Upload 12 files
Browse files- .gitattributes +2 -0
- adaptertestsfor2bitday2/creaprompt_W2A4_recovery_PHASE1B_FFN_AttnOut_ProjIO_rankmix_gamma_beta.safetensors +3 -0
- adaptertestsfor2bitday2/creaprompt_W2A4_recovery_PHASE1B_FFN_AttnOut_ProjIO_rankmix_gamma_beta_state.json +0 -0
- adaptertestsfor2bitday2/creaprompt_W2A4_recovery_PHASE2_CrossAttnKV_ProjOut_rankmix_gamma_beta.safetensors +3 -0
- adaptertestsfor2bitday2/creaprompt_W2A4_recovery_PHASE2_CrossAttnKV_ProjOut_rankmix_gamma_beta_state.json +0 -0
- adaptertestsfor2bitday2/download.png +3 -0
- adaptertestsfor2bitday2/orbitquant_W4_vs_W2_all_layers.csv +0 -0
- adaptertestsfor2bitday2/orbitquant_W4_vs_W2_by_suffix.csv +13 -0
- adaptertestsfor2bitday2/orbitquant_W4_vs_W2_diagnostic_report.json +1448 -0
- adaptertestsfor2bitday2/orbitquant_W4_vs_W2_diagnostic_report.xlsx +3 -0
- adaptertestsfor2bitday2/orbitquant_W4_vs_W2_next_training_plan.json +0 -0
- adaptertestsfor2bitday2/orbitquant_W4_vs_W2_top_training_targets.csv +0 -0
- adaptertestsfor2bitday2/w4_w2_compare_condition_cache_captions_only_random_latents_1024x1024_8.pt +3 -0
.gitattributes
CHANGED
|
@@ -35,3 +35,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
examples/630df8e8-f7b1-4225-889b-247dcb55e2f0.png filter=lfs diff=lfs merge=lfs -text
|
| 37 |
examples/download.png filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
examples/630df8e8-f7b1-4225-889b-247dcb55e2f0.png filter=lfs diff=lfs merge=lfs -text
|
| 37 |
examples/download.png filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
adaptertestsfor2bitday2/download.png filter=lfs diff=lfs merge=lfs -text
|
| 39 |
+
adaptertestsfor2bitday2/orbitquant_W4_vs_W2_diagnostic_report.xlsx filter=lfs diff=lfs merge=lfs -text
|
adaptertestsfor2bitday2/creaprompt_W2A4_recovery_PHASE1B_FFN_AttnOut_ProjIO_rankmix_gamma_beta.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7c9c61aa0b499e2a43e1703ef54bc67ed78bcccc75effa4e4b62502690c1ca29
|
| 3 |
+
size 404958616
|
adaptertestsfor2bitday2/creaprompt_W2A4_recovery_PHASE1B_FFN_AttnOut_ProjIO_rankmix_gamma_beta_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
adaptertestsfor2bitday2/creaprompt_W2A4_recovery_PHASE2_CrossAttnKV_ProjOut_rankmix_gamma_beta.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b43365440fa7ba1de07b263102b20b392e50e6dd71df25d805fd7c47949ee4a
|
| 3 |
+
size 163569480
|
adaptertestsfor2bitday2/creaprompt_W2A4_recovery_PHASE2_CrossAttnKV_ProjOut_rankmix_gamma_beta_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
adaptertestsfor2bitday2/download.png
ADDED
|
Git LFS Details
|
adaptertestsfor2bitday2/orbitquant_W4_vs_W2_all_layers.csv
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
adaptertestsfor2bitday2/orbitquant_W4_vs_W2_by_suffix.csv
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
suffix,count,old_rank128_count,phase1b_count,phase2_count,composite_count,w4_mse_mean,w2_raw_mse_mean,w2_old_rank128_mse_mean,w2_phase1b_mse_mean,w2_phase2_mse_mean,w2_composite_mse_mean,w2_raw_to_w4_ratio_mean,w2_composite_to_w4_ratio_mean,composite_improvement_vs_raw_mean,phase1b_improvement_vs_raw_mean,phase2_improvement_vs_raw_mean,remaining_gap_score_mean,remaining_gap_score_max
|
| 2 |
+
attn1.to_q,70,0,0,0,0,0.008834530145039174,0.06604109280267541,0.06604109280267541,0.06604109280267541,0.06604109280267541,0.06604109280267541,7.5747877444341,7.5747877444341,0.9999999999832079,0.9999999999832079,0.9999999999832079,0.42929220002638296,0.8156589509387167
|
| 3 |
+
attn1.to_k,70,0,0,0,0,0.00869125885917444,0.06352441829714772,0.06352441829714772,0.06352441829714772,0.06352441829714772,0.06352441829714772,7.395005113804398,7.395005113804398,0.9999999999825412,0.9999999999825412,0.9999999999825412,0.4027360624185252,0.7311920707443591
|
| 4 |
+
attn1.to_v,70,0,0,0,0,0.0078085727364238244,0.055820106880796354,0.055820106880796354,0.055820106880796354,0.055820106880796354,0.055820106880796354,7.1828383652128185,7.1828383652128185,0.999999999979725,0.999999999979725,0.999999999979725,0.34468865166305246,0.8469947038879494
|
| 5 |
+
proj_out,11,11,11,8,11,0.08516350478106238,0.5946647835556756,0.35843181815421715,0.21572238788637635,0.2251953121226329,0.20004504513866328,6.913556153067888,2.452973561882824,2.969396590218453,2.8728746957207307,2.388979858331337,0.2912881384368135,0.7227891036499625
|
| 6 |
+
attn2.to_q,70,0,0,0,0,0.004732320876079327,0.037790954292618806,0.037790954292618806,0.037790954292618806,0.037790954292618806,0.037790954292618806,8.31881567261658,8.31881567261658,0.9999999999712236,0.9999999999712236,0.9999999999712236,0.2711913227898678,0.6841755529628081
|
| 7 |
+
ff.net.2,70,70,70,0,70,0.008616097393292431,0.05920763128883756,0.04301157893223346,0.02601208476651441,0.05920763128883756,0.02601208476651441,6.932097684619555,3.2009256575082317,2.3014962630105096,2.3014962630105096,0.9999999999747406,0.0572061937220024,0.20667232622681672
|
| 8 |
+
ff.net.0.proj,70,70,70,0,70,0.006887027408722171,0.04879796355012424,0.03164028028936164,0.0176630408599336,0.04879796355012424,0.0176630408599336,7.171405831067578,2.504632836281881,2.998213235077801,2.998213235077801,0.9999999999754532,0.02951947878339631,0.1137982287591066
|
| 9 |
+
proj_in,11,11,11,0,11,0.008933692221838206,0.06406978043038157,0.03873854998329824,0.018474196746806285,0.06406978043038157,0.018474196746806285,7.050454051901254,1.9995457179267833,3.6405517751576513,3.6405517751576513,0.9999999999808292,0.020491349940082455,0.05859008625009349
|
| 10 |
+
attn2.to_out.0,70,70,70,0,70,0.0003375165737176238,0.002381115644698648,0.002395683516781632,0.00222385320224494,0.002381115644698648,0.00222385320224494,7.598942937360687,6.7306297362361756,1.8347576429611256,1.8347576429611256,0.9999999981489446,0.012840779013027317,0.05913571060692841
|
| 11 |
+
attn1.to_out.0,70,70,70,0,70,0.0042316013468252445,0.030931819595006883,0.01690724441420515,0.005962002335374044,0.030931819595006883,0.005962002335374044,7.476497691796118,1.4532056634829924,5.661566033087392,5.661566033087392,0.9999999999555428,0.0034474704133703057,0.021837045500065322
|
| 12 |
+
attn2.to_v,70,0,0,70,70,0.05053253279267145,0.33406722963388474,0.33406722963388474,0.33406722963388474,0.020737389160785824,0.020737389160785824,6.628456469971321,0.4083913035125052,16.325964238694546,0.9999999999957944,16.325964238694546,0.0,0.0
|
| 13 |
+
attn2.to_k,70,0,0,70,70,0.0519350185824026,0.42513413368059055,0.42513413368059055,0.42513413368059055,0.024206796191200346,0.024206796191200346,8.66438299826568,0.47265911793929566,18.609512818870797,0.9999999999969618,18.609512818870797,0.0,0.0
|
adaptertestsfor2bitday2/orbitquant_W4_vs_W2_diagnostic_report.json
ADDED
|
@@ -0,0 +1,1448 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"diagnostic_type": "FP16 teacher layer output vs W4A4 vs W2A4 raw/adapted variants",
|
| 3 |
+
"teacher": "/content/creaprompt_sdxl/creapromptLightning_creapromtHypersdxlV1.safetensors",
|
| 4 |
+
"w4_model": "/content/creaprompt_sdxl/orbitquant_full/creaprompt_FULL_OrbitQuant_UNet_W4A4_with_VAE_CLIPs.safetensors",
|
| 5 |
+
"w2_model": "/content/creaprompt_sdxl/orbitquant_full/creaprompt_FULL_OrbitQuant_UNet_W2A4_with_VAE_CLIPs.safetensors",
|
| 6 |
+
"old_rank128_adapter": "/content/creaprompt_sdxl/adapter_training/checkpoints/creaprompt_W2A4_recovery_FFN_AttnOut_rank128_from_rank8.safetensors",
|
| 7 |
+
"phase1b_adapter": "/content/creaprompt_sdxl/adapter_training/checkpoints/creaprompt_W2A4_recovery_PHASE1B_FFN_AttnOut_ProjIO_rankmix_gamma_beta.safetensors",
|
| 8 |
+
"phase2_adapter": "/content/creaprompt_sdxl/adapter_training/checkpoints/creaprompt_W2A4_recovery_PHASE2_CrossAttnKV_ProjOut_rankmix_gamma_beta.safetensors",
|
| 9 |
+
"dataset_mode": "captions_only_random_latents",
|
| 10 |
+
"diag_width": 1024,
|
| 11 |
+
"diag_height": 1024,
|
| 12 |
+
"diag_samples": 8,
|
| 13 |
+
"timesteps": [
|
| 14 |
+
999,
|
| 15 |
+
850,
|
| 16 |
+
700,
|
| 17 |
+
500,
|
| 18 |
+
300,
|
| 19 |
+
150,
|
| 20 |
+
50
|
| 21 |
+
],
|
| 22 |
+
"max_tokens_per_layer": 2048,
|
| 23 |
+
"total_layers": 722,
|
| 24 |
+
"outputs": {
|
| 25 |
+
"all_layers_csv": "/content/creaprompt_sdxl/diagnostics_w4_w2_compare/orbitquant_W4_vs_W2_all_layers.csv",
|
| 26 |
+
"top_training_targets_csv": "/content/creaprompt_sdxl/diagnostics_w4_w2_compare/orbitquant_W4_vs_W2_top_training_targets.csv",
|
| 27 |
+
"suffix_summary_csv": "/content/creaprompt_sdxl/diagnostics_w4_w2_compare/orbitquant_W4_vs_W2_by_suffix.csv",
|
| 28 |
+
"plan_json": "/content/creaprompt_sdxl/diagnostics_w4_w2_compare/orbitquant_W4_vs_W2_next_training_plan.json",
|
| 29 |
+
"report_json": "/content/creaprompt_sdxl/diagnostics_w4_w2_compare/orbitquant_W4_vs_W2_diagnostic_report.json",
|
| 30 |
+
"xlsx": "/content/creaprompt_sdxl/diagnostics_w4_w2_compare/orbitquant_W4_vs_W2_diagnostic_report.xlsx"
|
| 31 |
+
},
|
| 32 |
+
"suffix_summary": [
|
| 33 |
+
{
|
| 34 |
+
"suffix": "attn1.to_q",
|
| 35 |
+
"count": 70,
|
| 36 |
+
"old_rank128_count": 0,
|
| 37 |
+
"phase1b_count": 0,
|
| 38 |
+
"phase2_count": 0,
|
| 39 |
+
"composite_count": 0,
|
| 40 |
+
"w4_mse_mean": 0.008834530145039174,
|
| 41 |
+
"w2_raw_mse_mean": 0.06604109280267541,
|
| 42 |
+
"w2_old_rank128_mse_mean": 0.06604109280267541,
|
| 43 |
+
"w2_phase1b_mse_mean": 0.06604109280267541,
|
| 44 |
+
"w2_phase2_mse_mean": 0.06604109280267541,
|
| 45 |
+
"w2_composite_mse_mean": 0.06604109280267541,
|
| 46 |
+
"w2_raw_to_w4_ratio_mean": 7.5747877444341,
|
| 47 |
+
"w2_composite_to_w4_ratio_mean": 7.5747877444341,
|
| 48 |
+
"composite_improvement_vs_raw_mean": 0.9999999999832079,
|
| 49 |
+
"phase1b_improvement_vs_raw_mean": 0.9999999999832079,
|
| 50 |
+
"phase2_improvement_vs_raw_mean": 0.9999999999832079,
|
| 51 |
+
"remaining_gap_score_mean": 0.42929220002638296,
|
| 52 |
+
"remaining_gap_score_max": 0.8156589509387167
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"suffix": "attn1.to_k",
|
| 56 |
+
"count": 70,
|
| 57 |
+
"old_rank128_count": 0,
|
| 58 |
+
"phase1b_count": 0,
|
| 59 |
+
"phase2_count": 0,
|
| 60 |
+
"composite_count": 0,
|
| 61 |
+
"w4_mse_mean": 0.00869125885917444,
|
| 62 |
+
"w2_raw_mse_mean": 0.06352441829714772,
|
| 63 |
+
"w2_old_rank128_mse_mean": 0.06352441829714772,
|
| 64 |
+
"w2_phase1b_mse_mean": 0.06352441829714772,
|
| 65 |
+
"w2_phase2_mse_mean": 0.06352441829714772,
|
| 66 |
+
"w2_composite_mse_mean": 0.06352441829714772,
|
| 67 |
+
"w2_raw_to_w4_ratio_mean": 7.395005113804398,
|
| 68 |
+
"w2_composite_to_w4_ratio_mean": 7.395005113804398,
|
| 69 |
+
"composite_improvement_vs_raw_mean": 0.9999999999825412,
|
| 70 |
+
"phase1b_improvement_vs_raw_mean": 0.9999999999825412,
|
| 71 |
+
"phase2_improvement_vs_raw_mean": 0.9999999999825412,
|
| 72 |
+
"remaining_gap_score_mean": 0.4027360624185252,
|
| 73 |
+
"remaining_gap_score_max": 0.7311920707443591
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"suffix": "attn1.to_v",
|
| 77 |
+
"count": 70,
|
| 78 |
+
"old_rank128_count": 0,
|
| 79 |
+
"phase1b_count": 0,
|
| 80 |
+
"phase2_count": 0,
|
| 81 |
+
"composite_count": 0,
|
| 82 |
+
"w4_mse_mean": 0.0078085727364238244,
|
| 83 |
+
"w2_raw_mse_mean": 0.055820106880796354,
|
| 84 |
+
"w2_old_rank128_mse_mean": 0.055820106880796354,
|
| 85 |
+
"w2_phase1b_mse_mean": 0.055820106880796354,
|
| 86 |
+
"w2_phase2_mse_mean": 0.055820106880796354,
|
| 87 |
+
"w2_composite_mse_mean": 0.055820106880796354,
|
| 88 |
+
"w2_raw_to_w4_ratio_mean": 7.1828383652128185,
|
| 89 |
+
"w2_composite_to_w4_ratio_mean": 7.1828383652128185,
|
| 90 |
+
"composite_improvement_vs_raw_mean": 0.999999999979725,
|
| 91 |
+
"phase1b_improvement_vs_raw_mean": 0.999999999979725,
|
| 92 |
+
"phase2_improvement_vs_raw_mean": 0.999999999979725,
|
| 93 |
+
"remaining_gap_score_mean": 0.34468865166305246,
|
| 94 |
+
"remaining_gap_score_max": 0.8469947038879494
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"suffix": "proj_out",
|
| 98 |
+
"count": 11,
|
| 99 |
+
"old_rank128_count": 11,
|
| 100 |
+
"phase1b_count": 11,
|
| 101 |
+
"phase2_count": 8,
|
| 102 |
+
"composite_count": 11,
|
| 103 |
+
"w4_mse_mean": 0.08516350478106238,
|
| 104 |
+
"w2_raw_mse_mean": 0.5946647835556756,
|
| 105 |
+
"w2_old_rank128_mse_mean": 0.35843181815421715,
|
| 106 |
+
"w2_phase1b_mse_mean": 0.21572238788637635,
|
| 107 |
+
"w2_phase2_mse_mean": 0.2251953121226329,
|
| 108 |
+
"w2_composite_mse_mean": 0.20004504513866328,
|
| 109 |
+
"w2_raw_to_w4_ratio_mean": 6.913556153067888,
|
| 110 |
+
"w2_composite_to_w4_ratio_mean": 2.452973561882824,
|
| 111 |
+
"composite_improvement_vs_raw_mean": 2.969396590218453,
|
| 112 |
+
"phase1b_improvement_vs_raw_mean": 2.8728746957207307,
|
| 113 |
+
"phase2_improvement_vs_raw_mean": 2.388979858331337,
|
| 114 |
+
"remaining_gap_score_mean": 0.2912881384368135,
|
| 115 |
+
"remaining_gap_score_max": 0.7227891036499625
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"suffix": "attn2.to_q",
|
| 119 |
+
"count": 70,
|
| 120 |
+
"old_rank128_count": 0,
|
| 121 |
+
"phase1b_count": 0,
|
| 122 |
+
"phase2_count": 0,
|
| 123 |
+
"composite_count": 0,
|
| 124 |
+
"w4_mse_mean": 0.004732320876079327,
|
| 125 |
+
"w2_raw_mse_mean": 0.037790954292618806,
|
| 126 |
+
"w2_old_rank128_mse_mean": 0.037790954292618806,
|
| 127 |
+
"w2_phase1b_mse_mean": 0.037790954292618806,
|
| 128 |
+
"w2_phase2_mse_mean": 0.037790954292618806,
|
| 129 |
+
"w2_composite_mse_mean": 0.037790954292618806,
|
| 130 |
+
"w2_raw_to_w4_ratio_mean": 8.31881567261658,
|
| 131 |
+
"w2_composite_to_w4_ratio_mean": 8.31881567261658,
|
| 132 |
+
"composite_improvement_vs_raw_mean": 0.9999999999712236,
|
| 133 |
+
"phase1b_improvement_vs_raw_mean": 0.9999999999712236,
|
| 134 |
+
"phase2_improvement_vs_raw_mean": 0.9999999999712236,
|
| 135 |
+
"remaining_gap_score_mean": 0.2711913227898678,
|
| 136 |
+
"remaining_gap_score_max": 0.6841755529628081
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"suffix": "ff.net.2",
|
| 140 |
+
"count": 70,
|
| 141 |
+
"old_rank128_count": 70,
|
| 142 |
+
"phase1b_count": 70,
|
| 143 |
+
"phase2_count": 0,
|
| 144 |
+
"composite_count": 70,
|
| 145 |
+
"w4_mse_mean": 0.008616097393292431,
|
| 146 |
+
"w2_raw_mse_mean": 0.05920763128883756,
|
| 147 |
+
"w2_old_rank128_mse_mean": 0.04301157893223346,
|
| 148 |
+
"w2_phase1b_mse_mean": 0.02601208476651441,
|
| 149 |
+
"w2_phase2_mse_mean": 0.05920763128883756,
|
| 150 |
+
"w2_composite_mse_mean": 0.02601208476651441,
|
| 151 |
+
"w2_raw_to_w4_ratio_mean": 6.932097684619555,
|
| 152 |
+
"w2_composite_to_w4_ratio_mean": 3.2009256575082317,
|
| 153 |
+
"composite_improvement_vs_raw_mean": 2.3014962630105096,
|
| 154 |
+
"phase1b_improvement_vs_raw_mean": 2.3014962630105096,
|
| 155 |
+
"phase2_improvement_vs_raw_mean": 0.9999999999747406,
|
| 156 |
+
"remaining_gap_score_mean": 0.0572061937220024,
|
| 157 |
+
"remaining_gap_score_max": 0.20667232622681672
|
| 158 |
+
},
|
| 159 |
+
{
|
| 160 |
+
"suffix": "ff.net.0.proj",
|
| 161 |
+
"count": 70,
|
| 162 |
+
"old_rank128_count": 70,
|
| 163 |
+
"phase1b_count": 70,
|
| 164 |
+
"phase2_count": 0,
|
| 165 |
+
"composite_count": 70,
|
| 166 |
+
"w4_mse_mean": 0.006887027408722171,
|
| 167 |
+
"w2_raw_mse_mean": 0.04879796355012424,
|
| 168 |
+
"w2_old_rank128_mse_mean": 0.03164028028936164,
|
| 169 |
+
"w2_phase1b_mse_mean": 0.0176630408599336,
|
| 170 |
+
"w2_phase2_mse_mean": 0.04879796355012424,
|
| 171 |
+
"w2_composite_mse_mean": 0.0176630408599336,
|
| 172 |
+
"w2_raw_to_w4_ratio_mean": 7.171405831067578,
|
| 173 |
+
"w2_composite_to_w4_ratio_mean": 2.504632836281881,
|
| 174 |
+
"composite_improvement_vs_raw_mean": 2.998213235077801,
|
| 175 |
+
"phase1b_improvement_vs_raw_mean": 2.998213235077801,
|
| 176 |
+
"phase2_improvement_vs_raw_mean": 0.9999999999754532,
|
| 177 |
+
"remaining_gap_score_mean": 0.02951947878339631,
|
| 178 |
+
"remaining_gap_score_max": 0.1137982287591066
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"suffix": "proj_in",
|
| 182 |
+
"count": 11,
|
| 183 |
+
"old_rank128_count": 11,
|
| 184 |
+
"phase1b_count": 11,
|
| 185 |
+
"phase2_count": 0,
|
| 186 |
+
"composite_count": 11,
|
| 187 |
+
"w4_mse_mean": 0.008933692221838206,
|
| 188 |
+
"w2_raw_mse_mean": 0.06406978043038157,
|
| 189 |
+
"w2_old_rank128_mse_mean": 0.03873854998329824,
|
| 190 |
+
"w2_phase1b_mse_mean": 0.018474196746806285,
|
| 191 |
+
"w2_phase2_mse_mean": 0.06406978043038157,
|
| 192 |
+
"w2_composite_mse_mean": 0.018474196746806285,
|
| 193 |
+
"w2_raw_to_w4_ratio_mean": 7.050454051901254,
|
| 194 |
+
"w2_composite_to_w4_ratio_mean": 1.9995457179267833,
|
| 195 |
+
"composite_improvement_vs_raw_mean": 3.6405517751576513,
|
| 196 |
+
"phase1b_improvement_vs_raw_mean": 3.6405517751576513,
|
| 197 |
+
"phase2_improvement_vs_raw_mean": 0.9999999999808292,
|
| 198 |
+
"remaining_gap_score_mean": 0.020491349940082455,
|
| 199 |
+
"remaining_gap_score_max": 0.05859008625009349
|
| 200 |
+
},
|
| 201 |
+
{
|
| 202 |
+
"suffix": "attn2.to_out.0",
|
| 203 |
+
"count": 70,
|
| 204 |
+
"old_rank128_count": 70,
|
| 205 |
+
"phase1b_count": 70,
|
| 206 |
+
"phase2_count": 0,
|
| 207 |
+
"composite_count": 70,
|
| 208 |
+
"w4_mse_mean": 0.0003375165737176238,
|
| 209 |
+
"w2_raw_mse_mean": 0.002381115644698648,
|
| 210 |
+
"w2_old_rank128_mse_mean": 0.002395683516781632,
|
| 211 |
+
"w2_phase1b_mse_mean": 0.00222385320224494,
|
| 212 |
+
"w2_phase2_mse_mean": 0.002381115644698648,
|
| 213 |
+
"w2_composite_mse_mean": 0.00222385320224494,
|
| 214 |
+
"w2_raw_to_w4_ratio_mean": 7.598942937360687,
|
| 215 |
+
"w2_composite_to_w4_ratio_mean": 6.7306297362361756,
|
| 216 |
+
"composite_improvement_vs_raw_mean": 1.8347576429611256,
|
| 217 |
+
"phase1b_improvement_vs_raw_mean": 1.8347576429611256,
|
| 218 |
+
"phase2_improvement_vs_raw_mean": 0.9999999981489446,
|
| 219 |
+
"remaining_gap_score_mean": 0.012840779013027317,
|
| 220 |
+
"remaining_gap_score_max": 0.05913571060692841
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"suffix": "attn1.to_out.0",
|
| 224 |
+
"count": 70,
|
| 225 |
+
"old_rank128_count": 70,
|
| 226 |
+
"phase1b_count": 70,
|
| 227 |
+
"phase2_count": 0,
|
| 228 |
+
"composite_count": 70,
|
| 229 |
+
"w4_mse_mean": 0.0042316013468252445,
|
| 230 |
+
"w2_raw_mse_mean": 0.030931819595006883,
|
| 231 |
+
"w2_old_rank128_mse_mean": 0.01690724441420515,
|
| 232 |
+
"w2_phase1b_mse_mean": 0.005962002335374044,
|
| 233 |
+
"w2_phase2_mse_mean": 0.030931819595006883,
|
| 234 |
+
"w2_composite_mse_mean": 0.005962002335374044,
|
| 235 |
+
"w2_raw_to_w4_ratio_mean": 7.476497691796118,
|
| 236 |
+
"w2_composite_to_w4_ratio_mean": 1.4532056634829924,
|
| 237 |
+
"composite_improvement_vs_raw_mean": 5.661566033087392,
|
| 238 |
+
"phase1b_improvement_vs_raw_mean": 5.661566033087392,
|
| 239 |
+
"phase2_improvement_vs_raw_mean": 0.9999999999555428,
|
| 240 |
+
"remaining_gap_score_mean": 0.0034474704133703057,
|
| 241 |
+
"remaining_gap_score_max": 0.021837045500065322
|
| 242 |
+
},
|
| 243 |
+
{
|
| 244 |
+
"suffix": "attn2.to_v",
|
| 245 |
+
"count": 70,
|
| 246 |
+
"old_rank128_count": 0,
|
| 247 |
+
"phase1b_count": 0,
|
| 248 |
+
"phase2_count": 70,
|
| 249 |
+
"composite_count": 70,
|
| 250 |
+
"w4_mse_mean": 0.05053253279267145,
|
| 251 |
+
"w2_raw_mse_mean": 0.33406722963388474,
|
| 252 |
+
"w2_old_rank128_mse_mean": 0.33406722963388474,
|
| 253 |
+
"w2_phase1b_mse_mean": 0.33406722963388474,
|
| 254 |
+
"w2_phase2_mse_mean": 0.020737389160785824,
|
| 255 |
+
"w2_composite_mse_mean": 0.020737389160785824,
|
| 256 |
+
"w2_raw_to_w4_ratio_mean": 6.628456469971321,
|
| 257 |
+
"w2_composite_to_w4_ratio_mean": 0.4083913035125052,
|
| 258 |
+
"composite_improvement_vs_raw_mean": 16.325964238694546,
|
| 259 |
+
"phase1b_improvement_vs_raw_mean": 0.9999999999957944,
|
| 260 |
+
"phase2_improvement_vs_raw_mean": 16.325964238694546,
|
| 261 |
+
"remaining_gap_score_mean": 0.0,
|
| 262 |
+
"remaining_gap_score_max": 0.0
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"suffix": "attn2.to_k",
|
| 266 |
+
"count": 70,
|
| 267 |
+
"old_rank128_count": 0,
|
| 268 |
+
"phase1b_count": 0,
|
| 269 |
+
"phase2_count": 70,
|
| 270 |
+
"composite_count": 70,
|
| 271 |
+
"w4_mse_mean": 0.0519350185824026,
|
| 272 |
+
"w2_raw_mse_mean": 0.42513413368059055,
|
| 273 |
+
"w2_old_rank128_mse_mean": 0.42513413368059055,
|
| 274 |
+
"w2_phase1b_mse_mean": 0.42513413368059055,
|
| 275 |
+
"w2_phase2_mse_mean": 0.024206796191200346,
|
| 276 |
+
"w2_composite_mse_mean": 0.024206796191200346,
|
| 277 |
+
"w2_raw_to_w4_ratio_mean": 8.66438299826568,
|
| 278 |
+
"w2_composite_to_w4_ratio_mean": 0.47265911793929566,
|
| 279 |
+
"composite_improvement_vs_raw_mean": 18.609512818870797,
|
| 280 |
+
"phase1b_improvement_vs_raw_mean": 0.9999999999969618,
|
| 281 |
+
"phase2_improvement_vs_raw_mean": 18.609512818870797,
|
| 282 |
+
"remaining_gap_score_mean": 0.0,
|
| 283 |
+
"remaining_gap_score_max": 0.0
|
| 284 |
+
}
|
| 285 |
+
],
|
| 286 |
+
"top_40_training_targets": [
|
| 287 |
+
{
|
| 288 |
+
"training_priority_rank": 1,
|
| 289 |
+
"module_name": "up_blocks.1.attentions.0.transformer_blocks.1.attn1.to_v",
|
| 290 |
+
"ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.1.attn1.to_v.weight",
|
| 291 |
+
"suffix": "attn1.to_v",
|
| 292 |
+
"shape": [
|
| 293 |
+
640,
|
| 294 |
+
640
|
| 295 |
+
],
|
| 296 |
+
"has_old_rank128": false,
|
| 297 |
+
"has_phase1b": false,
|
| 298 |
+
"has_phase2": false,
|
| 299 |
+
"composite_source": "raw",
|
| 300 |
+
"w4_mse_mean": 0.016171995251040374,
|
| 301 |
+
"w2_raw_mse_mean": 0.125401715614966,
|
| 302 |
+
"w2_old_rank128_mse_mean": 0.125401715614966,
|
| 303 |
+
"w2_phase1b_mse_mean": 0.125401715614966,
|
| 304 |
+
"w2_phase2_mse_mean": 0.125401715614966,
|
| 305 |
+
"w2_composite_mse_mean": 0.125401715614966,
|
| 306 |
+
"w2_raw_to_w4_mse_ratio": 7.754251325243521,
|
| 307 |
+
"w2_composite_to_w4_mse_ratio": 7.754251325243521,
|
| 308 |
+
"old_rank128_improvement_vs_raw": 0.9999999999920256,
|
| 309 |
+
"phase1b_improvement_vs_raw": 0.9999999999920256,
|
| 310 |
+
"phase2_improvement_vs_raw": 0.9999999999920256,
|
| 311 |
+
"composite_improvement_vs_raw": 0.9999999999920256,
|
| 312 |
+
"remaining_w2_specific_gap_score": 0.8469947038879494,
|
| 313 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 314 |
+
"recommended_rank": 256
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"training_priority_rank": 2,
|
| 318 |
+
"module_name": "up_blocks.1.attentions.1.transformer_blocks.1.attn1.to_q",
|
| 319 |
+
"ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.1.attn1.to_q.weight",
|
| 320 |
+
"suffix": "attn1.to_q",
|
| 321 |
+
"shape": [
|
| 322 |
+
640,
|
| 323 |
+
640
|
| 324 |
+
],
|
| 325 |
+
"has_old_rank128": false,
|
| 326 |
+
"has_phase1b": false,
|
| 327 |
+
"has_phase2": false,
|
| 328 |
+
"composite_source": "raw",
|
| 329 |
+
"w4_mse_mean": 0.01798078600716378,
|
| 330 |
+
"w2_raw_mse_mean": 0.130427686364523,
|
| 331 |
+
"w2_old_rank128_mse_mean": 0.130427686364523,
|
| 332 |
+
"w2_phase1b_mse_mean": 0.130427686364523,
|
| 333 |
+
"w2_phase2_mse_mean": 0.130427686364523,
|
| 334 |
+
"w2_composite_mse_mean": 0.130427686364523,
|
| 335 |
+
"w2_raw_to_w4_mse_ratio": 7.253725521526435,
|
| 336 |
+
"w2_composite_to_w4_mse_ratio": 7.253725521526435,
|
| 337 |
+
"old_rank128_improvement_vs_raw": 0.9999999999923329,
|
| 338 |
+
"phase1b_improvement_vs_raw": 0.9999999999923329,
|
| 339 |
+
"phase2_improvement_vs_raw": 0.9999999999923329,
|
| 340 |
+
"composite_improvement_vs_raw": 0.9999999999923329,
|
| 341 |
+
"remaining_w2_specific_gap_score": 0.8156589509387167,
|
| 342 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 343 |
+
"recommended_rank": 256
|
| 344 |
+
},
|
| 345 |
+
{
|
| 346 |
+
"training_priority_rank": 3,
|
| 347 |
+
"module_name": "up_blocks.1.attentions.0.transformer_blocks.1.attn1.to_q",
|
| 348 |
+
"ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.1.attn1.to_q.weight",
|
| 349 |
+
"suffix": "attn1.to_q",
|
| 350 |
+
"shape": [
|
| 351 |
+
640,
|
| 352 |
+
640
|
| 353 |
+
],
|
| 354 |
+
"has_old_rank128": false,
|
| 355 |
+
"has_phase1b": false,
|
| 356 |
+
"has_phase2": false,
|
| 357 |
+
"composite_source": "raw",
|
| 358 |
+
"w4_mse_mean": 0.01650474433388029,
|
| 359 |
+
"w2_raw_mse_mean": 0.12441772927663156,
|
| 360 |
+
"w2_old_rank128_mse_mean": 0.12441772927663156,
|
| 361 |
+
"w2_phase1b_mse_mean": 0.12441772927663156,
|
| 362 |
+
"w2_phase2_mse_mean": 0.12441772927663156,
|
| 363 |
+
"w2_composite_mse_mean": 0.12441772927663156,
|
| 364 |
+
"w2_raw_to_w4_mse_ratio": 7.5383009122833515,
|
| 365 |
+
"w2_composite_to_w4_mse_ratio": 7.5383009122833515,
|
| 366 |
+
"old_rank128_improvement_vs_raw": 0.9999999999919625,
|
| 367 |
+
"phase1b_improvement_vs_raw": 0.9999999999919625,
|
| 368 |
+
"phase2_improvement_vs_raw": 0.9999999999919625,
|
| 369 |
+
"composite_improvement_vs_raw": 0.9999999999919625,
|
| 370 |
+
"remaining_w2_specific_gap_score": 0.8134805528411615,
|
| 371 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 372 |
+
"recommended_rank": 256
|
| 373 |
+
},
|
| 374 |
+
{
|
| 375 |
+
"training_priority_rank": 4,
|
| 376 |
+
"module_name": "down_blocks.1.attentions.1.transformer_blocks.1.attn1.to_q",
|
| 377 |
+
"ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_q.weight",
|
| 378 |
+
"suffix": "attn1.to_q",
|
| 379 |
+
"shape": [
|
| 380 |
+
640,
|
| 381 |
+
640
|
| 382 |
+
],
|
| 383 |
+
"has_old_rank128": false,
|
| 384 |
+
"has_phase1b": false,
|
| 385 |
+
"has_phase2": false,
|
| 386 |
+
"composite_source": "raw",
|
| 387 |
+
"w4_mse_mean": 0.016551085664624616,
|
| 388 |
+
"w2_raw_mse_mean": 0.12307428023112672,
|
| 389 |
+
"w2_old_rank128_mse_mean": 0.12307428023112672,
|
| 390 |
+
"w2_phase1b_mse_mean": 0.12307428023112672,
|
| 391 |
+
"w2_phase2_mse_mean": 0.12307428023112672,
|
| 392 |
+
"w2_composite_mse_mean": 0.12307428023112672,
|
| 393 |
+
"w2_raw_to_w4_mse_ratio": 7.436024603917248,
|
| 394 |
+
"w2_composite_to_w4_mse_ratio": 7.436024603917248,
|
| 395 |
+
"old_rank128_improvement_vs_raw": 0.9999999999918748,
|
| 396 |
+
"phase1b_improvement_vs_raw": 0.9999999999918748,
|
| 397 |
+
"phase2_improvement_vs_raw": 0.9999999999918748,
|
| 398 |
+
"composite_improvement_vs_raw": 0.9999999999918748,
|
| 399 |
+
"remaining_w2_specific_gap_score": 0.7921090956843738,
|
| 400 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 401 |
+
"recommended_rank": 256
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"training_priority_rank": 5,
|
| 405 |
+
"module_name": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_q",
|
| 406 |
+
"ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_q.weight",
|
| 407 |
+
"suffix": "attn1.to_q",
|
| 408 |
+
"shape": [
|
| 409 |
+
640,
|
| 410 |
+
640
|
| 411 |
+
],
|
| 412 |
+
"has_old_rank128": false,
|
| 413 |
+
"has_phase1b": false,
|
| 414 |
+
"has_phase2": false,
|
| 415 |
+
"composite_source": "raw",
|
| 416 |
+
"w4_mse_mean": 0.015716641160127307,
|
| 417 |
+
"w2_raw_mse_mean": 0.11642114937837635,
|
| 418 |
+
"w2_old_rank128_mse_mean": 0.11642114937837635,
|
| 419 |
+
"w2_phase1b_mse_mean": 0.11642114937837635,
|
| 420 |
+
"w2_phase2_mse_mean": 0.11642114937837635,
|
| 421 |
+
"w2_composite_mse_mean": 0.11642114937837635,
|
| 422 |
+
"w2_raw_to_w4_mse_ratio": 7.407508270044756,
|
| 423 |
+
"w2_composite_to_w4_mse_ratio": 7.407508270044756,
|
| 424 |
+
"old_rank128_improvement_vs_raw": 0.9999999999914104,
|
| 425 |
+
"phase1b_improvement_vs_raw": 0.9999999999914104,
|
| 426 |
+
"phase2_improvement_vs_raw": 0.9999999999914104,
|
| 427 |
+
"composite_improvement_vs_raw": 0.9999999999914104,
|
| 428 |
+
"remaining_w2_specific_gap_score": 0.7459694774574699,
|
| 429 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 430 |
+
"recommended_rank": 256
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"training_priority_rank": 6,
|
| 434 |
+
"module_name": "up_blocks.1.attentions.1.transformer_blocks.1.attn1.to_k",
|
| 435 |
+
"ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.1.attn1.to_k.weight",
|
| 436 |
+
"suffix": "attn1.to_k",
|
| 437 |
+
"shape": [
|
| 438 |
+
640,
|
| 439 |
+
640
|
| 440 |
+
],
|
| 441 |
+
"has_old_rank128": false,
|
| 442 |
+
"has_phase1b": false,
|
| 443 |
+
"has_phase2": false,
|
| 444 |
+
"composite_source": "raw",
|
| 445 |
+
"w4_mse_mean": 0.01650732191878238,
|
| 446 |
+
"w2_raw_mse_mean": 0.11842691459293876,
|
| 447 |
+
"w2_old_rank128_mse_mean": 0.11842691459293876,
|
| 448 |
+
"w2_phase1b_mse_mean": 0.11842691459293876,
|
| 449 |
+
"w2_phase2_mse_mean": 0.11842691459293876,
|
| 450 |
+
"w2_composite_mse_mean": 0.11842691459293876,
|
| 451 |
+
"w2_raw_to_w4_mse_ratio": 7.174205190183874,
|
| 452 |
+
"w2_composite_to_w4_mse_ratio": 7.174205190183874,
|
| 453 |
+
"old_rank128_improvement_vs_raw": 0.999999999991556,
|
| 454 |
+
"phase1b_improvement_vs_raw": 0.999999999991556,
|
| 455 |
+
"phase2_improvement_vs_raw": 0.999999999991556,
|
| 456 |
+
"composite_improvement_vs_raw": 0.999999999991556,
|
| 457 |
+
"remaining_w2_specific_gap_score": 0.7311920707443591,
|
| 458 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 459 |
+
"recommended_rank": 256
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"training_priority_rank": 7,
|
| 463 |
+
"module_name": "mid_block.attentions.0.proj_out",
|
| 464 |
+
"ldm_key": "model.diffusion_model.middle_block.1.proj_out.weight",
|
| 465 |
+
"suffix": "proj_out",
|
| 466 |
+
"shape": [
|
| 467 |
+
1280,
|
| 468 |
+
1280
|
| 469 |
+
],
|
| 470 |
+
"has_old_rank128": true,
|
| 471 |
+
"has_phase1b": true,
|
| 472 |
+
"has_phase2": true,
|
| 473 |
+
"composite_source": "phase2",
|
| 474 |
+
"w4_mse_mean": 0.18414856466863835,
|
| 475 |
+
"w2_raw_mse_mean": 1.3472472195114409,
|
| 476 |
+
"w2_old_rank128_mse_mean": 0.85862555035523,
|
| 477 |
+
"w2_phase1b_mse_mean": 0.4570868414427553,
|
| 478 |
+
"w2_phase2_mse_mean": 0.4683431349694729,
|
| 479 |
+
"w2_composite_mse_mean": 0.4683431349694729,
|
| 480 |
+
"w2_raw_to_w4_mse_ratio": 7.316088626204586,
|
| 481 |
+
"w2_composite_to_w4_mse_ratio": 2.543289630357305,
|
| 482 |
+
"old_rank128_improvement_vs_raw": 1.5690742244421791,
|
| 483 |
+
"phase1b_improvement_vs_raw": 2.9474644582986103,
|
| 484 |
+
"phase2_improvement_vs_raw": 2.8766242502868744,
|
| 485 |
+
"composite_improvement_vs_raw": 2.8766242502868744,
|
| 486 |
+
"remaining_w2_specific_gap_score": 0.7227891036499625,
|
| 487 |
+
"recommended_action": "train_block_level_or_rank256_plus_real_latents",
|
| 488 |
+
"recommended_rank": 256
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"training_priority_rank": 8,
|
| 492 |
+
"module_name": "up_blocks.0.attentions.1.proj_out",
|
| 493 |
+
"ldm_key": "model.diffusion_model.output_blocks.1.1.proj_out.weight",
|
| 494 |
+
"suffix": "proj_out",
|
| 495 |
+
"shape": [
|
| 496 |
+
1280,
|
| 497 |
+
1280
|
| 498 |
+
],
|
| 499 |
+
"has_old_rank128": true,
|
| 500 |
+
"has_phase1b": true,
|
| 501 |
+
"has_phase2": true,
|
| 502 |
+
"composite_source": "phase2",
|
| 503 |
+
"w4_mse_mean": 0.1777853944471904,
|
| 504 |
+
"w2_raw_mse_mean": 1.3340980570231165,
|
| 505 |
+
"w2_old_rank128_mse_mean": 0.7680485237921987,
|
| 506 |
+
"w2_phase1b_mse_mean": 0.6045318022370338,
|
| 507 |
+
"w2_phase2_mse_mean": 0.4564306310244969,
|
| 508 |
+
"w2_composite_mse_mean": 0.4564306310244969,
|
| 509 |
+
"w2_raw_to_w4_mse_ratio": 7.503980071950705,
|
| 510 |
+
"w2_composite_to_w4_mse_ratio": 2.5673123061720813,
|
| 511 |
+
"old_rank128_improvement_vs_raw": 1.7369970980926328,
|
| 512 |
+
"phase1b_improvement_vs_raw": 2.2068285772297824,
|
| 513 |
+
"phase2_improvement_vs_raw": 2.9228933518894173,
|
| 514 |
+
"composite_improvement_vs_raw": 2.9228933518894173,
|
| 515 |
+
"remaining_w2_specific_gap_score": 0.7153693449211498,
|
| 516 |
+
"recommended_action": "train_block_level_or_rank256_plus_real_latents",
|
| 517 |
+
"recommended_rank": 256
|
| 518 |
+
},
|
| 519 |
+
{
|
| 520 |
+
"training_priority_rank": 9,
|
| 521 |
+
"module_name": "up_blocks.1.attentions.0.transformer_blocks.1.attn1.to_k",
|
| 522 |
+
"ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.1.attn1.to_k.weight",
|
| 523 |
+
"suffix": "attn1.to_k",
|
| 524 |
+
"shape": [
|
| 525 |
+
640,
|
| 526 |
+
640
|
| 527 |
+
],
|
| 528 |
+
"has_old_rank128": false,
|
| 529 |
+
"has_phase1b": false,
|
| 530 |
+
"has_phase2": false,
|
| 531 |
+
"composite_source": "raw",
|
| 532 |
+
"w4_mse_mean": 0.01555688923690468,
|
| 533 |
+
"w2_raw_mse_mean": 0.11352902118648801,
|
| 534 |
+
"w2_old_rank128_mse_mean": 0.11352902118648801,
|
| 535 |
+
"w2_phase1b_mse_mean": 0.11352902118648801,
|
| 536 |
+
"w2_phase2_mse_mean": 0.11352902118648801,
|
| 537 |
+
"w2_composite_mse_mean": 0.11352902118648801,
|
| 538 |
+
"w2_raw_to_w4_mse_ratio": 7.29766854094919,
|
| 539 |
+
"w2_composite_to_w4_mse_ratio": 7.29766854094919,
|
| 540 |
+
"old_rank128_improvement_vs_raw": 0.9999999999911916,
|
| 541 |
+
"phase1b_improvement_vs_raw": 0.9999999999911916,
|
| 542 |
+
"phase2_improvement_vs_raw": 0.9999999999911916,
|
| 543 |
+
"composite_improvement_vs_raw": 0.9999999999911916,
|
| 544 |
+
"remaining_w2_specific_gap_score": 0.7149681452181973,
|
| 545 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 546 |
+
"recommended_rank": 256
|
| 547 |
+
},
|
| 548 |
+
{
|
| 549 |
+
"training_priority_rank": 10,
|
| 550 |
+
"module_name": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_q",
|
| 551 |
+
"ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_q.weight",
|
| 552 |
+
"suffix": "attn1.to_q",
|
| 553 |
+
"shape": [
|
| 554 |
+
640,
|
| 555 |
+
640
|
| 556 |
+
],
|
| 557 |
+
"has_old_rank128": false,
|
| 558 |
+
"has_phase1b": false,
|
| 559 |
+
"has_phase2": false,
|
| 560 |
+
"composite_source": "raw",
|
| 561 |
+
"w4_mse_mean": 0.015393487393989094,
|
| 562 |
+
"w2_raw_mse_mean": 0.11155489552766085,
|
| 563 |
+
"w2_old_rank128_mse_mean": 0.11155489552766085,
|
| 564 |
+
"w2_phase1b_mse_mean": 0.11155489552766085,
|
| 565 |
+
"w2_phase2_mse_mean": 0.11155489552766085,
|
| 566 |
+
"w2_composite_mse_mean": 0.11155489552766085,
|
| 567 |
+
"w2_raw_to_w4_mse_ratio": 7.246889068423464,
|
| 568 |
+
"w2_composite_to_w4_mse_ratio": 7.246889068423464,
|
| 569 |
+
"old_rank128_improvement_vs_raw": 0.9999999999910357,
|
| 570 |
+
"phase1b_improvement_vs_raw": 0.9999999999910357,
|
| 571 |
+
"phase2_improvement_vs_raw": 0.9999999999910357,
|
| 572 |
+
"composite_improvement_vs_raw": 0.9999999999910357,
|
| 573 |
+
"remaining_w2_specific_gap_score": 0.6968710574081131,
|
| 574 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 575 |
+
"recommended_rank": 256
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"training_priority_rank": 11,
|
| 579 |
+
"module_name": "down_blocks.1.attentions.1.transformer_blocks.1.attn1.to_k",
|
| 580 |
+
"ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_k.weight",
|
| 581 |
+
"suffix": "attn1.to_k",
|
| 582 |
+
"shape": [
|
| 583 |
+
640,
|
| 584 |
+
640
|
| 585 |
+
],
|
| 586 |
+
"has_old_rank128": false,
|
| 587 |
+
"has_phase1b": false,
|
| 588 |
+
"has_phase2": false,
|
| 589 |
+
"composite_source": "raw",
|
| 590 |
+
"w4_mse_mean": 0.015574711058954043,
|
| 591 |
+
"w2_raw_mse_mean": 0.11173469506736312,
|
| 592 |
+
"w2_old_rank128_mse_mean": 0.11173469506736312,
|
| 593 |
+
"w2_phase1b_mse_mean": 0.11173469506736312,
|
| 594 |
+
"w2_phase2_mse_mean": 0.11173469506736312,
|
| 595 |
+
"w2_composite_mse_mean": 0.11173469506736312,
|
| 596 |
+
"w2_raw_to_w4_mse_ratio": 7.174110302094608,
|
| 597 |
+
"w2_composite_to_w4_mse_ratio": 7.174110302094608,
|
| 598 |
+
"old_rank128_improvement_vs_raw": 0.9999999999910502,
|
| 599 |
+
"phase1b_improvement_vs_raw": 0.9999999999910502,
|
| 600 |
+
"phase2_improvement_vs_raw": 0.9999999999910502,
|
| 601 |
+
"composite_improvement_vs_raw": 0.9999999999910502,
|
| 602 |
+
"remaining_w2_specific_gap_score": 0.6898623319239804,
|
| 603 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 604 |
+
"recommended_rank": 256
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"training_priority_rank": 12,
|
| 608 |
+
"module_name": "up_blocks.0.attentions.2.transformer_blocks.9.attn2.to_q",
|
| 609 |
+
"ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn2.to_q.weight",
|
| 610 |
+
"suffix": "attn2.to_q",
|
| 611 |
+
"shape": [
|
| 612 |
+
1280,
|
| 613 |
+
1280
|
| 614 |
+
],
|
| 615 |
+
"has_old_rank128": false,
|
| 616 |
+
"has_phase1b": false,
|
| 617 |
+
"has_phase2": false,
|
| 618 |
+
"composite_source": "raw",
|
| 619 |
+
"w4_mse_mean": 0.006283481646927872,
|
| 620 |
+
"w2_raw_mse_mean": 0.06878376293129154,
|
| 621 |
+
"w2_old_rank128_mse_mean": 0.06878376293129154,
|
| 622 |
+
"w2_phase1b_mse_mean": 0.06878376293129154,
|
| 623 |
+
"w2_phase2_mse_mean": 0.06878376293129154,
|
| 624 |
+
"w2_composite_mse_mean": 0.06878376293129154,
|
| 625 |
+
"w2_raw_to_w4_mse_ratio": 10.94675958096808,
|
| 626 |
+
"w2_composite_to_w4_mse_ratio": 10.94675958096808,
|
| 627 |
+
"old_rank128_improvement_vs_raw": 0.9999999999854616,
|
| 628 |
+
"phase1b_improvement_vs_raw": 0.9999999999854616,
|
| 629 |
+
"phase2_improvement_vs_raw": 0.9999999999854616,
|
| 630 |
+
"composite_improvement_vs_raw": 0.9999999999854616,
|
| 631 |
+
"remaining_w2_specific_gap_score": 0.6841755529628081,
|
| 632 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 633 |
+
"recommended_rank": 256
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"training_priority_rank": 13,
|
| 637 |
+
"module_name": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_q",
|
| 638 |
+
"ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_q.weight",
|
| 639 |
+
"suffix": "attn1.to_q",
|
| 640 |
+
"shape": [
|
| 641 |
+
640,
|
| 642 |
+
640
|
| 643 |
+
],
|
| 644 |
+
"has_old_rank128": false,
|
| 645 |
+
"has_phase1b": false,
|
| 646 |
+
"has_phase2": false,
|
| 647 |
+
"composite_source": "raw",
|
| 648 |
+
"w4_mse_mean": 0.012902505862127458,
|
| 649 |
+
"w2_raw_mse_mean": 0.0985980766958424,
|
| 650 |
+
"w2_old_rank128_mse_mean": 0.0985980766958424,
|
| 651 |
+
"w2_phase1b_mse_mean": 0.0985980766958424,
|
| 652 |
+
"w2_phase2_mse_mean": 0.0985980766958424,
|
| 653 |
+
"w2_composite_mse_mean": 0.0985980766958424,
|
| 654 |
+
"w2_raw_to_w4_mse_ratio": 7.641777321536753,
|
| 655 |
+
"w2_composite_to_w4_mse_ratio": 7.641777321536753,
|
| 656 |
+
"old_rank128_improvement_vs_raw": 0.9999999999898578,
|
| 657 |
+
"phase1b_improvement_vs_raw": 0.9999999999898578,
|
| 658 |
+
"phase2_improvement_vs_raw": 0.9999999999898578,
|
| 659 |
+
"composite_improvement_vs_raw": 0.9999999999898578,
|
| 660 |
+
"remaining_w2_specific_gap_score": 0.6548664697532293,
|
| 661 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 662 |
+
"recommended_rank": 256
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"training_priority_rank": 14,
|
| 666 |
+
"module_name": "up_blocks.1.attentions.1.transformer_blocks.1.attn1.to_v",
|
| 667 |
+
"ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.1.attn1.to_v.weight",
|
| 668 |
+
"suffix": "attn1.to_v",
|
| 669 |
+
"shape": [
|
| 670 |
+
640,
|
| 671 |
+
640
|
| 672 |
+
],
|
| 673 |
+
"has_old_rank128": false,
|
| 674 |
+
"has_phase1b": false,
|
| 675 |
+
"has_phase2": false,
|
| 676 |
+
"composite_source": "raw",
|
| 677 |
+
"w4_mse_mean": 0.014045619937990392,
|
| 678 |
+
"w2_raw_mse_mean": 0.10288021394184657,
|
| 679 |
+
"w2_old_rank128_mse_mean": 0.10288021394184657,
|
| 680 |
+
"w2_phase1b_mse_mean": 0.10288021394184657,
|
| 681 |
+
"w2_phase2_mse_mean": 0.10288021394184657,
|
| 682 |
+
"w2_composite_mse_mean": 0.10288021394184657,
|
| 683 |
+
"w2_raw_to_w4_mse_ratio": 7.3247186232238075,
|
| 684 |
+
"w2_composite_to_w4_mse_ratio": 7.3247186232238075,
|
| 685 |
+
"old_rank128_improvement_vs_raw": 0.9999999999902799,
|
| 686 |
+
"phase1b_improvement_vs_raw": 0.9999999999902799,
|
| 687 |
+
"phase2_improvement_vs_raw": 0.9999999999902799,
|
| 688 |
+
"composite_improvement_vs_raw": 0.9999999999902799,
|
| 689 |
+
"remaining_w2_specific_gap_score": 0.6506884050865713,
|
| 690 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 691 |
+
"recommended_rank": 256
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"training_priority_rank": 15,
|
| 695 |
+
"module_name": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_k",
|
| 696 |
+
"ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_k.weight",
|
| 697 |
+
"suffix": "attn1.to_k",
|
| 698 |
+
"shape": [
|
| 699 |
+
640,
|
| 700 |
+
640
|
| 701 |
+
],
|
| 702 |
+
"has_old_rank128": false,
|
| 703 |
+
"has_phase1b": false,
|
| 704 |
+
"has_phase2": false,
|
| 705 |
+
"composite_source": "raw",
|
| 706 |
+
"w4_mse_mean": 0.015184641383322222,
|
| 707 |
+
"w2_raw_mse_mean": 0.10630956571549177,
|
| 708 |
+
"w2_old_rank128_mse_mean": 0.10630956571549177,
|
| 709 |
+
"w2_phase1b_mse_mean": 0.10630956571549177,
|
| 710 |
+
"w2_phase2_mse_mean": 0.10630956571549177,
|
| 711 |
+
"w2_composite_mse_mean": 0.10630956571549177,
|
| 712 |
+
"w2_raw_to_w4_mse_ratio": 7.0011245590069615,
|
| 713 |
+
"w2_composite_to_w4_mse_ratio": 7.0011245590069615,
|
| 714 |
+
"old_rank128_improvement_vs_raw": 0.9999999999905934,
|
| 715 |
+
"phase1b_improvement_vs_raw": 0.9999999999905934,
|
| 716 |
+
"phase2_improvement_vs_raw": 0.9999999999905934,
|
| 717 |
+
"composite_improvement_vs_raw": 0.9999999999905934,
|
| 718 |
+
"remaining_w2_specific_gap_score": 0.6379769456796033,
|
| 719 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 720 |
+
"recommended_rank": 256
|
| 721 |
+
},
|
| 722 |
+
{
|
| 723 |
+
"training_priority_rank": 16,
|
| 724 |
+
"module_name": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_k",
|
| 725 |
+
"ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_k.weight",
|
| 726 |
+
"suffix": "attn1.to_k",
|
| 727 |
+
"shape": [
|
| 728 |
+
640,
|
| 729 |
+
640
|
| 730 |
+
],
|
| 731 |
+
"has_old_rank128": false,
|
| 732 |
+
"has_phase1b": false,
|
| 733 |
+
"has_phase2": false,
|
| 734 |
+
"composite_source": "raw",
|
| 735 |
+
"w4_mse_mean": 0.015266332110123975,
|
| 736 |
+
"w2_raw_mse_mean": 0.10559878458401986,
|
| 737 |
+
"w2_old_rank128_mse_mean": 0.10559878458401986,
|
| 738 |
+
"w2_phase1b_mse_mean": 0.10559878458401986,
|
| 739 |
+
"w2_phase2_mse_mean": 0.10559878458401986,
|
| 740 |
+
"w2_composite_mse_mean": 0.10559878458401986,
|
| 741 |
+
"w2_raw_to_w4_mse_ratio": 6.917102537489944,
|
| 742 |
+
"w2_composite_to_w4_mse_ratio": 6.917102537489944,
|
| 743 |
+
"old_rank128_improvement_vs_raw": 0.9999999999905301,
|
| 744 |
+
"phase1b_improvement_vs_raw": 0.9999999999905301,
|
| 745 |
+
"phase2_improvement_vs_raw": 0.9999999999905301,
|
| 746 |
+
"composite_improvement_vs_raw": 0.9999999999905301,
|
| 747 |
+
"remaining_w2_specific_gap_score": 0.624838836224875,
|
| 748 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 749 |
+
"recommended_rank": 256
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"training_priority_rank": 17,
|
| 753 |
+
"module_name": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_k",
|
| 754 |
+
"ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_k.weight",
|
| 755 |
+
"suffix": "attn1.to_k",
|
| 756 |
+
"shape": [
|
| 757 |
+
640,
|
| 758 |
+
640
|
| 759 |
+
],
|
| 760 |
+
"has_old_rank128": false,
|
| 761 |
+
"has_phase1b": false,
|
| 762 |
+
"has_phase2": false,
|
| 763 |
+
"composite_source": "raw",
|
| 764 |
+
"w4_mse_mean": 0.0125871665908822,
|
| 765 |
+
"w2_raw_mse_mean": 0.09344679423208747,
|
| 766 |
+
"w2_old_rank128_mse_mean": 0.09344679423208747,
|
| 767 |
+
"w2_phase1b_mse_mean": 0.09344679423208747,
|
| 768 |
+
"w2_phase2_mse_mean": 0.09344679423208747,
|
| 769 |
+
"w2_composite_mse_mean": 0.09344679423208747,
|
| 770 |
+
"w2_raw_to_w4_mse_ratio": 7.423973739439805,
|
| 771 |
+
"w2_composite_to_w4_mse_ratio": 7.423973739439805,
|
| 772 |
+
"old_rank128_improvement_vs_raw": 0.9999999999892987,
|
| 773 |
+
"phase1b_improvement_vs_raw": 0.9999999999892987,
|
| 774 |
+
"phase2_improvement_vs_raw": 0.9999999999892987,
|
| 775 |
+
"composite_improvement_vs_raw": 0.9999999999892987,
|
| 776 |
+
"remaining_w2_specific_gap_score": 0.6002997521891889,
|
| 777 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 778 |
+
"recommended_rank": 256
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"training_priority_rank": 18,
|
| 782 |
+
"module_name": "down_blocks.1.attentions.1.transformer_blocks.1.attn1.to_v",
|
| 783 |
+
"ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_v.weight",
|
| 784 |
+
"suffix": "attn1.to_v",
|
| 785 |
+
"shape": [
|
| 786 |
+
640,
|
| 787 |
+
640
|
| 788 |
+
],
|
| 789 |
+
"has_old_rank128": false,
|
| 790 |
+
"has_phase1b": false,
|
| 791 |
+
"has_phase2": false,
|
| 792 |
+
"composite_source": "raw",
|
| 793 |
+
"w4_mse_mean": 0.011522587821153658,
|
| 794 |
+
"w2_raw_mse_mean": 0.08661618735641241,
|
| 795 |
+
"w2_old_rank128_mse_mean": 0.08661618735641241,
|
| 796 |
+
"w2_phase1b_mse_mean": 0.08661618735641241,
|
| 797 |
+
"w2_phase2_mse_mean": 0.08661618735641241,
|
| 798 |
+
"w2_composite_mse_mean": 0.08661618735641241,
|
| 799 |
+
"w2_raw_to_w4_mse_ratio": 7.517077647252264,
|
| 800 |
+
"w2_composite_to_w4_mse_ratio": 7.517077647252264,
|
| 801 |
+
"old_rank128_improvement_vs_raw": 0.9999999999884548,
|
| 802 |
+
"phase1b_improvement_vs_raw": 0.9999999999884548,
|
| 803 |
+
"phase2_improvement_vs_raw": 0.9999999999884548,
|
| 804 |
+
"composite_improvement_vs_raw": 0.9999999999884548,
|
| 805 |
+
"remaining_w2_specific_gap_score": 0.5644844185182065,
|
| 806 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 807 |
+
"recommended_rank": 256
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"training_priority_rank": 19,
|
| 811 |
+
"module_name": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_q",
|
| 812 |
+
"ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_q.weight",
|
| 813 |
+
"suffix": "attn1.to_q",
|
| 814 |
+
"shape": [
|
| 815 |
+
640,
|
| 816 |
+
640
|
| 817 |
+
],
|
| 818 |
+
"has_old_rank128": false,
|
| 819 |
+
"has_phase1b": false,
|
| 820 |
+
"has_phase2": false,
|
| 821 |
+
"composite_source": "raw",
|
| 822 |
+
"w4_mse_mean": 0.011824695875735156,
|
| 823 |
+
"w2_raw_mse_mean": 0.08770120383373328,
|
| 824 |
+
"w2_old_rank128_mse_mean": 0.08770120383373328,
|
| 825 |
+
"w2_phase1b_mse_mean": 0.08770120383373328,
|
| 826 |
+
"w2_phase2_mse_mean": 0.08770120383373328,
|
| 827 |
+
"w2_composite_mse_mean": 0.08770120383373328,
|
| 828 |
+
"w2_raw_to_w4_mse_ratio": 7.416783040169649,
|
| 829 |
+
"w2_composite_to_w4_mse_ratio": 7.416783040169649,
|
| 830 |
+
"old_rank128_improvement_vs_raw": 0.9999999999885976,
|
| 831 |
+
"phase1b_improvement_vs_raw": 0.9999999999885976,
|
| 832 |
+
"phase2_improvement_vs_raw": 0.9999999999885976,
|
| 833 |
+
"composite_improvement_vs_raw": 0.9999999999885976,
|
| 834 |
+
"remaining_w2_specific_gap_score": 0.5627595973701779,
|
| 835 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 836 |
+
"recommended_rank": 256
|
| 837 |
+
},
|
| 838 |
+
{
|
| 839 |
+
"training_priority_rank": 20,
|
| 840 |
+
"module_name": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_v",
|
| 841 |
+
"ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_v.weight",
|
| 842 |
+
"suffix": "attn1.to_v",
|
| 843 |
+
"shape": [
|
| 844 |
+
640,
|
| 845 |
+
640
|
| 846 |
+
],
|
| 847 |
+
"has_old_rank128": false,
|
| 848 |
+
"has_phase1b": false,
|
| 849 |
+
"has_phase2": false,
|
| 850 |
+
"composite_source": "raw",
|
| 851 |
+
"w4_mse_mean": 0.013221082560318922,
|
| 852 |
+
"w2_raw_mse_mean": 0.09296748986733812,
|
| 853 |
+
"w2_old_rank128_mse_mean": 0.09296748986733812,
|
| 854 |
+
"w2_phase1b_mse_mean": 0.09296748986733812,
|
| 855 |
+
"w2_phase2_mse_mean": 0.09296748986733812,
|
| 856 |
+
"w2_composite_mse_mean": 0.09296748986733812,
|
| 857 |
+
"w2_raw_to_w4_mse_ratio": 7.031760783291242,
|
| 858 |
+
"w2_composite_to_w4_mse_ratio": 7.031760783291242,
|
| 859 |
+
"old_rank128_improvement_vs_raw": 0.9999999999892435,
|
| 860 |
+
"phase1b_improvement_vs_raw": 0.9999999999892435,
|
| 861 |
+
"phase2_improvement_vs_raw": 0.9999999999892435,
|
| 862 |
+
"composite_improvement_vs_raw": 0.9999999999892435,
|
| 863 |
+
"remaining_w2_specific_gap_score": 0.5607576595098677,
|
| 864 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 865 |
+
"recommended_rank": 256
|
| 866 |
+
},
|
| 867 |
+
{
|
| 868 |
+
"training_priority_rank": 21,
|
| 869 |
+
"module_name": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_k",
|
| 870 |
+
"ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_k.weight",
|
| 871 |
+
"suffix": "attn1.to_k",
|
| 872 |
+
"shape": [
|
| 873 |
+
640,
|
| 874 |
+
640
|
| 875 |
+
],
|
| 876 |
+
"has_old_rank128": false,
|
| 877 |
+
"has_phase1b": false,
|
| 878 |
+
"has_phase2": false,
|
| 879 |
+
"composite_source": "raw",
|
| 880 |
+
"w4_mse_mean": 0.011692194467676538,
|
| 881 |
+
"w2_raw_mse_mean": 0.08697623320456062,
|
| 882 |
+
"w2_old_rank128_mse_mean": 0.08697623320456062,
|
| 883 |
+
"w2_phase1b_mse_mean": 0.08697623320456062,
|
| 884 |
+
"w2_phase2_mse_mean": 0.08697623320456062,
|
| 885 |
+
"w2_composite_mse_mean": 0.08697623320456062,
|
| 886 |
+
"w2_raw_to_w4_mse_ratio": 7.438828821875183,
|
| 887 |
+
"w2_composite_to_w4_mse_ratio": 7.438828821875183,
|
| 888 |
+
"old_rank128_improvement_vs_raw": 0.9999999999885025,
|
| 889 |
+
"phase1b_improvement_vs_raw": 0.9999999999885025,
|
| 890 |
+
"phase2_improvement_vs_raw": 0.9999999999885025,
|
| 891 |
+
"composite_improvement_vs_raw": 0.9999999999885025,
|
| 892 |
+
"remaining_w2_specific_gap_score": 0.560025077183101,
|
| 893 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 894 |
+
"recommended_rank": 256
|
| 895 |
+
},
|
| 896 |
+
{
|
| 897 |
+
"training_priority_rank": 22,
|
| 898 |
+
"module_name": "up_blocks.0.attentions.1.transformer_blocks.8.attn1.to_v",
|
| 899 |
+
"ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.8.attn1.to_v.weight",
|
| 900 |
+
"suffix": "attn1.to_v",
|
| 901 |
+
"shape": [
|
| 902 |
+
1280,
|
| 903 |
+
1280
|
| 904 |
+
],
|
| 905 |
+
"has_old_rank128": false,
|
| 906 |
+
"has_phase1b": false,
|
| 907 |
+
"has_phase2": false,
|
| 908 |
+
"composite_source": "raw",
|
| 909 |
+
"w4_mse_mean": 0.01207385812553444,
|
| 910 |
+
"w2_raw_mse_mean": 0.08799946933452572,
|
| 911 |
+
"w2_old_rank128_mse_mean": 0.08799946933452572,
|
| 912 |
+
"w2_phase1b_mse_mean": 0.08799946933452572,
|
| 913 |
+
"w2_phase2_mse_mean": 0.08799946933452572,
|
| 914 |
+
"w2_composite_mse_mean": 0.08799946933452572,
|
| 915 |
+
"w2_raw_to_w4_mse_ratio": 7.288429962675419,
|
| 916 |
+
"w2_composite_to_w4_mse_ratio": 7.288429962675419,
|
| 917 |
+
"old_rank128_improvement_vs_raw": 0.9999999999886362,
|
| 918 |
+
"phase1b_improvement_vs_raw": 0.9999999999886362,
|
| 919 |
+
"phase2_improvement_vs_raw": 0.9999999999886362,
|
| 920 |
+
"composite_improvement_vs_raw": 0.9999999999886362,
|
| 921 |
+
"remaining_w2_specific_gap_score": 0.5533784996700567,
|
| 922 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 923 |
+
"recommended_rank": 256
|
| 924 |
+
},
|
| 925 |
+
{
|
| 926 |
+
"training_priority_rank": 23,
|
| 927 |
+
"module_name": "down_blocks.1.attentions.0.transformer_blocks.1.attn1.to_q",
|
| 928 |
+
"ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_q.weight",
|
| 929 |
+
"suffix": "attn1.to_q",
|
| 930 |
+
"shape": [
|
| 931 |
+
640,
|
| 932 |
+
640
|
| 933 |
+
],
|
| 934 |
+
"has_old_rank128": false,
|
| 935 |
+
"has_phase1b": false,
|
| 936 |
+
"has_phase2": false,
|
| 937 |
+
"composite_source": "raw",
|
| 938 |
+
"w4_mse_mean": 0.012419876642525196,
|
| 939 |
+
"w2_raw_mse_mean": 0.08907664686973606,
|
| 940 |
+
"w2_old_rank128_mse_mean": 0.08907664686973606,
|
| 941 |
+
"w2_phase1b_mse_mean": 0.08907664686973606,
|
| 942 |
+
"w2_phase2_mse_mean": 0.08907664686973606,
|
| 943 |
+
"w2_composite_mse_mean": 0.08907664686973606,
|
| 944 |
+
"w2_raw_to_w4_mse_ratio": 7.1721039931724295,
|
| 945 |
+
"w2_composite_to_w4_mse_ratio": 7.1721039931724295,
|
| 946 |
+
"old_rank128_improvement_vs_raw": 0.9999999999887736,
|
| 947 |
+
"phase1b_improvement_vs_raw": 0.9999999999887736,
|
| 948 |
+
"phase2_improvement_vs_raw": 0.9999999999887736,
|
| 949 |
+
"composite_improvement_vs_raw": 0.9999999999887736,
|
| 950 |
+
"remaining_w2_specific_gap_score": 0.5497903278502804,
|
| 951 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 952 |
+
"recommended_rank": 256
|
| 953 |
+
},
|
| 954 |
+
{
|
| 955 |
+
"training_priority_rank": 24,
|
| 956 |
+
"module_name": "up_blocks.1.attentions.2.transformer_blocks.1.attn1.to_q",
|
| 957 |
+
"ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.1.attn1.to_q.weight",
|
| 958 |
+
"suffix": "attn1.to_q",
|
| 959 |
+
"shape": [
|
| 960 |
+
640,
|
| 961 |
+
640
|
| 962 |
+
],
|
| 963 |
+
"has_old_rank128": false,
|
| 964 |
+
"has_phase1b": false,
|
| 965 |
+
"has_phase2": false,
|
| 966 |
+
"composite_source": "raw",
|
| 967 |
+
"w4_mse_mean": 0.012714513893505293,
|
| 968 |
+
"w2_raw_mse_mean": 0.0901942936969655,
|
| 969 |
+
"w2_old_rank128_mse_mean": 0.0901942936969655,
|
| 970 |
+
"w2_phase1b_mse_mean": 0.0901942936969655,
|
| 971 |
+
"w2_phase2_mse_mean": 0.0901942936969655,
|
| 972 |
+
"w2_composite_mse_mean": 0.0901942936969655,
|
| 973 |
+
"w2_raw_to_w4_mse_ratio": 7.09380590129709,
|
| 974 |
+
"w2_composite_to_w4_mse_ratio": 7.09380590129709,
|
| 975 |
+
"old_rank128_improvement_vs_raw": 0.9999999999889128,
|
| 976 |
+
"phase1b_improvement_vs_raw": 0.9999999999889128,
|
| 977 |
+
"phase2_improvement_vs_raw": 0.9999999999889128,
|
| 978 |
+
"composite_improvement_vs_raw": 0.9999999999889128,
|
| 979 |
+
"remaining_w2_specific_gap_score": 0.549626519200985,
|
| 980 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 981 |
+
"recommended_rank": 256
|
| 982 |
+
},
|
| 983 |
+
{
|
| 984 |
+
"training_priority_rank": 25,
|
| 985 |
+
"module_name": "up_blocks.0.attentions.1.transformer_blocks.7.attn1.to_v",
|
| 986 |
+
"ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.7.attn1.to_v.weight",
|
| 987 |
+
"suffix": "attn1.to_v",
|
| 988 |
+
"shape": [
|
| 989 |
+
1280,
|
| 990 |
+
1280
|
| 991 |
+
],
|
| 992 |
+
"has_old_rank128": false,
|
| 993 |
+
"has_phase1b": false,
|
| 994 |
+
"has_phase2": false,
|
| 995 |
+
"composite_source": "raw",
|
| 996 |
+
"w4_mse_mean": 0.012551478842007262,
|
| 997 |
+
"w2_raw_mse_mean": 0.08876089937984943,
|
| 998 |
+
"w2_old_rank128_mse_mean": 0.08876089937984943,
|
| 999 |
+
"w2_phase1b_mse_mean": 0.08876089937984943,
|
| 1000 |
+
"w2_phase2_mse_mean": 0.08876089937984943,
|
| 1001 |
+
"w2_composite_mse_mean": 0.08876089937984943,
|
| 1002 |
+
"w2_raw_to_w4_mse_ratio": 7.071748316677466,
|
| 1003 |
+
"w2_composite_to_w4_mse_ratio": 7.071748316677466,
|
| 1004 |
+
"old_rank128_improvement_vs_raw": 0.9999999999887337,
|
| 1005 |
+
"phase1b_improvement_vs_raw": 0.9999999999887337,
|
| 1006 |
+
"phase2_improvement_vs_raw": 0.9999999999887337,
|
| 1007 |
+
"composite_improvement_vs_raw": 0.9999999999887337,
|
| 1008 |
+
"remaining_w2_specific_gap_score": 0.5389338414034505,
|
| 1009 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1010 |
+
"recommended_rank": 256
|
| 1011 |
+
},
|
| 1012 |
+
{
|
| 1013 |
+
"training_priority_rank": 26,
|
| 1014 |
+
"module_name": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_k",
|
| 1015 |
+
"ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_k.weight",
|
| 1016 |
+
"suffix": "attn1.to_k",
|
| 1017 |
+
"shape": [
|
| 1018 |
+
640,
|
| 1019 |
+
640
|
| 1020 |
+
],
|
| 1021 |
+
"has_old_rank128": false,
|
| 1022 |
+
"has_phase1b": false,
|
| 1023 |
+
"has_phase2": false,
|
| 1024 |
+
"composite_source": "raw",
|
| 1025 |
+
"w4_mse_mean": 0.011307834963580328,
|
| 1026 |
+
"w2_raw_mse_mean": 0.08250298537313938,
|
| 1027 |
+
"w2_old_rank128_mse_mean": 0.08250298537313938,
|
| 1028 |
+
"w2_phase1b_mse_mean": 0.08250298537313938,
|
| 1029 |
+
"w2_phase2_mse_mean": 0.08250298537313938,
|
| 1030 |
+
"w2_composite_mse_mean": 0.08250298537313938,
|
| 1031 |
+
"w2_raw_to_w4_mse_ratio": 7.296090333080073,
|
| 1032 |
+
"w2_composite_to_w4_mse_ratio": 7.296090333080073,
|
| 1033 |
+
"old_rank128_improvement_vs_raw": 0.9999999999878791,
|
| 1034 |
+
"phase1b_improvement_vs_raw": 0.9999999999878791,
|
| 1035 |
+
"phase2_improvement_vs_raw": 0.9999999999878791,
|
| 1036 |
+
"composite_improvement_vs_raw": 0.9999999999878791,
|
| 1037 |
+
"remaining_w2_specific_gap_score": 0.5194462486653656,
|
| 1038 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1039 |
+
"recommended_rank": 256
|
| 1040 |
+
},
|
| 1041 |
+
{
|
| 1042 |
+
"training_priority_rank": 27,
|
| 1043 |
+
"module_name": "down_blocks.1.attentions.0.transformer_blocks.1.attn1.to_k",
|
| 1044 |
+
"ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_k.weight",
|
| 1045 |
+
"suffix": "attn1.to_k",
|
| 1046 |
+
"shape": [
|
| 1047 |
+
640,
|
| 1048 |
+
640
|
| 1049 |
+
],
|
| 1050 |
+
"has_old_rank128": false,
|
| 1051 |
+
"has_phase1b": false,
|
| 1052 |
+
"has_phase2": false,
|
| 1053 |
+
"composite_source": "raw",
|
| 1054 |
+
"w4_mse_mean": 0.012032134258853537,
|
| 1055 |
+
"w2_raw_mse_mean": 0.08493821855102267,
|
| 1056 |
+
"w2_old_rank128_mse_mean": 0.08493821855102267,
|
| 1057 |
+
"w2_phase1b_mse_mean": 0.08493821855102267,
|
| 1058 |
+
"w2_phase2_mse_mean": 0.08493821855102267,
|
| 1059 |
+
"w2_composite_mse_mean": 0.08493821855102267,
|
| 1060 |
+
"w2_raw_to_w4_mse_ratio": 7.059281148019419,
|
| 1061 |
+
"w2_composite_to_w4_mse_ratio": 7.059281148019419,
|
| 1062 |
+
"old_rank128_improvement_vs_raw": 0.9999999999882266,
|
| 1063 |
+
"phase1b_improvement_vs_raw": 0.9999999999882266,
|
| 1064 |
+
"phase2_improvement_vs_raw": 0.9999999999882266,
|
| 1065 |
+
"composite_improvement_vs_raw": 0.9999999999882266,
|
| 1066 |
+
"remaining_w2_specific_gap_score": 0.5146645464196242,
|
| 1067 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1068 |
+
"recommended_rank": 256
|
| 1069 |
+
},
|
| 1070 |
+
{
|
| 1071 |
+
"training_priority_rank": 28,
|
| 1072 |
+
"module_name": "up_blocks.0.attentions.1.transformer_blocks.8.attn1.to_q",
|
| 1073 |
+
"ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.8.attn1.to_q.weight",
|
| 1074 |
+
"suffix": "attn1.to_q",
|
| 1075 |
+
"shape": [
|
| 1076 |
+
1280,
|
| 1077 |
+
1280
|
| 1078 |
+
],
|
| 1079 |
+
"has_old_rank128": false,
|
| 1080 |
+
"has_phase1b": false,
|
| 1081 |
+
"has_phase2": false,
|
| 1082 |
+
"composite_source": "raw",
|
| 1083 |
+
"w4_mse_mean": 0.01177227936152901,
|
| 1084 |
+
"w2_raw_mse_mean": 0.08393190681402173,
|
| 1085 |
+
"w2_old_rank128_mse_mean": 0.08393190681402173,
|
| 1086 |
+
"w2_phase1b_mse_mean": 0.08393190681402173,
|
| 1087 |
+
"w2_phase2_mse_mean": 0.08393190681402173,
|
| 1088 |
+
"w2_composite_mse_mean": 0.08393190681402173,
|
| 1089 |
+
"w2_raw_to_w4_mse_ratio": 7.129622414600161,
|
| 1090 |
+
"w2_composite_to_w4_mse_ratio": 7.129622414600161,
|
| 1091 |
+
"old_rank128_improvement_vs_raw": 0.9999999999880855,
|
| 1092 |
+
"phase1b_improvement_vs_raw": 0.9999999999880855,
|
| 1093 |
+
"phase2_improvement_vs_raw": 0.9999999999880855,
|
| 1094 |
+
"composite_improvement_vs_raw": 0.9999999999880855,
|
| 1095 |
+
"remaining_w2_specific_gap_score": 0.5144708973144891,
|
| 1096 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1097 |
+
"recommended_rank": 256
|
| 1098 |
+
},
|
| 1099 |
+
{
|
| 1100 |
+
"training_priority_rank": 29,
|
| 1101 |
+
"module_name": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_q",
|
| 1102 |
+
"ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_q.weight",
|
| 1103 |
+
"suffix": "attn1.to_q",
|
| 1104 |
+
"shape": [
|
| 1105 |
+
640,
|
| 1106 |
+
640
|
| 1107 |
+
],
|
| 1108 |
+
"has_old_rank128": false,
|
| 1109 |
+
"has_phase1b": false,
|
| 1110 |
+
"has_phase2": false,
|
| 1111 |
+
"composite_source": "raw",
|
| 1112 |
+
"w4_mse_mean": 0.011715796842638935,
|
| 1113 |
+
"w2_raw_mse_mean": 0.0831726472824812,
|
| 1114 |
+
"w2_old_rank128_mse_mean": 0.0831726472824812,
|
| 1115 |
+
"w2_phase1b_mse_mean": 0.0831726472824812,
|
| 1116 |
+
"w2_phase2_mse_mean": 0.0831726472824812,
|
| 1117 |
+
"w2_composite_mse_mean": 0.0831726472824812,
|
| 1118 |
+
"w2_raw_to_w4_mse_ratio": 7.09918824920898,
|
| 1119 |
+
"w2_composite_to_w4_mse_ratio": 7.09918824920898,
|
| 1120 |
+
"old_rank128_improvement_vs_raw": 0.9999999999879767,
|
| 1121 |
+
"phase1b_improvement_vs_raw": 0.9999999999879767,
|
| 1122 |
+
"phase2_improvement_vs_raw": 0.9999999999879767,
|
| 1123 |
+
"composite_improvement_vs_raw": 0.9999999999879767,
|
| 1124 |
+
"remaining_w2_specific_gap_score": 0.5072856329680118,
|
| 1125 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1126 |
+
"recommended_rank": 256
|
| 1127 |
+
},
|
| 1128 |
+
{
|
| 1129 |
+
"training_priority_rank": 30,
|
| 1130 |
+
"module_name": "up_blocks.1.attentions.2.transformer_blocks.1.attn1.to_k",
|
| 1131 |
+
"ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.1.attn1.to_k.weight",
|
| 1132 |
+
"suffix": "attn1.to_k",
|
| 1133 |
+
"shape": [
|
| 1134 |
+
640,
|
| 1135 |
+
640
|
| 1136 |
+
],
|
| 1137 |
+
"has_old_rank128": false,
|
| 1138 |
+
"has_phase1b": false,
|
| 1139 |
+
"has_phase2": false,
|
| 1140 |
+
"composite_source": "raw",
|
| 1141 |
+
"w4_mse_mean": 0.011052693789159613,
|
| 1142 |
+
"w2_raw_mse_mean": 0.0802028871008328,
|
| 1143 |
+
"w2_old_rank128_mse_mean": 0.0802028871008328,
|
| 1144 |
+
"w2_phase1b_mse_mean": 0.0802028871008328,
|
| 1145 |
+
"w2_phase2_mse_mean": 0.0802028871008328,
|
| 1146 |
+
"w2_composite_mse_mean": 0.0802028871008328,
|
| 1147 |
+
"w2_raw_to_w4_mse_ratio": 7.25641084639825,
|
| 1148 |
+
"w2_composite_to_w4_mse_ratio": 7.25641084639825,
|
| 1149 |
+
"old_rank128_improvement_vs_raw": 0.9999999999875315,
|
| 1150 |
+
"phase1b_improvement_vs_raw": 0.9999999999875315,
|
| 1151 |
+
"phase2_improvement_vs_raw": 0.9999999999875315,
|
| 1152 |
+
"composite_improvement_vs_raw": 0.9999999999875315,
|
| 1153 |
+
"remaining_w2_specific_gap_score": 0.501782212777361,
|
| 1154 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1155 |
+
"recommended_rank": 256
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"training_priority_rank": 31,
|
| 1159 |
+
"module_name": "up_blocks.0.attentions.0.transformer_blocks.6.attn1.to_q",
|
| 1160 |
+
"ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.6.attn1.to_q.weight",
|
| 1161 |
+
"suffix": "attn1.to_q",
|
| 1162 |
+
"shape": [
|
| 1163 |
+
1280,
|
| 1164 |
+
1280
|
| 1165 |
+
],
|
| 1166 |
+
"has_old_rank128": false,
|
| 1167 |
+
"has_phase1b": false,
|
| 1168 |
+
"has_phase2": false,
|
| 1169 |
+
"composite_source": "raw",
|
| 1170 |
+
"w4_mse_mean": 0.010030013425940914,
|
| 1171 |
+
"w2_raw_mse_mean": 0.076130835339427,
|
| 1172 |
+
"w2_old_rank128_mse_mean": 0.076130835339427,
|
| 1173 |
+
"w2_phase1b_mse_mean": 0.076130835339427,
|
| 1174 |
+
"w2_phase2_mse_mean": 0.076130835339427,
|
| 1175 |
+
"w2_composite_mse_mean": 0.076130835339427,
|
| 1176 |
+
"w2_raw_to_w4_mse_ratio": 7.590302435182919,
|
| 1177 |
+
"w2_composite_to_w4_mse_ratio": 7.590302435182919,
|
| 1178 |
+
"old_rank128_improvement_vs_raw": 0.9999999999868646,
|
| 1179 |
+
"phase1b_improvement_vs_raw": 0.9999999999868646,
|
| 1180 |
+
"phase2_improvement_vs_raw": 0.9999999999868646,
|
| 1181 |
+
"composite_improvement_vs_raw": 0.9999999999868646,
|
| 1182 |
+
"remaining_w2_specific_gap_score": 0.5017252295375259,
|
| 1183 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1184 |
+
"recommended_rank": 256
|
| 1185 |
+
},
|
| 1186 |
+
{
|
| 1187 |
+
"training_priority_rank": 32,
|
| 1188 |
+
"module_name": "up_blocks.0.attentions.1.transformer_blocks.7.attn1.to_k",
|
| 1189 |
+
"ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.7.attn1.to_k.weight",
|
| 1190 |
+
"suffix": "attn1.to_k",
|
| 1191 |
+
"shape": [
|
| 1192 |
+
1280,
|
| 1193 |
+
1280
|
| 1194 |
+
],
|
| 1195 |
+
"has_old_rank128": false,
|
| 1196 |
+
"has_phase1b": false,
|
| 1197 |
+
"has_phase2": false,
|
| 1198 |
+
"composite_source": "raw",
|
| 1199 |
+
"w4_mse_mean": 0.012142672136958157,
|
| 1200 |
+
"w2_raw_mse_mean": 0.08435673732310534,
|
| 1201 |
+
"w2_old_rank128_mse_mean": 0.08435673732310534,
|
| 1202 |
+
"w2_phase1b_mse_mean": 0.08435673732310534,
|
| 1203 |
+
"w2_phase2_mse_mean": 0.08435673732310534,
|
| 1204 |
+
"w2_composite_mse_mean": 0.08435673732310534,
|
| 1205 |
+
"w2_raw_to_w4_mse_ratio": 6.9471312709996536,
|
| 1206 |
+
"w2_composite_to_w4_mse_ratio": 6.9471312709996536,
|
| 1207 |
+
"old_rank128_improvement_vs_raw": 0.9999999999881455,
|
| 1208 |
+
"phase1b_improvement_vs_raw": 0.9999999999881455,
|
| 1209 |
+
"phase2_improvement_vs_raw": 0.9999999999881455,
|
| 1210 |
+
"composite_improvement_vs_raw": 0.9999999999881455,
|
| 1211 |
+
"remaining_w2_specific_gap_score": 0.5016805904606905,
|
| 1212 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1213 |
+
"recommended_rank": 256
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"training_priority_rank": 33,
|
| 1217 |
+
"module_name": "up_blocks.0.attentions.0.transformer_blocks.4.attn1.to_k",
|
| 1218 |
+
"ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.4.attn1.to_k.weight",
|
| 1219 |
+
"suffix": "attn1.to_k",
|
| 1220 |
+
"shape": [
|
| 1221 |
+
1280,
|
| 1222 |
+
1280
|
| 1223 |
+
],
|
| 1224 |
+
"has_old_rank128": false,
|
| 1225 |
+
"has_phase1b": false,
|
| 1226 |
+
"has_phase2": false,
|
| 1227 |
+
"composite_source": "raw",
|
| 1228 |
+
"w4_mse_mean": 0.010702501915927445,
|
| 1229 |
+
"w2_raw_mse_mean": 0.07881842686661653,
|
| 1230 |
+
"w2_old_rank128_mse_mean": 0.07881842686661653,
|
| 1231 |
+
"w2_phase1b_mse_mean": 0.07881842686661653,
|
| 1232 |
+
"w2_phase2_mse_mean": 0.07881842686661653,
|
| 1233 |
+
"w2_composite_mse_mean": 0.07881842686661653,
|
| 1234 |
+
"w2_raw_to_w4_mse_ratio": 7.364486124684041,
|
| 1235 |
+
"w2_composite_to_w4_mse_ratio": 7.364486124684041,
|
| 1236 |
+
"old_rank128_improvement_vs_raw": 0.9999999999873126,
|
| 1237 |
+
"phase1b_improvement_vs_raw": 0.9999999999873126,
|
| 1238 |
+
"phase2_improvement_vs_raw": 0.9999999999873126,
|
| 1239 |
+
"composite_improvement_vs_raw": 0.9999999999873126,
|
| 1240 |
+
"remaining_w2_specific_gap_score": 0.5016387841693692,
|
| 1241 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1242 |
+
"recommended_rank": 256
|
| 1243 |
+
},
|
| 1244 |
+
{
|
| 1245 |
+
"training_priority_rank": 34,
|
| 1246 |
+
"module_name": "up_blocks.0.attentions.0.transformer_blocks.4.attn1.to_q",
|
| 1247 |
+
"ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.4.attn1.to_q.weight",
|
| 1248 |
+
"suffix": "attn1.to_q",
|
| 1249 |
+
"shape": [
|
| 1250 |
+
1280,
|
| 1251 |
+
1280
|
| 1252 |
+
],
|
| 1253 |
+
"has_old_rank128": false,
|
| 1254 |
+
"has_phase1b": false,
|
| 1255 |
+
"has_phase2": false,
|
| 1256 |
+
"composite_source": "raw",
|
| 1257 |
+
"w4_mse_mean": 0.010684565009017075,
|
| 1258 |
+
"w2_raw_mse_mean": 0.0787019906565547,
|
| 1259 |
+
"w2_old_rank128_mse_mean": 0.0787019906565547,
|
| 1260 |
+
"w2_phase1b_mse_mean": 0.0787019906565547,
|
| 1261 |
+
"w2_phase2_mse_mean": 0.0787019906565547,
|
| 1262 |
+
"w2_composite_mse_mean": 0.0787019906565547,
|
| 1263 |
+
"w2_raw_to_w4_mse_ratio": 7.365951780233393,
|
| 1264 |
+
"w2_composite_to_w4_mse_ratio": 7.365951780233393,
|
| 1265 |
+
"old_rank128_improvement_vs_raw": 0.9999999999872937,
|
| 1266 |
+
"phase1b_improvement_vs_raw": 0.9999999999872937,
|
| 1267 |
+
"phase2_improvement_vs_raw": 0.9999999999872937,
|
| 1268 |
+
"composite_improvement_vs_raw": 0.9999999999872937,
|
| 1269 |
+
"remaining_w2_specific_gap_score": 0.5010130775353723,
|
| 1270 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1271 |
+
"recommended_rank": 256
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"training_priority_rank": 35,
|
| 1275 |
+
"module_name": "up_blocks.0.attentions.0.transformer_blocks.9.attn2.to_q",
|
| 1276 |
+
"ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.9.attn2.to_q.weight",
|
| 1277 |
+
"suffix": "attn2.to_q",
|
| 1278 |
+
"shape": [
|
| 1279 |
+
1280,
|
| 1280 |
+
1280
|
| 1281 |
+
],
|
| 1282 |
+
"has_old_rank128": false,
|
| 1283 |
+
"has_phase1b": false,
|
| 1284 |
+
"has_phase2": false,
|
| 1285 |
+
"composite_source": "raw",
|
| 1286 |
+
"w4_mse_mean": 0.004660304186732641,
|
| 1287 |
+
"w2_raw_mse_mean": 0.050641766084092,
|
| 1288 |
+
"w2_old_rank128_mse_mean": 0.050641766084092,
|
| 1289 |
+
"w2_phase1b_mse_mean": 0.050641766084092,
|
| 1290 |
+
"w2_phase2_mse_mean": 0.050641766084092,
|
| 1291 |
+
"w2_composite_mse_mean": 0.050641766084092,
|
| 1292 |
+
"w2_raw_to_w4_mse_ratio": 10.866622444388234,
|
| 1293 |
+
"w2_composite_to_w4_mse_ratio": 10.866622444388234,
|
| 1294 |
+
"old_rank128_improvement_vs_raw": 0.9999999999802535,
|
| 1295 |
+
"phase1b_improvement_vs_raw": 0.9999999999802535,
|
| 1296 |
+
"phase2_improvement_vs_raw": 0.9999999999802535,
|
| 1297 |
+
"composite_improvement_vs_raw": 0.9999999999802535,
|
| 1298 |
+
"remaining_w2_specific_gap_score": 0.49966318587962766,
|
| 1299 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1300 |
+
"recommended_rank": 256
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"training_priority_rank": 36,
|
| 1304 |
+
"module_name": "mid_block.attentions.0.transformer_blocks.4.attn1.to_v",
|
| 1305 |
+
"ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.4.attn1.to_v.weight",
|
| 1306 |
+
"suffix": "attn1.to_v",
|
| 1307 |
+
"shape": [
|
| 1308 |
+
1280,
|
| 1309 |
+
1280
|
| 1310 |
+
],
|
| 1311 |
+
"has_old_rank128": false,
|
| 1312 |
+
"has_phase1b": false,
|
| 1313 |
+
"has_phase2": false,
|
| 1314 |
+
"composite_source": "raw",
|
| 1315 |
+
"w4_mse_mean": 0.010012863469975335,
|
| 1316 |
+
"w2_raw_mse_mean": 0.07585773031626429,
|
| 1317 |
+
"w2_old_rank128_mse_mean": 0.07585773031626429,
|
| 1318 |
+
"w2_phase1b_mse_mean": 0.07585773031626429,
|
| 1319 |
+
"w2_phase2_mse_mean": 0.07585773031626429,
|
| 1320 |
+
"w2_composite_mse_mean": 0.07585773031626429,
|
| 1321 |
+
"w2_raw_to_w4_mse_ratio": 7.576027630473136,
|
| 1322 |
+
"w2_composite_to_w4_mse_ratio": 7.576027630473136,
|
| 1323 |
+
"old_rank128_improvement_vs_raw": 0.9999999999868173,
|
| 1324 |
+
"phase1b_improvement_vs_raw": 0.9999999999868173,
|
| 1325 |
+
"phase2_improvement_vs_raw": 0.9999999999868173,
|
| 1326 |
+
"composite_improvement_vs_raw": 0.9999999999868173,
|
| 1327 |
+
"remaining_w2_specific_gap_score": 0.49884253055230965,
|
| 1328 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1329 |
+
"recommended_rank": 256
|
| 1330 |
+
},
|
| 1331 |
+
{
|
| 1332 |
+
"training_priority_rank": 37,
|
| 1333 |
+
"module_name": "up_blocks.0.attentions.0.transformer_blocks.0.attn1.to_k",
|
| 1334 |
+
"ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.0.attn1.to_k.weight",
|
| 1335 |
+
"suffix": "attn1.to_k",
|
| 1336 |
+
"shape": [
|
| 1337 |
+
1280,
|
| 1338 |
+
1280
|
| 1339 |
+
],
|
| 1340 |
+
"has_old_rank128": false,
|
| 1341 |
+
"has_phase1b": false,
|
| 1342 |
+
"has_phase2": false,
|
| 1343 |
+
"composite_source": "raw",
|
| 1344 |
+
"w4_mse_mean": 0.006486877957026341,
|
| 1345 |
+
"w2_raw_mse_mean": 0.060104410363627334,
|
| 1346 |
+
"w2_old_rank128_mse_mean": 0.060104410363627334,
|
| 1347 |
+
"w2_phase1b_mse_mean": 0.060104410363627334,
|
| 1348 |
+
"w2_phase2_mse_mean": 0.060104410363627334,
|
| 1349 |
+
"w2_composite_mse_mean": 0.060104410363627334,
|
| 1350 |
+
"w2_raw_to_w4_mse_ratio": 9.265537405287388,
|
| 1351 |
+
"w2_composite_to_w4_mse_ratio": 9.265537405287388,
|
| 1352 |
+
"old_rank128_improvement_vs_raw": 0.9999999999833623,
|
| 1353 |
+
"phase1b_improvement_vs_raw": 0.9999999999833623,
|
| 1354 |
+
"phase2_improvement_vs_raw": 0.9999999999833623,
|
| 1355 |
+
"composite_improvement_vs_raw": 0.9999999999833623,
|
| 1356 |
+
"remaining_w2_specific_gap_score": 0.4967952520925702,
|
| 1357 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1358 |
+
"recommended_rank": 256
|
| 1359 |
+
},
|
| 1360 |
+
{
|
| 1361 |
+
"training_priority_rank": 38,
|
| 1362 |
+
"module_name": "up_blocks.0.attentions.1.transformer_blocks.7.attn1.to_q",
|
| 1363 |
+
"ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.7.attn1.to_q.weight",
|
| 1364 |
+
"suffix": "attn1.to_q",
|
| 1365 |
+
"shape": [
|
| 1366 |
+
1280,
|
| 1367 |
+
1280
|
| 1368 |
+
],
|
| 1369 |
+
"has_old_rank128": false,
|
| 1370 |
+
"has_phase1b": false,
|
| 1371 |
+
"has_phase2": false,
|
| 1372 |
+
"composite_source": "raw",
|
| 1373 |
+
"w4_mse_mean": 0.012249513346302723,
|
| 1374 |
+
"w2_raw_mse_mean": 0.08413445058145694,
|
| 1375 |
+
"w2_old_rank128_mse_mean": 0.08413445058145694,
|
| 1376 |
+
"w2_phase1b_mse_mean": 0.08413445058145694,
|
| 1377 |
+
"w2_phase2_mse_mean": 0.08413445058145694,
|
| 1378 |
+
"w2_composite_mse_mean": 0.08413445058145694,
|
| 1379 |
+
"w2_raw_to_w4_mse_ratio": 6.868391273681325,
|
| 1380 |
+
"w2_composite_to_w4_mse_ratio": 6.868391273681325,
|
| 1381 |
+
"old_rank128_improvement_vs_raw": 0.9999999999881142,
|
| 1382 |
+
"phase1b_improvement_vs_raw": 0.9999999999881142,
|
| 1383 |
+
"phase2_improvement_vs_raw": 0.9999999999881142,
|
| 1384 |
+
"composite_improvement_vs_raw": 0.9999999999881142,
|
| 1385 |
+
"remaining_w2_specific_gap_score": 0.49373387561506293,
|
| 1386 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1387 |
+
"recommended_rank": 256
|
| 1388 |
+
},
|
| 1389 |
+
{
|
| 1390 |
+
"training_priority_rank": 39,
|
| 1391 |
+
"module_name": "mid_block.attentions.0.transformer_blocks.4.attn1.to_q",
|
| 1392 |
+
"ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.4.attn1.to_q.weight",
|
| 1393 |
+
"suffix": "attn1.to_q",
|
| 1394 |
+
"shape": [
|
| 1395 |
+
1280,
|
| 1396 |
+
1280
|
| 1397 |
+
],
|
| 1398 |
+
"has_old_rank128": false,
|
| 1399 |
+
"has_phase1b": false,
|
| 1400 |
+
"has_phase2": false,
|
| 1401 |
+
"composite_source": "raw",
|
| 1402 |
+
"w4_mse_mean": 0.008935813060296434,
|
| 1403 |
+
"w2_raw_mse_mean": 0.07091100793331861,
|
| 1404 |
+
"w2_old_rank128_mse_mean": 0.07091100793331861,
|
| 1405 |
+
"w2_phase1b_mse_mean": 0.07091100793331861,
|
| 1406 |
+
"w2_phase2_mse_mean": 0.07091100793331861,
|
| 1407 |
+
"w2_composite_mse_mean": 0.07091100793331861,
|
| 1408 |
+
"w2_raw_to_w4_mse_ratio": 7.935596620799343,
|
| 1409 |
+
"w2_composite_to_w4_mse_ratio": 7.935596620799343,
|
| 1410 |
+
"old_rank128_improvement_vs_raw": 0.9999999999858977,
|
| 1411 |
+
"phase1b_improvement_vs_raw": 0.9999999999858977,
|
| 1412 |
+
"phase2_improvement_vs_raw": 0.9999999999858977,
|
| 1413 |
+
"composite_improvement_vs_raw": 0.9999999999858977,
|
| 1414 |
+
"remaining_w2_specific_gap_score": 0.49181014700773557,
|
| 1415 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1416 |
+
"recommended_rank": 256
|
| 1417 |
+
},
|
| 1418 |
+
{
|
| 1419 |
+
"training_priority_rank": 40,
|
| 1420 |
+
"module_name": "up_blocks.0.attentions.0.transformer_blocks.1.attn1.to_k",
|
| 1421 |
+
"ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.1.attn1.to_k.weight",
|
| 1422 |
+
"suffix": "attn1.to_k",
|
| 1423 |
+
"shape": [
|
| 1424 |
+
1280,
|
| 1425 |
+
1280
|
| 1426 |
+
],
|
| 1427 |
+
"has_old_rank128": false,
|
| 1428 |
+
"has_phase1b": false,
|
| 1429 |
+
"has_phase2": false,
|
| 1430 |
+
"composite_source": "raw",
|
| 1431 |
+
"w4_mse_mean": 0.007506641941810293,
|
| 1432 |
+
"w2_raw_mse_mean": 0.06461270240002445,
|
| 1433 |
+
"w2_old_rank128_mse_mean": 0.06461270240002445,
|
| 1434 |
+
"w2_phase1b_mse_mean": 0.06461270240002445,
|
| 1435 |
+
"w2_phase2_mse_mean": 0.06461270240002445,
|
| 1436 |
+
"w2_composite_mse_mean": 0.06461270240002445,
|
| 1437 |
+
"w2_raw_to_w4_mse_ratio": 8.60740433502482,
|
| 1438 |
+
"w2_composite_to_w4_mse_ratio": 8.60740433502482,
|
| 1439 |
+
"old_rank128_improvement_vs_raw": 0.999999999984523,
|
| 1440 |
+
"phase1b_improvement_vs_raw": 0.999999999984523,
|
| 1441 |
+
"phase2_improvement_vs_raw": 0.999999999984523,
|
| 1442 |
+
"composite_improvement_vs_raw": 0.999999999984523,
|
| 1443 |
+
"remaining_w2_specific_gap_score": 0.491534952344222,
|
| 1444 |
+
"recommended_action": "add_adapter_missing_layer",
|
| 1445 |
+
"recommended_rank": 256
|
| 1446 |
+
}
|
| 1447 |
+
]
|
| 1448 |
+
}
|
adaptertestsfor2bitday2/orbitquant_W4_vs_W2_diagnostic_report.xlsx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6090b8819e0936338bfd36a5e24f51be39770f8a6dad831b66d359e13632cc32
|
| 3 |
+
size 708266
|
adaptertestsfor2bitday2/orbitquant_W4_vs_W2_next_training_plan.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
adaptertestsfor2bitday2/orbitquant_W4_vs_W2_top_training_targets.csv
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
adaptertestsfor2bitday2/w4_w2_compare_condition_cache_captions_only_random_latents_1024x1024_8.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0ee428ea02342c6d3d918e67bad07ad166327f4a4000a2fd9048562d7b79a781
|
| 3 |
+
size 3605469
|