/256 [01:15<02:29,  1.15it/s, ppl=8.80]
Evaluating c4:  34%|███▎      | 86/256 [01:15<02:28,  1.15it/s, ppl=8.80]
Evaluating c4:  34%|███▎      | 86/256 [01:16<02:28,  1.15it/s, ppl=8.83]
Evaluating c4:  34%|███▍      | 87/256 [01:16<02:28,  1.14it/s, ppl=8.83]
Evaluating c4:  34%|███▍      | 87/256 [01:17<02:28,  1.14it/s, ppl=8.84]
Evaluating c4:  34%|███▍      | 88/256 [01:17<02:27,  1.14it/s, ppl=8.84]
Evaluating c4:  34%|███▍      | 88/256 [01:17<02:27,  1.14it/s, ppl=8.93]
Evaluating c4:  35%|███▍      | 89/256 [01:17<02:26,  1.14it/s, ppl=8.93]
Evaluating c4:  35%|███▍      | 89/256 [01:18<02:26,  1.14it/s, ppl=8.98]
Evaluating c4:  35%|███▌      | 90/256 [01:18<02:25,  1.14it/s, ppl=8.98]
Evaluating c4:  35%|███▌      | 90/256 [01:19<02:25,  1.14it/s, ppl=8.97]
Evaluating c4:  36%|███▌      | 91/256 [01:19<02:23,  1.15it/s, ppl=8.97]
Evaluating c4:  36%|███▌      | 91/256 [01:20<02:23,  1.15it/s, ppl=8.97]
Evaluating c4:  36%|███▌      | 92/256 [01:20<02:22,  1.15it/s, ppl=8.97]
Evaluating c4:  36%|███▌      | 92/256 [01:21<02:22,  1.15it/s, ppl=8.97]
Evaluating c4:  36%|███▋      | 93/256 [01:21<02:21,  1.15it/s, ppl=8.97]
Evaluating c4:  36%|███▋      | 93/256 [01:22<02:21,  1.15it/s, ppl=8.99]
Evaluating c4:  37%|███▋      | 94/256 [01:22<02:20,  1.15it/s, ppl=8.99]
Evaluating c4:  37%|███▋      | 94/256 [01:23<02:20,  1.15it/s, ppl=9.01]
Evaluating c4:  37%|███▋      | 95/256 [01:23<02:21,  1.14it/s, ppl=9.01]
Evaluating c4:  37%|███▋      | 95/256 [01:23<02:21,  1.14it/s, ppl=9.04]
Evaluating c4:  38%|███▊      | 96/256 [01:23<02:19,  1.14it/s, ppl=9.04]
Evaluating c4:  38%|███▊      | 96/256 [01:24<02:19,  1.14it/s, ppl=9.01]
Evaluating c4:  38%|███▊      | 97/256 [01:24<02:18,  1.15it/s, ppl=9.01]
Evaluating c4:  38%|███▊      | 97/256 [01:25<02:18,  1.15it/s, ppl=9.13]
Evaluating c4:  38%|███▊      | 98/256 [01:25<02:17,  1.15it/s, ppl=9.13]
Evaluating c4:  38%|███▊      | 98/256 [01:26<02:17,  1.15it/s, ppl=9.14]
Evaluating c4:  39%|███▊      | 99/256 [01:26<02:16,  1.15it/s, ppl=9.14]
Evaluating c4:  39%|███▊      | 99/256 [01:27<02:16,  1.15it/s, ppl=9.08]
Evaluating c4:  39%|███▉      | 100/256 [01:27<02:15,  1.15it/s, ppl=9.08]
Evaluating c4:  39%|███▉      | 100/256 [01:28<02:15,  1.15it/s, ppl=9.05]
Evaluating c4:  39%|███▉      | 101/256 [01:28<02:14,  1.15it/s, ppl=9.05]
Evaluating c4:  39%|███▉      | 101/256 [01:29<02:14,  1.15it/s, ppl=8.91]
Evaluating c4:  40%|███▉      | 102/256 [01:29<02:13,  1.15it/s, ppl=8.91]
Evaluating c4:  40%|███▉      | 102/256 [01:30<02:13,  1.15it/s, ppl=8.91]
Evaluating c4:  40%|████      | 103/256 [01:30<02:13,  1.15it/s, ppl=8.91]
Evaluating c4:  40%|████      | 103/256 [01:30<02:13,  1.15it/s, ppl=8.98]
Evaluating c4:  41%|████      | 104/256 [01:30<02:12,  1.15it/s, ppl=8.98]
Evaluating c4:  41%|████      | 104/256 [01:31<02:12,  1.15it/s, ppl=8.99]
Evaluating c4:  41%|████      | 105/256 [01:31<02:11,  1.15it/s, ppl=8.99]
Evaluating c4:  41%|████      | 105/256 [01:32<02:11,  1.15it/s, ppl=9.00]
Evaluating c4:  41%|████▏     | 106/256 [01:32<02:10,  1.15it/s, ppl=9.00]
Evaluating c4:  41%|████▏     | 106/256 [01:33<02:10,  1.15it/s, ppl=9.02]
Evaluating c4:  42%|████▏     | 107/256 [01:33<02:09,  1.15it/s, ppl=9.02]
Evaluating c4:  42%|████▏     | 107/256 [01:34<02:09,  1.15it/s, ppl=9.07]
Evaluating c4:  42%|████▏     | 108/256 [01:34<02:08,  1.15it/s, ppl=9.07]
Evaluating c4:  42%|████▏     | 108/256 [01:35<02:08,  1.15it/s, ppl=9.08]
Evaluating c4:  43%|████▎     | 109/256 [01:35<02:07,  1.16it/s, ppl=9.08]
Evaluating c4:  43%|████▎     | 109/256 [01:36<02:07,  1.16it/s, ppl=9.09]
Evaluating c4:  43%|████▎     | 110/256 [01:36<02:06,  1.16it/s, ppl=9.09]
Evaluating c4:  43%|████▎     | 110/256 [01:37<02:06,  1.16it/s, ppl=9.13]
Evaluating c4:  43%|████▎     | 111/256 [01:37<02:06,  1.14it/s, ppl=9.13]
Evaluating c4:  43%|████▎     | 111/256 [01:37<02:06,  1.14it/s, ppl=9.24]
Evaluating c4:  44%|████▍     | 112/256 [01:37<02:05,  1.15it/s, ppl=9.24]
Evaluating c4:  44%|████▍     | 112/256 [01:38<02:05,  1.15it/s, ppl=9.23]
Evaluating c4:  44%|████▍     | 113/256 [01:38<02:04,  1.15it/s, ppl=9.23]
Evaluating c4:  44%|████▍     | 113/256 [01:39<02:04,  1.15it/s, ppl=9.23]
Evaluating c4:  45%|████▍     | 114/256 [01:39<02:03,  1.15it/s, ppl=9.23]
Evaluating c4:  45%|████▍     | 114/256 [01:40<02:03,  1.15it/s, ppl=9.22]
Evaluating c4:  45%|████▍     | 115/256 [01:40<02:02,  1.15it/s, ppl=9.22]
Evaluating c4:  45%|████▍     | 115/256 [01:41<02:02,  1.15it/s, ppl=9.24]
Evaluating c4:  45%|████▌     | 116/256 [01:41<02:01,  1.15it/s, ppl=9.24]
Evaluating c4:  45%|████▌     | 116/256 [01:42<02:01,  1.15it/s, ppl=9.22]
Evaluating c4:  46%|████▌     | 117/256 [01:42<02:00,  1.15it/s, ppl=9.22]
Evaluating c4:  46%|████▌     | 117/256 [01:43<02:00,  1.15it/s, ppl=9.29]
Evaluating c4:  46%|████▌     | 118/256 [01:43<01:59,  1.16it/s, ppl=9.29]
Evaluating c4:  46%|████▌     | 118/256 [01:43<01:59,  1.16it/s, ppl=9.40]
Evaluating c4:  46%|████▋     | 119/256 [01:43<01:59,  1.14it/s, ppl=9.40]
Evaluating c4:  46%|████▋     | 119/256 [01:44<01:59,  1.14it/s, ppl=9.41]
Evaluating c4:  47%|████▋     | 120/256 [01:44<01:58,  1.15it/s, ppl=9.41]
Evaluating c4:  47%|████▋     | 120/256 [01:45<01:58,  1.15it/s, ppl=9.33]
Evaluating c4:  47%|████▋     | 121/256 [01:45<01:57,  1.15it/s, ppl=9.33]
Evaluating c4:  47%|████▋     | 121/256 [01:46<01:57,  1.15it/s, ppl=9.35]
Evaluating c4:  48%|████▊     | 122/256 [01:46<01:56,  1.15it/s, ppl=9.35]
Evaluating c4:  48%|████▊     | 122/256 [01:47<01:56,  1.15it/s, ppl=9.35]
Evaluating c4:  48%|████▊     | 123/256 [01:47<01:55,  1.15it/s, ppl=9.35]
Evaluating c4:  48%|████▊     | 123/256 [01:48<01:55,  1.15it/s, ppl=9.34]
Evaluating c4:  48%|████▊     | 124/256 [01:48<01:54,  1.15it/s, ppl=9.34]
Evaluating c4:  48%|████▊     | 124/256 [01:49<01:54,  1.15it/s, ppl=9.34]
Evaluating c4:  49%|████▉     | 125/256 [01:49<01:53,  1.15it/s, ppl=9.34]
Evaluating c4:  49%|████▉     | 125/256 [01:50<01:53,  1.15it/s, ppl=9.32]
Evaluating c4:  49%|████▉     | 126/256 [01:50<01:52,  1.16it/s, ppl=9.32]
Evaluating c4:  49%|████▉     | 126/256 [01:50<01:52,  1.16it/s, ppl=9.39]
Evaluating c4:  50%|████▉     | 127/256 [01:50<01:51,  1.16it/s, ppl=9.39]
Evaluating c4:  50%|████▉     | 127/256 [01:51<01:51,  1.16it/s, ppl=9.37]
Evaluating c4:  50%|█████     | 128/256 [01:51<01:52,  1.14it/s, ppl=9.37]
Evaluating c4:  50%|█████     | 128/256 [01:52<01:52,  1.14it/s, ppl=9.38]
Evaluating c4:  50%|█████     | 129/256 [01:52<01:51,  1.14it/s, ppl=9.38]
Evaluating c4:  50%|█████     | 129/256 [01:53<01:51,  1.14it/s, ppl=9.50]
Evaluating c4:  51%|█████     | 130/256 [01:53<01:50,  1.14it/s, ppl=9.50]
Evaluating c4:  51%|█████     | 130/256 [01:54<01:50,  1.14it/s, ppl=9.49]
Evaluating c4:  51%|█████     | 131/256 [01:54<01:49,  1.14it/s, ppl=9.49]
Evaluating c4:  51%|█████     | 131/256 [01:55<01:49,  1.14it/s, ppl=9.49]
Evaluating c4:  52%|█████▏    | 132/256 [01:55<01:48,  1.15it/s, ppl=9.49]
Evaluating c4:  52%|█████▏    | 132/256 [01:56<01:48,  1.15it/s, ppl=9.51]
Evaluating c4:  52%|█████▏    | 133/256 [01:56<01:47,  1.15it/s, ppl=9.51]
Evaluating c4:  52%|█████▏    | 133/256 [01:56<01:47,  1.15it/s, ppl=9.54]
Evaluating c4:  52%|█████▏    | 134/256 [01:56<01:45,  1.15it/s, ppl=9.54]
Evaluating c4:  52%|█████▏    | 134/256 [01:57<01:45,  1.15it/s, ppl=9.53]
Evaluating c4:  53%|█████▎    | 135/256 [01:57<01:45,  1.15it/s, ppl=9.53]
Evaluating c4:  53%|█████▎    | 135/256 [01:58<01:45,  1.15it/s, ppl=9.54]
Evaluating c4:  53%|█████▎    | 136/256 [01:58<01:45,  1.14it/s, ppl=9.54]
Evaluating c4:  53%|█████▎    | 136/256 [01:59<01:45,  1.14it/s, ppl=9.56]
Evaluating c4:  54%|█████▎    | 137/256 [01:59<01:44,  1.14it/s, ppl=9.56]
Evaluating c4:  54%|█████▎    | 137/256 [02:00<01:44,  1.14it/s, ppl=9.55]
Evaluating c4:  54%|█████▍    | 138/256 [02:00<01:42,  1.15it/s, ppl=9.55]
Evaluating c4:  54%|█████▍    | 138/256 [02:01<01:42,  1.15it/s, ppl=9.60]
Evaluating c4:  54%|█████▍    | 139/256 [02:01<01:41,  1.15it/s, ppl=9.60]
Evaluating c4:  54%|█████▍    | 139/256 [02:02<01:41,  1.15it/s, ppl=9.65]
Evaluating c4:  55%|█████▍    | 140/256 [02:02<01:40,  1.15it/s, ppl=9.65]
Evaluating c4:  55%|█████▍    | 140/256 [02:03<01:40,  1.15it/s, ppl=9.64]
Evaluating c4:  55%|█████▌    | 141/256 [02:03<01:39,  1.15it/s, ppl=9.64]
Evaluating c4:  55%|█████▌    | 141/256 [02:03<01:39,  1.15it/s, ppl=9.65]
Evaluating c4:  55%|█████▌    | 142/256 [02:03<01:38,  1.15it/s, ppl=9.65]
Evaluating c4:  55%|█████▌    | 142/256 [02:04<01:38,  1.15it/s, ppl=9.69]
Evaluating c4:  56%|█████▌    | 143/256 [02:04<01:37,  1.15it/s, ppl=9.69]
Evaluating c4:  56%|█████▌    | 143/256 [02:05<01:37,  1.15it/s, ppl=9.65]
Evaluating c4:  56%|█████▋    | 144/256 [02:05<01:37,  1.15it/s, ppl=9.65]
Evaluating c4:  56%|█████▋    | 144/256 [02:06<01:37,  1.15it/s, ppl=9.66]
Evaluating c4:  57%|█████▋    | 145/256 [02:06<01:36,  1.15it/s, ppl=9.66]
Evaluating c4:  57%|█████▋    | 145/256 [02:07<01:36,  1.15it/s, ppl=9.68]
Evaluating c4:  57%|█████▋    | 146/256 [02:07<01:35,  1.15it/s, ppl=9.68]
Evaluating c4:  57%|█████▋    | 146/256 [02:08<01:35,  1.15it/s, ppl=9.67]
Evaluating c4:  57%|█████▋    | 147/256 [02:08<01:34,  1.15it/s, ppl=9.67]
Evaluating c4:  57%|█████▋    | 147/256 [02:09<01:34,  1.15it/s, ppl=9.65]
Evaluating c4:  58%|█████▊    | 148/256 [02:09<01:33,  1.15it/s, ppl=9.65]
Evaluating c4:  58%|█████▊    | 148/256 [02:10<01:33,  1.15it/s, ppl=9.67]
Evaluating c4:  58%|█████▊    | 149/256 [02:10<01:32,  1.15it/s, ppl=9.67]
Evaluating c4:  58%|█████▊    | 149/256 [02:10<01:32,  1.15it/s, ppl=9.66]
Evaluating c4:  59%|█████▊    | 150/256 [02:10<01:31,  1.16it/s, ppl=9.66]
Evaluating c4:  59%|█████▊    | 150/256 [02:11<01:31,  1.16it/s, ppl=9.67]
Evaluating c4:  59%|█████▉    | 151/256 [02:11<01:30,  1.16it/s, ppl=9.67]
Evaluating c4:  59%|█████▉    | 151/256 [02:12<01:30,  1.16it/s, ppl=9.65]
Evaluating c4:  59%|█████▉    | 152/256 [02:12<01:30,  1.14it/s, ppl=9.65]
Evaluating c4:  59%|█████▉    | 152/256 [02:13<01:30,  1.14it/s, ppl=9.65]
Evaluating c4:  60%|█████▉    | 153/256 [02:13<01:29,  1.15it/s, ppl=9.65]
Evaluating c4:  60%|█████▉    | 153/256 [02:14<01:29,  1.15it/s, ppl=9.60]
Evaluating c4:  60%|██████    | 154/256 [02:14<01:28,  1.15it/s, ppl=9.60]
Evaluating c4:  60%|██████    | 154/256 [02:15<01:28,  1.15it/s, ppl=9.57]
Evaluating c4:  61%|██████    | 155/256 [02:15<01:27,  1.15it/s, ppl=9.57]
Evaluating c4:  61%|██████    | 155/256 [02:16<01:27,  1.15it/s, ppl=9.58]
Evaluating c4:  61%|██████    | 156/256 [02:16<01:26,  1.15it/s, ppl=9.58]
Evaluating c4:  61%|██████    | 156/256 [02:16<01:26,  1.15it/s, ppl=9.62]
Evaluating c4:  61%|██████▏   | 157/256 [02:16<01:25,  1.15it/s, ppl=9.62]
Evaluating c4:  61%|██████▏   | 157/256 [02:17<01:25,  1.15it/s, ppl=9.64]
Evaluating c4:  62%|██████▏   | 158/256 [02:17<01:24,  1.16it/s, ppl=9.64]
Evaluating c4:  62%|██████▏   | 158/256 [02:18<01:24,  1.16it/s, ppl=9.56]
Evaluating c4:  62%|██████▏   | 159/256 [02:18<01:23,  1.16it/s, ppl=9.56]
Evaluating c4:  62%|██████▏   | 159/256 [02:19<01:23,  1.16it/s, ppl=9.52]
Evaluating c4:  62%|██████▎   | 160/256 [02:19<01:23,  1.16it/s, ppl=9.52]
Evaluating c4:  62%|██████▎   | 160/256 [02:20<01:23,  1.16it/s, ppl=9.50]
Evaluating c4:  63%|██████▎   | 161/256 [02:20<01:23,  1.14it/s, ppl=9.50]
Evaluating c4:  63%|██████▎   | 161/256 [02:21<01:23,  1.14it/s, ppl=9.49]
Evaluating c4:  63%|██████▎   | 162/256 [02:21<01:21,  1.15it/s, ppl=9.49]
Evaluating c4:  63%|██████▎   | 162/256 [02:22<01:21,  1.15it/s, ppl=9.51]
Evaluating c4:  64%|██████▎   | 163/256 [02:22<01:20,  1.15it/s, ppl=9.51]
Evaluating c4:  64%|██████▎   | 163/256 [02:23<01:20,  1.15it/s, ppl=9.54]
Evaluating c4:  64%|██████▍   | 164/256 [02:23<01:19,  1.15it/s, ppl=9.54]
Evaluating c4:  64%|██████▍   | 164/256 [02:23<01:19,  1.15it/s, ppl=9.49]
Evaluating c4:  64%|██████▍   | 165/256 [02:23<01:18,  1.15it/s, ppl=9.49]
Evaluating c4:  64%|██████▍   | 165/256 [02:24<01:18,  1.15it/s, ppl=9.40]
Evaluating c4:  65%|██████▍   | 166/256 [02:24<01:17,  1.15it/s, ppl=9.40]
Evaluating c4:  65%|██████▍   | 166/256 [02:25<01:17,  1.15it/s, ppl=9.40]
Evaluating c4:  65%|██████▌   | 167/256 [02:25<01:17,  1.15it/s, ppl=9.40]
Evaluating c4:  65%|██████▌   | 167/256 [02:26<01:17,  1.15it/s, ppl=9.42]
Evaluating c4:  66%|██████▌   | 168/256 [02:26<01:16,  1.16it/s, ppl=9.42]
Evaluating c4:  66%|██████▌   | 168/256 [02:27<01:16,  1.16it/s, ppl=9.42]
Evaluating c4:  66%|██████▌   | 169/256 [02:27<01:16,  1.14it/s, ppl=9.42]
Evaluating c4:  66%|██████▌   | 169/256 [02:28<01:16,  1.14it/s, ppl=9.36]
Evaluating c4:  66%|██████▋   | 170/256 [02:28<01:15,  1.15it/s, ppl=9.36]
Evaluating c4:  66%|██████▋   | 170/256 [02:29<01:15,  1.15it/s, ppl=9.39]
Evaluating c4:  67%|██████▋   | 171/256 [02:29<01:13,  1.15it/s, ppl=9.39]
Evaluating c4:  67%|██████▋   | 171/256 [02:30<01:13,  1.15it/s, ppl=9.35]
Evaluating c4:  67%|██████▋   | 172/256 [02:30<01:12,  1.15it/s, ppl=9.35]
Evaluating c4:  67%|██████▋   | 172/256 [02:30<01:12,  1.15it/s, ppl=9.37]
Evaluating c4:  68%|██████▊   | 173/256 [02:30<01:11,  1.15it/s, ppl=9.37]
Evaluating c4:  68%|██████▊   | 173/256 [02:31<01:11,  1.15it/s, ppl=9.36]
Evaluating c4:  68%|██████▊   | 174/256 [02:31<01:11,  1.15it/s, ppl=9.36]
Evaluating c4:  68%|██████▊   | 174/256 [02:32<01:11,  1.15it/s, ppl=9.35]
Evaluating c4:  68%|██████▊   | 175/256 [02:32<01:10,  1.15it/s, ppl=9.35]
Evaluating c4:  68%|██████▊   | 175/256 [02:33<01:10,  1.15it/s, ppl=9.39]
Evaluating c4:  69%|██████▉   | 176/256 [02:33<01:09,  1.16it/s, ppl=9.39]
Evaluating c4:  69%|██████▉   | 176/256 [02:34<01:09,  1.16it/s, ppl=9.39]
Evaluating c4:  69%|██████▉   | 177/256 [02:34<01:08,  1.15it/s, ppl=9.39]
Evaluating c4:  69%|██████▉   | 177/256 [02:35<01:08,  1.15it/s, ppl=9.41]
Evaluating c4:  70%|██████▉   | 178/256 [02:35<01:07,  1.15it/s, ppl=9.41]
Evaluating c4:  70%|██████▉   | 178/256 [02:36<01:07,  1.15it/s, ppl=9.41]
Evaluating c4:  70%|██████▉   | 179/256 [02:36<01:06,  1.15it/s, ppl=9.41]
Evaluating c4:  70%|██████▉   | 179/256 [02:36<01:06,  1.15it/s, ppl=9.43]
Evaluating c4:  70%|███████   | 180/256 [02:36<01:05,  1.15it/s, ppl=9.43]
Evaluating c4:  70%|███████   | 180/256 [02:37<01:05,  1.15it/s, ppl=9.42]
Evaluating c4:  71%|███████   | 181/256 [02:37<01:05,  1.15it/s, ppl=9.42]
Evaluating c4:  71%|███████   | 181/256 [02:38<01:05,  1.15it/s, ppl=9.43]
Evaluating c4:  71%|███████   | 182/256 [02:38<01:04,  1.15it/s, ppl=9.43]
Evaluating c4:  71%|███████   | 182/256 [02:39<01:04,  1.15it/s, ppl=9.42]
Evaluating c4:  71%|███████▏  | 183/256 [02:39<01:03,  1.15it/s, ppl=9.42]
Evaluating c4:  71%|███████▏  | 183/256 [02:40<01:03,  1.15it/s, ppl=9.42]
Evaluating c4:  72%|███████▏  | 184/256 [02:40<01:02,  1.15it/s, ppl=9.42]
Evaluating c4:  72%|███████▏  | 184/256 [02:41<01:02,  1.15it/s, ppl=9.40]
Evaluating c4:  72%|███████▏  | 185/256 [02:41<01:02,  1.14it/s, ppl=9.40]
Evaluating c4:  72%|███████▏  | 185/256 [02:42<01:02,  1.14it/s, ppl=9.41]
Evaluating c4:  73%|███████▎  | 186/256 [02:42<01:01,  1.15it/s, ppl=9.41]
Evaluating c4:  73%|███████▎  | 186/256 [02:43<01:01,  1.15it/s, ppl=9.42]
Evaluating c4:  73%|███████▎  | 187/256 [02:43<01:00,  1.15it/s, ppl=9.42]
Evaluating c4:  73%|███████▎  | 187/256 [02:43<01:00,  1.15it/s, ppl=9.44]
Evaluating c4:  73%|███████▎  | 188/256 [02:43<00:59,  1.15it/s, ppl=9.44]
Evaluating c4:  73%|███████▎  | 188/256 [02:44<00:59,  1.15it/s, ppl=9.44]
Evaluating c4:  74%|███████▍  | 189/256 [02:44<00:58,  1.15it/s, ppl=9.44]
Evaluating c4:  74%|███████▍  | 189/256 [02:45<00:58,  1.15it/s, ppl=9.44]
Evaluating c4:  74%|███████▍  | 190/256 [02:45<00:57,  1.15it/s, ppl=9.44]
Evaluating c4:  74%|███████▍  | 190/256 [02:46<00:57,  1.15it/s, ppl=9.44]
Evaluating c4:  75%|███████▍  | 191/256 [02:46<00:56,  1.15it/s, ppl=9.44]
Evaluating c4:  75%|███████▍  | 191/256 [02:47<00:56,  1.15it/s, ppl=9.46]
Evaluating c4:  75%|███████▌  | 192/256 [02:47<00:55,  1.15it/s, ppl=9.46]
Evaluating c4:  75%|███████▌  | 192/256 [02:48<00:55,  1.15it/s, ppl=9.42]
Evaluating c4:  75%|███████▌  | 193/256 [02:48<00:55,  1.14it/s, ppl=9.42]
Evaluating c4:  75%|███████▌  | 193/256 [02:49<00:55,  1.14it/s, ppl=9.43]
Evaluating c4:  76%|███████▌  | 194/256 [02:49<00:54,  1.14it/s, ppl=9.43]
Evaluating c4:  76%|███████▌  | 194/256 [02:50<00:54,  1.14it/s, ppl=9.42]
Evaluating c4:  76%|███████▌  | 195/256 [02:50<00:53,  1.15it/s, ppl=9.42]
Evaluating c4:  76%|███████▌  | 195/256 [02:50<00:53,  1.15it/s, ppl=9.41]
Evaluating c4:  77%|███████▋  | 196/256 [02:50<00:52,  1.15it/s, ppl=9.41]
Evaluating c4:  77%|███████▋  | 196/256 [02:51<00:52,  1.15it/s, ppl=9.42]
Evaluating c4:  77%|███████▋  | 197/256 [02:51<00:51,  1.15it/s, ppl=9.42]
Evaluating c4:  77%|███████▋  | 197/256 [02:52<00:51,  1.15it/s, ppl=9.45]
Evaluating c4:  77%|███████▋  | 198/256 [02:52<00:50,  1.15it/s, ppl=9.45]
Evaluating c4:  77%|███████▋  | 198/256 [02:53<00:50,  1.15it/s, ppl=9.44]
Evaluating c4:  78%|███████▊  | 199/256 [02:53<00:49,  1.15it/s, ppl=9.44]
Evaluating c4:  78%|███████▊  | 199/256 [02:54<00:49,  1.15it/s, ppl=9.46]
Evaluating c4:  78%|███████▊  | 200/256 [02:54<00:48,  1.15it/s, ppl=9.46]
Evaluating c4:  78%|███████▊  | 200/256 [02:55<00:48,  1.15it/s, ppl=9.48]
Evaluating c4:  79%|███████▊  | 201/256 [02:55<00:48,  1.14it/s, ppl=9.48]
Evaluating c4:  79%|███████▊  | 201/256 [02:56<00:48,  1.14it/s, ppl=9.47]
Evaluating c4:  79%|███████▉  | 202/256 [02:56<00:47,  1.15it/s, ppl=9.47]
Evaluating c4:  79%|███████▉  | 202/256 [02:57<00:47,  1.15it/s, ppl=9.46]
Evaluating c4:  79%|███████▉  | 203/256 [02:57<00:46,  1.15it/s, ppl=9.46]
Evaluating c4:  79%|███████▉  | 203/256 [02:57<00:46,  1.15it/s, ppl=9.43]
Evaluating c4:  80%|███████▉  | 204/256 [02:57<00:45,  1.15it/s, ppl=9.43]
Evaluating c4:  80%|███████▉  | 204/256 [02:58<00:45,  1.15it/s, ppl=9.46]
Evaluating c4:  80%|████████  | 205/256 [02:58<00:44,  1.15it/s, ppl=9.46]
Evaluating c4:  80%|████████  | 205/256 [02:59<00:44,  1.15it/s, ppl=9.50]
Evaluating c4:  80%|████████  | 206/256 [02:59<00:43,  1.15it/s, ppl=9.50]
Evaluating c4:  80%|████████  | 206/256 [03:00<00:43,  1.15it/s, ppl=9.51]
Evaluating c4:  81%|████████  | 207/256 [03:00<00:42,  1.15it/s, ppl=9.51]
Evaluating c4:  81%|████████  | 207/256 [03:01<00:42,  1.15it/s, ppl=9.54]
Evaluating c4:  81%|████████▏ | 208/256 [03:01<00:41,  1.16it/s, ppl=9.54]
Evaluating c4:  81%|████████▏ | 208/256 [03:02<00:41,  1.16it/s, ppl=9.56]
Evaluating c4:  82%|████████▏ | 209/256 [03:02<00:41,  1.14it/s, ppl=9.56]
Evaluating c4:  82%|████████▏ | 209/256 [03:03<00:41,  1.14it/s, ppl=9.56]
Evaluating c4:  82%|████████▏ | 210/256 [03:03<00:40,  1.15it/s, ppl=9.56]
Evaluating c4:  82%|████████▏ | 210/256 [03:03<00:40,  1.15it/s, ppl=9.57]
Evaluating c4:  82%|████████▏ | 211/256 [03:03<00:39,  1.15it/s, ppl=9.57]
Evaluating c4:  82%|████████▏ | 211/256 [03:04<00:39,  1.15it/s, ppl=9.57]
Evaluating c4:  83%|████████▎ | 212/256 [03:04<00:38,  1.15it/s, ppl=9.57]
Evaluating c4:  83%|████████▎ | 212/256 [03:05<00:38,  1.15it/s, ppl=9.56]
Evaluating c4:  83%|████████▎ | 213/256 [03:05<00:37,  1.15it/s, ppl=9.56]
Evaluating c4:  83%|████████▎ | 213/256 [03:06<00:37,  1.15it/s, ppl=9.58]
Evaluating c4:  84%|████████▎ | 214/256 [03:06<00:36,  1.15it/s, ppl=9.58]
Evaluating c4:  84%|████████▎ | 214/256 [03:07<00:36,  1.15it/s, ppl=9.58]
Evaluating c4:  84%|████████▍ | 215/256 [03:07<00:35,  1.15it/s, ppl=9.58]
Evaluating c4:  84%|████████▍ | 215/256 [03:08<00:35,  1.15it/s, ppl=9.59]
Evaluating c4:  84%|████████▍ | 216/256 [03:08<00:34,  1.15it/s, ppl=9.59]
Evaluating c4:  84%|████████▍ | 216/256 [03:09<00:34,  1.15it/s, ppl=9.58]
Evaluating c4:  85%|████████▍ | 217/256 [03:09<00:34,  1.14it/s, ppl=9.58]
Evaluating c4:  85%|████████▍ | 217/256 [03:10<00:34,  1.14it/s, ppl=9.60]
Evaluating c4:  85%|████████▌ | 218/256 [03:10<00:33,  1.15it/s, ppl=9.60]
Evaluating c4:  85%|████████▌ | 218/256 [03:10<00:33,  1.15it/s, ppl=9.57]
Evaluating c4:  86%|████████▌ | 219/256 [03:10<00:32,  1.15it/s, ppl=9.57]
Evaluating c4:  86%|████████▌ | 219/256 [03:11<00:32,  1.15it/s, ppl=9.58]
Evaluating c4:  86%|████████▌ | 220/256 [03:11<00:31,  1.15it/s, ppl=9.58]
Evaluating c4:  86%|████████▌ | 220/256 [03:12<00:31,  1.15it/s, ppl=9.57]
Evaluating c4:  86%|████████▋ | 221/256 [03:12<00:30,  1.15it/s, ppl=9.57]
Evaluating c4:  86%|████████▋ | 221/256 [03:13<00:30,  1.15it/s, ppl=9.56]
Evaluating c4:  87%|████████▋ | 222/256 [03:13<00:29,  1.15it/s, ppl=9.56]
Evaluating c4:  87%|████████▋ | 222/256 [03:14<00:29,  1.15it/s, ppl=9.52]
Evaluating c4:  87%|████████▋ | 223/256 [03:14<00:28,  1.15it/s, ppl=9.52]
Evaluating c4:  87%|████████▋ | 223/256 [03:15<00:28,  1.15it/s, ppl=9.53]
Evaluating c4:  88%|████████▊ | 224/256 [03:15<00:27,  1.16it/s, ppl=9.53]
Evaluating c4:  88%|████████▊ | 224/256 [03:16<00:27,  1.16it/s, ppl=9.51]
Evaluating c4:  88%|████████▊ | 225/256 [03:16<00:27,  1.15it/s, ppl=9.51]
Evaluating c4:  88%|████████▊ | 225/256 [03:16<00:27,  1.15it/s, ppl=9.52]
Evaluating c4:  88%|████████▊ | 226/256 [03:16<00:26,  1.15it/s, ppl=9.52]
Evaluating c4:  88%|████████▊ | 226/256 [03:17<00:26,  1.15it/s, ppl=9.52]
Evaluating c4:  89%|████████▊ | 227/256 [03:17<00:25,  1.15it/s, ppl=9.52]
Evaluating c4:  89%|████████▊ | 227/256 [03:18<00:25,  1.15it/s, ppl=9.52]
Evaluating c4:  89%|████████▉ | 228/256 [03:18<00:24,  1.15it/s, ppl=9.52]
Evaluating c4:  89%|████████▉ | 228/256 [03:19<00:24,  1.15it/s, ppl=9.52]
Evaluating c4:  89%|████████▉ | 229/256 [03:19<00:23,  1.15it/s, ppl=9.52]
Evaluating c4:  89%|████████▉ | 229/256 [03:20<00:23,  1.15it/s, ppl=9.54]
Evaluating c4:  90%|████████▉ | 230/256 [03:20<00:22,  1.15it/s, ppl=9.54]
Evaluating c4:  90%|████████▉ | 230/256 [03:21<00:22,  1.15it/s, ppl=9.54]
Evaluating c4:  90%|█████████ | 231/256 [03:21<00:21,  1.16it/s, ppl=9.54]
Evaluating c4:  90%|█████████ | 231/256 [03:22<00:21,  1.16it/s, ppl=9.54]
Evaluating c4:  91%|█████████ | 232/256 [03:22<00:20,  1.16it/s, ppl=9.54]
Evaluating c4:  91%|█████████ | 232/256 [03:23<00:20,  1.16it/s, ppl=9.58]
Evaluating c4:  91%|█████████ | 233/256 [03:23<00:20,  1.14it/s, ppl=9.58]
Evaluating c4:  91%|█████████ | 233/256 [03:23<00:20,  1.14it/s, ppl=9.58]
Evaluating c4:  91%|█████████▏| 234/256 [03:23<00:19,  1.15it/s, ppl=9.58]
Evaluating c4:  91%|█████████▏| 234/256 [03:24<00:19,  1.15it/s, ppl=9.58]
Evaluating c4:  92%|█████████▏| 235/256 [03:24<00:18,  1.15it/s, ppl=9.58]
Evaluating c4:  92%|█████████▏| 235/256 [03:25<00:18,  1.15it/s, ppl=9.63]
Evaluating c4:  92%|█████████▏| 236/256 [03:25<00:17,  1.15it/s, ppl=9.63]
Evaluating c4:  92%|█████████▏| 236/256 [03:26<00:17,  1.15it/s, ppl=9.64]
Evaluating c4:  93%|█████████▎| 237/256 [03:26<00:16,  1.14it/s, ppl=9.64]
Evaluating c4:  93%|█████████▎| 237/256 [03:27<00:16,  1.14it/s, ppl=9.62]
Evaluating c4:  93%|█████████▎| 238/256 [03:27<00:15,  1.14it/s, ppl=9.62]
Evaluating c4:  93%|█████████▎| 238/256 [03:28<00:15,  1.14it/s, ppl=9.60]
Evaluating c4:  93%|█████████▎| 239/256 [03:28<00:14,  1.14it/s, ppl=9.60]
Evaluating c4:  93%|█████████▎| 239/256 [03:29<00:14,  1.14it/s, ppl=9.60]
Evaluating c4:  94%|█████████▍| 240/256 [03:29<00:13,  1.15it/s, ppl=9.60]
Evaluating c4:  94%|█████████▍| 240/256 [03:30<00:13,  1.15it/s, ppl=9.62]
Evaluating c4:  94%|█████████▍| 241/256 [03:30<00:13,  1.14it/s, ppl=9.62]
Evaluating c4:  94%|█████████▍| 241/256 [03:30<00:13,  1.14it/s, ppl=9.62]
Evaluating c4:  95%|█████████▍| 242/256 [03:30<00:12,  1.14it/s, ppl=9.62]
Evaluating c4:  95%|█████████▍| 242/256 [03:31<00:12,  1.14it/s, ppl=9.63]
Evaluating c4:  95%|█████████▍| 243/256 [03:31<00:11,  1.15it/s, ppl=9.63]
Evaluating c4:  95%|█████████▍| 243/256 [03:32<00:11,  1.15it/s, ppl=9.60]
Evaluating c4:  95%|█████████▌| 244/256 [03:32<00:10,  1.15it/s, ppl=9.60]
Evaluating c4:  95%|█████████▌| 244/256 [03:33<00:10,  1.15it/s, ppl=9.61]
Evaluating c4:  96%|█████████▌| 245/256 [03:33<00:09,  1.15it/s, ppl=9.61]
Evaluating c4:  96%|█████████▌| 245/256 [03:34<00:09,  1.15it/s, ppl=9.61]
Evaluating c4:  96%|█████████▌| 246/256 [03:34<00:08,  1.15it/s, ppl=9.61]
Evaluating c4:  96%|█████████▌| 246/256 [03:35<00:08,  1.15it/s, ppl=9.62]
Evaluating c4:  96%|█████████▋| 247/256 [03:35<00:07,  1.15it/s, ppl=9.62]
Evaluating c4:  96%|█████████▋| 247/256 [03:36<00:07,  1.15it/s, ppl=9.61]
Evaluating c4:  97%|█████████▋| 248/256 [03:36<00:06,  1.15it/s, ppl=9.61]
Evaluating c4:  97%|█████████▋| 248/256 [03:37<00:06,  1.15it/s, ppl=9.60]
Evaluating c4:  97%|█████████▋| 249/256 [03:37<00:06,  1.14it/s, ppl=9.60]
Evaluating c4:  97%|█████████▋| 249/256 [03:37<00:06,  1.14it/s, ppl=9.72]
Evaluating c4:  98%|█████████▊| 250/256 [03:37<00:05,  1.15it/s, ppl=9.72]
Evaluating c4:  98%|█████████▊| 250/256 [03:38<00:05,  1.15it/s, ppl=9.71]
Evaluating c4:  98%|█████████▊| 251/256 [03:38<00:04,  1.15it/s, ppl=9.71]
Evaluating c4:  98%|█████████▊| 251/256 [03:39<00:04,  1.15it/s, ppl=9.73]
Evaluating c4:  98%|█████████▊| 252/256 [03:39<00:03,  1.15it/s, ppl=9.73]
Evaluating c4:  98%|█████████▊| 252/256 [03:40<00:03,  1.15it/s, ppl=9.75]
Evaluating c4:  99%|█████████▉| 253/256 [03:40<00:02,  1.15it/s, ppl=9.75]
Evaluating c4:  99%|█████████▉| 253/256 [03:41<00:02,  1.15it/s, ppl=9.71]
Evaluating c4:  99%|█████████▉| 254/256 [03:41<00:01,  1.15it/s, ppl=9.71]
Evaluating c4:  99%|█████████▉| 254/256 [03:42<00:01,  1.15it/s, ppl=9.78]
Evaluating c4: 100%|█████████▉| 255/256 [03:42<00:00,  1.16it/s, ppl=9.78]
Evaluating c4: 100%|█████████▉| 255/256 [03:43<00:00,  1.16it/s, ppl=9.79]
Evaluating c4: 100%|██████████| 256/256 [03:43<00:00,  1.14it/s, ppl=9.79]
Evaluating c4: 100%|██████████| 256/256 [03:43<00:00,  1.15it/s, ppl=9.79]
+c4 perplexity: 9.7912
+
Evaluating wikitext2:   0%|          | 0/141 [00:00<?, ?it/s]
Evaluating wikitext2:   0%|          | 0/141 [00:00<?, ?it/s, ppl=5.53]
Evaluating wikitext2:   1%|          | 1/141 [00:00<02:01,  1.15it/s, ppl=5.53]
Evaluating wikitext2:   1%|          | 1/141 [00:01<02:01,  1.15it/s, ppl=7.26]
Evaluating wikitext2:   1%|▏         | 2/141 [00:01<02:00,  1.16it/s, ppl=7.26]
Evaluating wikitext2:   1%|▏         | 2/141 [00:02<02:00,  1.16it/s, ppl=8.17]
Evaluating wikitext2:   2%|▏         | 3/141 [00:02<01:59,  1.16it/s, ppl=8.17]
Evaluating wikitext2:   2%|▏         | 3/141 [00:03<01:59,  1.16it/s, ppl=7.77]
Evaluating wikitext2:   3%|▎         | 4/141 [00:03<01:58,  1.16it/s, ppl=7.77]
Evaluating wikitext2:   3%|▎         | 4/141 [00:04<01:58,  1.16it/s, ppl=7.35]
Evaluating wikitext2:   4%|▎         | 5/141 [00:04<01:57,  1.16it/s, ppl=7.35]
Evaluating wikitext2:   4%|▎         | 5/141 [00:05<01:57,  1.16it/s, ppl=6.66]
Evaluating wikitext2:   4%|▍         | 6/141 [00:05<01:56,  1.16it/s, ppl=6.66]
Evaluating wikitext2:   4%|▍         | 6/141 [00:06<01:56,  1.16it/s, ppl=6.27]
Evaluating wikitext2:   5%|▍         | 7/141 [00:06<01:57,  1.14it/s, ppl=6.27]
Evaluating wikitext2:   5%|▍         | 7/141 [00:06<01:57,  1.14it/s, ppl=6.24]
Evaluating wikitext2:   6%|▌         | 8/141 [00:06<01:55,  1.15it/s, ppl=6.24]
Evaluating wikitext2:   6%|▌         | 8/141 [00:07<01:55,  1.15it/s, ppl=6.46]
Evaluating wikitext2:   6%|▋         | 9/141 [00:07<01:54,  1.15it/s, ppl=6.46]
Evaluating wikitext2:   6%|▋         | 9/141 [00:08<01:54,  1.15it/s, ppl=6.47]
Evaluating wikitext2:   7%|▋         | 10/141 [00:08<01:53,  1.15it/s, ppl=6.47]
Evaluating wikitext2:   7%|▋         | 10/141 [00:09<01:53,  1.15it/s, ppl=6.43]
Evaluating wikitext2:   8%|▊         | 11/141 [00:09<01:52,  1.15it/s, ppl=6.43]
Evaluating wikitext2:   8%|▊         | 11/141 [00:10<01:52,  1.15it/s, ppl=6.57]
Evaluating wikitext2:   9%|▊         | 12/141 [00:10<01:51,  1.15it/s, ppl=6.57]
Evaluating wikitext2:   9%|▊         | 12/141 [00:11<01:51,  1.15it/s, ppl=6.65]
Evaluating wikitext2:   9%|▉         | 13/141 [00:11<01:50,  1.16it/s, ppl=6.65]
Evaluating wikitext2:   9%|▉         | 13/141 [00:12<01:50,  1.16it/s, ppl=6.76]
Evaluating wikitext2:  10%|▉         | 14/141 [00:12<01:51,  1.14it/s, ppl=6.76]
Evaluating wikitext2:  10%|▉         | 14/141 [00:13<01:51,  1.14it/s, ppl=6.84]
Evaluating wikitext2:  11%|█         | 15/141 [00:13<01:49,  1.15it/s, ppl=6.84]
Evaluating wikitext2:  11%|█         | 15/141 [00:13<01:49,  1.15it/s, ppl=7.00]
Evaluating wikitext2:  11%|█▏        | 16/141 [00:13<01:48,  1.15it/s, ppl=7.00]
Evaluating wikitext2:  11%|█▏        | 16/141 [00:14<01:48,  1.15it/s, ppl=7.02]
Evaluating wikitext2:  12%|█▏        | 17/141 [00:14<01:47,  1.15it/s, ppl=7.02]
Evaluating wikitext2:  12%|█▏        | 17/141 [00:15<01:47,  1.15it/s, ppl=7.05]
Evaluating wikitext2:  13%|█▎        | 18/141 [00:15<01:46,  1.15it/s, ppl=7.05]
Evaluating wikitext2:  13%|█▎        | 18/141 [00:16<01:46,  1.15it/s, ppl=6.97]
Evaluating wikitext2:  13%|█▎        | 19/141 [00:16<01:45,  1.15it/s, ppl=6.97]
Evaluating wikitext2:  13%|█▎        | 19/141 [00:17<01:45,  1.15it/s, ppl=6.95]
Evaluating wikitext2:  14%|█▍        | 20/141 [00:17<01:44,  1.15it/s, ppl=6.95]
Evaluating wikitext2:  14%|█▍        | 20/141 [00:18<01:44,  1.15it/s, ppl=6.98]
Evaluating wikitext2:  15%|█▍        | 21/141 [00:18<01:43,  1.16it/s, ppl=6.98]
Evaluating wikitext2:  15%|█▍        | 21/141 [00:19<01:43,  1.16it/s, ppl=7.00]
Evaluating wikitext2:  16%|█▌        | 22/141 [00:19<01:42,  1.16it/s, ppl=7.00]
Evaluating wikitext2:  16%|█▌        | 22/141 [00:20<01:42,  1.16it/s, ppl=7.01]
Evaluating wikitext2:  16%|█▋        | 23/141 [00:20<01:43,  1.14it/s, ppl=7.01]
Evaluating wikitext2:  16%|█▋        | 23/141 [00:20<01:43,  1.14it/s, ppl=7.04]
Evaluating wikitext2:  17%|█▋        | 24/141 [00:20<01:42,  1.14it/s, ppl=7.04]
Evaluating wikitext2:  17%|█▋        | 24/141 [00:21<01:42,  1.14it/s, ppl=7.09]
Evaluating wikitext2:  18%|█▊        | 25/141 [00:21<01:42,  1.13it/s, ppl=7.09]
Evaluating wikitext2:  18%|█▊        | 25/141 [00:22<01:42,  1.13it/s, ppl=7.16]
Evaluating wikitext2:  18%|█▊        | 26/141 [00:22<01:41,  1.13it/s, ppl=7.16]
Evaluating wikitext2:  18%|█▊        | 26/141 [00:23<01:41,  1.13it/s, ppl=7.25]
Evaluating wikitext2:  19%|█▉        | 27/141 [00:23<01:40,  1.14it/s, ppl=7.25]
Evaluating wikitext2:  19%|█▉        | 27/141 [00:24<01:40,  1.14it/s, ppl=7.29]
Evaluating wikitext2:  20%|█▉        | 28/141 [00:24<01:38,  1.14it/s, ppl=7.29]
Evaluating wikitext2:  20%|█▉        | 28/141 [00:25<01:38,  1.14it/s, ppl=7.27]
Evaluating wikitext2:  21%|██        | 29/141 [00:25<01:37,  1.15it/s, ppl=7.27]
Evaluating wikitext2:  21%|██        | 29/141 [00:26<01:37,  1.15it/s, ppl=7.20]
Evaluating wikitext2:  21%|██▏       | 30/141 [00:26<01:36,  1.15it/s, ppl=7.20]
Evaluating wikitext2:  21%|██▏       | 30/141 [00:27<01:36,  1.15it/s, ppl=7.10]
Evaluating wikitext2:  22%|██▏       | 31/141 [00:27<01:36,  1.14it/s, ppl=7.10]
Evaluating wikitext2:  22%|██▏       | 31/141 [00:27<01:36,  1.14it/s, ppl=7.02]
Evaluating wikitext2:  23%|██▎       | 32/141 [00:27<01:35,  1.14it/s, ppl=7.02]
Evaluating wikitext2:  23%|██▎       | 32/141 [00:28<01:35,  1.14it/s, ppl=7.01]
Evaluating wikitext2:  23%|██▎       | 33/141 [00:28<01:34,  1.15it/s, ppl=7.01]
Evaluating wikitext2:  23%|██▎       | 33/141 [00:29<01:34,  1.15it/s, ppl=7.00]
Evaluating wikitext2:  24%|██▍       | 34/141 [00:29<01:33,  1.15it/s, ppl=7.00]
Evaluating wikitext2:  24%|██▍       | 34/141 [00:30<01:33,  1.15it/s, ppl=7.02]
Evaluating wikitext2:  25%|██▍       | 35/141 [00:30<01:32,  1.15it/s, ppl=7.02]
Evaluating wikitext2:  25%|██▍       | 35/141 [00:31<01:32,  1.15it/s, ppl=6.90]
Evaluating wikitext2:  26%|██▌       | 36/141 [00:31<01:31,  1.15it/s, ppl=6.90]
Evaluating wikitext2:  26%|██▌       | 36/141 [00:32<01:31,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  26%|██▌       | 37/141 [00:32<01:30,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  26%|██▌       | 37/141 [00:33<01:30,  1.15it/s, ppl=6.69]
Evaluating wikitext2:  27%|██▋       | 38/141 [00:33<01:29,  1.15it/s, ppl=6.69]
Evaluating wikitext2:  27%|██▋       | 38/141 [00:33<01:29,  1.15it/s, ppl=6.59]
Evaluating wikitext2:  28%|██▊       | 39/141 [00:33<01:29,  1.14it/s, ppl=6.59]
Evaluating wikitext2:  28%|██▊       | 39/141 [00:34<01:29,  1.14it/s, ppl=6.53]
Evaluating wikitext2:  28%|██▊       | 40/141 [00:34<01:28,  1.14it/s, ppl=6.53]
Evaluating wikitext2:  28%|██▊       | 40/141 [00:35<01:28,  1.14it/s, ppl=6.57]
Evaluating wikitext2:  29%|██▉       | 41/141 [00:35<01:27,  1.15it/s, ppl=6.57]
Evaluating wikitext2:  29%|██▉       | 41/141 [00:36<01:27,  1.15it/s, ppl=6.68]
Evaluating wikitext2:  30%|██▉       | 42/141 [00:36<01:26,  1.15it/s, ppl=6.68]
Evaluating wikitext2:  30%|██▉       | 42/141 [00:37<01:26,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  30%|███       | 43/141 [00:37<01:25,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  30%|███       | 43/141 [00:38<01:25,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  31%|███       | 44/141 [00:38<01:24,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  31%|███       | 44/141 [00:39<01:24,  1.15it/s, ppl=6.85]
Evaluating wikitext2:  32%|███▏      | 45/141 [00:39<01:23,  1.15it/s, ppl=6.85]
Evaluating wikitext2:  32%|███▏      | 45/141 [00:40<01:23,  1.15it/s, ppl=6.88]
Evaluating wikitext2:  33%|███▎      | 46/141 [00:40<01:22,  1.15it/s, ppl=6.88]
Evaluating wikitext2:  33%|███▎      | 46/141 [00:40<01:22,  1.15it/s, ppl=6.93]
Evaluating wikitext2:  33%|███▎      | 47/141 [00:40<01:22,  1.14it/s, ppl=6.93]
Evaluating wikitext2:  33%|███▎      | 47/141 [00:41<01:22,  1.14it/s, ppl=6.94]
Evaluating wikitext2:  34%|███▍      | 48/141 [00:41<01:21,  1.15it/s, ppl=6.94]
Evaluating wikitext2:  34%|███▍      | 48/141 [00:42<01:21,  1.15it/s, ppl=6.93]
Evaluating wikitext2:  35%|███▍      | 49/141 [00:42<01:19,  1.15it/s, ppl=6.93]
Evaluating wikitext2:  35%|███▍      | 49/141 [00:43<01:19,  1.15it/s, ppl=6.93]
Evaluating wikitext2:  35%|███▌      | 50/141 [00:43<01:18,  1.15it/s, ppl=6.93]
Evaluating wikitext2:  35%|███▌      | 50/141 [00:44<01:18,  1.15it/s, ppl=6.89]
Evaluating wikitext2:  36%|███▌      | 51/141 [00:44<01:18,  1.15it/s, ppl=6.89]
Evaluating wikitext2:  36%|███▌      | 51/141 [00:45<01:18,  1.15it/s, ppl=6.86]
Evaluating wikitext2:  37%|███▋      | 52/141 [00:45<01:17,  1.15it/s, ppl=6.86]
Evaluating wikitext2:  37%|███▋      | 52/141 [00:46<01:17,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  38%|███▊      | 53/141 [00:46<01:16,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  38%|███▊      | 53/141 [00:47<01:16,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  38%|███▊      | 54/141 [00:47<01:16,  1.14it/s, ppl=6.79]
Evaluating wikitext2:  38%|███▊      | 54/141 [00:47<01:16,  1.14it/s, ppl=6.74]
Evaluating wikitext2:  39%|███▉      | 55/141 [00:47<01:14,  1.15it/s, ppl=6.74]
Evaluating wikitext2:  39%|███▉      | 55/141 [00:48<01:14,  1.15it/s, ppl=6.69]
Evaluating wikitext2:  40%|███▉      | 56/141 [00:48<01:13,  1.15it/s, ppl=6.69]
Evaluating wikitext2:  40%|███▉      | 56/141 [00:49<01:13,  1.15it/s, ppl=6.66]
Evaluating wikitext2:  40%|████      | 57/141 [00:49<01:12,  1.15it/s, ppl=6.66]
Evaluating wikitext2:  40%|████      | 57/141 [00:50<01:12,  1.15it/s, ppl=6.64]
Evaluating wikitext2:  41%|████      | 58/141 [00:50<01:11,  1.15it/s, ppl=6.64]
Evaluating wikitext2:  41%|████      | 58/141 [00:51<01:11,  1.15it/s, ppl=6.65]
Evaluating wikitext2:  42%|████▏     | 59/141 [00:51<01:11,  1.15it/s, ppl=6.65]
Evaluating wikitext2:  42%|████▏     | 59/141 [00:52<01:11,  1.15it/s, ppl=6.69]
Evaluating wikitext2:  43%|████▎     | 60/141 [00:52<01:10,  1.15it/s, ppl=6.69]
Evaluating wikitext2:  43%|████▎     | 60/141 [00:53<01:10,  1.15it/s, ppl=6.73]
Evaluating wikitext2:  43%|████▎     | 61/141 [00:53<01:09,  1.16it/s, ppl=6.73]
Evaluating wikitext2:  43%|████▎     | 61/141 [00:53<01:09,  1.16it/s, ppl=6.78]
Evaluating wikitext2:  44%|████▍     | 62/141 [00:53<01:09,  1.14it/s, ppl=6.78]
Evaluating wikitext2:  44%|████▍     | 62/141 [00:54<01:09,  1.14it/s, ppl=6.72]
Evaluating wikitext2:  45%|████▍     | 63/141 [00:54<01:07,  1.15it/s, ppl=6.72]
Evaluating wikitext2:  45%|████▍     | 63/141 [00:55<01:07,  1.15it/s, ppl=6.70]
Evaluating wikitext2:  45%|████▌     | 64/141 [00:55<01:06,  1.15it/s, ppl=6.70]
Evaluating wikitext2:  45%|████▌     | 64/141 [00:56<01:06,  1.15it/s, ppl=6.68]
Evaluating wikitext2:  46%|████▌     | 65/141 [00:56<01:05,  1.15it/s, ppl=6.68]
Evaluating wikitext2:  46%|████▌     | 65/141 [00:57<01:05,  1.15it/s, ppl=6.64]
Evaluating wikitext2:  47%|████▋     | 66/141 [00:57<01:05,  1.15it/s, ppl=6.64]
Evaluating wikitext2:  47%|████▋     | 66/141 [00:58<01:05,  1.15it/s, ppl=6.62]
Evaluating wikitext2:  48%|████▊     | 67/141 [00:58<01:04,  1.15it/s, ppl=6.62]
Evaluating wikitext2:  48%|████▊     | 67/141 [00:59<01:04,  1.15it/s, ppl=6.60]
Evaluating wikitext2:  48%|████▊     | 68/141 [00:59<01:03,  1.15it/s, ppl=6.60]
Evaluating wikitext2:  48%|████▊     | 68/141 [01:00<01:03,  1.15it/s, ppl=6.58]
Evaluating wikitext2:  49%|████▉     | 69/141 [01:00<01:02,  1.16it/s, ppl=6.58]
Evaluating wikitext2:  49%|████▉     | 69/141 [01:00<01:02,  1.16it/s, ppl=6.59]
Evaluating wikitext2:  50%|████▉     | 70/141 [01:00<01:02,  1.14it/s, ppl=6.59]
Evaluating wikitext2:  50%|████▉     | 70/141 [01:01<01:02,  1.14it/s, ppl=6.61]
Evaluating wikitext2:  50%|█████     | 71/141 [01:01<01:01,  1.15it/s, ppl=6.61]
Evaluating wikitext2:  50%|█████     | 71/141 [01:02<01:01,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  51%|█████     | 72/141 [01:02<01:00,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  51%|█████     | 72/141 [01:03<01:00,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  52%|█████▏    | 73/141 [01:03<00:59,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  52%|█████▏    | 73/141 [01:04<00:59,  1.15it/s, ppl=6.64]
Evaluating wikitext2:  52%|█████▏    | 74/141 [01:04<00:58,  1.15it/s, ppl=6.64]
Evaluating wikitext2:  52%|█████▏    | 74/141 [01:05<00:58,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  53%|█████▎    | 75/141 [01:05<00:57,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  53%|█████▎    | 75/141 [01:06<00:57,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  54%|█████▍    | 76/141 [01:06<00:56,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  54%|█████▍    | 76/141 [01:06<00:56,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  55%|█████▍    | 77/141 [01:06<00:55,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  55%|█████▍    | 77/141 [01:07<00:55,  1.15it/s, ppl=6.65]
Evaluating wikitext2:  55%|█████▌    | 78/141 [01:07<00:55,  1.14it/s, ppl=6.65]
Evaluating wikitext2:  55%|█████▌    | 78/141 [01:08<00:55,  1.14it/s, ppl=6.68]
Evaluating wikitext2:  56%|█████▌    | 79/141 [01:08<00:54,  1.15it/s, ppl=6.68]
Evaluating wikitext2:  56%|█████▌    | 79/141 [01:09<00:54,  1.15it/s, ppl=6.65]
Evaluating wikitext2:  57%|█████▋    | 80/141 [01:09<00:53,  1.15it/s, ppl=6.65]
Evaluating wikitext2:  57%|█████▋    | 80/141 [01:10<00:53,  1.15it/s, ppl=6.65]
Evaluating wikitext2:  57%|█████▋    | 81/141 [01:10<00:52,  1.15it/s, ppl=6.65]
Evaluating wikitext2:  57%|█████▋    | 81/141 [01:11<00:52,  1.15it/s, ppl=6.64]
Evaluating wikitext2:  58%|█████▊    | 82/141 [01:11<00:51,  1.15it/s, ppl=6.64]
Evaluating wikitext2:  58%|█████▊    | 82/141 [01:12<00:51,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  59%|█████▉    | 83/141 [01:12<00:50,  1.15it/s, ppl=6.63]
Evaluating wikitext2:  59%|█████▉    | 83/141 [01:13<00:50,  1.15it/s, ppl=6.61]
Evaluating wikitext2:  60%|█████▉    | 84/141 [01:13<00:49,  1.15it/s, ppl=6.61]
Evaluating wikitext2:  60%|█████▉    | 84/141 [01:13<00:49,  1.15it/s, ppl=6.59]
Evaluating wikitext2:  60%|██████    | 85/141 [01:13<00:48,  1.14it/s, ppl=6.59]
Evaluating wikitext2:  60%|██████    | 85/141 [01:14<00:48,  1.14it/s, ppl=6.60]
Evaluating wikitext2:  61%|██████    | 86/141 [01:14<00:47,  1.15it/s, ppl=6.60]
Evaluating wikitext2:  61%|██████    | 86/141 [01:15<00:47,  1.15it/s, ppl=6.62]
Evaluating wikitext2:  62%|██████▏   | 87/141 [01:15<00:46,  1.15it/s, ppl=6.62]
Evaluating wikitext2:  62%|██████▏   | 87/141 [01:16<00:46,  1.15it/s, ppl=6.62]
Evaluating wikitext2:  62%|██████▏   | 88/141 [01:16<00:45,  1.15it/s, ppl=6.62]
Evaluating wikitext2:  62%|██████▏   | 88/141 [01:17<00:45,  1.15it/s, ppl=6.68]
Evaluating wikitext2:  63%|██████▎   | 89/141 [01:17<00:45,  1.15it/s, ppl=6.68]
Evaluating wikitext2:  63%|██████▎   | 89/141 [01:18<00:45,  1.15it/s, ppl=6.73]
Evaluating wikitext2:  64%|██████▍   | 90/141 [01:18<00:44,  1.15it/s, ppl=6.73]
Evaluating wikitext2:  64%|██████▍   | 90/141 [01:19<00:44,  1.15it/s, ppl=6.75]
Evaluating wikitext2:  65%|██████▍   | 91/141 [01:19<00:43,  1.15it/s, ppl=6.75]
Evaluating wikitext2:  65%|██████▍   | 91/141 [01:20<00:43,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  65%|██████▌   | 92/141 [01:20<00:42,  1.14it/s, ppl=6.79]
Evaluating wikitext2:  65%|██████▌   | 92/141 [01:20<00:42,  1.14it/s, ppl=6.82]
Evaluating wikitext2:  66%|██████▌   | 93/141 [01:20<00:41,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  66%|██████▌   | 93/141 [01:21<00:41,  1.15it/s, ppl=6.83]
Evaluating wikitext2:  67%|██████▋   | 94/141 [01:21<00:40,  1.15it/s, ppl=6.83]
Evaluating wikitext2:  67%|██████▋   | 94/141 [01:22<00:40,  1.15it/s, ppl=6.85]
Evaluating wikitext2:  67%|██████▋   | 95/141 [01:22<00:39,  1.15it/s, ppl=6.85]
Evaluating wikitext2:  67%|██████▋   | 95/141 [01:23<00:39,  1.15it/s, ppl=6.86]
Evaluating wikitext2:  68%|██████▊   | 96/141 [01:23<00:39,  1.15it/s, ppl=6.86]
Evaluating wikitext2:  68%|██████▊   | 96/141 [01:24<00:39,  1.15it/s, ppl=6.87]
Evaluating wikitext2:  69%|██████▉   | 97/141 [01:24<00:38,  1.15it/s, ppl=6.87]
Evaluating wikitext2:  69%|██████▉   | 97/141 [01:25<00:38,  1.15it/s, ppl=6.85]
Evaluating wikitext2:  70%|██████▉   | 98/141 [01:25<00:37,  1.14it/s, ppl=6.85]
Evaluating wikitext2:  70%|██████▉   | 98/141 [01:26<00:37,  1.14it/s, ppl=6.86]
Evaluating wikitext2:  70%|███████   | 99/141 [01:26<00:36,  1.14it/s, ppl=6.86]
Evaluating wikitext2:  70%|███████   | 99/141 [01:26<00:36,  1.14it/s, ppl=6.87]
Evaluating wikitext2:  71%|███████   | 100/141 [01:26<00:35,  1.15it/s, ppl=6.87]
Evaluating wikitext2:  71%|███████   | 100/141 [01:27<00:35,  1.15it/s, ppl=6.86]
Evaluating wikitext2:  72%|███████▏  | 101/141 [01:27<00:34,  1.15it/s, ppl=6.86]
Evaluating wikitext2:  72%|███████▏  | 101/141 [01:28<00:34,  1.15it/s, ppl=6.87]
Evaluating wikitext2:  72%|███████▏  | 102/141 [01:28<00:33,  1.15it/s, ppl=6.87]
Evaluating wikitext2:  72%|███████▏  | 102/141 [01:29<00:33,  1.15it/s, ppl=6.86]
Evaluating wikitext2:  73%|███████▎  | 103/141 [01:29<00:32,  1.15it/s, ppl=6.86]
Evaluating wikitext2:  73%|███████▎  | 103/141 [01:30<00:32,  1.15it/s, ppl=6.88]
Evaluating wikitext2:  74%|███████▍  | 104/141 [01:30<00:32,  1.15it/s, ppl=6.88]
Evaluating wikitext2:  74%|███████▍  | 104/141 [01:31<00:32,  1.15it/s, ppl=6.87]
Evaluating wikitext2:  74%|███████▍  | 105/141 [01:31<00:31,  1.15it/s, ppl=6.87]
Evaluating wikitext2:  74%|███████▍  | 105/141 [01:32<00:31,  1.15it/s, ppl=6.87]
Evaluating wikitext2:  75%|███████▌  | 106/141 [01:32<00:30,  1.14it/s, ppl=6.87]
Evaluating wikitext2:  75%|███████▌  | 106/141 [01:33<00:30,  1.14it/s, ppl=6.85]
Evaluating wikitext2:  76%|███████▌  | 107/141 [01:33<00:29,  1.15it/s, ppl=6.85]
Evaluating wikitext2:  76%|███████▌  | 107/141 [01:33<00:29,  1.15it/s, ppl=6.84]
Evaluating wikitext2:  77%|███████▋  | 108/141 [01:33<00:28,  1.15it/s, ppl=6.84]
Evaluating wikitext2:  77%|███████▋  | 108/141 [01:34<00:28,  1.15it/s, ppl=6.83]
Evaluating wikitext2:  77%|███████▋  | 109/141 [01:34<00:27,  1.15it/s, ppl=6.83]
Evaluating wikitext2:  77%|███████▋  | 109/141 [01:35<00:27,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  78%|███████▊  | 110/141 [01:35<00:26,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  78%|███████▊  | 110/141 [01:36<00:26,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  79%|███████▊  | 111/141 [01:36<00:25,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  79%|███████▊  | 111/141 [01:37<00:25,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  79%|███████▉  | 112/141 [01:37<00:25,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  79%|███████▉  | 112/141 [01:38<00:25,  1.15it/s, ppl=6.83]
Evaluating wikitext2:  80%|████████  | 113/141 [01:38<00:24,  1.14it/s, ppl=6.83]
Evaluating wikitext2:  80%|████████  | 113/141 [01:39<00:24,  1.14it/s, ppl=6.84]
Evaluating wikitext2:  81%|████████  | 114/141 [01:39<00:23,  1.15it/s, ppl=6.84]
Evaluating wikitext2:  81%|████████  | 114/141 [01:40<00:23,  1.15it/s, ppl=6.83]
Evaluating wikitext2:  82%|████████▏ | 115/141 [01:40<00:22,  1.15it/s, ppl=6.83]
Evaluating wikitext2:  82%|████████▏ | 115/141 [01:40<00:22,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  82%|████████▏ | 116/141 [01:40<00:21,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  82%|████████▏ | 116/141 [01:41<00:21,  1.15it/s, ppl=6.84]
Evaluating wikitext2:  83%|████████▎ | 117/141 [01:41<00:20,  1.15it/s, ppl=6.84]
Evaluating wikitext2:  83%|████████▎ | 117/141 [01:42<00:20,  1.15it/s, ppl=6.83]
Evaluating wikitext2:  84%|████████▎ | 118/141 [01:42<00:19,  1.15it/s, ppl=6.83]
Evaluating wikitext2:  84%|████████▎ | 118/141 [01:43<00:19,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  84%|████████▍ | 119/141 [01:43<00:19,  1.15it/s, ppl=6.82]
Evaluating wikitext2:  84%|████████▍ | 119/141 [01:44<00:19,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  85%|████████▌ | 120/141 [01:44<00:18,  1.16it/s, ppl=6.81]
Evaluating wikitext2:  85%|████████▌ | 120/141 [01:45<00:18,  1.16it/s, ppl=6.81]
Evaluating wikitext2:  86%|████████▌ | 121/141 [01:45<00:17,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  86%|████████▌ | 121/141 [01:46<00:17,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  87%|████████▋ | 122/141 [01:46<00:16,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  87%|████████▋ | 122/141 [01:46<00:16,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  87%|████████▋ | 123/141 [01:46<00:15,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  87%|████████▋ | 123/141 [01:47<00:15,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  88%|████████▊ | 124/141 [01:47<00:14,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  88%|████████▊ | 124/141 [01:48<00:14,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  89%|████████▊ | 125/141 [01:48<00:13,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  89%|████████▊ | 125/141 [01:49<00:13,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  89%|████████▉ | 126/141 [01:49<00:13,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  89%|████████▉ | 126/141 [01:50<00:13,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  90%|█████████ | 127/141 [01:50<00:12,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  90%|█████████ | 127/141 [01:51<00:12,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  91%|█████████ | 128/141 [01:51<00:11,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  91%|█████████ | 128/141 [01:52<00:11,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  91%|█████████▏| 129/141 [01:52<00:10,  1.14it/s, ppl=6.79]
Evaluating wikitext2:  91%|█████████▏| 129/141 [01:53<00:10,  1.14it/s, ppl=6.80]
Evaluating wikitext2:  92%|█████████▏| 130/141 [01:53<00:09,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  92%|█████████▏| 130/141 [01:53<00:09,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  93%|█████████▎| 131/141 [01:53<00:08,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  93%|█████████▎| 131/141 [01:54<00:08,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  94%|█████████▎| 132/141 [01:54<00:07,  1.15it/s, ppl=6.81]
Evaluating wikitext2:  94%|█████████▎| 132/141 [01:55<00:07,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  94%|█████████▍| 133/141 [01:55<00:06,  1.15it/s, ppl=6.79]
Evaluating wikitext2:  94%|█████████▍| 133/141 [01:56<00:06,  1.15it/s, ppl=6.76]
Evaluating wikitext2:  95%|█████████▌| 134/141 [01:56<00:06,  1.15it/s, ppl=6.76]
Evaluating wikitext2:  95%|█████████▌| 134/141 [01:57<00:06,  1.15it/s, ppl=6.77]
Evaluating wikitext2:  96%|█████████▌| 135/141 [01:57<00:05,  1.15it/s, ppl=6.77]
Evaluating wikitext2:  96%|█████████▌| 135/141 [01:58<00:05,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  96%|█████████▋| 136/141 [01:58<00:04,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  96%|█████████▋| 136/141 [01:59<00:04,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  97%|█████████▋| 137/141 [01:59<00:03,  1.14it/s, ppl=6.80]
Evaluating wikitext2:  97%|█████████▋| 137/141 [02:00<00:03,  1.14it/s, ppl=6.80]
Evaluating wikitext2:  98%|█████████▊| 138/141 [02:00<00:02,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  98%|█████████▊| 138/141 [02:00<00:02,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  99%|█████████▊| 139/141 [02:00<00:01,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  99%|█████████▊| 139/141 [02:01<00:01,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  99%|█████████▉| 140/141 [02:01<00:00,  1.15it/s, ppl=6.80]
Evaluating wikitext2:  99%|█████████▉| 140/141 [02:02<00:00,  1.15it/s, ppl=6.81]
Evaluating wikitext2: 100%|██████████| 141/141 [02:02<00:00,  1.15it/s, ppl=6.81]
Evaluating wikitext2: 100%|██████████| 141/141 [02:02<00:00,  1.15it/s, ppl=6.81]
+wikitext2 perplexity: 6.8087
+INFO 12-22 21:39:15 [__init__.py:241] Automatically detected platform cuda.
+`torch_dtype` is deprecated! Use `dtype` instead!
+Loading model and tokenizer...
+
Loading checkpoint shards:   0%|          | 0/4 [00:00<?, ?it/s]
Loading checkpoint shards:  25%|██▌       | 1/4 [00:00<00:02,  1.17it/s]
Loading checkpoint shards:  50%|█████     | 2/4 [00:01<00:01,  1.20it/s]
Loading checkpoint shards:  75%|███████▌  | 3/4 [00:02<00:00,  1.10it/s]
Loading checkpoint shards: 100%|██████████| 4/4 [00:02<00:00,  1.57it/s]
Loading checkpoint shards: 100%|██████████| 4/4 [00:02<00:00,  1.38it/s]
+Replacing weights with pseudo-dequantized versions...
+
Layers:   0%|          | 0/32 [00:00<?, ?it/s]
Layers:   3%|▎         | 1/32 [00:03<01:40,  3.25s/it]
Layers:   6%|▋         | 2/32 [00:04<00:58,  1.93s/it]
Layers:   9%|▉         | 3/32 [00:05<00:43,  1.51s/it]
Layers:  12%|█▎        | 4/32 [00:06<00:36,  1.32s/it]
Layers:  16%|█▌        | 5/32 [00:07<00:33,  1.23s/it]
Layers:  19%|█▉        | 6/32 [00:08<00:31,  1.21s/it]
Layers:  22%|██▏       | 7/32 [00:09<00:29,  1.20s/it]
Layers:  25%|██▌       | 8/32 [00:10<00:28,  1.19s/it]
Layers:  28%|██▊       | 9/32 [00:11<00:25,  1.13s/it]
Layers:  31%|███▏      | 10/32 [00:12<00:23,  1.08s/it]
Layers:  34%|███▍      | 11/32 [00:13<00:22,  1.06s/it]
Layers:  38%|███▊      | 12/32 [00:14<00:20,  1.04s/it]
Layers:  41%|████      | 13/32 [00:15<00:19,  1.03s/it]
Layers:  44%|████▍     | 14/32 [00:16<00:18,  1.01s/it]
Layers:  47%|████▋     | 15/32 [00:17<00:17,  1.01s/it]
Layers:  50%|█████     | 16/32 [00:18<00:16,  1.00s/it]
Layers:  53%|█████▎    | 17/32 [00:19<00:14,  1.00it/s]
Layers:  56%|█████▋    | 18/32 [00:20<00:13,  1.01it/s]
Layers:  59%|█████▉    | 19/32 [00:21<00:13,  1.01s/it]
Layers:  62%|██████▎   | 20/32 [00:22<00:11,  1.01it/s]
Layers:  66%|██████▌   | 21/32 [00:23<00:11,  1.01s/it]
Layers:  69%|██████▉   | 22/32 [00:24<00:10,  1.00s/it]
Layers:  72%|███████▏  | 23/32 [00:25<00:08,  1.00it/s]
Layers:  75%|███████▌  | 24/32 [00:26<00:07,  1.00it/s]
Layers:  78%|███████▊  | 25/32 [00:27<00:07,  1.01s/it]
Layers:  81%|████████▏ | 26/32 [00:28<00:06,  1.02s/it]
Layers:  84%|████████▍ | 27/32 [00:29<00:05,  1.00s/it]
Layers:  88%|████████▊ | 28/32 [00:30<00:04,  1.00s/it]
Layers:  91%|█████████ | 29/32 [00:31<00:03,  1.03s/it]
Layers:  94%|█████████▍| 30/32 [00:32<00:02,  1.03s/it]
Layers:  97%|█████████▋| 31/32 [00:34<00:01,  1.04s/it]
Layers: 100%|██████████| 32/32 [00:35<00:00,  1.05s/it]
Layers: 100%|██████████| 32/32 [00:35<00:00,  1.10s/it]
+`pretrained` model kwarg is not of type `str`. Many other model arguments may be ignored. Please do not launch via accelerate or use `parallelize=True` if passing an existing model this way.
+Passed an already-initialized model through `pretrained`, assuming single-process call to evaluate() or custom distributed integration
+[Task: boolq] metric acc is defined, but aggregation is not. using default aggregation=mean
+[Task: boolq] metric acc is defined, but higher_is_better is not. using default higher_is_better=True
+Overwriting default num_fewshot of boolq from None to 0
+Overwriting default num_fewshot of hellaswag from None to 0
+Overwriting default num_fewshot of arc_challenge from None to 0
+Overwriting default num_fewshot of arc_easy from None to 0
+
  0%|          | 0/3270 [00:00<?, ?it/s]
  6%|▌         | 191/3270 [00:00<00:01, 1907.33it/s]
 12%|█▏        | 385/3270 [00:00<00:01, 1922.45it/s]
 18%|█▊        | 578/3270 [00:00<00:01, 1924.85it/s]
 24%|██▎       | 773/3270 [00:00<00:01, 1932.31it/s]
 30%|██▉       | 967/3270 [00:00<00:01, 1928.05it/s]
 36%|███▌      | 1161/3270 [00:00<00:01, 1929.40it/s]
 42%|████▏     | 1359/3270 [00:00<00:00, 1944.68it/s]
 48%|████▊     | 1554/3270 [00:00<00:00, 1941.96it/s]
 54%|█████▎    | 1753/3270 [00:00<00:00, 1956.33it/s]
 60%|█████▉    | 1949/3270 [00:01<00:00, 1950.71it/s]
 66%|██████▌   | 2146/3270 [00:01<00:00, 1956.15it/s]
 72%|███████▏  | 2344/3270 [00:01<00:00, 1960.41it/s]
 78%|███████▊  | 2543/3270 [00:01<00:00, 1968.46it/s]
 84%|████████▍ | 2742/3270 [00:01<00:00, 1974.11it/s]
 90%|████████▉ | 2940/3270 [00:01<00:00, 1975.46it/s]
 96%|█████████▌| 3138/3270 [00:01<00:00, 1966.06it/s]
100%|██████████| 3270/3270 [00:01<00:00, 1952.93it/s]
+
  0%|          | 0/10042 [00:00<?, ?it/s]
  3%|▎         | 272/10042 [00:00<00:03, 2714.11it/s]
  5%|▌         | 548/10042 [00:00<00:03, 2740.62it/s]
  8%|▊         | 825/10042 [00:00<00:03, 2750.28it/s]
 11%|█         | 1101/10042 [00:00<00:03, 2731.70it/s]
 14%|█▎        | 1375/10042 [00:00<00:08, 1054.23it/s]
 16%|█▋        | 1653/10042 [00:01<00:06, 1337.98it/s]
 19%|█▉        | 1924/10042 [00:01<00:05, 1598.47it/s]
 22%|██▏       | 2204/10042 [00:01<00:04, 1854.04it/s]
 25%|██▍       | 2481/10042 [00:01<00:03, 2067.73it/s]
 27%|██▋       | 2757/10042 [00:01<00:03, 2240.62it/s]
 30%|███       | 3036/10042 [00:01<00:02, 2384.78it/s]
 33%|███▎      | 3312/10042 [00:01<00:02, 2484.98it/s]
 36%|███▌      | 3590/10042 [00:01<00:02, 2565.62it/s]
 39%|███▉      | 3917/10042 [00:01<00:02, 2766.63it/s]
 43%|████▎     | 4349/10042 [00:01<00:01, 3216.46it/s]
 48%|████▊     | 4791/10042 [00:02<00:01, 3568.69it/s]
 52%|█████▏    | 5204/10042 [00:02<00:01, 3732.09it/s]
 56%|█████▌    | 5636/10042 [00:02<00:01, 3905.62it/s]
 60%|██████    | 6067/10042 [00:02<00:00, 4024.08it/s]
 65%|██████▍   | 6485/10042 [00:02<00:00, 4069.92it/s]
 69%|██████▉   | 6907/10042 [00:02<00:00, 4112.13it/s]
 73%|███████▎  | 7328/10042 [00:02<00:00, 4139.04it/s]
 77%|███████▋  | 7771/10042 [00:02<00:00, 4225.67it/s]
 82%|████████▏ | 8199/10042 [00:02<00:00, 4239.97it/s]
 86%|████████▌ | 8635/10042 [00:02<00:00, 4275.37it/s]
 90%|█████████ | 9066/10042 [00:03<00:00, 4282.83it/s]
 95%|█████████▍| 9496/10042 [00:03<00:00, 4286.45it/s]
 99%|█████████▉| 9931/10042 [00:03<00:00, 4303.71it/s]
100%|██████████| 10042/10042 [00:03<00:00, 3064.44it/s]
+
  0%|          | 0/1172 [00:00<?, ?it/s]
 14%|█▎        | 159/1172 [00:00<00:00, 1586.30it/s]
 27%|██▋       | 318/1172 [00:00<00:00, 1517.10it/s]
 40%|████      | 470/1172 [00:00<00:00, 1298.42it/s]
 55%|█████▍    | 640/1172 [00:00<00:00, 1437.44it/s]
 70%|██████▉   | 815/1172 [00:00<00:00, 1539.97it/s]
 84%|████████▍ | 988/1172 [00:00<00:00, 1599.54it/s]
 99%|█████████▉| 1166/1172 [00:00<00:00, 1654.11it/s]
100%|██████████| 1172/1172 [00:00<00:00, 1559.38it/s]
+
  0%|          | 0/2376 [00:00<?, ?it/s]
  8%|▊         | 179/2376 [00:00<00:01, 1786.83it/s]
 15%|█▌        | 358/2376 [00:00<00:01, 1765.85it/s]
 23%|██▎       | 535/2376 [00:00<00:01, 1738.37it/s]
 30%|██▉       | 712/2376 [00:00<00:00, 1748.36it/s]
 37%|███▋      | 890/2376 [00:00<00:00, 1759.03it/s]
 45%|████▍     | 1068/2376 [00:00<00:00, 1764.06it/s]
 52%|█████▏    | 1245/2376 [00:00<00:00, 1761.53it/s]
 60%|█████▉    | 1422/2376 [00:00<00:00, 1753.36it/s]
 68%|██████▊   | 1610/2376 [00:00<00:00, 1789.95it/s]
 75%|███████▌  | 1792/2376 [00:01<00:00, 1795.78it/s]
 83%|████████▎ | 1975/2376 [00:01<00:00, 1803.35it/s]
 91%|█████████ | 2159/2376 [00:01<00:00, 1811.66it/s]
 99%|█████████▊| 2341/2376 [00:01<00:00, 1790.99it/s]
100%|██████████| 2376/2376 [00:01<00:00, 1778.27it/s]
+
Running loglikelihood requests:   0%|          | 0/60896 [00:00<?, ?it/s]
Running loglikelihood requests:   0%|          | 1/60896 [00:15<254:21:18, 15.04s/it]
Running loglikelihood requests:   0%|          | 65/60896 [00:19<4:01:17,  4.20it/s] 
Running loglikelihood requests:   0%|          | 129/60896 [00:23<2:17:17,  7.38it/s]
Running loglikelihood requests:   0%|          | 193/60896 [00:27<1:41:46,  9.94it/s]
Running loglikelihood requests:   0%|          | 257/60896 [00:31<1:24:01, 12.03it/s]
Running loglikelihood requests:   1%|          | 321/60896 [00:34<1:13:40, 13.70it/s]
Running loglikelihood requests:   1%|          | 385/60896 [00:38<1:06:09, 15.24it/s]
Running loglikelihood requests:   1%|          | 449/60896 [00:41<1:00:59, 16.52it/s]
Running loglikelihood requests:   1%|          | 513/60896 [00:44<57:13, 17.59it/s]  
Running loglikelihood requests:   1%|          | 577/60896 [00:47<54:36, 18.41it/s]
Running loglikelihood requests:   1%|          | 641/60896 [00:50<52:09, 19.25it/s]
Running loglikelihood requests:   1%|          | 705/60896 [00:53<49:53, 20.11it/s]
Running loglikelihood requests:   1%|▏         | 769/60896 [00:56<48:09, 20.81it/s]
Running loglikelihood requests:   1%|▏         | 833/60896 [00:59<46:38, 21.46it/s]
Running loglikelihood requests:   1%|▏         | 897/60896 [01:01<45:29, 21.98it/s]
Running loglikelihood requests:   2%|▏         | 961/60896 [01:04<44:11, 22.60it/s]
Running loglikelihood requests:   2%|▏         | 1025/60896 [01:07<43:17, 23.05it/s]
Running loglikelihood requests:   2%|▏         | 1089/60896 [01:09<42:18, 23.56it/s]
Running loglikelihood requests:   2%|▏         | 1153/60896 [01:12<41:17, 24.11it/s]
Running loglikelihood requests:   2%|▏         | 1217/60896 [01:14<40:22, 24.64it/s]
Running loglikelihood requests:   2%|▏         | 1281/60896 [01:17<39:27, 25.18it/s]
Running loglikelihood requests:   2%|▏         | 1345/60896 [01:19<38:44, 25.62it/s]
Running loglikelihood requests:   2%|▏         | 1409/60896 [01:21<38:09, 25.98it/s]
Running loglikelihood requests:   2%|▏         | 1473/60896 [01:24<37:39, 26.30it/s]
Running loglikelihood requests:   3%|▎         | 1537/60896 [01:26<37:06, 26.66it/s]
Running loglikelihood requests:   3%|▎         | 1601/60896 [01:28<36:36, 27.00it/s]
Running loglikelihood requests:   3%|▎         | 1661/60896 [01:31<36:55, 26.73it/s]
Running loglikelihood requests:   3%|▎         | 1725/60896 [01:33<36:25, 27.08it/s]
Running loglikelihood requests:   3%|▎         | 1789/60896 [01:35<35:46, 27.54it/s]
Running loglikelihood requests:   3%|▎         | 1853/60896 [01:37<35:16, 27.90it/s]
Running loglikelihood requests:   3%|▎         | 1917/60896 [01:39<34:21, 28.61it/s]
Running loglikelihood requests:   3%|▎         | 1978/60896 [01:42<34:17, 28.64it/s]
Running loglikelihood requests:   3%|▎         | 2041/60896 [01:44<33:44, 29.07it/s]
Running loglikelihood requests:   3%|▎         | 2105/60896 [01:46<33:00, 29.68it/s]
Running loglikelihood requests:   4%|▎         | 2169/60896 [01:48<32:29, 30.13it/s]
Running loglikelihood requests:   4%|▎         | 2233/60896 [01:50<31:58, 30.58it/s]
Running loglikelihood requests:   4%|▍         | 2297/60896 [01:52<31:35, 30.91it/s]
Running loglikelihood requests:   4%|▍         | 2361/60896 [01:54<31:19, 31.14it/s]
Running loglikelihood requests:   4%|▍         | 2425/60896 [01:56<31:02, 31.39it/s]
Running loglikelihood requests:   4%|▍         | 2488/60896 [01:58<30:45, 31.64it/s]
Running loglikelihood requests:   4%|▍         | 2549/60896 [02:00<30:49, 31.55it/s]
Running loglikelihood requests:   4%|▍         | 2613/60896 [02:02<30:28, 31.87it/s]
Running loglikelihood requests:   4%|▍         | 2677/60896 [02:04<29:57, 32.40it/s]
Running loglikelihood requests:   4%|▍         | 2736/60896 [02:05<30:16, 32.02it/s]
Running loglikelihood requests:   5%|▍         | 2799/60896 [02:07<29:46, 32.52it/s]
Running loglikelihood requests:   5%|▍         | 2863/60896 [02:09<29:15, 33.06it/s]
Running loglikelihood requests:   5%|▍         | 2926/60896 [02:11<29:03, 33.26it/s]
Running loglikelihood requests:   5%|▍         | 2988/60896 [02:13<29:02, 33.23it/s]
Running loglikelihood requests:   5%|▌         | 3051/60896 [02:15<28:42, 33.57it/s]
Running loglikelihood requests:   5%|▌         | 3112/60896 [02:17<28:42, 33.54it/s]
Running loglikelihood requests:   5%|▌         | 3173/60896 [02:18<28:39, 33.56it/s]
Running loglikelihood requests:   5%|▌         | 3227/60896 [02:20<29:03, 33.08it/s]
Running loglikelihood requests:   5%|▌         | 3286/60896 [02:22<28:30, 33.69it/s]
Running loglikelihood requests:   5%|▌         | 3344/60896 [02:23<28:14, 33.97it/s]
Running loglikelihood requests:   6%|▌         | 3398/60896 [02:25<28:35, 33.51it/s]
Running loglikelihood requests:   6%|▌         | 3447/60896 [02:27<29:41, 32.25it/s]
Running loglikelihood requests:   6%|▌         | 3492/60896 [02:28<31:07, 30.73it/s]
Running loglikelihood requests:   6%|▌         | 3537/60896 [02:30<32:12, 29.68it/s]
Running loglikelihood requests:   6%|▌         | 3576/60896 [02:32<34:15, 27.88it/s]
Running loglikelihood requests:   6%|▌         | 3621/60896 [02:33<34:37, 27.56it/s]
Running loglikelihood requests:   6%|▌         | 3653/60896 [02:35<37:57, 25.13it/s]
Running loglikelihood requests:   6%|▌         | 3698/60896 [02:37<36:58, 25.78it/s]
Running loglikelihood requests:   6%|▌         | 3741/60896 [02:38<36:48, 25.88it/s]
Running loglikelihood requests:   6%|▌         | 3775/60896 [02:40<39:09, 24.32it/s]
Running loglikelihood requests:   6%|▋         | 3810/60896 [02:42<40:49, 23.30it/s]
Running loglikelihood requests:   6%|▋         | 3852/60896 [02:43<39:38, 23.98it/s]
Running loglikelihood requests:   6%|▋         | 3886/60896 [02:45<41:18, 23.01it/s]
Running loglikelihood requests:   6%|▋         | 3918/60896 [02:47<43:14, 21.96it/s]
Running loglikelihood requests:   6%|▋         | 3953/60896 [02:48<43:41, 21.72it/s]
Running loglikelihood requests:   7%|▋         | 3989/60896 [02:50<43:31, 21.79it/s]
Running loglikelihood requests:   7%|▋         | 4033/60896 [02:52<40:39, 23.31it/s]
Running loglikelihood requests:   7%|▋         | 4068/60896 [02:53<41:39, 22.73it/s]
Running loglikelihood requests:   7%|▋         | 4100/60896 [02:55<43:27, 21.78it/s]
Running loglikelihood requests:   7%|▋         | 4132/60896 [02:56<45:00, 21.02it/s]
Running loglikelihood requests:   7%|▋         | 4168/60896 [02:58<44:22, 21.31it/s]
Running loglikelihood requests:   7%|▋         | 4210/60896 [03:00<41:45, 22.62it/s]
Running loglikelihood requests:   7%|▋         | 4245/60896 [03:01<42:24, 22.27it/s]
Running loglikelihood requests:   7%|▋         | 4280/60896 [03:03<42:57, 21.97it/s]
Running loglikelihood requests:   7%|▋         | 4312/60896 [03:05<44:20, 21.27it/s]
Running loglikelihood requests:   7%|▋         | 4344/60896 [03:06<45:22, 20.77it/s]
Running loglikelihood requests:   7%|▋         | 4377/60896 [03:08<45:42, 20.61it/s]
Running loglikelihood requests:   7%|▋         | 4411/60896 [03:10<45:43, 20.59it/s]
Running loglikelihood requests:   7%|▋         | 4458/60896 [03:11<40:48, 23.05it/s]
Running loglikelihood requests:   7%|▋         | 4492/60896 [03:13<41:57, 22.40it/s]
Running loglikelihood requests:   7%|▋         | 4527/60896 [03:14<42:27, 22.12it/s]
Running loglikelihood requests:   7%|▋         | 4560/60896 [03:16<43:31, 21.57it/s]
Running loglikelihood requests:   8%|▊         | 4592/60896 [03:18<44:53, 20.90it/s]
Running loglikelihood requests:   8%|▊         | 4627/60896 [03:19<44:29, 21.08it/s]
Running loglikelihood requests:   8%|▊         | 4662/60896 [03:21<44:12, 21.20it/s]
Running loglikelihood requests:   8%|▊         | 4703/60896 [03:23<41:29, 22.57it/s]
Running loglikelihood requests:   8%|▊         | 4737/60896 [03:24<42:20, 22.11it/s]
Running loglikelihood requests:   8%|▊         | 4771/60896 [03:26<42:41, 21.91it/s]
Running loglikelihood requests:   8%|▊         | 4806/60896 [03:27<42:34, 21.96it/s]
Running loglikelihood requests:   8%|▊         | 4838/60896 [03:29<43:36, 21.42it/s]
Running loglikelihood requests:   8%|▊         | 4870/60896 [03:31<44:27, 21.00it/s]
Running loglikelihood requests:   8%|▊         | 4903/60896 [03:32<44:33, 20.95it/s]
Running loglikelihood requests:   8%|▊         | 4938/60896 [03:34<43:49, 21.28it/s]
Running loglikelihood requests:   8%|▊         | 4986/60896 [03:35<38:51, 23.98it/s]
Running loglikelihood requests:   8%|▊         | 5021/60896 [03:37<39:45, 23.42it/s]
Running loglikelihood requests:   8%|▊         | 5054/60896 [03:39<41:12, 22.59it/s]
Running loglikelihood requests:   8%|▊         | 5093/60896 [03:40<40:05, 23.20it/s]
Running loglikelihood requests:   8%|▊         | 5125/60896 [03:42<41:39, 22.31it/s]
Running loglikelihood requests:   8%|▊         | 5157/60896 [03:43<42:49, 21.69it/s]
Running loglikelihood requests:   9%|▊         | 5189/60896 [03:45<43:40, 21.26it/s]
Running loglikelihood requests:   9%|▊         | 5222/60896 [03:46<43:58, 21.10it/s]
Running loglikelihood requests:   9%|▊         | 5255/60896 [03:48<44:06, 21.02it/s]
Running loglikelihood requests:   9%|▊         | 5289/60896 [03:50<43:47, 21.17it/s]
Running loglikelihood requests:   9%|▉         | 5335/60896 [03:51<39:18, 23.56it/s]
Running loglikelihood requests:   9%|▉         | 5372/60896 [03:53<39:19, 23.53it/s]
Running loglikelihood requests:   9%|▉         | 5404/60896 [03:54<41:10, 22.46it/s]
Running loglikelihood requests:   9%|▉         | 5439/60896 [03:56<41:18, 22.38it/s]
Running loglikelihood requests:   9%|▉         | 5471/60896 [03:58<42:27, 21.75it/s]
Running loglikelihood requests:   9%|▉         | 5503/60896 [03:59<43:18, 21.32it/s]
Running loglikelihood requests:   9%|▉         | 5535/60896 [04:01<44:08, 20.90it/s]
Running loglikelihood requests:   9%|▉         | 5567/60896 [04:02<44:29, 20.72it/s]
Running loglikelihood requests:   9%|▉         | 5601/60896 [04:04<43:56, 20.98it/s]
Running loglikelihood requests:   9%|▉         | 5636/60896 [04:05<43:09, 21.34it/s]
Running loglikelihood requests:   9%|▉         | 5670/60896 [04:07<42:59, 21.41it/s]
Running loglikelihood requests:   9%|▉         | 5717/60896 [04:09<38:37, 23.81it/s]
Running loglikelihood requests:   9%|▉         | 5754/60896 [04:10<38:44, 23.72it/s]
Running loglikelihood requests:  10%|▉         | 5788/60896 [04:12<39:45, 23.10it/s]
Running loglikelihood requests:  10%|▉         | 5820/60896 [04:13<41:12, 22.28it/s]
Running loglikelihood requests:  10%|▉         | 5852/60896 [04:15<42:31, 21.57it/s]
Running loglikelihood requests:  10%|▉         | 5887/60896 [04:16<42:06, 21.77it/s]
Running loglikelihood requests:  10%|▉         | 5919/60896 [04:18<42:55, 21.35it/s]
Running loglikelihood requests:  10%|▉         | 5951/60896 [04:20<43:30, 21.05it/s]
Running loglikelihood requests:  10%|▉         | 5983/60896 [04:21<44:08, 20.73it/s]
Running loglikelihood requests:  10%|▉         | 6015/60896 [04:23<44:21, 20.62it/s]
Running loglikelihood requests:  10%|▉         | 6051/60896 [04:24<42:54, 21.30it/s]
Running loglikelihood requests:  10%|▉         | 6084/60896 [04:26<43:04, 21.21it/s]
Running loglikelihood requests:  10%|█         | 6129/60896 [04:28<38:56, 23.44it/s]
Running loglikelihood requests:  10%|█         | 6173/60896 [04:29<36:56, 24.69it/s]
Running loglikelihood requests:  10%|█         | 6206/60896 [04:31<38:36, 23.60it/s]
Running loglikelihood requests:  10%|█         | 6238/60896 [04:32<40:13, 22.65it/s]
Running loglikelihood requests:  10%|█         | 6273/60896 [04:34<40:22, 22.54it/s]
Running loglikelihood requests:  10%|█         | 6306/60896 [04:35<41:21, 22.00it/s]
Running loglikelihood requests:  10%|█         | 6338/60896 [04:37<42:15, 21.51it/s]
Running loglikelihood requests:  10%|█         | 6370/60896 [04:39<42:54, 21.18it/s]
Running loglikelihood requests:  11%|█         | 6402/60896 [04:40<43:21, 20.95it/s]
Running loglikelihood requests:  11%|█         | 6434/60896 [04:42<43:40, 20.78it/s]
Running loglikelihood requests:  11%|█         | 6467/60896 [04:43<43:38, 20.78it/s]
Running loglikelihood requests:  11%|█         | 6500/60896 [04:45<43:28, 20.85it/s]
Running loglikelihood requests:  11%|█         | 6532/60896 [04:46<43:44, 20.72it/s]
Running loglikelihood requests:  11%|█         | 6579/60896 [04:48<38:32, 23.49it/s]
Running loglikelihood requests:  11%|█         | 6612/60896 [04:50<39:51, 22.70it/s]
Running loglikelihood requests:  11%|█         | 6644/60896 [04:51<41:02, 22.03it/s]
Running loglikelihood requests:  11%|█         | 6678/60896 [04:53<41:10, 21.95it/s]
Running loglikelihood requests:  11%|█         | 6710/60896 [04:54<41:59, 21.51it/s]
Running loglikelihood requests:  11%|█         | 6744/60896 [04:56<41:49, 21.58it/s]
Running loglikelihood requests:  11%|█         | 6776/60896 [04:57<42:40, 21.13it/s]
Running loglikelihood requests:  11%|█         | 6808/60896 [04:59<43:03, 20.94it/s]
Running loglikelihood requests:  11%|█         | 6840/60896 [05:01<43:19, 20.80it/s]
Running loglikelihood requests:  11%|█▏        | 6872/60896 [05:02<43:29, 20.70it/s]
Running loglikelihood requests:  11%|█▏        | 6906/60896 [05:04<42:56, 20.96it/s]
Running loglikelihood requests:  11%|█▏        | 6940/60896 [05:05<42:26, 21.19it/s]
Running loglikelihood requests:  11%|█▏        | 6981/60896 [05:07<39:33, 22.72it/s]
Running loglikelihood requests:  12%|█▏        | 7014/60896 [05:08<40:21, 22.25it/s]
Running loglikelihood requests:  12%|█▏        | 7048/60896 [05:10<40:41, 22.05it/s]
Running loglikelihood requests:  12%|█▏        | 7080/60896 [05:12<41:32, 21.59it/s]
Running loglikelihood requests:  12%|█▏        | 7113/60896 [05:13<41:46, 21.46it/s]
Running loglikelihood requests:  12%|█▏        | 7148/60896 [05:15<41:10, 21.75it/s]
Running loglikelihood requests:  12%|█▏        | 7182/60896 [05:16<41:22, 21.63it/s]
Running loglikelihood requests:  12%|█▏        | 7214/60896 [05:18<41:59, 21.30it/s]
Running loglikelihood requests:  12%|█▏        | 7246/60896 [05:19<42:25, 21.07it/s]
Running loglikelihood requests:  12%|█▏        | 7278/60896 [05:21<42:44, 20.91it/s]
Running loglikelihood requests:  12%|█▏        | 7310/60896 [05:22<43:08, 20.70it/s]
Running loglikelihood requests:  12%|█▏        | 7342/60896 [05:24<43:13, 20.65it/s]
Running loglikelihood requests:  12%|█▏        | 7378/60896 [05:26<41:43, 21.37it/s]
Running loglikelihood requests:  12%|█▏        | 7410/60896 [05:27<42:12, 21.12it/s]
Running loglikelihood requests:  12%|█▏        | 7446/60896 [05:29<41:00, 21.73it/s]
Running loglikelihood requests:  12%|█▏        | 7479/60896 [05:30<41:26, 21.48it/s]
Running loglikelihood requests:  12%|█▏        | 7514/60896 [05:32<40:50, 21.79it/s]
Running loglikelihood requests:  12%|█▏        | 7546/60896 [05:33<41:29, 21.43it/s]
Running loglikelihood requests:  12%|█▏        | 7578/60896 [05:35<41:57, 21.17it/s]
Running loglikelihood requests:  12%|█▎        | 7612/60896 [05:37<41:32, 21.38it/s]
Running loglikelihood requests:  13%|█▎        | 7647/60896 [05:38<41:06, 21.59it/s]
Running loglikelihood requests:  13%|█▎        | 7679/60896 [05:40<41:39, 21.29it/s]
Running loglikelihood requests:  13%|█▎        | 7711/60896 [05:41<42:03, 21.08it/s]
Running loglikelihood requests:  13%|█▎        | 7743/60896 [05:43<42:19, 20.93it/s]
Running loglikelihood requests:  13%|█▎        | 7775/60896 [05:44<42:42, 20.73it/s]
Running loglikelihood requests:  13%|█▎        | 7807/60896 [05:46<42:46, 20.69it/s]
Running loglikelihood requests:  13%|█▎        | 7839/60896 [05:47<42:49, 20.65it/s]
Running loglikelihood requests:  13%|█▎        | 7875/60896 [05:49<41:19, 21.39it/s]
Running loglikelihood requests:  13%|█▎        | 7909/60896 [05:51<41:09, 21.45it/s]
Running loglikelihood requests:  13%|█▎        | 7942/60896 [05:52<41:16, 21.38it/s]
Running loglikelihood requests:  13%|█▎        | 7988/60896 [05:54<36:32, 24.13it/s]
Running loglikelihood requests:  13%|█▎        | 8023/60896 [05:55<36:48, 23.94it/s]
Running loglikelihood requests:  13%|█▎        | 8055/60896 [05:57<37:55, 23.22it/s]
Running loglikelihood requests:  13%|█▎        | 8087/60896 [05:58<38:58, 22.58it/s]
Running loglikelihood requests:  13%|█▎        | 8119/60896 [06:00<39:32, 22.25it/s]
Running loglikelihood requests:  13%|█▎        | 8155/60896 [06:01<38:32, 22.81it/s]
Running loglikelihood requests:  13%|█▎        | 8187/60896 [06:03<39:13, 22.39it/s]
Running loglikelihood requests:  13%|█▎        | 8219/60896 [06:04<39:41, 22.12it/s]
Running loglikelihood requests:  14%|█▎        | 8251/60896 [06:06<40:08, 21.86it/s]
Running loglikelihood requests:  14%|█▎        | 8283/60896 [06:07<40:19, 21.75it/s]
Running loglikelihood requests:  14%|█▎        | 8315/60896 [06:09<40:27, 21.66it/s]
Running loglikelihood requests:  14%|█▎        | 8349/60896 [06:10<39:48, 22.00it/s]
Running loglikelihood requests:  14%|█▍        | 8383/60896 [06:12<39:32, 22.14it/s]
Running loglikelihood requests:  14%|█▍        | 8415/60896 [06:13<39:52, 21.93it/s]
Running loglikelihood requests:  14%|█▍        | 8452/60896 [06:15<38:18, 22.81it/s]
Running loglikelihood requests:  14%|█▍        | 8497/60896 [06:16<34:53, 25.03it/s]
Running loglikelihood requests:  14%|█▍        | 8529/60896 [06:18<36:25, 23.97it/s]
Running loglikelihood requests:  14%|█▍        | 8562/60896 [06:19<37:23, 23.32it/s]
Running loglikelihood requests:  14%|█▍        | 8594/60896 [06:21<38:16, 22.78it/s]
Running loglikelihood requests:  14%|█▍        | 8626/60896 [06:22<38:53, 22.40it/s]
Running loglikelihood requests:  14%|█▍        | 8658/60896 [06:24<39:20, 22.13it/s]
Running loglikelihood requests:  14%|█▍        | 8692/60896 [06:25<38:56, 22.35it/s]
Running loglikelihood requests:  14%|█▍        | 8724/60896 [06:26<39:29, 22.02it/s]
Running loglikelihood requests:  14%|█▍        | 8756/60896 [06:28<39:43, 21.87it/s]
Running loglikelihood requests:  14%|█▍        | 8788/60896 [06:29<39:54, 21.76it/s]
Running loglikelihood requests:  14%|█▍        | 8820/60896 [06:31<40:00, 21.69it/s]
Running loglikelihood requests:  15%|█▍        | 8854/60896 [06:32<39:36, 21.90it/s]
Running loglikelihood requests:  15%|█▍        | 8890/60896 [06:34<38:22, 22.59it/s]
Running loglikelihood requests:  15%|█▍        | 8922/60896 [06:35<38:53, 22.28it/s]
Running loglikelihood requests:  15%|█▍        | 8957/60896 [06:37<38:12, 22.66it/s]
Running loglikelihood requests:  15%|█▍        | 8997/60896 [06:38<36:07, 23.94it/s]
Running loglikelihood requests:  15%|█▍        | 9033/60896 [06:40<36:10, 23.90it/s]
Running loglikelihood requests:  15%|█▍        | 9067/60896 [06:41<36:35, 23.60it/s]
Running loglikelihood requests:  15%|█▍        | 9099/60896 [06:43<37:32, 22.99it/s]
Running loglikelihood requests:  15%|█▍        | 9131/60896 [06:44<38:13, 22.57it/s]
Running loglikelihood requests:  15%|█▌        | 9163/60896 [06:46<38:42, 22.27it/s]
Running loglikelihood requests:  15%|█▌        | 9201/60896 [06:47<37:09, 23.19it/s]
Running loglikelihood requests:  15%|█▌        | 9236/60896 [06:49<36:56, 23.31it/s]
Running loglikelihood requests:  15%|█▌        | 9268/60896 [06:50<37:45, 22.79it/s]
Running loglikelihood requests:  15%|█▌        | 9300/60896 [06:52<38:21, 22.42it/s]
Running loglikelihood requests:  15%|█▌        | 9333/60896 [06:53<38:31, 22.31it/s]
Running loglikelihood requests:  15%|█▌        | 9365/60896 [06:55<38:51, 22.10it/s]
Running loglikelihood requests:  15%|█▌        | 9398/60896 [06:56<38:45, 22.14it/s]
Running loglikelihood requests:  15%|█▌        | 9431/60896 [06:58<38:41, 22.17it/s]
Running loglikelihood requests:  16%|█▌        | 9463/60896 [06:59<38:58, 21.99it/s]
Running loglikelihood requests:  16%|█▌        | 9498/60896 [07:01<38:07, 22.47it/s]
Running loglikelihood requests:  16%|█▌        | 9538/60896 [07:02<36:06, 23.70it/s]
Running loglikelihood requests:  16%|█▌        | 9571/60896 [07:04<36:46, 23.26it/s]
Running loglikelihood requests:  16%|█▌        | 9606/60896 [07:05<36:35, 23.36it/s]
Running loglikelihood requests:  16%|█▌        | 9638/60896 [07:07<37:24, 22.84it/s]
Running loglikelihood requests:  16%|█▌        | 9670/60896 [07:08<38:11, 22.35it/s]
Running loglikelihood requests:  16%|█▌        | 9706/60896 [07:10<37:13, 22.92it/s]
Running loglikelihood requests:  16%|█▌        | 9738/60896 [07:11<37:53, 22.50it/s]
Running loglikelihood requests:  16%|█▌        | 9771/60896 [07:13<37:59, 22.43it/s]
Running loglikelihood requests:  16%|█▌        | 9803/60896 [07:14<38:24, 22.17it/s]
Running loglikelihood requests:  16%|█▌        | 9835/60896 [07:16<38:52, 21.89it/s]
Running loglikelihood requests:  16%|█▌        | 9867/60896 [07:17<39:02, 21.79it/s]
Running loglikelihood requests:  16%|█▋        | 9899/60896 [07:19<39:07, 21.72it/s]
Running loglikelihood requests:  16%|█▋        | 9931/60896 [07:20<39:12, 21.66it/s]
Running loglikelihood requests:  16%|█▋        | 9963/60896 [07:22<39:15, 21.62it/s]
Running loglikelihood requests:  16%|█▋        | 9995/60896 [07:23<39:28, 21.49it/s]
Running loglikelihood requests:  16%|█▋        | 10029/60896 [07:25<38:41, 21.91it/s]
Running loglikelihood requests:  17%|█▋        | 10062/60896 [07:26<38:27, 22.03it/s]
Running loglikelihood requests:  17%|█▋        | 10101/60896 [07:28<36:27, 23.22it/s]
Running loglikelihood requests:  17%|█▋        | 10139/60896 [07:29<35:27, 23.86it/s]
Running loglikelihood requests:  17%|█▋        | 10172/60896 [07:31<36:27, 23.19it/s]
Running loglikelihood requests:  17%|█▋        | 10205/60896 [07:32<36:59, 22.84it/s]
Running loglikelihood requests:  17%|█▋        | 10237/60896 [07:34<37:42, 22.39it/s]
Running loglikelihood requests:  17%|█▋        | 10269/60896 [07:35<38:13, 22.08it/s]
Running loglikelihood requests:  17%|█▋        | 10303/60896 [07:37<37:58, 22.20it/s]
Running loglikelihood requests:  17%|█▋        | 10335/60896 [07:38<38:23, 21.95it/s]
Running loglikelihood requests:  17%|█▋        | 10369/60896 [07:40<37:59, 22.16it/s]
Running loglikelihood requests:  17%|█▋        | 10401/60896 [07:41<38:24, 21.91it/s]
Running loglikelihood requests:  17%|█▋        | 10433/60896 [07:43<38:40, 21.75it/s]
Running loglikelihood requests:  17%|█▋        | 10465/60896 [07:44<39:01, 21.54it/s]
Running loglikelihood requests:  17%|█▋        | 10497/60896 [07:46<39:06, 21.48it/s]
Running loglikelihood requests:  17%|█▋        | 10529/60896 [07:47<39:11, 21.42it/s]
Running loglikelihood requests:  17%|█▋        | 10561/60896 [07:49<39:12, 21.40it/s]
Running loglikelihood requests:  17%|█▋        | 10593/60896 [07:50<39:12, 21.38it/s]
Running loglikelihood requests:  17%|█▋        | 10626/60896 [07:52<39:02, 21.46it/s]
Running loglikelihood requests:  18%|█▊        | 10660/60896 [07:53<38:22, 21.82it/s]
Running loglikelihood requests:  18%|█▊        | 10692/60896 [07:55<38:35, 21.68it/s]
Running loglikelihood requests:  18%|█▊        | 10737/60896 [07:56<34:34, 24.18it/s]
Running loglikelihood requests:  18%|█▊        | 10769/60896 [07:58<35:52, 23.28it/s]
Running loglikelihood requests:  18%|█▊        | 10803/60896 [07:59<36:07, 23.11it/s]
Running loglikelihood requests:  18%|█▊        | 10836/60896 [08:01<36:36, 22.79it/s]
Running loglikelihood requests:  18%|█▊        | 10868/60896 [08:02<37:16, 22.37it/s]
Running loglikelihood requests:  18%|█▊        | 10903/60896 [08:04<36:45, 22.67it/s]
Running loglikelihood requests:  18%|█▊        | 10935/60896 [08:05<37:30, 22.20it/s]
Running loglikelihood requests:  18%|█▊        | 10968/60896 [08:07<37:34, 22.15it/s]
Running loglikelihood requests:  18%|█▊        | 11001/60896 [08:08<37:35, 22.12it/s]
Running loglikelihood requests:  18%|█▊        | 11033/60896 [08:10<37:57, 21.89it/s]
Running loglikelihood requests:  18%|█▊        | 11065/60896 [08:11<38:24, 21.62it/s]
Running loglikelihood requests:  18%|█▊        | 11097/60896 [08:13<38:31, 21.55it/s]
Running loglikelihood requests:  18%|█▊        | 11129/60896 [08:14<38:35, 21.50it/s]
Running loglikelihood requests:  18%|█▊        | 11161/60896 [08:16<38:37, 21.46it/s]
Running loglikelihood requests:  18%|█▊        | 11193/60896 [08:17<38:38, 21.43it/s]
Running loglikelihood requests:  18%|█▊        | 11225/60896 [08:19<38:53, 21.28it/s]
Running loglikelihood requests:  18%|█▊        | 11258/60896 [08:20<38:28, 21.51it/s]
Running loglikelihood requests:  19%|█▊        | 11298/60896 [08:22<35:49, 23.08it/s]
Running loglikelihood requests:  19%|█▊        | 11334/60896 [08:23<35:19, 23.38it/s]
Running loglikelihood requests:  19%|█▊        | 11367/60896 [08:25<35:55, 22.97it/s]
Running loglikelihood requests:  19%|█▊        | 11399/60896 [08:26<36:38, 22.51it/s]
Running loglikelihood requests:  19%|█▉        | 11432/60896 [08:28<36:49, 22.39it/s]
Running loglikelihood requests:  19%|█▉        | 11464/60896 [08:29<37:16, 22.11it/s]
Running loglikelihood requests:  19%|█▉        | 11496/60896 [08:31<37:35, 21.90it/s]
Running loglikelihood requests:  19%|█▉        | 11529/60896 [08:32<37:38, 21.86it/s]
Running loglikelihood requests:  19%|█▉        | 11561/60896 [08:34<37:50, 21.73it/s]
Running loglikelihood requests:  19%|█▉        | 11594/60896 [08:35<37:37, 21.84it/s]
Running loglikelihood requests:  19%|█▉        | 11626/60896 [08:37<37:48, 21.72it/s]
Running loglikelihood requests:  19%|█▉        | 11658/60896 [08:38<37:55, 21.64it/s]
Running loglikelihood requests:  19%|█▉        | 11690/60896 [08:40<38:14, 21.44it/s]
Running loglikelihood requests:  19%|█▉        | 11722/60896 [08:41<38:15, 21.43it/s]
Running loglikelihood requests:  19%|█▉        | 11754/60896 [08:43<38:14, 21.42it/s]
Running loglikelihood requests:  19%|█▉        | 11786/60896 [08:44<38:12, 21.42it/s]
Running loglikelihood requests:  19%|█▉        | 11819/60896 [08:46<37:49, 21.63it/s]
Running loglikelihood requests:  19%|█▉        | 11851/60896 [08:47<38:06, 21.45it/s]
Running loglikelihood requests:  20%|█▉        | 11886/60896 [08:49<37:03, 22.05it/s]
Running loglikelihood requests:  20%|█▉        | 11921/60896 [08:50<36:18, 22.48it/s]
Running loglikelihood requests:  20%|█▉        | 11968/60896 [08:52<32:21, 25.21it/s]
Running loglikelihood requests:  20%|█▉        | 12000/60896 [08:53<33:49, 24.10it/s]
Running loglikelihood requests:  20%|█▉        | 12032/60896 [08:55<35:01, 23.25it/s]
Running loglikelihood requests:  20%|█▉        | 12064/60896 [08:56<35:48, 22.73it/s]
Running loglikelihood requests:  20%|█▉        | 12096/60896 [08:58<36:21, 22.37it/s]
Running loglikelihood requests:  20%|█▉        | 12133/60896 [08:59<35:09, 23.12it/s]
Running loglikelihood requests:  20%|█▉        | 12167/60896 [09:01<35:22, 22.96it/s]
Running loglikelihood requests:  20%|██        | 12199/60896 [09:02<36:02, 22.52it/s]
Running loglikelihood requests:  20%|██        | 12231/60896 [09:04<36:30, 22.22it/s]
Running loglikelihood requests:  20%|██        | 12263/60896 [09:05<36:49, 22.01it/s]
Running loglikelihood requests:  20%|██        | 12295/60896 [09:07<37:03, 21.86it/s]
Running loglikelihood requests:  20%|██        | 12327/60896 [09:08<37:24, 21.64it/s]
Running loglikelihood requests:  20%|██        | 12359/60896 [09:10<37:26, 21.61it/s]
Running loglikelihood requests:  20%|██        | 12391/60896 [09:11<37:28, 21.57it/s]
Running loglikelihood requests:  20%|██        | 12423/60896 [09:13<37:28, 21.56it/s]
Running loglikelihood requests:  20%|██        | 12455/60896 [09:14<37:28, 21.54it/s]
Running loglikelihood requests:  21%|██        | 12489/60896 [09:15<36:52, 21.88it/s]
Running loglikelihood requests:  21%|██        | 12521/60896 [09:17<37:02, 21.76it/s]
Running loglikelihood requests:  21%|██        | 12555/60896 [09:18<36:07, 22.30it/s]
Running loglikelihood requests:  21%|██        | 12593/60896 [09:20<34:16, 23.48it/s]
Running loglikelihood requests:  21%|██        | 12627/60896 [09:21<34:25, 23.37it/s]
Running loglikelihood requests:  21%|██        | 12662/60896 [09:23<34:01, 23.63it/s]
Running loglikelihood requests:  21%|██        | 12694/60896 [09:24<34:38, 23.19it/s]
Running loglikelihood requests:  21%|██        | 12726/60896 [09:26<35:05, 22.88it/s]
Running loglikelihood requests:  21%|██        | 12758/60896 [09:27<35:23, 22.67it/s]
Running loglikelihood requests:  21%|██        | 12791/60896 [09:29<35:22, 22.67it/s]
Running loglikelihood requests:  21%|██        | 12826/60896 [09:30<34:38, 23.13it/s]
Running loglikelihood requests:  21%|██        | 12858/60896 [09:31<35:03, 22.84it/s]
Running loglikelihood requests:  21%|██        | 12890/60896 [09:33<35:20, 22.64it/s]
Running loglikelihood requests:  21%|██        | 12922/60896 [09:34<35:32, 22.50it/s]
Running loglikelihood requests:  21%|██▏       | 12954/60896 [09:36<35:52, 22.28it/s]
Running loglikelihood requests:  21%|██▏       | 12986/60896 [09:37<35:54, 22.24it/s]
Running loglikelihood requests:  21%|██▏       | 13018/60896 [09:39<35:55, 22.22it/s]
Running loglikelihood requests:  21%|██▏       | 13050/60896 [09:40<35:56, 22.19it/s]
Running loglikelihood requests:  21%|██▏       | 13085/60896 [09:42<34:57, 22.80it/s]
Running loglikelihood requests:  22%|██▏       | 13119/60896 [09:43<34:42, 22.94it/s]
Running loglikelihood requests:  22%|██▏       | 13155/60896 [09:44<33:48, 23.54it/s]
Running loglikelihood requests:  22%|██▏       | 13187/60896 [09:46<34:23, 23.12it/s]
Running loglikelihood requests:  22%|██▏       | 13226/60896 [09:47<32:42, 24.29it/s]
Running loglikelihood requests:  22%|██▏       | 13260/60896 [09:49<32:57, 24.09it/s]
Running loglikelihood requests:  22%|██▏       | 13294/60896 [09:50<33:16, 23.84it/s]
Running loglikelihood requests:  22%|██▏       | 13326/60896 [09:52<33:55, 23.37it/s]
Running loglikelihood requests:  22%|██▏       | 13358/60896 [09:53<34:24, 23.03it/s]
Running loglikelihood requests:  22%|██▏       | 13390/60896 [09:55<34:44, 22.79it/s]
Running loglikelihood requests:  22%|██▏       | 13424/60896 [09:56<34:21, 23.03it/s]
Running loglikelihood requests:  22%|██▏       | 13456/60896 [09:58<34:53, 22.66it/s]
Running loglikelihood requests:  22%|██▏       | 13490/60896 [09:59<34:25, 22.95it/s]
Running loglikelihood requests:  22%|██▏       | 13522/60896 [10:00<34:43, 22.74it/s]
Running loglikelihood requests:  22%|██▏       | 13556/60896 [10:02<34:18, 23.00it/s]
Running loglikelihood requests:  22%|██▏       | 13588/60896 [10:03<34:46, 22.68it/s]
Running loglikelihood requests:  22%|██▏       | 13620/60896 [10:05<34:56, 22.55it/s]
Running loglikelihood requests:  22%|██▏       | 13653/60896 [10:06<34:44, 22.66it/s]
Running loglikelihood requests:  22%|██▏       | 13685/60896 [10:08<34:55, 22.53it/s]
Running loglikelihood requests:  23%|██▎       | 13717/60896 [10:09<35:01, 22.45it/s]
Running loglikelihood requests:  23%|██▎       | 13750/60896 [10:11<34:55, 22.50it/s]
Running loglikelihood requests:  23%|██▎       | 13783/60896 [10:12<34:42, 22.62it/s]
Running loglikelihood requests:  23%|██▎       | 13815/60896 [10:13<34:52, 22.50it/s]
Running loglikelihood requests:  23%|██▎       | 13849/60896 [10:15<34:20, 22.83it/s]
Running loglikelihood requests:  23%|██▎       | 13882/60896 [10:16<34:24, 22.77it/s]
Running loglikelihood requests:  23%|██▎       | 13927/60896 [10:18<30:53, 25.35it/s]
Running loglikelihood requests:  23%|██▎       | 13962/60896 [10:19<31:10, 25.09it/s]
Running loglikelihood requests:  23%|██▎       | 13994/60896 [10:21<32:11, 24.28it/s]
Running loglikelihood requests:  23%|██▎       | 14029/60896 [10:22<32:05, 24.34it/s]
Running loglikelihood requests:  23%|██▎       | 14061/60896 [10:23<32:58, 23.67it/s]
Running loglikelihood requests:  23%|██▎       | 14093/60896 [10:25<33:30, 23.28it/s]
Running loglikelihood requests:  23%|██▎       | 14125/60896 [10:26<33:52, 23.02it/s]
Running loglikelihood requests:  23%|██▎       | 14161/60896 [10:28<32:54, 23.66it/s]
Running loglikelihood requests:  23%|██▎       | 14193/60896 [10:29<33:25, 23.28it/s]
Running loglikelihood requests:  23%|██▎       | 14225/60896 [10:31<34:02, 22.86it/s]
Running loglikelihood requests:  23%|██▎       | 14258/60896 [10:32<33:54, 22.93it/s]
Running loglikelihood requests:  23%|██▎       | 14290/60896 [10:33<34:07, 22.77it/s]
Running loglikelihood requests:  24%|██▎       | 14322/60896 [10:35<34:15, 22.65it/s]
Running loglikelihood requests:  24%|██▎       | 14354/60896 [10:36<34:21, 22.58it/s]
Running loglikelihood requests:  24%|██▎       | 14386/60896 [10:38<34:33, 22.43it/s]
Running loglikelihood requests:  24%|██▎       | 14418/60896 [10:39<34:32, 22.42it/s]
Running loglikelihood requests:  24%|██▎       | 14452/60896 [10:41<33:55, 22.82it/s]
Running loglikelihood requests:  24%|██▍       | 14485/60896 [10:42<33:46, 22.90it/s]
Running loglikelihood requests:  24%|██▍       | 14518/60896 [10:44<33:40, 22.96it/s]
Running loglikelihood requests:  24%|██▍       | 14552/60896 [10:45<33:24, 23.12it/s]
Running loglikelihood requests:  24%|██▍       | 14584/60896 [10:46<33:42, 22.90it/s]
Running loglikelihood requests:  24%|██▍       | 14616/60896 [10:48<33:55, 22.73it/s]
Running loglikelihood requests:  24%|██▍       | 14660/60896 [10:49<30:38, 25.15it/s]
Running loglikelihood requests:  24%|██▍       | 14692/60896 [10:51<31:37, 24.35it/s]
Running loglikelihood requests:  24%|██▍       | 14725/60896 [10:52<32:10, 23.91it/s]
Running loglikelihood requests:  24%|██▍       | 14757/60896 [10:54<32:45, 23.48it/s]
Running loglikelihood requests:  24%|██▍       | 14789/60896 [10:55<33:09, 23.18it/s]
Running loglikelihood requests:  24%|██▍       | 14821/60896 [10:56<33:26, 22.96it/s]
Running loglikelihood requests:  24%|██▍       | 14854/60896 [10:58<33:19, 23.02it/s]
Running loglikelihood requests:  24%|██▍       | 14886/60896 [10:59<33:44, 22.73it/s]
Running loglikelihood requests:  24%|██▍       | 14918/60896 [11:01<33:50, 22.65it/s]
Running loglikelihood requests:  25%|██▍       | 14950/60896 [11:02<33:53, 22.59it/s]
Running loglikelihood requests:  25%|██▍       | 14982/60896 [11:04<33:55, 22.55it/s]
Running loglikelihood requests:  25%|██▍       | 15014/60896 [11:05<34:00, 22.48it/s]
Running loglikelihood requests:  25%|██▍       | 15047/60896 [11:06<33:41, 22.68it/s]
Running loglikelihood requests:  25%|██▍       | 15079/60896 [11:08<33:46, 22.61it/s]
Running loglikelihood requests:  25%|██▍       | 15111/60896 [11:09<33:48, 22.57it/s]
Running loglikelihood requests:  25%|██▍       | 15143/60896 [11:11<34:00, 22.42it/s]
Running loglikelihood requests:  25%|██▍       | 15175/60896 [11:12<33:58, 22.43it/s]
Running loglikelihood requests:  25%|██▍       | 15207/60896 [11:14<33:56, 22.43it/s]
Running loglikelihood requests:  25%|██▌       | 15240/60896 [11:15<33:35, 22.65it/s]
Running loglikelihood requests:  25%|██▌       | 15272/60896 [11:16<33:39, 22.59it/s]
Running loglikelihood requests:  25%|██▌       | 15305/60896 [11:18<33:31, 22.66it/s]
Running loglikelihood requests:  25%|██▌       | 15338/60896 [11:19<33:17, 22.81it/s]
Running loglikelihood requests:  25%|██▌       | 15377/60896 [11:21<31:06, 24.38it/s]
Running loglikelihood requests:  25%|██▌       | 15411/60896 [11:22<31:01, 24.44it/s]
Running loglikelihood requests:  25%|██▌       | 15446/60896 [11:23<30:48, 24.59it/s]
Running loglikelihood requests:  25%|██▌       | 15478/60896 [11:25<31:21, 24.14it/s]
Running loglikelihood requests:  25%|██▌       | 15511/60896 [11:26<31:27, 24.05it/s]
Running loglikelihood requests:  26%|██▌       | 15543/60896 [11:28<31:47, 23.77it/s]
Running loglikelihood requests:  26%|██▌       | 15575/60896 [11:29<32:01, 23.58it/s]
Running loglikelihood requests:  26%|██▌       | 15607/60896 [11:30<32:19, 23.36it/s]
Running loglikelihood requests:  26%|██▌       | 15642/60896 [11:32<31:31, 23.93it/s]
Running loglikelihood requests:  26%|██▌       | 15674/60896 [11:33<31:49, 23.69it/s]
Running loglikelihood requests:  26%|██▌       | 15706/60896 [11:35<32:01, 23.51it/s]
Running loglikelihood requests:  26%|██▌       | 15738/60896 [11:36<32:10, 23.39it/s]
Running loglikelihood requests:  26%|██▌       | 15770/60896 [11:37<32:24, 23.20it/s]
Running loglikelihood requests:  26%|██▌       | 15802/60896 [11:39<32:25, 23.18it/s]
Running loglikelihood requests:  26%|██▌       | 15834/60896 [11:40<32:25, 23.16it/s]
Running loglikelihood requests:  26%|██▌       | 15866/60896 [11:41<32:25, 23.15it/s]
Running loglikelihood requests:  26%|██▌       | 15898/60896 [11:43<32:24, 23.14it/s]
Running loglikelihood requests:  26%|██▌       | 15931/60896 [11:44<32:19, 23.19it/s]
Running loglikelihood requests:  26%|██▌       | 15966/60896 [11:46<31:26, 23.81it/s]
Running loglikelihood requests:  26%|██▋       | 16000/60896 [11:47<31:09, 24.02it/s]
Running loglikelihood requests:  26%|██▋       | 16033/60896 [11:48<31:11, 23.97it/s]
Running loglikelihood requests:  26%|██▋       | 16072/60896 [11:50<29:36, 25.22it/s]
Running loglikelihood requests:  26%|██▋       | 16105/60896 [11:51<30:18, 24.63it/s]
Running loglikelihood requests:  27%|██▋       | 16141/60896 [11:53<29:47, 25.04it/s]
Running loglikelihood requests:  27%|██▋       | 16173/60896 [11:54<30:30, 24.43it/s]
Running loglikelihood requests:  27%|██▋       | 16205/60896 [11:55<31:00, 24.02it/s]
Running loglikelihood requests:  27%|██▋       | 16237/60896 [11:57<31:21, 23.74it/s]
Running loglikelihood requests:  27%|██▋       | 16269/60896 [11:58<31:43, 23.45it/s]
Running loglikelihood requests:  27%|██▋       | 16305/60896 [12:00<30:43, 24.18it/s]
Running loglikelihood requests:  27%|██▋       | 16342/60896 [12:01<29:45, 24.95it/s]
Running loglikelihood requests:  27%|██▋       | 16374/60896 [12:02<30:26, 24.37it/s]
Running loglikelihood requests:  27%|██▋       | 16406/60896 [12:04<30:56, 23.96it/s]
Running loglikelihood requests:  27%|██▋       | 16438/60896 [12:05<31:24, 23.59it/s]
Running loglikelihood requests:  27%|██▋       | 16470/60896 [12:07<31:35, 23.44it/s]
Running loglikelihood requests:  27%|██▋       | 16502/60896 [12:08<31:42, 23.33it/s]
Running loglikelihood requests:  27%|██▋       | 16534/60896 [12:09<31:47, 23.25it/s]
Running loglikelihood requests:  27%|██▋       | 16566/60896 [12:11<31:50, 23.21it/s]
Running loglikelihood requests:  27%|██▋       | 16598/60896 [12:12<31:52, 23.16it/s]
Running loglikelihood requests:  27%|██▋       | 16630/60896 [12:14<32:05, 22.99it/s]
Running loglikelihood requests:  27%|██▋       | 16663/60896 [12:15<31:45, 23.21it/s]
Running loglikelihood requests:  27%|██▋       | 16696/60896 [12:16<31:31, 23.37it/s]
Running loglikelihood requests:  27%|██▋       | 16730/60896 [12:18<31:01, 23.72it/s]
Running loglikelihood requests:  28%|██▊       | 16766/60896 [12:19<29:42, 24.75it/s]
Running loglikelihood requests:  28%|██▊       | 16806/60896 [12:20<28:02, 26.21it/s]
Running loglikelihood requests:  28%|██▊       | 16838/60896 [12:22<28:38, 25.63it/s]
Running loglikelihood requests:  28%|██▊       | 16872/60896 [12:23<28:34, 25.67it/s]
Running loglikelihood requests:  28%|██▊       | 16904/60896 [12:24<29:01, 25.26it/s]
Running loglikelihood requests:  28%|██▊       | 16936/60896 [12:26<29:20, 24.97it/s]
Running loglikelihood requests:  28%|██▊       | 16968/60896 [12:27<29:34, 24.75it/s]
Running loglikelihood requests:  28%|██▊       | 17004/60896 [12:28<28:48, 25.39it/s]
Running loglikelihood requests:  28%|██▊       | 17038/60896 [12:30<28:39, 25.51it/s]
Running loglikelihood requests:  28%|██▊       | 17070/60896 [12:31<29:03, 25.13it/s]
Running loglikelihood requests:  28%|██▊       | 17102/60896 [12:32<29:20, 24.87it/s]
Running loglikelihood requests:  28%|██▊       | 17134/60896 [12:34<29:32, 24.69it/s]
Running loglikelihood requests:  28%|██▊       | 17166/60896 [12:35<29:49, 24.44it/s]
Running loglikelihood requests:  28%|██▊       | 17199/60896 [12:36<29:35, 24.61it/s]
Running loglikelihood requests:  28%|██▊       | 17231/60896 [12:38<29:42, 24.50it/s]
Running loglikelihood requests:  28%|██▊       | 17263/60896 [12:39<29:46, 24.43it/s]
Running loglikelihood requests:  28%|██▊       | 17295/60896 [12:40<29:48, 24.38it/s]
Running loglikelihood requests:  28%|██▊       | 17328/60896 [12:42<29:41, 24.45it/s]
Running loglikelihood requests:  29%|██▊       | 17361/60896 [12:43<29:28, 24.62it/s]
Running loglikelihood requests:  29%|██▊       | 17394/60896 [12:44<29:18, 24.73it/s]
Running loglikelihood requests:  29%|██▊       | 17426/60896 [12:45<29:28, 24.58it/s]
Running loglikelihood requests:  29%|██▊       | 17464/60896 [12:47<27:59, 25.86it/s]
Running loglikelihood requests:  29%|██▊       | 17502/60896 [12:48<27:12, 26.58it/s]
Running loglikelihood requests:  29%|██▉       | 17534/60896 [12:49<27:54, 25.90it/s]
Running loglikelihood requests:  29%|██▉       | 17567/60896 [12:51<28:08, 25.66it/s]
Running loglikelihood requests:  29%|██▉       | 17599/60896 [12:52<28:33, 25.26it/s]
Running loglikelihood requests:  29%|██▉       | 17631/60896 [12:53<29:01, 24.85it/s]
Running loglikelihood requests:  29%|██▉       | 17663/60896 [12:55<29:10, 24.70it/s]
Running loglikelihood requests:  29%|██▉       | 17696/60896 [12:56<29:01, 24.81it/s]
Running loglikelihood requests:  29%|██▉       | 17728/60896 [12:57<29:10, 24.66it/s]
Running loglikelihood requests:  29%|██▉       | 17761/60896 [12:59<29:00, 24.78it/s]
Running loglikelihood requests:  29%|██▉       | 17793/60896 [13:00<29:14, 24.56it/s]
Running loglikelihood requests:  29%|██▉       | 17825/60896 [13:01<29:18, 24.49it/s]
Running loglikelihood requests:  29%|██▉       | 17857/60896 [13:03<29:20, 24.44it/s]
Running loglikelihood requests:  29%|██▉       | 17890/60896 [13:04<29:05, 24.63it/s]
Running loglikelihood requests:  29%|██▉       | 17922/60896 [13:05<29:11, 24.54it/s]
Running loglikelihood requests:  29%|██▉       | 17954/60896 [13:07<29:14, 24.47it/s]
Running loglikelihood requests:  30%|██▉       | 17986/60896 [13:08<29:28, 24.26it/s]
Running loglikelihood requests:  30%|██▉       | 18018/60896 [13:09<29:26, 24.28it/s]
Running loglikelihood requests:  30%|██▉       | 18054/60896 [13:11<28:20, 25.20it/s]
Running loglikelihood requests:  30%|██▉       | 18089/60896 [13:12<27:51, 25.62it/s]
Running loglikelihood requests:  30%|██▉       | 18122/60896 [13:13<28:00, 25.45it/s]
Running loglikelihood requests:  30%|██▉       | 18162/60896 [13:15<26:21, 27.02it/s]
Running loglikelihood requests:  30%|██▉       | 18198/60896 [13:16<26:01, 27.34it/s]
Running loglikelihood requests:  30%|██▉       | 18232/60896 [13:17<26:13, 27.11it/s]
Running loglikelihood requests:  30%|██▉       | 18265/60896 [13:18<26:36, 26.71it/s]
Running loglikelihood requests:  30%|███       | 18297/60896 [13:20<27:05, 26.20it/s]
Running loglikelihood requests:  30%|███       | 18329/60896 [13:21<27:27, 25.84it/s]
Running loglikelihood requests:  30%|███       | 18361/60896 [13:22<27:48, 25.49it/s]
Running loglikelihood requests:  30%|███       | 18395/60896 [13:23<27:26, 25.81it/s]
Running loglikelihood requests:  30%|███       | 18427/60896 [13:25<27:40, 25.57it/s]
Running loglikelihood requests:  30%|███       | 18462/60896 [13:26<27:05, 26.10it/s]
Running loglikelihood requests:  30%|███       | 18494/60896 [13:27<27:25, 25.76it/s]
Running loglikelihood requests:  30%|███       | 18526/60896 [13:29<27:49, 25.38it/s]
Running loglikelihood requests:  30%|███       | 18558/60896 [13:30<27:55, 25.27it/s]
Running loglikelihood requests:  31%|███       | 18591/60896 [13:31<27:44, 25.42it/s]
Running loglikelihood requests:  31%|███       | 18623/60896 [13:32<27:51, 25.29it/s]
Running loglikelihood requests:  31%|███       | 18655/60896 [13:34<27:55, 25.21it/s]
Running loglikelihood requests:  31%|███       | 18688/60896 [13:35<27:42, 25.38it/s]
Running loglikelihood requests:  31%|███       | 18721/60896 [13:36<27:47, 25.29it/s]
Running loglikelihood requests:  31%|███       | 18753/60896 [13:38<27:51, 25.21it/s]
Running loglikelihood requests:  31%|███       | 18788/60896 [13:39<27:09, 25.85it/s]
Running loglikelihood requests:  31%|███       | 18820/60896 [13:40<27:23, 25.60it/s]
Running loglikelihood requests:  31%|███       | 18854/60896 [13:41<27:04, 25.88it/s]
Running loglikelihood requests:  31%|███       | 18890/60896 [13:43<26:20, 26.57it/s]
Running loglikelihood requests:  31%|███       | 18931/60896 [13:44<24:54, 28.09it/s]
Running loglikelihood requests:  31%|███       | 18963/60896 [13:45<25:42, 27.19it/s]
Running loglikelihood requests:  31%|███       | 18999/60896 [13:47<25:23, 27.50it/s]
Running loglikelihood requests:  31%|███▏      | 19031/60896 [13:48<26:03, 26.78it/s]
Running loglikelihood requests:  31%|███▏      | 19063/60896 [13:49<26:32, 26.27it/s]
Running loglikelihood requests:  31%|███▏      | 19095/60896 [13:50<26:58, 25.83it/s]
Running loglikelihood requests:  31%|███▏      | 19131/60896 [13:52<26:13, 26.54it/s]
Running loglikelihood requests:  31%|███▏      | 19163/60896 [13:53<26:38, 26.11it/s]
Running loglikelihood requests:  32%|███▏      | 19196/60896 [13:54<26:41, 26.03it/s]
Running loglikelihood requests:  32%|███▏      | 19228/60896 [13:56<26:58, 25.75it/s]
Running loglikelihood requests:  32%|███▏      | 19260/60896 [13:57<27:16, 25.44it/s]
Running loglikelihood requests:  32%|███▏      | 19292/60896 [13:58<27:22, 25.33it/s]
Running loglikelihood requests:  32%|███▏      | 19324/60896 [13:59<27:25, 25.26it/s]
Running loglikelihood requests:  32%|███▏      | 19356/60896 [14:01<27:27, 25.21it/s]
Running loglikelihood requests:  32%|███▏      | 19388/60896 [14:02<27:29, 25.17it/s]
Running loglikelihood requests:  32%|███▏      | 19420/60896 [14:03<27:38, 25.02it/s]
Running loglikelihood requests:  32%|███▏      | 19452/60896 [14:05<27:35, 25.04it/s]
Running loglikelihood requests:  32%|███▏      | 19484/60896 [14:06<27:32, 25.05it/s]
Running loglikelihood requests:  32%|███▏      | 19518/60896 [14:07<27:00, 25.53it/s]
Running loglikelihood requests:  32%|███▏      | 19552/60896 [14:08<26:50, 25.67it/s]
Running loglikelihood requests:  32%|███▏      | 19587/60896 [14:10<26:16, 26.20it/s]
Running loglikelihood requests:  32%|███▏      | 19619/60896 [14:11<26:35, 25.86it/s]
Running loglikelihood requests:  32%|███▏      | 19661/60896 [14:12<24:31, 28.02it/s]
Running loglikelihood requests:  32%|███▏      | 19695/60896 [14:13<24:49, 27.66it/s]
Running loglikelihood requests:  32%|███▏      | 19727/60896 [14:15<25:36, 26.80it/s]
Running loglikelihood requests:  32%|███▏      | 19760/60896 [14:16<25:48, 26.56it/s]
Running loglikelihood requests:  33%|███▎      | 19795/60896 [14:17<25:29, 26.87it/s]
Running loglikelihood requests:  33%|███▎      | 19827/60896 [14:19<25:57, 26.38it/s]
Running loglikelihood requests:  33%|███▎      | 19859/60896 [14:20<26:16, 26.03it/s]
Running loglikelihood requests:  33%|███▎      | 19892/60896 [14:21<26:26, 25.85it/s]
Running loglikelihood requests:  33%|███▎      | 19925/60896 [14:22<26:22, 25.89it/s]
Running loglikelihood requests:  33%|███▎      | 19960/60896 [14:24<25:50, 26.40it/s]
Running loglikelihood requests:  33%|███▎      | 19992/60896 [14:25<26:10, 26.04it/s]
Running loglikelihood requests:  33%|███▎      | 20024/60896 [14:26<26:24, 25.79it/s]
Running loglikelihood requests:  33%|███▎      | 20056/60896 [14:27<26:34, 25.62it/s]
Running loglikelihood requests:  33%|███▎      | 20088/60896 [14:29<26:48, 25.38it/s]
Running loglikelihood requests:  33%|███▎      | 20121/60896 [14:30<26:34, 25.57it/s]
Running loglikelihood requests:  33%|███▎      | 20153/60896 [14:31<26:40, 25.46it/s]
Running loglikelihood requests:  33%|███▎      | 20185/60896 [14:33<26:43, 25.39it/s]
Running loglikelihood requests:  33%|███▎      | 20218/60896 [14:34<26:30, 25.58it/s]
Running loglikelihood requests:  33%|███▎      | 20251/60896 [14:35<26:30, 25.55it/s]
Running loglikelihood requests:  33%|███▎      | 20284/60896 [14:36<26:21, 25.68it/s]
Running loglikelihood requests:  33%|███▎      | 20318/60896 [14:38<25:59, 26.01it/s]
Running loglikelihood requests:  33%|███▎      | 20350/60896 [14:39<26:12, 25.78it/s]
Running loglikelihood requests:  33%|███▎      | 20385/60896 [14:40<25:43, 26.24it/s]
Running loglikelihood requests:  34%|███▎      | 20424/60896 [14:41<24:29, 27.54it/s]
Running loglikelihood requests:  34%|███▎      | 20456/60896 [14:43<25:11, 26.76it/s]
Running loglikelihood requests:  34%|███▎      | 20488/60896 [14:44<25:37, 26.28it/s]
Running loglikelihood requests:  34%|███▎      | 20522/60896 [14:45<25:25, 26.46it/s]
Running loglikelihood requests:  34%|███▍      | 20554/60896 [14:47<25:51, 26.00it/s]
Running loglikelihood requests:  34%|███▍      | 20586/60896 [14:48<26:02, 25.79it/s]
Running loglikelihood requests:  34%|███▍      | 20618/60896 [14:49<26:10, 25.64it/s]
Running loglikelihood requests:  34%|███▍      | 20653/60896 [14:50<25:32, 26.25it/s]
Running loglikelihood requests:  34%|███▍      | 20685/60896 [14:52<25:48, 25.97it/s]
Running loglikelihood requests:  34%|███▍      | 20722/60896 [14:53<24:57, 26.82it/s]
Running loglikelihood requests:  34%|███▍      | 20755/60896 [14:54<25:08, 26.60it/s]
Running loglikelihood requests:  34%|███▍      | 20787/60896 [14:55<25:30, 26.21it/s]
Running loglikelihood requests:  34%|███▍      | 20819/60896 [14:57<25:44, 25.94it/s]
Running loglikelihood requests:  34%|███▍      | 20851/60896 [14:58<25:55, 25.75it/s]
Running loglikelihood requests:  34%|███▍      | 20884/60896 [14:59<25:56, 25.70it/s]
Running loglikelihood requests:  34%|███▍      | 20916/60896 [15:01<26:02, 25.58it/s]
Running loglikelihood requests:  34%|███▍      | 20948/60896 [15:02<26:06, 25.50it/s]
Running loglikelihood requests:  34%|███▍      | 20980/60896 [15:03<26:09, 25.44it/s]
Running loglikelihood requests:  35%|███▍      | 21012/60896 [15:04<26:10, 25.40it/s]
Running loglikelihood requests:  35%|███▍      | 21048/60896 [15:06<25:18, 26.25it/s]
Running loglikelihood requests:  35%|███▍      | 21080/60896 [15:07<25:33, 25.96it/s]
Running loglikelihood requests:  35%|███▍      | 21113/60896 [15:08<25:30, 26.00it/s]
Running loglikelihood requests:  35%|███▍      | 21145/60896 [15:09<25:41, 25.79it/s]
Running loglikelihood requests:  35%|███▍      | 21186/60896 [15:11<23:35, 28.06it/s]
Running loglikelihood requests:  35%|███▍      | 21219/60896 [15:12<23:49, 27.75it/s]
Running loglikelihood requests:  35%|███▍      | 21252/60896 [15:13<24:09, 27.34it/s]
Running loglikelihood requests:  35%|███▍      | 21287/60896 [15:14<23:48, 27.73it/s]
Running loglikelihood requests:  35%|███▌      | 21319/60896 [15:16<24:10, 27.28it/s]
Running loglikelihood requests:  35%|███▌      | 21351/60896 [15:17<24:26, 26.96it/s]
Running loglikelihood requests:  35%|███▌      | 21388/60896 [15:18<23:33, 27.96it/s]
Running loglikelihood requests:  35%|███▌      | 21420/60896 [15:19<24:09, 27.23it/s]
Running loglikelihood requests:  35%|███▌      | 21455/60896 [15:20<23:46, 27.65it/s]
Running loglikelihood requests:  35%|███▌      | 21487/60896 [15:22<24:07, 27.22it/s]
Running loglikelihood requests:  35%|███▌      | 21519/60896 [15:23<24:22, 26.92it/s]
Running loglikelihood requests:  35%|███▌      | 21551/60896 [15:24<24:33, 26.70it/s]
Running loglikelihood requests:  35%|███▌      | 21583/60896 [15:25<24:40, 26.56it/s]
Running loglikelihood requests:  35%|███▌      | 21615/60896 [15:27<24:52, 26.32it/s]
Running loglikelihood requests:  36%|███▌      | 21647/60896 [15:28<24:52, 26.29it/s]
Running loglikelihood requests:  36%|███▌      | 21679/60896 [15:29<24:53, 26.27it/s]
Running loglikelihood requests:  36%|███▌      | 21713/60896 [15:30<24:25, 26.74it/s]
Running loglikelihood requests:  36%|███▌      | 21745/60896 [15:31<24:33, 26.57it/s]
Running loglikelihood requests:  36%|███▌      | 21780/60896 [15:33<23:58, 27.20it/s]
Running loglikelihood requests:  36%|███▌      | 21813/60896 [15:34<23:59, 27.15it/s]
Running loglikelihood requests:  36%|███▌      | 21845/60896 [15:35<24:20, 26.74it/s]
Running loglikelihood requests:  36%|███▌      | 21881/60896 [15:36<23:34, 27.59it/s]
Running loglikelihood requests:  36%|███▌      | 21918/60896 [15:38<22:50, 28.43it/s]
Running loglikelihood requests:  36%|███▌      | 21950/60896 [15:39<23:20, 27.80it/s]
Running loglikelihood requests:  36%|███▌      | 21984/60896 [15:40<23:17, 27.85it/s]
Running loglikelihood requests:  36%|███▌      | 22016/60896 [15:41<23:45, 27.27it/s]
Running loglikelihood requests:  36%|███▌      | 22048/60896 [15:42<23:59, 26.98it/s]
Running loglikelihood requests:  36%|███▋      | 22080/60896 [15:44<24:09, 26.78it/s]
Running loglikelihood requests:  36%|███▋      | 22114/60896 [15:45<23:49, 27.14it/s]
Running loglikelihood requests:  36%|███▋      | 22146/60896 [15:46<24:01, 26.89it/s]
Running loglikelihood requests:  36%|███▋      | 22179/60896 [15:47<24:03, 26.83it/s]
Running loglikelihood requests:  36%|███▋      | 22211/60896 [15:49<24:10, 26.68it/s]
Running loglikelihood requests:  37%|███▋      | 22243/60896 [15:50<24:14, 26.57it/s]
Running loglikelihood requests:  37%|███▋      | 22275/60896 [15:51<24:18, 26.49it/s]
Running loglikelihood requests:  37%|███▋      | 22307/60896 [15:52<24:19, 26.44it/s]
Running loglikelihood requests:  37%|███▋      | 22339/60896 [15:53<24:20, 26.40it/s]
Running loglikelihood requests:  37%|███▋      | 22371/60896 [15:55<24:25, 26.28it/s]
Running loglikelihood requests:  37%|███▋      | 22403/60896 [15:56<24:24, 26.29it/s]
Running loglikelihood requests:  37%|███▋      | 22435/60896 [15:57<24:22, 26.30it/s]
Running loglikelihood requests:  37%|███▋      | 22467/60896 [15:58<24:20, 26.30it/s]
Running loglikelihood requests:  37%|███▋      | 22500/60896 [16:00<24:05, 26.56it/s]
Running loglikelihood requests:  37%|███▋      | 22534/60896 [16:01<23:41, 26.98it/s]
Running loglikelihood requests:  37%|███▋      | 22566/60896 [16:02<23:51, 26.78it/s]
Running loglikelihood requests:  37%|███▋      | 22610/60896 [16:03<21:42, 29.38it/s]
Running loglikelihood requests:  37%|███▋      | 22642/60896 [16:04<22:22, 28.50it/s]
Running loglikelihood requests:  37%|███▋      | 22676/60896 [16:06<22:27, 28.37it/s]
Running loglikelihood requests:  37%|███▋      | 22708/60896 [16:07<22:54, 27.78it/s]
Running loglikelihood requests:  37%|███▋      | 22740/60896 [16:08<23:13, 27.38it/s]
Running loglikelihood requests:  37%|███▋      | 22772/60896 [16:09<23:37, 26.90it/s]
Running loglikelihood requests:  37%|███▋      | 22804/60896 [16:11<23:43, 26.76it/s]
Running loglikelihood requests:  38%|███▊      | 22838/60896 [16:12<23:22, 27.14it/s]
Running loglikelihood requests:  38%|███▊      | 22870/60896 [16:13<23:32, 26.92it/s]
Running loglikelihood requests:  38%|███▊      | 22902/60896 [16:14<23:39, 26.77it/s]
Running loglikelihood requests:  38%|███▊      | 22934/60896 [16:15<23:44, 26.66it/s]
Running loglikelihood requests:  38%|███▊      | 22966/60896 [16:17<23:55, 26.42it/s]
Running loglikelihood requests:  38%|███▊      | 22998/60896 [16:18<23:54, 26.42it/s]
Running loglikelihood requests:  38%|███▊      | 23030/60896 [16:19<23:53, 26.41it/s]
Running loglikelihood requests:  38%|███▊      | 23062/60896 [16:20<23:52, 26.41it/s]
Running loglikelihood requests:  38%|███▊      | 23094/60896 [16:21<23:51, 26.41it/s]
Running loglikelihood requests:  38%|███▊      | 23127/60896 [16:23<23:36, 26.66it/s]
Running loglikelihood requests:  38%|███▊      | 23159/60896 [16:24<23:39, 26.58it/s]
Running loglikelihood requests:  38%|███▊      | 23192/60896 [16:25<23:39, 26.56it/s]
Running loglikelihood requests:  38%|███▊      | 23224/60896 [16:26<23:40, 26.52it/s]
Running loglikelihood requests:  38%|███▊      | 23258/60896 [16:28<23:15, 26.98it/s]
Running loglikelihood requests:  38%|███▊      | 23291/60896 [16:29<23:08, 27.08it/s]
Running loglikelihood requests:  38%|███▊      | 23329/60896 [16:30<22:03, 28.39it/s]
Running loglikelihood requests:  38%|███▊      | 23361/60896 [16:31<22:29, 27.82it/s]
Running loglikelihood requests:  38%|███▊      | 23397/60896 [16:32<22:07, 28.26it/s]
Running loglikelihood requests:  38%|███▊      | 23429/60896 [16:34<22:30, 27.74it/s]
Running loglikelihood requests:  39%|███▊      | 23462/60896 [16:35<22:35, 27.62it/s]
Running loglikelihood requests:  39%|███▊      | 23494/60896 [16:36<22:51, 27.28it/s]
Running loglikelihood requests:  39%|███▊      | 23526/60896 [16:37<23:01, 27.05it/s]
Running loglikelihood requests:  39%|███▊      | 23561/60896 [16:38<22:40, 27.44it/s]
Running loglikelihood requests:  39%|███▊      | 23593/60896 [16:40<22:53, 27.17it/s]
Running loglikelihood requests:  39%|███▉      | 23628/60896 [16:41<22:25, 27.70it/s]
Running loglikelihood requests:  39%|███▉      | 23660/60896 [16:42<22:41, 27.34it/s]
Running loglikelihood requests:  39%|███▉      | 23692/60896 [16:43<22:53, 27.09it/s]
Running loglikelihood requests:  39%|███▉      | 23724/60896 [16:44<23:01, 26.91it/s]
Running loglikelihood requests:  39%|███▉      | 23756/60896 [16:46<23:12, 26.67it/s]
Running loglikelihood requests:  39%|███▉      | 23788/60896 [16:47<23:14, 26.62it/s]
Running loglikelihood requests:  39%|███▉      | 23820/60896 [16:48<23:14, 26.59it/s]
Running loglikelihood requests:  39%|███▉      | 23852/60896 [16:49<23:14, 26.56it/s]
Running loglikelihood requests:  39%|███▉      | 23886/60896 [16:51<22:48, 27.04it/s]
Running loglikelihood requests:  39%|███▉      | 23920/60896 [16:52<22:30, 27.37it/s]
Running loglikelihood requests:  39%|███▉      | 23952/60896 [16:53<22:50, 26.96it/s]
Running loglikelihood requests:  39%|███▉      | 23988/60896 [16:54<22:07, 27.81it/s]
Running loglikelihood requests:  39%|███▉      | 24020/60896 [16:55<22:24, 27.42it/s]
Running loglikelihood requests:  40%|███▉      | 24059/60896 [16:57<21:12, 28.95it/s]
Running loglikelihood requests:  40%|███▉      | 24092/60896 [16:58<21:29, 28.55it/s]
Running loglikelihood requests:  40%|███▉      | 24129/60896 [16:59<20:56, 29.25it/s]
Running loglikelihood requests:  40%|███▉      | 24161/60896 [17:00<21:35, 28.35it/s]
Running loglikelihood requests:  40%|███▉      | 24193/60896 [17:01<21:56, 27.87it/s]
Running loglikelihood requests:  40%|███▉      | 24225/60896 [17:03<22:11, 27.53it/s]
Running loglikelihood requests:  40%|███▉      | 24261/60896 [17:04<21:34, 28.29it/s]
Running loglikelihood requests:  40%|███▉      | 24293/60896 [17:05<21:55, 27.83it/s]
Running loglikelihood requests:  40%|███▉      | 24325/60896 [17:06<22:10, 27.49it/s]
Running loglikelihood requests:  40%|███▉      | 24357/60896 [17:07<22:31, 27.04it/s]
Running loglikelihood requests:  40%|████      | 24389/60896 [17:09<22:34, 26.95it/s]
Running loglikelihood requests:  40%|████      | 24421/60896 [17:10<22:36, 26.89it/s]
Running loglikelihood requests:  40%|████      | 24453/60896 [17:11<22:37, 26.84it/s]
Running loglikelihood requests:  40%|████      | 24485/60896 [17:12<22:38, 26.81it/s]
Running loglikelihood requests:  40%|████      | 24517/60896 [17:13<22:46, 26.62it/s]
Running loglikelihood requests:  40%|████      | 24551/60896 [17:15<22:18, 27.15it/s]
Running loglikelihood requests:  40%|████      | 24583/60896 [17:16<22:23, 27.03it/s]
Running loglikelihood requests:  40%|████      | 24615/60896 [17:17<22:26, 26.94it/s]
Running loglikelihood requests:  40%|████      | 24648/60896 [17:18<22:16, 27.12it/s]
Running loglikelihood requests:  41%|████      | 24684/60896 [17:19<21:31, 28.03it/s]
Running loglikelihood requests:  41%|████      | 24719/60896 [17:21<21:21, 28.24it/s]
Running loglikelihood requests:  41%|████      | 24751/60896 [17:22<21:39, 27.82it/s]
Running loglikelihood requests:  41%|████      | 24783/60896 [17:23<21:52, 27.52it/s]
Running loglikelihood requests:  41%|████      | 24815/60896 [17:24<22:00, 27.32it/s]
Running loglikelihood requests:  41%|████      | 24847/60896 [17:25<22:06, 27.17it/s]
Running loglikelihood requests:  41%|████      | 24879/60896 [17:27<22:10, 27.07it/s]
Running loglikelihood requests:  41%|████      | 24912/60896 [17:28<22:08, 27.09it/s]
Running loglikelihood requests:  41%|████      | 24945/60896 [17:29<21:58, 27.27it/s]
Running loglikelihood requests:  41%|████      | 24977/60896 [17:30<22:03, 27.14it/s]
Running loglikelihood requests:  41%|████      | 25009/60896 [17:31<22:06, 27.05it/s]
Running loglikelihood requests:  41%|████      | 25041/60896 [17:33<22:08, 26.99it/s]
Running loglikelihood requests:  41%|████      | 25073/60896 [17:34<22:09, 26.94it/s]
Running loglikelihood requests:  41%|████      | 25105/60896 [17:35<22:18, 26.74it/s]
Running loglikelihood requests:  41%|████▏     | 25137/60896 [17:36<22:15, 26.77it/s]
Running loglikelihood requests:  41%|████▏     | 25169/60896 [17:37<22:13, 26.79it/s]
Running loglikelihood requests:  41%|████▏     | 25205/60896 [17:39<21:23, 27.80it/s]
Running loglikelihood requests:  41%|████▏     | 25238/60896 [17:40<21:24, 27.76it/s]
Running loglikelihood requests:  41%|████▏     | 25270/60896 [17:41<21:36, 27.48it/s]
Running loglikelihood requests:  42%|████▏     | 25303/60896 [17:42<21:38, 27.40it/s]
Running loglikelihood requests:  42%|████▏     | 25344/60896 [17:43<20:04, 29.52it/s]
Running loglikelihood requests:  42%|████▏     | 25379/60896 [17:45<20:03, 29.50it/s]
Running loglikelihood requests:  42%|████▏     | 25416/60896 [17:46<19:43, 29.99it/s]
Running loglikelihood requests:  42%|████▏     | 25448/60896 [17:47<20:19, 29.07it/s]
Running loglikelihood requests:  42%|████▏     | 25480/60896 [17:48<20:45, 28.44it/s]
Running loglikelihood requests:  42%|████▏     | 25512/60896 [17:49<21:14, 27.77it/s]
Running loglikelihood requests:  42%|████▏     | 25545/60896 [17:51<21:12, 27.77it/s]
Running loglikelihood requests:  42%|████▏     | 25578/60896 [17:52<21:11, 27.77it/s]
Running loglikelihood requests:  42%|████▏     | 25610/60896 [17:53<21:22, 27.52it/s]
Running loglikelihood requests:  42%|████▏     | 25642/60896 [17:54<21:29, 27.34it/s]
Running loglikelihood requests:  42%|████▏     | 25674/60896 [17:55<21:42, 27.04it/s]
Running loglikelihood requests:  42%|████▏     | 25706/60896 [17:56<21:42, 27.01it/s]
Running loglikelihood requests:  42%|████▏     | 25738/60896 [17:58<21:42, 26.98it/s]
Running loglikelihood requests:  42%|████▏     | 25770/60896 [17:59<21:42, 26.97it/s]
Running loglikelihood requests:  42%|████▏     | 25803/60896 [18:00<21:29, 27.21it/s]
Running loglikelihood requests:  42%|████▏     | 25835/60896 [18:01<21:39, 26.97it/s]
Running loglikelihood requests:  42%|████▏     | 25870/60896 [18:02<21:03, 27.71it/s]
Running loglikelihood requests:  43%|████▎     | 25902/60896 [18:04<21:13, 27.48it/s]
Running loglikelihood requests:  43%|████▎     | 25938/60896 [18:05<20:33, 28.35it/s]
Running loglikelihood requests:  43%|████▎     | 25971/60896 [18:06<20:38, 28.20it/s]
Running loglikelihood requests:  43%|████▎     | 26006/60896 [18:07<20:19, 28.61it/s]
Running loglikelihood requests:  43%|████▎     | 26038/60896 [18:08<20:48, 27.91it/s]
Running loglikelihood requests:  43%|████▎     | 26070/60896 [18:10<20:59, 27.65it/s]
Running loglikelihood requests:  43%|████▎     | 26102/60896 [18:11<21:06, 27.47it/s]
Running loglikelihood requests:  43%|████▎     | 26142/60896 [18:12<19:44, 29.35it/s]
Running loglikelihood requests:  43%|████▎     | 26174/60896 [18:13<20:11, 28.65it/s]
Running loglikelihood requests:  43%|████▎     | 26206/60896 [18:14<20:40, 27.96it/s]
Running loglikelihood requests:  43%|████▎     | 26238/60896 [18:16<20:51, 27.68it/s]
Running loglikelihood requests:  43%|████▎     | 26270/60896 [18:17<20:59, 27.49it/s]
Running loglikelihood requests:  43%|████▎     | 26302/60896 [18:18<21:04, 27.35it/s]
Running loglikelihood requests:  43%|████▎     | 26334/60896 [18:19<21:08, 27.26it/s]
Running loglikelihood requests:  43%|████▎     | 26366/60896 [18:20<21:10, 27.19it/s]
Running loglikelihood requests:  43%|████▎     | 26398/60896 [18:21<21:20, 26.95it/s]
Running loglikelihood requests:  43%|████▎     | 26430/60896 [18:23<21:17, 26.97it/s]
Running loglikelihood requests:  43%|████▎     | 26463/60896 [18:24<21:03, 27.24it/s]
Running loglikelihood requests:  44%|████▎     | 26496/60896 [18:25<20:54, 27.43it/s]
Running loglikelihood requests:  44%|████▎     | 26529/60896 [18:26<20:33, 27.87it/s]
Running loglikelihood requests:  44%|████▎     | 26572/60896 [18:27<18:35, 30.76it/s]
Running loglikelihood requests:  44%|████▎     | 26604/60896 [18:28<19:14, 29.71it/s]
Running loglikelihood requests:  44%|████▎     | 26636/60896 [18:30<19:32, 29.23it/s]
Running loglikelihood requests:  44%|████▍     | 26669/60896 [18:31<19:34, 29.15it/s]
Running loglikelihood requests:  44%|████▍     | 26701/60896 [18:32<19:46, 28.83it/s]
Running loglikelihood requests:  44%|████▍     | 26733/60896 [18:33<19:54, 28.60it/s]
Running loglikelihood requests:  44%|████▍     | 26767/60896 [18:34<19:38, 28.97it/s]
Running loglikelihood requests:  44%|████▍     | 26801/60896 [18:35<19:36, 28.98it/s]
Running loglikelihood requests:  44%|████▍     | 26833/60896 [18:36<19:46, 28.71it/s]
Running loglikelihood requests:  44%|████▍     | 26865/60896 [18:38<19:53, 28.52it/s]
Running loglikelihood requests:  44%|████▍     | 26897/60896 [18:39<19:57, 28.39it/s]
Running loglikelihood requests:  44%|████▍     | 26929/60896 [18:40<20:00, 28.30it/s]
Running loglikelihood requests:  44%|████▍     | 26961/60896 [18:41<20:02, 28.22it/s]
Running loglikelihood requests:  44%|████▍     | 26995/60896 [18:42<19:46, 28.56it/s]
Running loglikelihood requests:  44%|████▍     | 27027/60896 [18:43<19:51, 28.42it/s]
Running loglikelihood requests:  44%|████▍     | 27059/60896 [18:44<19:54, 28.32it/s]
Running loglikelihood requests:  44%|████▍     | 27091/60896 [18:46<19:57, 28.24it/s]
Running loglikelihood requests:  45%|████▍     | 27124/60896 [18:47<19:45, 28.48it/s]
Running loglikelihood requests:  45%|████▍     | 27161/60896 [18:48<19:02, 29.53it/s]
Running loglikelihood requests:  45%|████▍     | 27194/60896 [18:49<19:06, 29.39it/s]
Running loglikelihood requests:  45%|████▍     | 27227/60896 [18:50<19:09, 29.29it/s]
Running loglikelihood requests:  45%|████▍     | 27259/60896 [18:51<19:21, 28.96it/s]
Running loglikelihood requests:  45%|████▍     | 27291/60896 [18:52<19:29, 28.73it/s]
Running loglikelihood requests:  45%|████▍     | 27324/60896 [18:54<19:24, 28.83it/s]
Running loglikelihood requests:  45%|████▍     | 27357/60896 [18:55<19:29, 28.68it/s]
Running loglikelihood requests:  45%|████▍     | 27390/60896 [18:56<19:23, 28.79it/s]
Running loglikelihood requests:  45%|████▌     | 27422/60896 [18:57<19:29, 28.61it/s]
Running loglikelihood requests:  45%|████▌     | 27454/60896 [18:58<19:34, 28.48it/s]
Running loglikelihood requests:  45%|████▌     | 27486/60896 [18:59<19:36, 28.39it/s]
Running loglikelihood requests:  45%|████▌     | 27518/60896 [19:00<19:47, 28.10it/s]
Running loglikelihood requests:  45%|████▌     | 27551/60896 [19:02<19:34, 28.38it/s]
Running loglikelihood requests:  45%|████▌     | 27583/60896 [19:03<19:35, 28.33it/s]
Running loglikelihood requests:  45%|████▌     | 27616/60896 [19:04<19:25, 28.55it/s]
Running loglikelihood requests:  45%|████▌     | 27650/60896 [19:05<19:07, 28.97it/s]
Running loglikelihood requests:  45%|████▌     | 27684/60896 [19:06<18:55, 29.25it/s]
Running loglikelihood requests:  46%|████▌     | 27726/60896 [19:07<17:37, 31.38it/s]
Running loglikelihood requests:  46%|████▌     | 27759/60896 [19:08<17:55, 30.80it/s]
Running loglikelihood requests:  46%|████▌     | 27792/60896 [19:10<18:09, 30.39it/s]
Running loglikelihood requests:  46%|████▌     | 27825/60896 [19:11<18:18, 30.10it/s]
Running loglikelihood requests:  46%|████▌     | 27857/60896 [19:12<18:34, 29.63it/s]
Running loglikelihood requests:  46%|████▌     | 27889/60896 [19:13<18:46, 29.30it/s]
Running loglikelihood requests:  46%|████▌     | 27922/60896 [19:14<18:49, 29.19it/s]
Running loglikelihood requests:  46%|████▌     | 27954/60896 [19:15<18:56, 28.99it/s]
Running loglikelihood requests:  46%|████▌     | 27986/60896 [19:16<19:00, 28.85it/s]
Running loglikelihood requests:  46%|████▌     | 28018/60896 [19:17<19:03, 28.75it/s]
Running loglikelihood requests:  46%|████▌     | 28050/60896 [19:19<19:05, 28.68it/s]
Running loglikelihood requests:  46%|████▌     | 28082/60896 [19:20<19:05, 28.64it/s]
Running loglikelihood requests:  46%|████▌     | 28114/60896 [19:21<19:11, 28.46it/s]
Running loglikelihood requests:  46%|████▌     | 28148/60896 [19:22<18:49, 28.99it/s]
Running loglikelihood requests:  46%|████▋     | 28181/60896 [19:23<18:43, 29.11it/s]
Running loglikelihood requests:  46%|████▋     | 28213/60896 [19:24<18:49, 28.93it/s]
Running loglikelihood requests:  46%|████▋     | 28247/60896 [19:25<18:31, 29.37it/s]
Running loglikelihood requests:  46%|████▋     | 28285/60896 [19:26<17:47, 30.54it/s]
Running loglikelihood requests:  47%|████▋     | 28317/60896 [19:28<18:06, 29.98it/s]
Running loglikelihood requests:  47%|████▋     | 28350/60896 [19:29<18:10, 29.85it/s]
Running loglikelihood requests:  47%|████▋     | 28383/60896 [19:30<18:12, 29.75it/s]
Running loglikelihood requests:  47%|████▋     | 28415/60896 [19:31<18:24, 29.42it/s]
Running loglikelihood requests:  47%|████▋     | 28449/60896 [19:32<18:12, 29.71it/s]
Running loglikelihood requests:  47%|████▋     | 28482/60896 [19:33<18:22, 29.39it/s]
Running loglikelihood requests:  47%|████▋     | 28514/60896 [19:34<18:30, 29.16it/s]
Running loglikelihood requests:  47%|████▋     | 28546/60896 [19:35<18:35, 29.01it/s]
Running loglikelihood requests:  47%|████▋     | 28578/60896 [19:37<18:38, 28.90it/s]
Running loglikelihood requests:  47%|████▋     | 28610/60896 [19:38<18:40, 28.83it/s]
Running loglikelihood requests:  47%|████▋     | 28642/60896 [19:39<18:40, 28.77it/s]
Running loglikelihood requests:  47%|████▋     | 28675/60896 [19:40<18:37, 28.84it/s]
Running loglikelihood requests:  47%|████▋     | 28708/60896 [19:41<18:28, 29.05it/s]
Running loglikelihood requests:  47%|████▋     | 28744/60896 [19:42<17:51, 30.00it/s]
Running loglikelihood requests:  47%|████▋     | 28782/60896 [19:43<17:05, 31.31it/s]
Running loglikelihood requests:  47%|████▋     | 28818/60896 [19:44<16:49, 31.77it/s]
Running loglikelihood requests:  47%|████▋     | 28853/60896 [19:45<16:52, 31.64it/s]
Running loglikelihood requests:  47%|████▋     | 28885/60896 [19:47<17:17, 30.86it/s]
Running loglikelihood requests:  47%|████▋     | 28917/60896 [19:48<17:34, 30.33it/s]
Running loglikelihood requests:  48%|████▊     | 28951/60896 [19:49<17:27, 30.49it/s]
Running loglikelihood requests:  48%|████▊     | 28983/60896 [19:50<17:41, 30.05it/s]
Running loglikelihood requests:  48%|████▊     | 29015/60896 [19:51<17:51, 29.77it/s]
Running loglikelihood requests:  48%|████▊     | 29048/60896 [19:52<17:56, 29.58it/s]
Running loglikelihood requests:  48%|████▊     | 29080/60896 [19:53<18:00, 29.44it/s]
Running loglikelihood requests:  48%|████▊     | 29112/60896 [19:54<18:03, 29.35it/s]
Running loglikelihood requests:  48%|████▊     | 29144/60896 [19:55<18:05, 29.24it/s]
Running loglikelihood requests:  48%|████▊     | 29180/60896 [19:56<17:26, 30.30it/s]
Running loglikelihood requests:  48%|████▊     | 29213/60896 [19:58<17:32, 30.09it/s]
Running loglikelihood requests:  48%|████▊     | 29246/60896 [19:59<17:32, 30.06it/s]
Running loglikelihood requests:  48%|████▊     | 29283/60896 [20:00<16:55, 31.14it/s]
Running loglikelihood requests:  48%|████▊     | 29316/60896 [20:01<17:03, 30.85it/s]
Running loglikelihood requests:  48%|████▊     | 29351/60896 [20:02<16:52, 31.17it/s]
Running loglikelihood requests:  48%|████▊     | 29387/60896 [20:03<16:34, 31.68it/s]
Running loglikelihood requests:  48%|████▊     | 29419/60896 [20:04<17:03, 30.76it/s]
Running loglikelihood requests:  48%|████▊     | 29451/60896 [20:05<17:18, 30.29it/s]
Running loglikelihood requests:  48%|████▊     | 29483/60896 [20:06<17:29, 29.94it/s]
Running loglikelihood requests:  48%|████▊     | 29517/60896 [20:08<17:17, 30.24it/s]
Running loglikelihood requests:  49%|████▊     | 29549/60896 [20:09<17:27, 29.92it/s]
Running loglikelihood requests:  49%|████▊     | 29581/60896 [20:10<17:41, 29.51it/s]
Running loglikelihood requests:  49%|████▊     | 29613/60896 [20:11<17:44, 29.38it/s]
Running loglikelihood requests:  49%|████▊     | 29645/60896 [20:12<17:47, 29.29it/s]
Running loglikelihood requests:  49%|████▊     | 29679/60896 [20:13<17:27, 29.79it/s]
Running loglikelihood requests:  49%|████▉     | 29711/60896 [20:14<17:34, 29.57it/s]
Running loglikelihood requests:  49%|████▉     | 29745/60896 [20:15<17:19, 29.96it/s]
Running loglikelihood requests:  49%|████▉     | 29786/60896 [20:16<16:13, 31.95it/s]
Running loglikelihood requests:  49%|████▉     | 29821/60896 [20:17<16:13, 31.92it/s]
Running loglikelihood requests:  49%|████▉     | 29853/60896 [20:19<16:38, 31.09it/s]
Running loglikelihood requests:  49%|████▉     | 29885/60896 [20:20<16:56, 30.51it/s]
Running loglikelihood requests:  49%|████▉     | 29917/60896 [20:21<17:08, 30.13it/s]
Running loglikelihood requests:  49%|████▉     | 29952/60896 [20:22<16:46, 30.73it/s]
Running loglikelihood requests:  49%|████▉     | 29984/60896 [20:23<17:06, 30.11it/s]
Running loglikelihood requests:  49%|████▉     | 30016/60896 [20:24<17:14, 29.85it/s]
Running loglikelihood requests:  49%|████▉     | 30048/60896 [20:25<17:19, 29.67it/s]
Running loglikelihood requests:  49%|████▉     | 30080/60896 [20:26<17:25, 29.49it/s]
Running loglikelihood requests:  49%|████▉     | 30114/60896 [20:27<17:07, 29.95it/s]
Running loglikelihood requests:  50%|████▉     | 30146/60896 [20:28<17:13, 29.75it/s]
Running loglikelihood requests:  50%|████▉     | 30182/60896 [20:30<16:48, 30.45it/s]
Running loglikelihood requests:  50%|████▉     | 30215/60896 [20:31<16:50, 30.35it/s]
Running loglikelihood requests:  50%|████▉     | 30253/60896 [20:32<16:11, 31.53it/s]
Running loglikelihood requests:  50%|████▉     | 30285/60896 [20:33<16:36, 30.71it/s]
Running loglikelihood requests:  50%|████▉     | 30317/60896 [20:34<16:54, 30.15it/s]
Running loglikelihood requests:  50%|████▉     | 30349/60896 [20:35<17:06, 29.76it/s]
Running loglikelihood requests:  50%|████▉     | 30383/60896 [20:36<17:01, 29.86it/s]
Running loglikelihood requests:  50%|████▉     | 30418/60896 [20:37<16:44, 30.36it/s]
Running loglikelihood requests:  50%|█████     | 30450/60896 [20:38<16:58, 29.90it/s]
Running loglikelihood requests:  50%|█████     | 30482/60896 [20:40<17:06, 29.62it/s]
Running loglikelihood requests:  50%|█████     | 30514/60896 [20:41<17:13, 29.40it/s]
Running loglikelihood requests:  50%|█████     | 30549/60896 [20:42<16:49, 30.06it/s]
Running loglikelihood requests:  50%|█████     | 30582/60896 [20:43<16:50, 29.99it/s]
Running loglikelihood requests:  50%|█████     | 30618/60896 [20:44<16:29, 30.60it/s]
Running loglikelihood requests:  50%|█████     | 30657/60896 [20:45<15:43, 32.04it/s]
Running loglikelihood requests:  50%|█████     | 30690/60896 [20:46<16:00, 31.43it/s]
Running loglikelihood requests:  50%|█████     | 30723/60896 [20:47<16:13, 31.00it/s]
Running loglikelihood requests:  51%|█████     | 30755/60896 [20:48<16:30, 30.43it/s]
Running loglikelihood requests:  51%|█████     | 30788/60896 [20:49<16:33, 30.31it/s]
Running loglikelihood requests:  51%|█████     | 30821/60896 [20:51<16:45, 29.91it/s]
Running loglikelihood requests:  51%|█████     | 30853/60896 [20:52<16:52, 29.66it/s]
Running loglikelihood requests:  51%|█████     | 30885/60896 [20:53<16:57, 29.48it/s]
Running loglikelihood requests:  51%|█████     | 30919/60896 [20:54<16:42, 29.90it/s]
Running loglikelihood requests:  51%|█████     | 30953/60896 [20:55<16:31, 30.20it/s]
Running loglikelihood requests:  51%|█████     | 30985/60896 [20:56<16:49, 29.63it/s]
Running loglikelihood requests:  51%|█████     | 31029/60896 [20:57<15:13, 32.70it/s]
Running loglikelihood requests:  51%|█████     | 31066/60896 [20:58<15:03, 33.01it/s]
Running loglikelihood requests:  51%|█████     | 31098/60896 [20:59<15:34, 31.88it/s]
Running loglikelihood requests:  51%|█████     | 31130/60896 [21:01<15:57, 31.09it/s]
Running loglikelihood requests:  51%|█████     | 31166/60896 [21:02<15:46, 31.42it/s]
Running loglikelihood requests:  51%|█████     | 31200/60896 [21:03<15:48, 31.31it/s]
Running loglikelihood requests:  51%|█████▏    | 31232/60896 [21:04<16:06, 30.70it/s]
Running loglikelihood requests:  51%|█████▏    | 31264/60896 [21:05<16:19, 30.26it/s]
Running loglikelihood requests:  51%|█████▏    | 31296/60896 [21:06<16:28, 29.95it/s]
Running loglikelihood requests:  51%|█████▏    | 31329/60896 [21:07<16:25, 30.01it/s]
Running loglikelihood requests:  52%|█████▏    | 31363/60896 [21:08<16:22, 30.07it/s]
Running loglikelihood requests:  52%|█████▏    | 31399/60896 [21:09<15:48, 31.11it/s]
Running loglikelihood requests:  52%|█████▏    | 31439/60896 [21:10<14:54, 32.95it/s]
Running loglikelihood requests:  52%|█████▏    | 31473/60896 [21:11<15:02, 32.59it/s]
Running loglikelihood requests:  52%|█████▏    | 31507/60896 [21:13<15:08, 32.34it/s]
Running loglikelihood requests:  52%|█████▏    | 31539/60896 [21:14<15:36, 31.34it/s]
Running loglikelihood requests:  52%|█████▏    | 31575/60896 [21:15<15:15, 32.01it/s]
Running loglikelihood requests:  52%|█████▏    | 31607/60896 [21:16<15:33, 31.38it/s]
Running loglikelihood requests:  52%|█████▏    | 31639/60896 [21:17<15:45, 30.93it/s]
Running loglikelihood requests:  52%|█████▏    | 31671/60896 [21:18<15:54, 30.62it/s]
Running loglikelihood requests:  52%|█████▏    | 31703/60896 [21:19<16:00, 30.40it/s]
Running loglikelihood requests:  52%|█████▏    | 31736/60896 [21:20<15:55, 30.53it/s]
Running loglikelihood requests:  52%|█████▏    | 31770/60896 [21:21<15:50, 30.64it/s]
Running loglikelihood requests:  52%|█████▏    | 31809/60896 [21:22<14:57, 32.40it/s]
Running loglikelihood requests:  52%|█████▏    | 31842/60896 [21:23<15:08, 31.97it/s]
Running loglikelihood requests:  52%|█████▏    | 31874/60896 [21:24<15:24, 31.39it/s]
Running loglikelihood requests:  52%|█████▏    | 31906/60896 [21:25<15:35, 30.97it/s]
Running loglikelihood requests:  52%|█████▏    | 31939/60896 [21:26<15:34, 30.97it/s]
Running loglikelihood requests:  53%|█████▎    | 31971/60896 [21:28<15:42, 30.69it/s]
Running loglikelihood requests:  53%|█████▎    | 32003/60896 [21:29<15:55, 30.24it/s]
Running loglikelihood requests:  53%|█████▎    | 32035/60896 [21:30<15:56, 30.17it/s]
Running loglikelihood requests:  53%|█████▎    | 32069/60896 [21:31<15:39, 30.69it/s]
Running loglikelihood requests:  53%|█████▎    | 32103/60896 [21:32<15:21, 31.26it/s]
Running loglikelihood requests:  53%|█████▎    | 32141/60896 [21:33<14:36, 32.81it/s]
Running loglikelihood requests:  53%|█████▎    | 32175/60896 [21:34<14:36, 32.76it/s]
Running loglikelihood requests:  53%|█████▎    | 32210/60896 [21:35<14:28, 33.01it/s]
Running loglikelihood requests:  53%|█████▎    | 32242/60896 [21:36<14:54, 32.03it/s]
Running loglikelihood requests:  53%|█████▎    | 32274/60896 [21:37<15:04, 31.64it/s]
Running loglikelihood requests:  53%|█████▎    | 32306/60896 [21:38<15:11, 31.37it/s]
Running loglikelihood requests:  53%|█████▎    | 32338/60896 [21:39<15:16, 31.18it/s]
Running loglikelihood requests:  53%|█████▎    | 32370/60896 [21:40<15:19, 31.04it/s]
Running loglikelihood requests:  53%|█████▎    | 32407/60896 [21:41<14:39, 32.38it/s]
Running loglikelihood requests:  53%|█████▎    | 32441/60896 [21:42<14:41, 32.27it/s]
Running loglikelihood requests:  53%|█████▎    | 32480/60896 [21:43<13:59, 33.83it/s]
Running loglikelihood requests:  53%|█████▎    | 32512/60896 [21:44<14:21, 32.95it/s]
Running loglikelihood requests:  53%|█████▎    | 32545/60896 [21:45<14:29, 32.62it/s]
Running loglikelihood requests:  53%|█████▎    | 32577/60896 [21:46<14:42, 32.10it/s]
Running loglikelihood requests:  54%|█████▎    | 32610/60896 [21:47<14:43, 32.03it/s]
Running loglikelihood requests:  54%|█████▎    | 32642/60896 [21:49<14:51, 31.69it/s]
Running loglikelihood requests:  54%|█████▎    | 32674/60896 [21:50<15:01, 31.29it/s]
Running loglikelihood requests:  54%|█████▎    | 32708/60896 [21:51<14:47, 31.74it/s]
Running loglikelihood requests:  54%|█████▍    | 32741/60896 [21:52<14:46, 31.77it/s]
Running loglikelihood requests:  54%|█████▍    | 32775/60896 [21:53<14:36, 32.08it/s]
Running loglikelihood requests:  54%|█████▍    | 32817/60896 [21:54<13:28, 34.75it/s]
Running loglikelihood requests:  54%|█████▍    | 32850/60896 [21:55<13:44, 34.02it/s]
Running loglikelihood requests:  54%|█████▍    | 32885/60896 [21:56<13:41, 34.09it/s]
Running loglikelihood requests:  54%|█████▍    | 32917/60896 [21:57<14:07, 33.00it/s]
Running loglikelihood requests:  54%|█████▍    | 32950/60896 [21:58<14:12, 32.79it/s]
Running loglikelihood requests:  54%|█████▍    | 32982/60896 [21:59<14:22, 32.35it/s]
Running loglikelihood requests:  54%|█████▍    | 33014/60896 [22:00<14:30, 32.04it/s]
Running loglikelihood requests:  54%|█████▍    | 33048/60896 [22:01<14:19, 32.40it/s]
Running loglikelihood requests:  54%|█████▍    | 33084/60896 [22:02<13:56, 33.24it/s]
Running loglikelihood requests:  54%|█████▍    | 33124/60896 [22:03<13:19, 34.73it/s]
Running loglikelihood requests:  54%|█████▍    | 33159/60896 [22:04<13:20, 34.65it/s]
Running loglikelihood requests:  55%|█████▍    | 33194/60896 [22:05<13:21, 34.58it/s]
Running loglikelihood requests:  55%|█████▍    | 33228/60896 [22:06<13:28, 34.23it/s]
Running loglikelihood requests:  55%|█████▍    | 33264/60896 [22:07<13:19, 34.58it/s]
Running loglikelihood requests:  55%|█████▍    | 33296/60896 [22:08<13:40, 33.65it/s]
Running loglikelihood requests:  55%|█████▍    | 33332/60896 [22:09<13:26, 34.17it/s]
Running loglikelihood requests:  55%|█████▍    | 33371/60896 [22:10<13:03, 35.14it/s]
Running loglikelihood requests:  55%|█████▍    | 33407/60896 [22:11<12:59, 35.25it/s]
Running loglikelihood requests:  55%|█████▍    | 33444/60896 [22:12<12:50, 35.64it/s]
Running loglikelihood requests:  55%|█████▍    | 33476/60896 [22:13<13:16, 34.44it/s]
Running loglikelihood requests:  55%|█████▌    | 33509/60896 [22:14<13:28, 33.89it/s]
Running loglikelihood requests:  55%|█████▌    | 33543/60896 [22:15<13:29, 33.80it/s]
Running loglikelihood requests:  55%|█████▌    | 33575/60896 [22:16<13:50, 32.90it/s]
Running loglikelihood requests:  55%|█████▌    | 33608/60896 [22:17<13:51, 32.81it/s]
Running loglikelihood requests:  55%|█████▌    | 33644/60896 [22:18<13:30, 33.63it/s]
Running loglikelihood requests:  55%|█████▌    | 33683/60896 [22:19<12:54, 35.15it/s]
Running loglikelihood requests:  55%|█████▌    | 33719/60896 [22:20<12:49, 35.33it/s]
Running loglikelihood requests:  55%|█████▌    | 33753/60896 [22:21<12:58, 34.86it/s]
Running loglikelihood requests:  55%|█████▌    | 33790/60896 [22:22<12:53, 35.07it/s]
Running loglikelihood requests:  56%|█████▌    | 33824/60896 [22:23<13:00, 34.68it/s]
Running loglikelihood requests:  56%|█████▌    | 33856/60896 [22:24<13:20, 33.79it/s]
Running loglikelihood requests:  56%|█████▌    | 33892/60896 [22:25<13:05, 34.37it/s]
Running loglikelihood requests:  56%|█████▌    | 33934/60896 [22:26<11:51, 37.89it/s]
Running loglikelihood requests:  56%|█████▌    | 33969/60896 [22:27<11:42, 38.31it/s]
Running loglikelihood requests:  56%|█████▌    | 34001/60896 [22:28<11:54, 37.65it/s]
Running loglikelihood requests:  56%|█████▌    | 34037/60896 [22:29<11:38, 38.48it/s]
Running loglikelihood requests:  56%|█████▌    | 34070/60896 [22:30<11:44, 38.08it/s]
Running loglikelihood requests:  56%|█████▌    | 34103/60896 [22:31<11:55, 37.43it/s]
Running loglikelihood requests:  56%|█████▌    | 34140/60896 [22:32<11:31, 38.71it/s]
Running loglikelihood requests:  56%|█████▌    | 34175/60896 [22:32<11:25, 38.96it/s]
Running loglikelihood requests:  56%|█████▌    | 34208/60896 [22:33<11:33, 38.47it/s]
Running loglikelihood requests:  56%|█████▌    | 34243/60896 [22:34<11:27, 38.79it/s]
Running loglikelihood requests:  56%|█████▋    | 34279/60896 [22:35<11:16, 39.36it/s]
Running loglikelihood requests:  56%|█████▋    | 34312/60896 [22:36<11:26, 38.74it/s]
Running loglikelihood requests:  56%|█████▋    | 34345/60896 [22:37<11:33, 38.31it/s]
Running loglikelihood requests:  56%|█████▋    | 34381/60896 [22:38<11:18, 39.11it/s]
Running loglikelihood requests:  57%|█████▋    | 34417/60896 [22:39<11:00, 40.08it/s]
Running loglikelihood requests:  57%|█████▋    | 34449/60896 [22:39<11:11, 39.40it/s]
Running loglikelihood requests:  57%|█████▋    | 34485/60896 [22:40<10:56, 40.23it/s]
Running loglikelihood requests:  57%|█████▋    | 34518/60896 [22:41<11:02, 39.84it/s]
Running loglikelihood requests:  57%|█████▋    | 34550/60896 [22:42<11:11, 39.21it/s]
Running loglikelihood requests:  57%|█████▋    | 34583/60896 [22:43<11:12, 39.12it/s]
Running loglikelihood requests:  57%|█████▋    | 34619/60896 [22:44<10:53, 40.19it/s]
Running loglikelihood requests:  57%|█████▋    | 34657/60896 [22:45<10:38, 41.09it/s]
Running loglikelihood requests:  57%|█████▋    | 34689/60896 [22:45<10:52, 40.16it/s]
Running loglikelihood requests:  57%|█████▋    | 34722/60896 [22:46<10:57, 39.81it/s]
Running loglikelihood requests:  57%|█████▋    | 34757/60896 [22:47<10:49, 40.22it/s]
Running loglikelihood requests:  57%|█████▋    | 34790/60896 [22:48<10:54, 39.90it/s]
Running loglikelihood requests:  57%|█████▋    | 34824/60896 [22:49<10:51, 40.02it/s]
Running loglikelihood requests:  57%|█████▋    | 34858/60896 [22:50<10:48, 40.17it/s]
Running loglikelihood requests:  57%|█████▋    | 34894/60896 [22:51<10:41, 40.50it/s]
Running loglikelihood requests:  57%|█████▋    | 34932/60896 [22:51<10:19, 41.92it/s]
Running loglikelihood requests:  57%|█████▋    | 34966/60896 [22:52<10:24, 41.50it/s]
Running loglikelihood requests:  57%|█████▋    | 34998/60896 [22:53<10:39, 40.48it/s]
Running loglikelihood requests:  58%|█████▊    | 35032/60896 [22:54<10:38, 40.50it/s]
Running loglikelihood requests:  58%|█████▊    | 35068/60896 [22:55<10:26, 41.22it/s]
Running loglikelihood requests:  58%|█████▊    | 35104/60896 [22:56<10:18, 41.71it/s]
Running loglikelihood requests:  58%|█████▊    | 35138/60896 [22:56<10:21, 41.43it/s]
Running loglikelihood requests:  58%|█████▊    | 35174/60896 [22:57<10:20, 41.44it/s]
Running loglikelihood requests:  58%|█████▊    | 35207/60896 [22:58<10:28, 40.88it/s]
Running loglikelihood requests:  58%|█████▊    | 35244/60896 [22:59<10:12, 41.91it/s]
Running loglikelihood requests:  58%|█████▊    | 35277/60896 [23:00<10:21, 41.20it/s]
Running loglikelihood requests:  58%|█████▊    | 35309/60896 [23:01<10:34, 40.35it/s]
Running loglikelihood requests:  58%|█████▊    | 35343/60896 [23:01<10:31, 40.46it/s]
Running loglikelihood requests:  58%|█████▊    | 35377/60896 [23:02<10:26, 40.71it/s]
Running loglikelihood requests:  58%|█████▊    | 35416/60896 [23:03<09:56, 42.69it/s]
Running loglikelihood requests:  58%|█████▊    | 35449/60896 [23:04<10:13, 41.46it/s]
Running loglikelihood requests:  58%|█████▊    | 35481/60896 [23:05<10:24, 40.70it/s]
Running loglikelihood requests:  58%|█████▊    | 35515/60896 [23:06<10:21, 40.87it/s]
Running loglikelihood requests:  58%|█████▊    | 35548/60896 [23:06<10:23, 40.64it/s]
Running loglikelihood requests:  58%|█████▊    | 35581/60896 [23:07<10:25, 40.47it/s]
Running loglikelihood requests:  58%|█████▊    | 35620/60896 [23:08<09:55, 42.41it/s]
Running loglikelihood requests:  59%|█████▊    | 35654/60896 [23:09<09:59, 42.09it/s]
Running loglikelihood requests:  59%|█████▊    | 35686/60896 [23:10<10:20, 40.62it/s]
Running loglikelihood requests:  59%|█████▊    | 35718/60896 [23:11<10:26, 40.17it/s]
Running loglikelihood requests:  59%|█████▊    | 35750/60896 [23:11<10:30, 39.86it/s]
Running loglikelihood requests:  59%|█████▉    | 35784/60896 [23:12<10:22, 40.34it/s]
Running loglikelihood requests:  59%|█████▉    | 35816/60896 [23:13<10:27, 39.97it/s]
Running loglikelihood requests:  59%|█████▉    | 35851/60896 [23:14<10:14, 40.79it/s]
Running loglikelihood requests:  59%|█████▉    | 35884/60896 [23:15<10:14, 40.72it/s]
Running loglikelihood requests:  59%|█████▉    | 35920/60896 [23:15<10:03, 41.36it/s]
Running loglikelihood requests:  59%|█████▉    | 35955/60896 [23:16<09:56, 41.84it/s]
Running loglikelihood requests:  59%|█████▉    | 35987/60896 [23:17<10:06, 41.09it/s]
Running loglikelihood requests:  59%|█████▉    | 36019/60896 [23:18<10:13, 40.56it/s]
Running loglikelihood requests:  59%|█████▉    | 36053/60896 [23:19<10:07, 40.92it/s]
Running loglikelihood requests:  59%|█████▉    | 36085/60896 [23:20<10:13, 40.44it/s]
Running loglikelihood requests:  59%|█████▉    | 36118/60896 [23:20<10:12, 40.47it/s]
Running loglikelihood requests:  59%|█████▉    | 36152/60896 [23:21<10:04, 40.91it/s]
Running loglikelihood requests:  59%|█████▉    | 36186/60896 [23:22<09:59, 41.23it/s]
Running loglikelihood requests:  59%|█████▉    | 36222/60896 [23:23<09:51, 41.73it/s]
Running loglikelihood requests:  60%|█████▉    | 36254/60896 [23:24<09:59, 41.07it/s]
Running loglikelihood requests:  60%|█████▉    | 36289/60896 [23:24<09:50, 41.66it/s]
Running loglikelihood requests:  60%|█████▉    | 36322/60896 [23:25<09:53, 41.39it/s]
Running loglikelihood requests:  60%|█████▉    | 36354/60896 [23:26<10:01, 40.83it/s]
Running loglikelihood requests:  60%|█████▉    | 36387/60896 [23:27<10:00, 40.80it/s]
Running loglikelihood requests:  60%|█████▉    | 36420/60896 [23:28<10:00, 40.77it/s]
Running loglikelihood requests:  60%|█████▉    | 36452/60896 [23:28<09:51, 41.32it/s]
Running loglikelihood requests:  60%|█████▉    | 36489/60896 [23:29<09:26, 43.10it/s]
Running loglikelihood requests:  60%|█████▉    | 36521/60896 [23:30<09:26, 43.00it/s]
Running loglikelihood requests:  60%|██████    | 36555/60896 [23:31<09:17, 43.69it/s]
Running loglikelihood requests:  60%|██████    | 36588/60896 [23:31<09:15, 43.80it/s]
Running loglikelihood requests:  60%|██████    | 36621/60896 [23:32<09:13, 43.89it/s]
Running loglikelihood requests:  60%|██████    | 36654/60896 [23:33<09:11, 43.95it/s]
Running loglikelihood requests:  60%|██████    | 36689/60896 [23:34<09:00, 44.78it/s]
Running loglikelihood requests:  60%|██████    | 36723/60896 [23:34<08:56, 45.05it/s]
Running loglikelihood requests:  60%|██████    | 36756/60896 [23:35<08:58, 44.85it/s]
Running loglikelihood requests:  60%|██████    | 36788/60896 [23:36<09:04, 44.29it/s]
Running loglikelihood requests:  60%|██████    | 36821/60896 [23:37<09:11, 43.69it/s]
Running loglikelihood requests:  61%|██████    | 36858/60896 [23:37<08:48, 45.45it/s]
Running loglikelihood requests:  61%|██████    | 36894/60896 [23:38<08:38, 46.32it/s]
Running loglikelihood requests:  61%|██████    | 36927/60896 [23:39<08:44, 45.71it/s]
Running loglikelihood requests:  61%|██████    | 36960/60896 [23:40<08:48, 45.31it/s]
Running loglikelihood requests:  61%|██████    | 36992/60896 [23:40<08:56, 44.59it/s]
Running loglikelihood requests:  61%|██████    | 37026/60896 [23:41<08:50, 44.98it/s]
Running loglikelihood requests:  61%|██████    | 37059/60896 [23:42<08:51, 44.87it/s]
Running loglikelihood requests:  61%|██████    | 37091/60896 [23:43<08:56, 44.39it/s]
Running loglikelihood requests:  61%|██████    | 37123/60896 [23:43<09:01, 43.91it/s]
Running loglikelihood requests:  61%|██████    | 37159/60896 [23:44<08:46, 45.12it/s]
Running loglikelihood requests:  61%|██████    | 37191/60896 [23:45<08:52, 44.51it/s]
Running loglikelihood requests:  61%|██████    | 37223/60896 [23:46<08:56, 44.12it/s]
Running loglikelihood requests:  61%|██████    | 37255/60896 [23:46<09:00, 43.77it/s]
Running loglikelihood requests:  61%|██████    | 37289/60896 [23:47<08:51, 44.41it/s]
Running loglikelihood requests:  61%|██████▏   | 37321/60896 [23:48<08:55, 44.05it/s]
Running loglikelihood requests:  61%|██████▏   | 37354/60896 [23:49<08:51, 44.26it/s]
Running loglikelihood requests:  61%|██████▏   | 37386/60896 [23:49<08:53, 44.03it/s]
Running loglikelihood requests:  61%|██████▏   | 37421/60896 [23:50<08:40, 45.08it/s]
Running loglikelihood requests:  62%|██████▏   | 37454/60896 [23:51<08:42, 44.85it/s]
Running loglikelihood requests:  62%|██████▏   | 37486/60896 [23:52<08:46, 44.42it/s]
Running loglikelihood requests:  62%|██████▏   | 37529/60896 [23:52<08:00, 48.59it/s]
Running loglikelihood requests:  62%|██████▏   | 37563/60896 [23:53<08:07, 47.87it/s]
Running loglikelihood requests:  62%|██████▏   | 37595/60896 [23:54<08:28, 45.79it/s]
Running loglikelihood requests:  62%|██████▏   | 37627/60896 [23:55<08:39, 44.82it/s]
Running loglikelihood requests:  62%|██████▏   | 37660/60896 [23:55<08:41, 44.58it/s]
Running loglikelihood requests:  62%|██████▏   | 37693/60896 [23:56<08:38, 44.76it/s]
Running loglikelihood requests:  62%|██████▏   | 37725/60896 [23:57<08:39, 44.57it/s]
Running loglikelihood requests:  62%|██████▏   | 37758/60896 [23:58<08:44, 44.08it/s]
Running loglikelihood requests:  62%|██████▏   | 37793/60896 [23:58<08:44, 44.06it/s]
Running loglikelihood requests:  62%|██████▏   | 37825/60896 [23:59<08:48, 43.65it/s]
Running loglikelihood requests:  62%|██████▏   | 37863/60896 [24:00<08:16, 46.40it/s]
Running loglikelihood requests:  62%|██████▏   | 37899/60896 [24:01<08:03, 47.56it/s]
Running loglikelihood requests:  62%|██████▏   | 37933/60896 [24:01<08:02, 47.57it/s]
Running loglikelihood requests:  62%|██████▏   | 37965/60896 [24:02<08:10, 46.76it/s]
Running loglikelihood requests:  62%|██████▏   | 37999/60896 [24:03<08:07, 47.02it/s]
Running loglikelihood requests:  62%|██████▏   | 38032/60896 [24:03<08:10, 46.66it/s]
Running loglikelihood requests:  63%|██████▎   | 38065/60896 [24:04<08:09, 46.61it/s]
Running loglikelihood requests:  63%|██████▎   | 38099/60896 [24:05<08:05, 46.99it/s]
Running loglikelihood requests:  63%|██████▎   | 38133/60896 [24:05<08:01, 47.26it/s]
Running loglikelihood requests:  63%|██████▎   | 38167/60896 [24:06<07:58, 47.46it/s]
Running loglikelihood requests:  63%|██████▎   | 38200/60896 [24:07<08:00, 47.19it/s]
Running loglikelihood requests:  63%|██████▎   | 38234/60896 [24:08<07:57, 47.41it/s]
Running loglikelihood requests:  63%|██████▎   | 38267/60896 [24:08<07:59, 47.15it/s]
Running loglikelihood requests:  63%|██████▎   | 38299/60896 [24:09<08:05, 46.54it/s]
Running loglikelihood requests:  63%|██████▎   | 38333/60896 [24:10<08:00, 46.96it/s]
Running loglikelihood requests:  63%|██████▎   | 38365/60896 [24:10<08:10, 45.89it/s]
Running loglikelihood requests:  63%|██████▎   | 38398/60896 [24:11<08:08, 46.08it/s]
Running loglikelihood requests:  63%|██████▎   | 38430/60896 [24:12<08:10, 45.79it/s]
Running loglikelihood requests:  63%|██████▎   | 38463/60896 [24:13<08:07, 46.06it/s]
Running loglikelihood requests:  63%|██████▎   | 38495/60896 [24:13<08:08, 45.81it/s]
Running loglikelihood requests:  63%|██████▎   | 38527/60896 [24:14<08:10, 45.64it/s]
Running loglikelihood requests:  63%|██████▎   | 38560/60896 [24:15<08:06, 45.93it/s]
Running loglikelihood requests:  63%|██████▎   | 38592/60896 [24:15<08:07, 45.72it/s]
Running loglikelihood requests:  63%|██████▎   | 38631/60896 [24:16<07:38, 48.53it/s]
Running loglikelihood requests:  63%|██████▎   | 38663/60896 [24:17<07:47, 47.55it/s]
Running loglikelihood requests:  64%|██████▎   | 38704/60896 [24:18<07:18, 50.65it/s]
Running loglikelihood requests:  64%|██████▎   | 38736/60896 [24:18<07:39, 48.23it/s]
Running loglikelihood requests:  64%|██████▎   | 38769/60896 [24:19<07:43, 47.76it/s]
Running loglikelihood requests:  64%|██████▎   | 38801/60896 [24:20<07:49, 47.01it/s]
Running loglikelihood requests:  64%|██████▍   | 38835/60896 [24:20<07:46, 47.33it/s]
Running loglikelihood requests:  64%|██████▍   | 38868/60896 [24:21<07:47, 47.11it/s]
Running loglikelihood requests:  64%|██████▍   | 38900/60896 [24:22<07:51, 46.64it/s]
Running loglikelihood requests:  64%|██████▍   | 38932/60896 [24:23<07:54, 46.31it/s]
Running loglikelihood requests:  64%|██████▍   | 38965/60896 [24:23<07:51, 46.50it/s]
Running loglikelihood requests:  64%|██████▍   | 38997/60896 [24:24<07:54, 46.11it/s]
Running loglikelihood requests:  64%|██████▍   | 39033/60896 [24:25<07:38, 47.64it/s]
Running loglikelihood requests:  64%|██████▍   | 39068/60896 [24:25<07:40, 47.43it/s]
Running loglikelihood requests:  64%|██████▍   | 39100/60896 [24:26<07:45, 46.87it/s]
Running loglikelihood requests:  64%|██████▍   | 39133/60896 [24:27<07:44, 46.89it/s]
Running loglikelihood requests:  64%|██████▍   | 39165/60896 [24:28<07:47, 46.48it/s]
Running loglikelihood requests:  64%|██████▍   | 39197/60896 [24:28<07:49, 46.18it/s]
Running loglikelihood requests:  64%|██████▍   | 39231/60896 [24:29<07:42, 46.83it/s]
Running loglikelihood requests:  64%|██████▍   | 39263/60896 [24:30<07:45, 46.44it/s]
Running loglikelihood requests:  65%|██████▍   | 39296/60896 [24:30<07:43, 46.57it/s]
Running loglikelihood requests:  65%|██████▍   | 39330/60896 [24:31<07:33, 47.52it/s]
Running loglikelihood requests:  65%|██████▍   | 39363/60896 [24:32<07:30, 47.78it/s]
Running loglikelihood requests:  65%|██████▍   | 39395/60896 [24:32<07:39, 46.76it/s]
Running loglikelihood requests:  65%|██████▍   | 39428/60896 [24:33<07:34, 47.24it/s]
Running loglikelihood requests:  65%|██████▍   | 39462/60896 [24:34<07:26, 48.02it/s]
Running loglikelihood requests:  65%|██████▍   | 39494/60896 [24:34<07:28, 47.71it/s]
Running loglikelihood requests:  65%|██████▍   | 39527/60896 [24:35<07:25, 47.92it/s]
Running loglikelihood requests:  65%|██████▍   | 39569/60896 [24:36<06:50, 51.99it/s]
Running loglikelihood requests:  65%|██████▌   | 39601/60896 [24:37<07:01, 50.48it/s]
Running loglikelihood requests:  65%|██████▌   | 39635/60896 [24:37<07:02, 50.29it/s]
Running loglikelihood requests:  65%|██████▌   | 39667/60896 [24:38<07:10, 49.29it/s]
Running loglikelihood requests:  65%|██████▌   | 39699/60896 [24:39<07:16, 48.59it/s]
Running loglikelihood requests:  65%|██████▌   | 39731/60896 [24:39<07:25, 47.46it/s]
Running loglikelihood requests:  65%|██████▌   | 39763/60896 [24:40<07:26, 47.29it/s]
Running loglikelihood requests:  65%|██████▌   | 39795/60896 [24:41<07:27, 47.10it/s]
Running loglikelihood requests:  65%|██████▌   | 39827/60896 [24:41<07:27, 47.12it/s]
Running loglikelihood requests:  65%|██████▌   | 39859/60896 [24:42<07:26, 47.14it/s]
Running loglikelihood requests:  66%|██████▌   | 39892/60896 [24:43<07:21, 47.60it/s]
Running loglikelihood requests:  66%|██████▌   | 39927/60896 [24:43<07:09, 48.77it/s]
Running loglikelihood requests:  66%|██████▌   | 39959/60896 [24:44<07:13, 48.31it/s]
Running loglikelihood requests:  66%|██████▌   | 39992/60896 [24:45<07:11, 48.43it/s]
Running loglikelihood requests:  66%|██████▌   | 40024/60896 [24:45<07:14, 48.06it/s]
Running loglikelihood requests:  66%|██████▌   | 40061/60896 [24:46<06:56, 50.01it/s]
Running loglikelihood requests:  66%|██████▌   | 40093/60896 [24:47<07:11, 48.20it/s]
Running loglikelihood requests:  66%|██████▌   | 40125/60896 [24:47<07:13, 47.90it/s]
Running loglikelihood requests:  66%|██████▌   | 40157/60896 [24:48<07:14, 47.72it/s]
Running loglikelihood requests:  66%|██████▌   | 40189/60896 [24:49<07:15, 47.58it/s]
Running loglikelihood requests:  66%|██████▌   | 40221/60896 [24:49<07:15, 47.48it/s]
Running loglikelihood requests:  66%|██████▌   | 40253/60896 [24:50<07:15, 47.42it/s]
Running loglikelihood requests:  66%|██████▌   | 40287/60896 [24:51<07:07, 48.25it/s]
Running loglikelihood requests:  66%|██████▌   | 40319/60896 [24:51<07:01, 48.86it/s]
Running loglikelihood requests:  66%|██████▋   | 40351/60896 [24:52<06:56, 49.30it/s]
Running loglikelihood requests:  66%|██████▋   | 40384/60896 [24:53<06:49, 50.09it/s]
Running loglikelihood requests:  66%|██████▋   | 40417/60896 [24:53<06:50, 49.92it/s]
Running loglikelihood requests:  66%|██████▋   | 40451/60896 [24:54<06:40, 50.99it/s]
Running loglikelihood requests:  66%|██████▋   | 40486/60896 [24:55<06:30, 52.22it/s]
Running loglikelihood requests:  67%|██████▋   | 40520/60896 [24:55<06:27, 52.62it/s]
Running loglikelihood requests:  67%|██████▋   | 40552/60896 [24:56<06:31, 51.96it/s]
Running loglikelihood requests:  67%|██████▋   | 40587/60896 [24:57<06:23, 52.92it/s]
Running loglikelihood requests:  67%|██████▋   | 40619/60896 [24:57<06:28, 52.19it/s]
Running loglikelihood requests:  67%|██████▋   | 40651/60896 [24:58<06:31, 51.67it/s]
Running loglikelihood requests:  67%|██████▋   | 40683/60896 [24:58<06:34, 51.30it/s]
Running loglikelihood requests:  67%|██████▋   | 40715/60896 [24:59<06:35, 51.05it/s]
Running loglikelihood requests:  67%|██████▋   | 40747/60896 [25:00<06:42, 50.03it/s]
Running loglikelihood requests:  67%|██████▋   | 40779/60896 [25:00<06:41, 50.15it/s]
Running loglikelihood requests:  67%|██████▋   | 40813/60896 [25:01<06:32, 51.12it/s]
Running loglikelihood requests:  67%|██████▋   | 40846/60896 [25:02<06:29, 51.45it/s]
Running loglikelihood requests:  67%|██████▋   | 40878/60896 [25:02<06:30, 51.22it/s]
Running loglikelihood requests:  67%|██████▋   | 40911/60896 [25:03<06:27, 51.52it/s]
Running loglikelihood requests:  67%|██████▋   | 40945/60896 [25:04<06:22, 52.20it/s]
Running loglikelihood requests:  67%|██████▋   | 40978/60896 [25:04<06:21, 52.22it/s]
Running loglikelihood requests:  67%|██████▋   | 41013/60896 [25:05<06:13, 53.18it/s]
Running loglikelihood requests:  67%|██████▋   | 41047/60896 [25:05<06:11, 53.37it/s]
Running loglikelihood requests:  67%|██████▋   | 41080/60896 [25:06<06:13, 53.04it/s]
Running loglikelihood requests:  68%|██████▊   | 41112/60896 [25:07<06:24, 51.43it/s]
Running loglikelihood requests:  68%|██████▊   | 41144/60896 [25:07<06:25, 51.20it/s]
Running loglikelihood requests:  68%|██████▊   | 41178/60896 [25:08<06:19, 51.96it/s]
Running loglikelihood requests:  68%|██████▊   | 41210/60896 [25:09<06:21, 51.58it/s]
Running loglikelihood requests:  68%|██████▊   | 41242/60896 [25:09<06:22, 51.32it/s]
Running loglikelihood requests:  68%|██████▊   | 41274/60896 [25:10<06:23, 51.13it/s]
Running loglikelihood requests:  68%|██████▊   | 41306/60896 [25:11<06:24, 50.99it/s]
Running loglikelihood requests:  68%|██████▊   | 41338/60896 [25:11<06:24, 50.88it/s]
Running loglikelihood requests:  68%|██████▊   | 41371/60896 [25:12<06:20, 51.28it/s]
Running loglikelihood requests:  68%|██████▊   | 41405/60896 [25:12<06:09, 52.76it/s]
Running loglikelihood requests:  68%|██████▊   | 41437/60896 [25:13<06:08, 52.85it/s]
Running loglikelihood requests:  68%|██████▊   | 41470/60896 [25:14<06:10, 52.45it/s]
Running loglikelihood requests:  68%|██████▊   | 41503/60896 [25:14<06:04, 53.14it/s]
Running loglikelihood requests:  68%|██████▊   | 41535/60896 [25:15<06:04, 53.14it/s]
Running loglikelihood requests:  68%|██████▊   | 41568/60896 [25:15<06:00, 53.61it/s]
Running loglikelihood requests:  68%|██████▊   | 41605/60896 [25:16<05:44, 55.94it/s]
Running loglikelihood requests:  68%|██████▊   | 41637/60896 [25:17<05:49, 55.10it/s]
Running loglikelihood requests:  68%|██████▊   | 41675/60896 [25:17<05:34, 57.49it/s]
Running loglikelihood requests:  68%|██████▊   | 41711/60896 [25:18<05:29, 58.16it/s]
Running loglikelihood requests:  69%|██████▊   | 41748/60896 [25:19<05:24, 59.10it/s]
Running loglikelihood requests:  69%|██████▊   | 41780/60896 [25:19<05:33, 57.31it/s]
Running loglikelihood requests:  69%|██████▊   | 41813/60896 [25:20<05:37, 56.57it/s]
Running loglikelihood requests:  69%|██████▊   | 41845/60896 [25:20<05:50, 54.36it/s]
Running loglikelihood requests:  69%|██████▉   | 41877/60896 [25:21<05:52, 53.99it/s]
Running loglikelihood requests:  69%|██████▉   | 41909/60896 [25:22<05:53, 53.73it/s]
Running loglikelihood requests:  69%|██████▉   | 41941/60896 [25:22<05:54, 53.54it/s]
Running loglikelihood requests:  69%|██████▉   | 41973/60896 [25:23<05:54, 53.43it/s]
Running loglikelihood requests:  69%|██████▉   | 42005/60896 [25:23<05:53, 53.42it/s]
Running loglikelihood requests:  69%|██████▉   | 42038/60896 [25:24<05:50, 53.78it/s]
Running loglikelihood requests:  69%|██████▉   | 42070/60896 [25:25<05:50, 53.66it/s]
Running loglikelihood requests:  69%|██████▉   | 42102/60896 [25:25<05:50, 53.58it/s]
Running loglikelihood requests:  69%|██████▉   | 42134/60896 [25:26<05:50, 53.51it/s]
Running loglikelihood requests:  69%|██████▉   | 42166/60896 [25:26<05:50, 53.49it/s]
Running loglikelihood requests:  69%|██████▉   | 42204/60896 [25:27<05:31, 56.44it/s]
Running loglikelihood requests:  69%|██████▉   | 42239/60896 [25:28<05:33, 55.97it/s]
Running loglikelihood requests:  69%|██████▉   | 42275/60896 [25:28<05:25, 57.18it/s]
Running loglikelihood requests:  69%|██████▉   | 42317/60896 [25:29<05:04, 61.02it/s]
Running loglikelihood requests:  70%|██████▉   | 42352/60896 [25:29<05:07, 60.25it/s]
Running loglikelihood requests:  70%|██████▉   | 42385/60896 [25:30<05:15, 58.70it/s]
Running loglikelihood requests:  70%|██████▉   | 42417/60896 [25:31<05:23, 57.12it/s]
Running loglikelihood requests:  70%|██████▉   | 42449/60896 [25:31<05:29, 56.02it/s]
Running loglikelihood requests:  70%|██████▉   | 42481/60896 [25:32<05:33, 55.24it/s]
Running loglikelihood requests:  70%|██████▉   | 42513/60896 [25:32<05:36, 54.71it/s]
Running loglikelihood requests:  70%|██████▉   | 42545/60896 [25:33<05:37, 54.33it/s]
Running loglikelihood requests:  70%|██████▉   | 42577/60896 [25:34<05:38, 54.07it/s]
Running loglikelihood requests:  70%|██████▉   | 42611/60896 [25:34<05:38, 53.95it/s]
Running loglikelihood requests:  70%|███████   | 42643/60896 [25:35<05:38, 53.87it/s]
Running loglikelihood requests:  70%|███████   | 42676/60896 [25:35<05:35, 54.30it/s]
Running loglikelihood requests:  70%|███████   | 42708/60896 [25:36<05:36, 54.10it/s]
Running loglikelihood requests:  70%|███████   | 42740/60896 [25:37<05:36, 53.94it/s]
Running loglikelihood requests:  70%|███████   | 42774/60896 [25:37<05:30, 54.83it/s]
Running loglikelihood requests:  70%|███████   | 42807/60896 [25:38<05:29, 54.96it/s]
Running loglikelihood requests:  70%|███████   | 42844/60896 [25:38<05:16, 57.05it/s]
Running loglikelihood requests:  70%|███████   | 42876/60896 [25:39<05:21, 56.02it/s]
Running loglikelihood requests:  70%|███████   | 42915/60896 [25:40<05:05, 58.81it/s]
Running loglikelihood requests:  71%|███████   | 42947/60896 [25:40<05:14, 57.16it/s]
Running loglikelihood requests:  71%|███████   | 42981/60896 [25:41<05:13, 57.08it/s]
Running loglikelihood requests:  71%|███████   | 43016/60896 [25:41<05:15, 56.59it/s]
Running loglikelihood requests:  71%|███████   | 43053/60896 [25:42<05:06, 58.19it/s]
Running loglikelihood requests:  71%|███████   | 43085/60896 [25:43<05:13, 56.84it/s]
Running loglikelihood requests:  71%|███████   | 43117/60896 [25:43<05:18, 55.90it/s]
Running loglikelihood requests:  71%|███████   | 43149/60896 [25:44<05:21, 55.23it/s]
Running loglikelihood requests:  71%|███████   | 43181/60896 [25:44<05:23, 54.71it/s]
Running loglikelihood requests:  71%|███████   | 43213/60896 [25:45<05:25, 54.40it/s]
Running loglikelihood requests:  71%|███████   | 43245/60896 [25:46<05:25, 54.16it/s]
Running loglikelihood requests:  71%|███████   | 43277/60896 [25:46<05:27, 53.85it/s]
Running loglikelihood requests:  71%|███████   | 43309/60896 [25:47<05:28, 53.54it/s]
Running loglikelihood requests:  71%|███████   | 43342/60896 [25:47<05:24, 54.12it/s]
Running loglikelihood requests:  71%|███████   | 43374/60896 [25:48<05:24, 54.03it/s]
Running loglikelihood requests:  71%|███████▏  | 43407/60896 [25:49<05:27, 53.42it/s]
Running loglikelihood requests:  71%|███████▏  | 43440/60896 [25:49<05:23, 54.03it/s]
Running loglikelihood requests:  71%|███████▏  | 43472/60896 [25:50<05:22, 53.99it/s]
Running loglikelihood requests:  71%|███████▏  | 43507/60896 [25:50<05:13, 55.43it/s]
Running loglikelihood requests:  71%|███████▏  | 43540/60896 [25:51<05:13, 55.45it/s]
Running loglikelihood requests:  72%|███████▏  | 43578/60896 [25:52<04:58, 57.95it/s]
Running loglikelihood requests:  72%|███████▏  | 43610/60896 [25:52<05:05, 56.66it/s]
Running loglikelihood requests:  72%|███████▏  | 43642/60896 [25:53<05:09, 55.79it/s]
Running loglikelihood requests:  72%|███████▏  | 43674/60896 [25:53<05:11, 55.21it/s]
Running loglikelihood requests:  72%|███████▏  | 43709/60896 [25:54<05:05, 56.30it/s]
Running loglikelihood requests:  72%|███████▏  | 43741/60896 [25:55<05:08, 55.53it/s]
Running loglikelihood requests:  72%|███████▏  | 43774/60896 [25:55<05:13, 54.67it/s]
Running loglikelihood requests:  72%|███████▏  | 43806/60896 [25:56<05:14, 54.39it/s]
Running loglikelihood requests:  72%|███████▏  | 43838/60896 [25:56<05:14, 54.20it/s]
Running loglikelihood requests:  72%|███████▏  | 43871/60896 [25:57<05:11, 54.57it/s]
Running loglikelihood requests:  72%|███████▏  | 43903/60896 [25:58<05:12, 54.29it/s]
Running loglikelihood requests:  72%|███████▏  | 43937/60896 [25:58<05:07, 55.10it/s]
Running loglikelihood requests:  72%|███████▏  | 43969/60896 [25:59<05:09, 54.67it/s]
Running loglikelihood requests:  72%|███████▏  | 44002/60896 [25:59<05:09, 54.63it/s]
Running loglikelihood requests:  72%|███████▏  | 44034/60896 [26:00<05:11, 54.14it/s]
Running loglikelihood requests:  72%|███████▏  | 44067/60896 [26:01<05:10, 54.25it/s]
Running loglikelihood requests:  72%|███████▏  | 44100/60896 [26:01<05:09, 54.35it/s]
Running loglikelihood requests:  72%|███████▏  | 44133/60896 [26:02<05:07, 54.43it/s]
Running loglikelihood requests:  73%|███████▎  | 44165/60896 [26:02<05:15, 52.95it/s]
Running loglikelihood requests:  73%|███████▎  | 44205/60896 [26:03<04:53, 56.87it/s]
Running loglikelihood requests:  73%|███████▎  | 44243/60896 [26:04<04:51, 57.20it/s]
Running loglikelihood requests:  73%|███████▎  | 44286/60896 [26:04<04:37, 59.79it/s]
Running loglikelihood requests:  73%|███████▎  | 44318/60896 [26:05<04:53, 56.43it/s]
Running loglikelihood requests:  73%|███████▎  | 44350/60896 [26:06<05:03, 54.53it/s]
Running loglikelihood requests:  73%|███████▎  | 44382/60896 [26:06<05:05, 54.01it/s]
Running loglikelihood requests:  73%|███████▎  | 44414/60896 [26:07<05:07, 53.64it/s]
Running loglikelihood requests:  73%|███████▎  | 44446/60896 [26:07<05:08, 53.40it/s]
Running loglikelihood requests:  73%|███████▎  | 44478/60896 [26:08<05:08, 53.22it/s]
Running loglikelihood requests:  73%|███████▎  | 44510/60896 [26:09<05:08, 53.08it/s]
Running loglikelihood requests:  73%|███████▎  | 44542/60896 [26:09<05:13, 52.17it/s]
Running loglikelihood requests:  73%|███████▎  | 44574/60896 [26:10<05:11, 52.39it/s]
Running loglikelihood requests:  73%|███████▎  | 44606/60896 [26:11<05:09, 52.57it/s]
Running loglikelihood requests:  73%|███████▎  | 44639/60896 [26:11<05:05, 53.18it/s]
Running loglikelihood requests:  73%|███████▎  | 44672/60896 [26:12<05:02, 53.70it/s]
Running loglikelihood requests:  73%|███████▎  | 44705/60896 [26:12<04:59, 54.03it/s]
Running loglikelihood requests:  73%|███████▎  | 44738/60896 [26:13<04:57, 54.24it/s]
Running loglikelihood requests:  74%|███████▎  | 44770/60896 [26:14<04:59, 53.86it/s]
Running loglikelihood requests:  74%|███████▎  | 44802/60896 [26:14<04:59, 53.66it/s]
Running loglikelihood requests:  74%|███████▎  | 44834/60896 [26:15<05:00, 53.50it/s]
Running loglikelihood requests:  74%|███████▎  | 44868/60896 [26:15<04:54, 54.41it/s]
Running loglikelihood requests:  74%|███████▎  | 44900/60896 [26:16<04:55, 54.04it/s]
Running loglikelihood requests:  74%|███████▍  | 44941/60896 [26:17<04:38, 57.27it/s]
Running loglikelihood requests:  74%|███████▍  | 44974/60896 [26:17<04:41, 56.53it/s]
Running loglikelihood requests:  74%|███████▍  | 45006/60896 [26:18<04:46, 55.55it/s]
Running loglikelihood requests:  74%|███████▍  | 45042/60896 [26:18<04:39, 56.77it/s]
Running loglikelihood requests:  74%|███████▍  | 45074/60896 [26:19<04:44, 55.70it/s]
Running loglikelihood requests:  74%|███████▍  | 45106/60896 [26:20<04:47, 54.93it/s]
Running loglikelihood requests:  74%|███████▍  | 45138/60896 [26:20<04:49, 54.41it/s]
Running loglikelihood requests:  74%|███████▍  | 45170/60896 [26:21<04:51, 54.04it/s]
Running loglikelihood requests:  74%|███████▍  | 45202/60896 [26:21<04:51, 53.79it/s]
Running loglikelihood requests:  74%|███████▍  | 45234/60896 [26:22<04:52, 53.63it/s]
Running loglikelihood requests:  74%|███████▍  | 45266/60896 [26:23<04:52, 53.51it/s]
Running loglikelihood requests:  74%|███████▍  | 45298/60896 [26:23<04:51, 53.45it/s]
Running loglikelihood requests:  74%|███████▍  | 45330/60896 [26:24<04:51, 53.34it/s]
Running loglikelihood requests:  74%|███████▍  | 45363/60896 [26:24<04:53, 52.83it/s]
Running loglikelihood requests:  75%|███████▍  | 45395/60896 [26:25<04:51, 53.20it/s]
Running loglikelihood requests:  75%|███████▍  | 45427/60896 [26:26<04:49, 53.45it/s]
Running loglikelihood requests:  75%|███████▍  | 45459/60896 [26:26<04:47, 53.63it/s]
Running loglikelihood requests:  75%|███████▍  | 45492/60896 [26:27<04:43, 54.31it/s]
Running loglikelihood requests:  75%|███████▍  | 45525/60896 [26:27<04:40, 54.75it/s]
Running loglikelihood requests:  75%|███████▍  | 45558/60896 [26:28<04:38, 55.08it/s]
Running loglikelihood requests:  75%|███████▍  | 45595/60896 [26:29<04:27, 57.28it/s]
Running loglikelihood requests:  75%|███████▍  | 45628/60896 [26:29<04:28, 56.87it/s]
Running loglikelihood requests:  75%|███████▍  | 45662/60896 [26:30<04:27, 57.05it/s]
Running loglikelihood requests:  75%|███████▌  | 45694/60896 [26:30<04:30, 56.14it/s]
Running loglikelihood requests:  75%|███████▌  | 45726/60896 [26:31<04:33, 55.50it/s]
Running loglikelihood requests:  75%|███████▌  | 45758/60896 [26:32<04:39, 54.15it/s]
Running loglikelihood requests:  75%|███████▌  | 45794/60896 [26:32<04:29, 56.12it/s]
Running loglikelihood requests:  75%|███████▌  | 45827/60896 [26:33<04:28, 56.03it/s]
Running loglikelihood requests:  75%|███████▌  | 45859/60896 [26:33<04:31, 55.45it/s]
Running loglikelihood requests:  75%|███████▌  | 45891/60896 [26:34<04:32, 55.06it/s]
Running loglikelihood requests:  75%|███████▌  | 45923/60896 [26:35<04:33, 54.78it/s]
Running loglikelihood requests:  75%|███████▌  | 45955/60896 [26:35<04:33, 54.60it/s]
Running loglikelihood requests:  76%|███████▌  | 45987/60896 [26:36<04:33, 54.45it/s]
Running loglikelihood requests:  76%|███████▌  | 46020/60896 [26:36<04:31, 54.87it/s]
Running loglikelihood requests:  76%|███████▌  | 46052/60896 [26:37<04:31, 54.64it/s]
Running loglikelihood requests:  76%|███████▌  | 46084/60896 [26:38<04:31, 54.48it/s]
Running loglikelihood requests:  76%|███████▌  | 46116/60896 [26:38<04:31, 54.45it/s]
Running loglikelihood requests:  76%|███████▌  | 46148/60896 [26:39<04:36, 53.41it/s]
Running loglikelihood requests:  76%|███████▌  | 46181/60896 [26:39<04:31, 54.23it/s]
Running loglikelihood requests:  76%|███████▌  | 46214/60896 [26:40<04:28, 54.78it/s]
Running loglikelihood requests:  76%|███████▌  | 46246/60896 [26:40<04:28, 54.48it/s]
Running loglikelihood requests:  76%|███████▌  | 46278/60896 [26:41<04:28, 54.44it/s]
Running loglikelihood requests:  76%|███████▌  | 46310/60896 [26:42<04:28, 54.41it/s]
Running loglikelihood requests:  76%|███████▌  | 46346/60896 [26:42<04:18, 56.39it/s]
Running loglikelihood requests:  76%|███████▌  | 46385/60896 [26:43<04:04, 59.32it/s]
Running loglikelihood requests:  76%|███████▌  | 46417/60896 [26:43<04:10, 57.78it/s]
Running loglikelihood requests:  76%|███████▋  | 46452/60896 [26:44<04:07, 58.29it/s]
Running loglikelihood requests:  76%|███████▋  | 46484/60896 [26:45<04:16, 56.13it/s]
Running loglikelihood requests:  76%|███████▋  | 46519/60896 [26:45<04:11, 57.14it/s]
Running loglikelihood requests:  76%|███████▋  | 46554/60896 [26:46<04:08, 57.82it/s]
Running loglikelihood requests:  77%|███████▋  | 46588/60896 [26:46<04:08, 57.50it/s]
Running loglikelihood requests:  77%|███████▋  | 46620/60896 [26:47<04:13, 56.26it/s]
Running loglikelihood requests:  77%|███████▋  | 46652/60896 [26:48<04:15, 55.67it/s]
Running loglikelihood requests:  77%|███████▋  | 46684/60896 [26:48<04:17, 55.25it/s]
Running loglikelihood requests:  77%|███████▋  | 46716/60896 [26:49<04:17, 54.99it/s]
Running loglikelihood requests:  77%|███████▋  | 46750/60896 [26:49<04:13, 55.83it/s]
Running loglikelihood requests:  77%|███████▋  | 46782/60896 [26:50<04:14, 55.40it/s]
Running loglikelihood requests:  77%|███████▋  | 46815/60896 [26:51<04:13, 55.62it/s]
Running loglikelihood requests:  77%|███████▋  | 46847/60896 [26:51<04:19, 54.23it/s]
Running loglikelihood requests:  77%|███████▋  | 46879/60896 [26:52<04:09, 56.25it/s]
Running loglikelihood requests:  77%|███████▋  | 46911/60896 [26:52<04:02, 57.74it/s]
Running loglikelihood requests:  77%|███████▋  | 46943/60896 [26:53<03:56, 58.90it/s]
Running loglikelihood requests:  77%|███████▋  | 46976/60896 [26:53<03:50, 60.31it/s]
Running loglikelihood requests:  77%|███████▋  | 47008/60896 [26:54<03:48, 60.70it/s]
Running loglikelihood requests:  77%|███████▋  | 47044/60896 [26:54<03:38, 63.28it/s]
Running loglikelihood requests:  77%|███████▋  | 47082/60896 [26:55<03:28, 66.17it/s]
Running loglikelihood requests:  77%|███████▋  | 47118/60896 [26:55<03:25, 67.16it/s]
Running loglikelihood requests:  77%|███████▋  | 47156/60896 [26:56<03:19, 68.97it/s]
Running loglikelihood requests:  77%|███████▋  | 47190/60896 [26:56<03:21, 67.90it/s]
Running loglikelihood requests:  78%|███████▊  | 47226/60896 [26:57<03:19, 68.35it/s]
Running loglikelihood requests:  78%|███████▊  | 47261/60896 [26:57<03:24, 66.68it/s]
Running loglikelihood requests:  78%|███████▊  | 47293/60896 [26:58<03:28, 65.25it/s]
Running loglikelihood requests:  78%|███████▊  | 47325/60896 [26:58<03:31, 64.15it/s]
Running loglikelihood requests:  78%|███████▊  | 47358/60896 [26:59<03:31, 63.92it/s]
Running loglikelihood requests:  78%|███████▊  | 47390/60896 [27:00<03:33, 63.24it/s]
Running loglikelihood requests:  78%|███████▊  | 47422/60896 [27:00<03:34, 62.77it/s]
Running loglikelihood requests:  78%|███████▊  | 47454/60896 [27:01<03:35, 62.41it/s]
Running loglikelihood requests:  78%|███████▊  | 47486/60896 [27:01<03:36, 62.00it/s]
Running loglikelihood requests:  78%|███████▊  | 47518/60896 [27:02<03:36, 61.87it/s]
Running loglikelihood requests:  78%|███████▊  | 47550/60896 [27:02<03:35, 61.81it/s]
Running loglikelihood requests:  78%|███████▊  | 47583/60896 [27:03<03:33, 62.32it/s]
Running loglikelihood requests:  78%|███████▊  | 47616/60896 [27:03<03:37, 61.18it/s]
Running loglikelihood requests:  78%|███████▊  | 47648/60896 [27:04<03:35, 61.37it/s]
Running loglikelihood requests:  78%|███████▊  | 47680/60896 [27:04<03:34, 61.49it/s]
Running loglikelihood requests:  78%|███████▊  | 47713/60896 [27:05<03:31, 62.19it/s]
Running loglikelihood requests:  78%|███████▊  | 47746/60896 [27:05<03:30, 62.61it/s]
Running loglikelihood requests:  78%|███████▊  | 47780/60896 [27:06<03:26, 63.54it/s]
Running loglikelihood requests:  79%|███████▊  | 47812/60896 [27:06<03:27, 62.93it/s]
Running loglikelihood requests:  79%|███████▊  | 47854/60896 [27:07<03:11, 68.24it/s]
Running loglikelihood requests:  79%|███████▊  | 47886/60896 [27:07<03:16, 66.14it/s]
Running loglikelihood requests:  79%|███████▊  | 47919/60896 [27:08<03:18, 65.42it/s]
Running loglikelihood requests:  79%|███████▉  | 47956/60896 [27:08<03:12, 67.16it/s]
Running loglikelihood requests:  79%|███████▉  | 47997/60896 [27:09<03:02, 70.72it/s]
Running loglikelihood requests:  79%|███████▉  | 48031/60896 [27:09<03:05, 69.17it/s]
Running loglikelihood requests:  79%|███████▉  | 48063/60896 [27:10<03:11, 66.97it/s]
Running loglikelihood requests:  79%|███████▉  | 48095/60896 [27:11<03:20, 63.99it/s]
Running loglikelihood requests:  79%|███████▉  | 48127/60896 [27:11<03:21, 63.39it/s]
Running loglikelihood requests:  79%|███████▉  | 48159/60896 [27:12<03:22, 62.93it/s]
Running loglikelihood requests:  79%|███████▉  | 48191/60896 [27:12<03:22, 62.64it/s]
Running loglikelihood requests:  79%|███████▉  | 48223/60896 [27:13<03:23, 62.36it/s]
Running loglikelihood requests:  79%|███████▉  | 48256/60896 [27:13<03:21, 62.63it/s]
Running loglikelihood requests:  79%|███████▉  | 48288/60896 [27:14<03:22, 62.35it/s]
Running loglikelihood requests:  79%|███████▉  | 48320/60896 [27:14<03:22, 62.25it/s]
Running loglikelihood requests:  79%|███████▉  | 48352/60896 [27:15<03:21, 62.20it/s]
Running loglikelihood requests:  79%|███████▉  | 48384/60896 [27:15<03:21, 62.15it/s]
Running loglikelihood requests:  80%|███████▉  | 48420/60896 [27:16<03:13, 64.41it/s]
Running loglikelihood requests:  80%|███████▉  | 48458/60896 [27:16<03:05, 67.16it/s]
Running loglikelihood requests:  80%|███████▉  | 48491/60896 [27:17<03:07, 66.20it/s]
Running loglikelihood requests:  80%|███████▉  | 48523/60896 [27:17<03:15, 63.28it/s]
Running loglikelihood requests:  80%|███████▉  | 48555/60896 [27:18<03:16, 62.87it/s]
Running loglikelihood requests:  80%|███████▉  | 48587/60896 [27:18<03:16, 62.54it/s]
Running loglikelihood requests:  80%|███████▉  | 48624/60896 [27:19<03:08, 65.18it/s]
Running loglikelihood requests:  80%|███████▉  | 48656/60896 [27:19<03:11, 64.02it/s]
Running loglikelihood requests:  80%|███████▉  | 48688/60896 [27:20<03:12, 63.36it/s]
Running loglikelihood requests:  80%|████████  | 48720/60896 [27:20<03:13, 62.93it/s]
Running loglikelihood requests:  80%|████████  | 48752/60896 [27:21<03:14, 62.59it/s]
Running loglikelihood requests:  80%|████████  | 48784/60896 [27:21<03:14, 62.22it/s]
Running loglikelihood requests:  80%|████████  | 48816/60896 [27:22<03:14, 62.00it/s]
Running loglikelihood requests:  80%|████████  | 48850/60896 [27:22<03:11, 63.04it/s]
Running loglikelihood requests:  80%|████████  | 48882/60896 [27:23<03:11, 62.77it/s]
Running loglikelihood requests:  80%|████████  | 48914/60896 [27:24<03:12, 62.39it/s]
Running loglikelihood requests:  80%|████████  | 48946/60896 [27:24<03:12, 62.06it/s]
Running loglikelihood requests:  80%|████████  | 48978/60896 [27:25<03:16, 60.74it/s]
Running loglikelihood requests:  80%|████████  | 49010/60896 [27:25<03:14, 61.13it/s]
Running loglikelihood requests:  81%|████████  | 49050/60896 [27:26<02:59, 65.91it/s]
Running loglikelihood requests:  81%|████████  | 49085/60896 [27:26<02:57, 66.49it/s]
Running loglikelihood requests:  81%|████████  | 49120/60896 [27:27<02:56, 66.87it/s]
Running loglikelihood requests:  81%|████████  | 49156/60896 [27:27<02:53, 67.84it/s]
Running loglikelihood requests:  81%|████████  | 49191/60896 [27:28<02:52, 67.96it/s]
Running loglikelihood requests:  81%|████████  | 49229/60896 [27:28<02:47, 69.74it/s]
Running loglikelihood requests:  81%|████████  | 49261/60896 [27:29<02:52, 67.51it/s]
Running loglikelihood requests:  81%|████████  | 49300/60896 [27:29<02:45, 70.06it/s]
Running loglikelihood requests:  81%|████████  | 49346/60896 [27:30<02:32, 75.89it/s]
Running loglikelihood requests:  81%|████████  | 49378/60896 [27:30<02:40, 71.85it/s]
Running loglikelihood requests:  81%|████████  | 49410/60896 [27:31<02:46, 69.02it/s]
Running loglikelihood requests:  81%|████████  | 49442/60896 [27:31<02:51, 66.95it/s]
Running loglikelihood requests:  81%|████████  | 49474/60896 [27:32<02:54, 65.58it/s]
Running loglikelihood requests:  81%|████████▏ | 49506/60896 [27:32<02:59, 63.45it/s]
Running loglikelihood requests:  81%|████████▏ | 49538/60896 [27:33<02:59, 63.17it/s]
Running loglikelihood requests:  81%|████████▏ | 49570/60896 [27:33<02:57, 63.65it/s]
Running loglikelihood requests:  81%|████████▏ | 49602/60896 [27:34<02:56, 64.01it/s]
Running loglikelihood requests:  82%|████████▏ | 49634/60896 [27:34<02:55, 64.19it/s]
Running loglikelihood requests:  82%|████████▏ | 49666/60896 [27:35<02:54, 64.30it/s]
Running loglikelihood requests:  82%|████████▏ | 49701/60896 [27:35<02:49, 66.22it/s]
Running loglikelihood requests:  82%|████████▏ | 49741/60896 [27:36<02:38, 70.53it/s]
Running loglikelihood requests:  82%|████████▏ | 49775/60896 [27:36<02:39, 69.85it/s]
Running loglikelihood requests:  82%|████████▏ | 49813/60896 [27:37<02:34, 71.79it/s]
Running loglikelihood requests:  82%|████████▏ | 49845/60896 [27:37<02:39, 69.47it/s]
Running loglikelihood requests:  82%|████████▏ | 49880/60896 [27:38<02:37, 69.79it/s]
Running loglikelihood requests:  82%|████████▏ | 49912/60896 [27:38<02:41, 68.22it/s]
Running loglikelihood requests:  82%|████████▏ | 49957/60896 [27:39<02:25, 74.95it/s]
Running loglikelihood requests:  82%|████████▏ | 49991/60896 [27:39<02:32, 71.45it/s]
Running loglikelihood requests:  82%|████████▏ | 50023/60896 [27:40<02:36, 69.56it/s]
Running loglikelihood requests:  82%|████████▏ | 50055/60896 [27:40<02:39, 67.85it/s]
Running loglikelihood requests:  82%|████████▏ | 50087/60896 [27:41<02:41, 66.91it/s]
Running loglikelihood requests:  82%|████████▏ | 50119/60896 [27:41<02:42, 66.23it/s]
Running loglikelihood requests:  82%|████████▏ | 50151/60896 [27:42<02:43, 65.74it/s]
Running loglikelihood requests:  82%|████████▏ | 50183/60896 [27:42<02:43, 65.37it/s]
Running loglikelihood requests:  82%|████████▏ | 50215/60896 [27:43<02:43, 65.19it/s]
Running loglikelihood requests:  83%|████████▎ | 50247/60896 [27:43<02:43, 65.02it/s]
Running loglikelihood requests:  83%|████████▎ | 50279/60896 [27:44<02:43, 64.95it/s]
Running loglikelihood requests:  83%|████████▎ | 50314/60896 [27:44<02:38, 66.74it/s]
Running loglikelihood requests:  83%|████████▎ | 50347/60896 [27:45<02:37, 66.79it/s]
Running loglikelihood requests:  83%|████████▎ | 50385/60896 [27:45<02:30, 69.89it/s]
Running loglikelihood requests:  83%|████████▎ | 50418/60896 [27:46<02:31, 68.96it/s]
Running loglikelihood requests:  83%|████████▎ | 50459/60896 [27:46<02:26, 71.31it/s]
Running loglikelihood requests:  83%|████████▎ | 50494/60896 [27:47<02:26, 71.07it/s]
Running loglikelihood requests:  83%|████████▎ | 50530/60896 [27:47<02:24, 71.62it/s]
Running loglikelihood requests:  83%|████████▎ | 50562/60896 [27:48<02:28, 69.63it/s]
Running loglikelihood requests:  83%|████████▎ | 50597/60896 [27:48<02:27, 69.95it/s]
Running loglikelihood requests:  83%|████████▎ | 50633/60896 [27:49<02:25, 70.78it/s]
Running loglikelihood requests:  83%|████████▎ | 50671/60896 [27:49<02:20, 72.63it/s]
Running loglikelihood requests:  83%|████████▎ | 50703/60896 [27:50<02:24, 70.32it/s]
Running loglikelihood requests:  83%|████████▎ | 50735/60896 [27:50<02:27, 68.73it/s]
Running loglikelihood requests:  83%|████████▎ | 50767/60896 [27:51<02:29, 67.59it/s]
Running loglikelihood requests:  83%|████████▎ | 50799/60896 [27:51<02:31, 66.76it/s]
Running loglikelihood requests:  83%|████████▎ | 50831/60896 [27:52<02:32, 66.14it/s]
Running loglikelihood requests:  84%|████████▎ | 50863/60896 [27:52<02:35, 64.35it/s]
Running loglikelihood requests:  84%|████████▎ | 50895/60896 [27:53<02:34, 64.63it/s]
Running loglikelihood requests:  84%|████████▎ | 50927/60896 [27:53<02:33, 64.84it/s]
Running loglikelihood requests:  84%|████████▎ | 50959/60896 [27:54<02:32, 64.95it/s]
Running loglikelihood requests:  84%|████████▎ | 50994/60896 [27:54<02:28, 66.83it/s]
Running loglikelihood requests:  84%|████████▍ | 51035/60896 [27:55<02:17, 71.74it/s]
Running loglikelihood requests:  84%|████████▍ | 51067/60896 [27:55<02:20, 69.78it/s]
Running loglikelihood requests:  84%|████████▍ | 51102/60896 [27:56<02:19, 70.22it/s]
Running loglikelihood requests:  84%|████████▍ | 51144/60896 [27:56<02:10, 74.72it/s]
Running loglikelihood requests:  84%|████████▍ | 51181/60896 [27:57<02:09, 74.89it/s]
Running loglikelihood requests:  84%|████████▍ | 51213/60896 [27:57<02:14, 71.96it/s]
Running loglikelihood requests:  84%|████████▍ | 51246/60896 [27:58<02:16, 70.55it/s]
Running loglikelihood requests:  84%|████████▍ | 51282/60896 [27:58<02:14, 71.31it/s]
Running loglikelihood requests:  84%|████████▍ | 51317/60896 [27:59<02:14, 71.20it/s]
Running loglikelihood requests:  84%|████████▍ | 51349/60896 [27:59<02:20, 67.78it/s]
Running loglikelihood requests:  84%|████████▍ | 51381/60896 [28:00<02:21, 67.01it/s]
Running loglikelihood requests:  84%|████████▍ | 51413/60896 [28:00<02:22, 66.46it/s]
Running loglikelihood requests:  84%|████████▍ | 51445/60896 [28:01<02:23, 66.03it/s]
Running loglikelihood requests:  85%|████████▍ | 51477/60896 [28:01<02:23, 65.61it/s]
Running loglikelihood requests:  85%|████████▍ | 51509/60896 [28:02<02:23, 65.58it/s]
Running loglikelihood requests:  85%|████████▍ | 51541/60896 [28:02<02:22, 65.62it/s]
Running loglikelihood requests:  85%|████████▍ | 51573/60896 [28:03<02:22, 65.63it/s]
Running loglikelihood requests:  85%|████████▍ | 51605/60896 [28:03<02:21, 65.51it/s]
Running loglikelihood requests:  85%|████████▍ | 51642/60896 [28:04<02:15, 68.53it/s]
Running loglikelihood requests:  85%|████████▍ | 51684/60896 [28:04<02:04, 73.75it/s]
Running loglikelihood requests:  85%|████████▍ | 51716/60896 [28:05<02:08, 71.28it/s]
Running loglikelihood requests:  85%|████████▍ | 51749/60896 [28:05<02:10, 70.14it/s]
Running loglikelihood requests:  85%|████████▌ | 51783/60896 [28:06<02:12, 68.65it/s]
Running loglikelihood requests:  85%|████████▌ | 51822/60896 [28:06<02:06, 71.88it/s]
Running loglikelihood requests:  85%|████████▌ | 51854/60896 [28:07<02:09, 69.87it/s]
Running loglikelihood requests:  85%|████████▌ | 51886/60896 [28:07<02:11, 68.49it/s]
Running loglikelihood requests:  85%|████████▌ | 51918/60896 [28:08<02:12, 67.52it/s]
Running loglikelihood requests:  85%|████████▌ | 51956/60896 [28:08<02:06, 70.53it/s]
Running loglikelihood requests:  85%|████████▌ | 51988/60896 [28:09<02:09, 68.99it/s]
Running loglikelihood requests:  85%|████████▌ | 52020/60896 [28:09<02:10, 67.92it/s]
Running loglikelihood requests:  85%|████████▌ | 52052/60896 [28:09<02:11, 67.15it/s]
Running loglikelihood requests:  86%|████████▌ | 52084/60896 [28:10<02:12, 66.61it/s]
Running loglikelihood requests:  86%|████████▌ | 52116/60896 [28:10<02:07, 68.83it/s]
Running loglikelihood requests:  86%|████████▌ | 52148/60896 [28:11<02:03, 70.57it/s]
Running loglikelihood requests:  86%|████████▌ | 52180/60896 [28:11<02:04, 69.84it/s]
Running loglikelihood requests:  86%|████████▌ | 52212/60896 [28:12<02:01, 71.35it/s]
Running loglikelihood requests:  86%|████████▌ | 52253/60896 [28:12<01:49, 78.58it/s]
Running loglikelihood requests:  86%|████████▌ | 52285/60896 [28:13<01:50, 77.58it/s]
Running loglikelihood requests:  86%|████████▌ | 52320/60896 [28:13<01:48, 78.93it/s]
Running loglikelihood requests:  86%|████████▌ | 52358/60896 [28:13<01:44, 81.87it/s]
Running loglikelihood requests:  86%|████████▌ | 52394/60896 [28:14<01:42, 82.59it/s]
Running loglikelihood requests:  86%|████████▌ | 52426/60896 [28:14<01:45, 80.38it/s]
Running loglikelihood requests:  86%|████████▌ | 52458/60896 [28:15<01:46, 78.88it/s]
Running loglikelihood requests:  86%|████████▌ | 52490/60896 [28:15<01:47, 77.86it/s]
Running loglikelihood requests:  86%|████████▋ | 52523/60896 [28:16<01:47, 77.73it/s]
Running loglikelihood requests:  86%|████████▋ | 52558/60896 [28:16<01:45, 79.02it/s]
Running loglikelihood requests:  86%|████████▋ | 52590/60896 [28:16<01:46, 77.79it/s]
Running loglikelihood requests:  86%|████████▋ | 52622/60896 [28:17<01:47, 77.02it/s]
Running loglikelihood requests:  86%|████████▋ | 52654/60896 [28:17<01:51, 74.21it/s]
Running loglikelihood requests:  87%|████████▋ | 52686/60896 [28:18<01:50, 74.63it/s]
Running loglikelihood requests:  87%|████████▋ | 52718/60896 [28:18<01:49, 74.91it/s]
Running loglikelihood requests:  87%|████████▋ | 52750/60896 [28:19<01:48, 75.11it/s]
Running loglikelihood requests:  87%|████████▋ | 52782/60896 [28:19<01:47, 75.28it/s]
Running loglikelihood requests:  87%|████████▋ | 52822/60896 [28:19<01:39, 81.01it/s]
Running loglikelihood requests:  87%|████████▋ | 52859/60896 [28:20<01:36, 82.92it/s]
Running loglikelihood requests:  87%|████████▋ | 52892/60896 [28:20<01:38, 81.51it/s]
Running loglikelihood requests:  87%|████████▋ | 52925/60896 [28:21<01:38, 80.61it/s]
Running loglikelihood requests:  87%|████████▋ | 52957/60896 [28:21<01:40, 79.19it/s]
Running loglikelihood requests:  87%|████████▋ | 52993/60896 [28:22<01:37, 81.03it/s]
Running loglikelihood requests:  87%|████████▋ | 53033/60896 [28:22<01:32, 85.06it/s]
Running loglikelihood requests:  87%|████████▋ | 53067/60896 [28:22<01:33, 83.65it/s]
Running loglikelihood requests:  87%|████████▋ | 53102/60896 [28:23<01:33, 83.42it/s]
Running loglikelihood requests:  87%|████████▋ | 53134/60896 [28:23<01:38, 78.97it/s]
Running loglikelihood requests:  87%|████████▋ | 53168/60896 [28:24<01:37, 79.37it/s]
Running loglikelihood requests:  87%|████████▋ | 53203/60896 [28:24<01:35, 80.23it/s]
Running loglikelihood requests:  87%|████████▋ | 53245/60896 [28:25<01:29, 85.92it/s]
Running loglikelihood requests:  87%|████████▋ | 53280/60896 [28:25<01:29, 84.96it/s]
Running loglikelihood requests:  88%|████████▊ | 53312/60896 [28:25<01:32, 82.19it/s]
Running loglikelihood requests:  88%|████████▊ | 53344/60896 [28:26<01:34, 80.24it/s]
Running loglikelihood requests:  88%|████████▊ | 53376/60896 [28:26<01:35, 78.90it/s]
Running loglikelihood requests:  88%|████████▊ | 53408/60896 [28:27<01:35, 78.26it/s]
Running loglikelihood requests:  88%|████████▊ | 53440/60896 [28:27<01:35, 77.83it/s]
Running loglikelihood requests:  88%|████████▊ | 53472/60896 [28:27<01:35, 77.48it/s]
Running loglikelihood requests:  88%|████████▊ | 53515/60896 [28:28<01:26, 85.11it/s]
Running loglikelihood requests:  88%|████████▊ | 53547/60896 [28:28<01:28, 82.60it/s]
Running loglikelihood requests:  88%|████████▊ | 53579/60896 [28:29<01:30, 80.79it/s]
Running loglikelihood requests:  88%|████████▊ | 53617/60896 [28:29<01:26, 84.04it/s]
Running loglikelihood requests:  88%|████████▊ | 53650/60896 [28:30<01:30, 80.25it/s]
Running loglikelihood requests:  88%|████████▊ | 53683/60896 [28:30<01:30, 80.08it/s]
Running loglikelihood requests:  88%|████████▊ | 53725/60896 [28:30<01:23, 86.21it/s]
Running loglikelihood requests:  88%|████████▊ | 53760/60896 [28:31<01:23, 85.61it/s]
Running loglikelihood requests:  88%|████████▊ | 53792/60896 [28:31<01:25, 83.09it/s]
Running loglikelihood requests:  88%|████████▊ | 53828/60896 [28:32<01:24, 84.10it/s]
Running loglikelihood requests:  88%|████████▊ | 53863/60896 [28:32<01:23, 84.09it/s]
Running loglikelihood requests:  89%|████████▊ | 53897/60896 [28:33<01:23, 83.36it/s]
Running loglikelihood requests:  89%|████████▊ | 53939/60896 [28:33<01:18, 88.57it/s]
Running loglikelihood requests:  89%|████████▊ | 53976/60896 [28:33<01:18, 88.59it/s]
Running loglikelihood requests:  89%|████████▊ | 54008/60896 [28:34<01:20, 85.06it/s]
Running loglikelihood requests:  89%|████████▊ | 54040/60896 [28:34<01:23, 82.52it/s]
Running loglikelihood requests:  89%|████████▉ | 54072/60896 [28:35<01:24, 80.81it/s]
Running loglikelihood requests:  89%|████████▉ | 54104/60896 [28:35<01:25, 79.64it/s]
Running loglikelihood requests:  89%|████████▉ | 54136/60896 [28:35<01:28, 76.42it/s]
Running loglikelihood requests:  89%|████████▉ | 54168/60896 [28:36<01:27, 76.69it/s]
Running loglikelihood requests:  89%|████████▉ | 54207/60896 [28:36<01:21, 81.73it/s]
Running loglikelihood requests:  89%|████████▉ | 54241/60896 [28:37<01:21, 81.88it/s]
Running loglikelihood requests:  89%|████████▉ | 54276/60896 [28:37<01:20, 82.70it/s]
Running loglikelihood requests:  89%|████████▉ | 54312/60896 [28:38<01:18, 84.02it/s]
Running loglikelihood requests:  89%|████████▉ | 54346/60896 [28:38<01:18, 83.58it/s]
Running loglikelihood requests:  89%|████████▉ | 54378/60896 [28:38<01:19, 81.73it/s]
Running loglikelihood requests:  89%|████████▉ | 54412/60896 [28:39<01:19, 81.90it/s]
Running loglikelihood requests:  89%|████████▉ | 54446/60896 [28:39<01:18, 82.09it/s]
Running loglikelihood requests:  89%|████████▉ | 54481/60896 [28:40<01:17, 83.08it/s]
Running loglikelihood requests:  90%|████████▉ | 54513/60896 [28:40<01:18, 81.45it/s]
Running loglikelihood requests:  90%|████████▉ | 54549/60896 [28:40<01:16, 82.92it/s]
Running loglikelihood requests:  90%|████████▉ | 54582/60896 [28:41<01:16, 82.02it/s]
Running loglikelihood requests:  90%|████████▉ | 54616/60896 [28:41<01:16, 82.16it/s]
Running loglikelihood requests:  90%|████████▉ | 54648/60896 [28:42<01:17, 80.75it/s]
Running loglikelihood requests:  90%|████████▉ | 54684/60896 [28:42<01:15, 82.62it/s]
Running loglikelihood requests:  90%|████████▉ | 54716/60896 [28:43<01:18, 78.69it/s]
Running loglikelihood requests:  90%|████████▉ | 54748/60896 [28:43<01:18, 78.25it/s]
Running loglikelihood requests:  90%|████████▉ | 54780/60896 [28:43<01:18, 77.87it/s]
Running loglikelihood requests:  90%|█████████ | 54812/60896 [28:44<01:18, 77.58it/s]
Running loglikelihood requests:  90%|█████████ | 54844/60896 [28:44<01:17, 78.25it/s]
Running loglikelihood requests:  90%|█████████ | 54876/60896 [28:45<01:16, 78.69it/s]
Running loglikelihood requests:  90%|█████████ | 54908/60896 [28:45<01:15, 79.12it/s]
Running loglikelihood requests:  90%|█████████ | 54943/60896 [28:45<01:12, 81.59it/s]
Running loglikelihood requests:  90%|█████████ | 54978/60896 [28:46<01:11, 83.30it/s]
Running loglikelihood requests:  90%|█████████ | 55011/60896 [28:46<01:11, 82.42it/s]
Running loglikelihood requests:  90%|█████████ | 55047/60896 [28:47<01:09, 84.44it/s]
Running loglikelihood requests:  90%|█████████ | 55081/60896 [28:47<01:08, 84.59it/s]
Running loglikelihood requests:  91%|█████████ | 55116/60896 [28:47<01:07, 85.33it/s]
Running loglikelihood requests:  91%|█████████ | 55154/60896 [28:48<01:05, 88.08it/s]
Running loglikelihood requests:  91%|█████████ | 55192/60896 [28:48<01:03, 90.06it/s]
Running loglikelihood requests:  91%|█████████ | 55227/60896 [28:49<01:03, 89.12it/s]
Running loglikelihood requests:  91%|█████████ | 55262/60896 [28:49<01:03, 88.53it/s]
Running loglikelihood requests:  91%|█████████ | 55294/60896 [28:49<01:06, 83.75it/s]
Running loglikelihood requests:  91%|█████████ | 55330/60896 [28:50<01:05, 85.55it/s]
Running loglikelihood requests:  91%|█████████ | 55369/60896 [28:50<01:01, 89.17it/s]
Running loglikelihood requests:  91%|█████████ | 55404/60896 [28:51<01:01, 88.76it/s]
Running loglikelihood requests:  91%|█████████ | 55436/60896 [28:51<01:03, 86.21it/s]
Running loglikelihood requests:  91%|█████████ | 55468/60896 [28:51<01:04, 84.40it/s]
Running loglikelihood requests:  91%|█████████ | 55500/60896 [28:52<01:05, 82.96it/s]
Running loglikelihood requests:  91%|█████████ | 55532/60896 [28:52<01:05, 81.86it/s]
Running loglikelihood requests:  91%|█████████ | 55564/60896 [28:53<01:05, 81.38it/s]
Running loglikelihood requests:  91%|█████████▏| 55596/60896 [28:53<01:05, 81.08it/s]
Running loglikelihood requests:  91%|█████████▏| 55630/60896 [28:53<01:03, 82.33it/s]
Running loglikelihood requests:  91%|█████████▏| 55669/60896 [28:54<01:00, 86.99it/s]
Running loglikelihood requests:  91%|█████████▏| 55703/60896 [28:54<01:00, 86.47it/s]
Running loglikelihood requests:  92%|█████████▏| 55736/60896 [28:55<01:00, 85.42it/s]
Running loglikelihood requests:  92%|█████████▏| 55768/60896 [28:55<01:01, 83.95it/s]
Running loglikelihood requests:  92%|█████████▏| 55800/60896 [28:55<01:03, 80.66it/s]
Running loglikelihood requests:  92%|█████████▏| 55837/60896 [28:56<00:59, 84.33it/s]
Running loglikelihood requests:  92%|█████████▏| 55874/60896 [28:56<00:57, 86.85it/s]
Running loglikelihood requests:  92%|█████████▏| 55911/60896 [28:57<00:56, 88.55it/s]
Running loglikelihood requests:  92%|█████████▏| 55946/60896 [28:57<00:56, 88.32it/s]
Running loglikelihood requests:  92%|█████████▏| 55978/60896 [28:57<00:57, 86.09it/s]
Running loglikelihood requests:  92%|█████████▏| 56016/60896 [28:58<00:54, 88.87it/s]
Running loglikelihood requests:  92%|█████████▏| 56052/60896 [28:58<00:54, 89.24it/s]
Running loglikelihood requests:  92%|█████████▏| 56088/60896 [28:59<00:53, 89.49it/s]
Running loglikelihood requests:  92%|█████████▏| 56120/60896 [28:59<00:55, 86.69it/s]
Running loglikelihood requests:  92%|█████████▏| 56155/60896 [28:59<00:54, 86.89it/s]
Running loglikelihood requests:  92%|█████████▏| 56188/60896 [29:00<00:54, 85.74it/s]
Running loglikelihood requests:  92%|█████████▏| 56220/60896 [29:00<00:55, 83.81it/s]
Running loglikelihood requests:  92%|█████████▏| 56252/60896 [29:01<00:56, 82.80it/s]
Running loglikelihood requests:  92%|█████████▏| 56284/60896 [29:01<00:56, 81.56it/s]
Running loglikelihood requests:  92%|█████████▏| 56316/60896 [29:01<00:56, 81.63it/s]
Running loglikelihood requests:  93%|█████████▎| 56348/60896 [29:02<00:55, 81.78it/s]
Running loglikelihood requests:  93%|█████████▎| 56390/60896 [29:02<00:52, 86.46it/s]
Running loglikelihood requests:  93%|█████████▎| 56425/60896 [29:03<00:51, 87.33it/s]
Running loglikelihood requests:  93%|█████████▎| 56463/60896 [29:03<00:49, 90.24it/s]
Running loglikelihood requests:  93%|█████████▎| 56500/60896 [29:03<00:48, 91.38it/s]
Running loglikelihood requests:  93%|█████████▎| 56539/60896 [29:04<00:46, 93.77it/s]
Running loglikelihood requests:  93%|█████████▎| 56577/60896 [29:04<00:45, 94.78it/s]
Running loglikelihood requests:  93%|█████████▎| 56617/60896 [29:05<00:44, 97.04it/s]
Running loglikelihood requests:  93%|█████████▎| 56653/60896 [29:05<00:44, 95.69it/s]
Running loglikelihood requests:  93%|█████████▎| 56689/60896 [29:05<00:44, 94.47it/s]
Running loglikelihood requests:  93%|█████████▎| 56726/60896 [29:06<00:44, 94.50it/s]
Running loglikelihood requests:  93%|█████████▎| 56763/60896 [29:06<00:43, 94.53it/s]
Running loglikelihood requests:  93%|█████████▎| 56799/60896 [29:07<00:43, 93.65it/s]
Running loglikelihood requests:  93%|█████████▎| 56832/60896 [29:07<00:44, 90.74it/s]
Running loglikelihood requests:  93%|█████████▎| 56866/60896 [29:07<00:45, 89.47it/s]
Running loglikelihood requests:  93%|█████████▎| 56904/60896 [29:08<00:43, 91.79it/s]
Running loglikelihood requests:  94%|█████████▎| 56944/60896 [29:08<00:42, 93.34it/s]
Running loglikelihood requests:  94%|█████████▎| 56982/60896 [29:09<00:42, 93.04it/s]
Running loglikelihood requests:  94%|█████████▎| 57014/60896 [29:09<00:43, 89.79it/s]
Running loglikelihood requests:  94%|█████████▎| 57046/60896 [29:09<00:43, 87.65it/s]
Running loglikelihood requests:  94%|█████████▎| 57080/60896 [29:10<00:43, 87.53it/s]
Running loglikelihood requests:  94%|█████████▍| 57116/60896 [29:10<00:42, 88.94it/s]
Running loglikelihood requests:  94%|█████████▍| 57148/60896 [29:11<00:43, 86.97it/s]
Running loglikelihood requests:  94%|█████████▍| 57183/60896 [29:11<00:42, 87.89it/s]
Running loglikelihood requests:  94%|█████████▍| 57227/60896 [29:11<00:38, 95.39it/s]
Running loglikelihood requests:  94%|█████████▍| 57264/60896 [29:12<00:38, 95.16it/s]
Running loglikelihood requests:  94%|█████████▍| 57301/60896 [29:12<00:37, 95.06it/s]
Running loglikelihood requests:  94%|█████████▍| 57340/60896 [29:12<00:36, 96.47it/s]
Running loglikelihood requests:  94%|█████████▍| 57372/60896 [29:13<00:38, 92.08it/s]
Running loglikelihood requests:  94%|█████████▍| 57417/60896 [29:13<00:35, 99.01it/s]
Running loglikelihood requests:  94%|█████████▍| 57450/60896 [29:14<00:36, 94.66it/s]
Running loglikelihood requests:  94%|█████████▍| 57489/60896 [29:14<00:35, 96.40it/s]
Running loglikelihood requests:  94%|█████████▍| 57527/60896 [29:14<00:34, 96.75it/s]
Running loglikelihood requests:  95%|█████████▍| 57561/60896 [29:15<00:35, 93.92it/s]
Running loglikelihood requests:  95%|█████████▍| 57594/60896 [29:15<00:37, 88.35it/s]
Running loglikelihood requests:  95%|█████████▍| 57626/60896 [29:16<00:37, 86.44it/s]
Running loglikelihood requests:  95%|█████████▍| 57658/60896 [29:16<00:37, 85.25it/s]
Running loglikelihood requests:  95%|█████████▍| 57690/60896 [29:16<00:37, 85.02it/s]
Running loglikelihood requests:  95%|█████████▍| 57722/60896 [29:17<00:37, 85.16it/s]
Running loglikelihood requests:  95%|█████████▍| 57757/60896 [29:17<00:35, 87.46it/s]
Running loglikelihood requests:  95%|█████████▍| 57792/60896 [29:18<00:34, 89.04it/s]
Running loglikelihood requests:  95%|█████████▍| 57826/60896 [29:18<00:34, 89.41it/s]
Running loglikelihood requests:  95%|█████████▌| 57864/60896 [29:18<00:32, 92.90it/s]
Running loglikelihood requests:  95%|█████████▌| 57899/60896 [29:19<00:32, 92.92it/s]
Running loglikelihood requests:  95%|█████████▌| 57935/60896 [29:19<00:31, 93.78it/s]
Running loglikelihood requests:  95%|█████████▌| 57975/60896 [29:19<00:29, 97.60it/s]
Running loglikelihood requests:  95%|█████████▌| 58007/60896 [29:20<00:30, 93.71it/s]
Running loglikelihood requests:  95%|█████████▌| 58044/60896 [29:20<00:29, 95.21it/s]
Running loglikelihood requests:  95%|█████████▌| 58080/60896 [29:21<00:29, 95.38it/s]
Running loglikelihood requests:  95%|█████████▌| 58119/60896 [29:21<00:28, 97.85it/s]
Running loglikelihood requests:  96%|█████████▌| 58158/60896 [29:21<00:28, 96.36it/s]
Running loglikelihood requests:  96%|█████████▌| 58195/60896 [29:22<00:27, 97.04it/s]
Running loglikelihood requests:  96%|█████████▌| 58233/60896 [29:22<00:27, 98.01it/s]
Running loglikelihood requests:  96%|█████████▌| 58265/60896 [29:22<00:27, 94.13it/s]
Running loglikelihood requests:  96%|█████████▌| 58298/60896 [29:23<00:28, 92.47it/s]
Running loglikelihood requests:  96%|█████████▌| 58336/60896 [29:23<00:26, 95.32it/s]
Running loglikelihood requests:  96%|█████████▌| 58370/60896 [29:24<00:26, 94.10it/s]
Running loglikelihood requests:  96%|█████████▌| 58406/60896 [29:24<00:26, 94.57it/s]
Running loglikelihood requests:  96%|█████████▌| 58440/60896 [29:24<00:26, 93.44it/s]
Running loglikelihood requests:  96%|█████████▌| 58474/60896 [29:25<00:26, 92.75it/s]
Running loglikelihood requests:  96%|█████████▌| 58510/60896 [29:25<00:25, 93.91it/s]
Running loglikelihood requests:  96%|█████████▌| 58559/60896 [29:25<00:22, 105.12it/s]
Running loglikelihood requests:  96%|█████████▌| 58601/60896 [29:26<00:21, 107.30it/s]
Running loglikelihood requests:  96%|█████████▋| 58643/60896 [29:26<00:20, 108.70it/s]
Running loglikelihood requests:  96%|█████████▋| 58686/60896 [29:27<00:19, 110.66it/s]
Running loglikelihood requests:  96%|█████████▋| 58723/60896 [29:27<00:20, 107.26it/s]
Running loglikelihood requests:  96%|█████████▋| 58759/60896 [29:27<00:21, 101.56it/s]
Running loglikelihood requests:  97%|█████████▋| 58793/60896 [29:28<00:21, 98.53it/s] 
Running loglikelihood requests:  97%|█████████▋| 58826/60896 [29:28<00:20, 99.23it/s]
Running loglikelihood requests:  97%|█████████▋| 58862/60896 [29:28<00:19, 102.23it/s]
Running loglikelihood requests:  97%|█████████▋| 58894/60896 [29:29<00:19, 101.08it/s]
Running loglikelihood requests:  97%|█████████▋| 58935/60896 [29:29<00:18, 108.27it/s]
Running loglikelihood requests:  97%|█████████▋| 58968/60896 [29:29<00:18, 106.19it/s]
Running loglikelihood requests:  97%|█████████▋| 59004/60896 [29:30<00:17, 107.57it/s]
Running loglikelihood requests:  97%|█████████▋| 59046/60896 [29:30<00:16, 113.95it/s]
Running loglikelihood requests:  97%|█████████▋| 59090/60896 [29:30<00:15, 119.97it/s]
Running loglikelihood requests:  97%|█████████▋| 59140/60896 [29:31<00:13, 129.82it/s]
Running loglikelihood requests:  97%|█████████▋| 59179/60896 [29:31<00:13, 126.88it/s]
Running loglikelihood requests:  97%|█████████▋| 59229/60896 [29:31<00:12, 134.72it/s]
Running loglikelihood requests:  97%|█████████▋| 59268/60896 [29:32<00:12, 130.26it/s]
Running loglikelihood requests:  97%|█████████▋| 59313/60896 [29:32<00:11, 132.64it/s]
Running loglikelihood requests:  97%|█████████▋| 59353/60896 [29:32<00:11, 129.62it/s]
Running loglikelihood requests:  98%|█████████▊| 59394/60896 [29:33<00:11, 128.47it/s]
Running loglikelihood requests:  98%|█████████▊| 59428/60896 [29:33<00:12, 121.54it/s]
Running loglikelihood requests:  98%|█████████▊| 59470/60896 [29:33<00:11, 123.97it/s]
Running loglikelihood requests:  98%|█████████▊| 59515/60896 [29:34<00:10, 128.41it/s]
Running loglikelihood requests:  98%|█████████▊| 59559/60896 [29:34<00:10, 130.58it/s]
Running loglikelihood requests:  98%|█████████▊| 59605/60896 [29:34<00:10, 128.63it/s]
Running loglikelihood requests:  98%|█████████▊| 59640/60896 [29:35<00:10, 122.61it/s]
Running loglikelihood requests:  98%|█████████▊| 59679/60896 [29:35<00:09, 121.86it/s]
Running loglikelihood requests:  98%|█████████▊| 59722/60896 [29:35<00:09, 125.09it/s]
Running loglikelihood requests:  98%|█████████▊| 59769/60896 [29:36<00:08, 131.07it/s]
Running loglikelihood requests:  98%|█████████▊| 59810/60896 [29:37<00:12, 83.87it/s] 
Running loglikelihood requests:  98%|█████████▊| 59850/60896 [29:37<00:11, 91.42it/s]
Running loglikelihood requests:  98%|█████████▊| 59892/60896 [29:37<00:09, 101.26it/s]
Running loglikelihood requests:  98%|█████████▊| 59935/60896 [29:37<00:08, 110.25it/s]
Running loglikelihood requests:  99%|█████████▊| 59985/60896 [29:38<00:07, 123.51it/s]
Running loglikelihood requests:  99%|█████████▊| 60032/60896 [29:38<00:06, 130.92it/s]
Running loglikelihood requests:  99%|█████████▊| 60081/60896 [29:38<00:05, 138.55it/s]
Running loglikelihood requests:  99%|█████████▊| 60125/60896 [29:39<00:05, 139.61it/s]
Running loglikelihood requests:  99%|█████████▉| 60161/60896 [29:39<00:05, 133.64it/s]
Running loglikelihood requests:  99%|█████████▉| 60199/60896 [29:39<00:05, 130.88it/s]
Running loglikelihood requests:  99%|█████████▉| 60240/60896 [29:40<00:04, 131.84it/s]
Running loglikelihood requests:  99%|█████████▉| 60299/60896 [29:40<00:03, 149.72it/s]
Running loglikelihood requests:  99%|█████████▉| 60338/60896 [29:40<00:04, 135.95it/s]
Running loglikelihood requests:  99%|█████████▉| 60395/60896 [29:41<00:03, 150.27it/s]
Running loglikelihood requests:  99%|█████████▉| 60436/60896 [29:41<00:03, 145.81it/s]
Running loglikelihood requests:  99%|█████████▉| 60475/60896 [29:41<00:02, 140.55it/s]
Running loglikelihood requests:  99%|█████████▉| 60523/60896 [29:41<00:02, 147.80it/s]
Running loglikelihood requests:  99%|█████████▉| 60579/60896 [29:42<00:01, 161.25it/s]
Running loglikelihood requests: 100%|█████████▉| 60644/60896 [29:42<00:01, 179.76it/s]
Running loglikelihood requests: 100%|█████████▉| 60679/60896 [29:42<00:01, 163.20it/s]
Running loglikelihood requests: 100%|█████████▉| 60720/60896 [29:43<00:01, 157.05it/s]
Running loglikelihood requests: 100%|█████████▉| 60784/60896 [29:43<00:00, 176.06it/s]
Running loglikelihood requests: 100%|█████████▉| 60830/60896 [29:43<00:00, 176.35it/s]
Running loglikelihood requests: 100%|█████████▉| 60888/60896 [29:43<00:00, 208.99it/s]
Running loglikelihood requests: 100%|██████████| 60896/60896 [29:43<00:00, 34.14it/s] 
+{'arc_challenge': {'alias': 'arc_challenge', 'acc,none': 0.4974402730375427, 'acc_stderr,none': 0.014611199329843774, 'acc_norm,none': 0.5051194539249146, 'acc_norm_stderr,none': 0.01461062489030916}, 'arc_easy': {'alias': 'arc_easy', 'acc,none': 0.7723063973063973, 'acc_stderr,none': 0.008604753300503747, 'acc_norm,none': 0.7525252525252525, 'acc_norm_stderr,none': 0.00885511441483471}, 'boolq': {'alias': 'boolq', 'acc,none': 0.7981651376146789, 'acc_stderr,none': 0.007019998324744636}, 'hellaswag': {'alias': 'hellaswag', 'acc,none': 0.5871340370444135, 'acc_stderr,none': 0.004913429010559091, 'acc_norm,none': 0.7765385381398128, 'acc_norm_stderr,none': 0.004157137038885681}}
diff --git a/nvfp/Meta-Llama-3-8B/weight_mse_comparison.png b/nvfp/Meta-Llama-3-8B/weight_mse_comparison.png
new file mode 100644
index 0000000000000000000000000000000000000000..fcdc06a3785825515d156bad0323e91bb0c2df1b
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/weight_mse_comparison.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6404c62e8739dbe7e110120f40a2758274eaf1bf46146f09a01d71faa372694b
+size 418090
diff --git a/nvfp/Meta-Llama-3-8B/weight_mse_comparison_rotation_only.png b/nvfp/Meta-Llama-3-8B/weight_mse_comparison_rotation_only.png
new file mode 100644
index 0000000000000000000000000000000000000000..2a1833bcd2fb768d5028f49b2f7f5896bd1856dc
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/weight_mse_comparison_rotation_only.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2cc867d5c48a718cf0667bd0126227756def727b690492bb924ee9f5c4a391b4
+size 844538
diff --git a/nvfp/Meta-Llama-3-8B/weight_mse_comparison_scale_only.png b/nvfp/Meta-Llama-3-8B/weight_mse_comparison_scale_only.png
new file mode 100644
index 0000000000000000000000000000000000000000..fe73fe033d4c4ebef306558bbed3903b8a33b09c
--- /dev/null
+++ b/nvfp/Meta-Llama-3-8B/weight_mse_comparison_scale_only.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:efc50fb8df1d4dbcc232571d492c220911d923aaf05c5a23bd7797c9b6ee4bec
+size 627355