diff --git "a/adaptertestsfor2bitday2/orbitquant_W4_vs_W2_next_training_plan.json" "b/adaptertestsfor2bitday2/orbitquant_W4_vs_W2_next_training_plan.json" new file mode 100644--- /dev/null +++ "b/adaptertestsfor2bitday2/orbitquant_W4_vs_W2_next_training_plan.json" @@ -0,0 +1,6066 @@ +{ + "plan_type": "next W2A4 recovery training based on W4A4 ceiling comparison", + "key_interpretation": { + "w4_mse_mean": "How close W4A4 gets to FP16 teacher at this layer.", + "w2_raw_mse_mean": "How bad W2A4 is before adapters.", + "w2_composite_mse_mean": "Best current W2A4 stack: phase2 if present, else phase1b, else old rank128, else raw.", + "w2_composite_to_w4_mse_ratio": "Remaining W2A4 gap relative to W4A4. High means W2 is still much worse than W4.", + "composite_improvement_vs_raw": "How much current adapters helped over raw W2. Near 1.0 means adapter method failed on that layer." + }, + "training_plan": [ + { + "training_priority_rank": 1, + "module_name": "up_blocks.1.attentions.0.transformer_blocks.1.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.1.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.016171995251040374, + "w2_raw_mse_mean": 0.125401715614966, + "w2_old_rank128_mse_mean": 0.125401715614966, + "w2_phase1b_mse_mean": 0.125401715614966, + "w2_phase2_mse_mean": 0.125401715614966, + "w2_composite_mse_mean": 0.125401715614966, + "w2_raw_to_w4_mse_ratio": 7.754251325243521, + "w2_composite_to_w4_mse_ratio": 7.754251325243521, + "old_rank128_improvement_vs_raw": 0.9999999999920256, + "phase1b_improvement_vs_raw": 0.9999999999920256, + "phase2_improvement_vs_raw": 0.9999999999920256, + "composite_improvement_vs_raw": 0.9999999999920256, + "remaining_w2_specific_gap_score": 0.8469947038879494, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 2, + "module_name": "up_blocks.1.attentions.1.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01798078600716378, + "w2_raw_mse_mean": 0.130427686364523, + "w2_old_rank128_mse_mean": 0.130427686364523, + "w2_phase1b_mse_mean": 0.130427686364523, + "w2_phase2_mse_mean": 0.130427686364523, + "w2_composite_mse_mean": 0.130427686364523, + "w2_raw_to_w4_mse_ratio": 7.253725521526435, + "w2_composite_to_w4_mse_ratio": 7.253725521526435, + "old_rank128_improvement_vs_raw": 0.9999999999923329, + "phase1b_improvement_vs_raw": 0.9999999999923329, + "phase2_improvement_vs_raw": 0.9999999999923329, + "composite_improvement_vs_raw": 0.9999999999923329, + "remaining_w2_specific_gap_score": 0.8156589509387167, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 3, + "module_name": "up_blocks.1.attentions.0.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01650474433388029, + "w2_raw_mse_mean": 0.12441772927663156, + "w2_old_rank128_mse_mean": 0.12441772927663156, + "w2_phase1b_mse_mean": 0.12441772927663156, + "w2_phase2_mse_mean": 0.12441772927663156, + "w2_composite_mse_mean": 0.12441772927663156, + "w2_raw_to_w4_mse_ratio": 7.5383009122833515, + "w2_composite_to_w4_mse_ratio": 7.5383009122833515, + "old_rank128_improvement_vs_raw": 0.9999999999919625, + "phase1b_improvement_vs_raw": 0.9999999999919625, + "phase2_improvement_vs_raw": 0.9999999999919625, + "composite_improvement_vs_raw": 0.9999999999919625, + "remaining_w2_specific_gap_score": 0.8134805528411615, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 4, + "module_name": "down_blocks.1.attentions.1.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.016551085664624616, + "w2_raw_mse_mean": 0.12307428023112672, + "w2_old_rank128_mse_mean": 0.12307428023112672, + "w2_phase1b_mse_mean": 0.12307428023112672, + "w2_phase2_mse_mean": 0.12307428023112672, + "w2_composite_mse_mean": 0.12307428023112672, + "w2_raw_to_w4_mse_ratio": 7.436024603917248, + "w2_composite_to_w4_mse_ratio": 7.436024603917248, + "old_rank128_improvement_vs_raw": 0.9999999999918748, + "phase1b_improvement_vs_raw": 0.9999999999918748, + "phase2_improvement_vs_raw": 0.9999999999918748, + "composite_improvement_vs_raw": 0.9999999999918748, + "remaining_w2_specific_gap_score": 0.7921090956843738, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 5, + "module_name": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.015716641160127307, + "w2_raw_mse_mean": 0.11642114937837635, + "w2_old_rank128_mse_mean": 0.11642114937837635, + "w2_phase1b_mse_mean": 0.11642114937837635, + "w2_phase2_mse_mean": 0.11642114937837635, + "w2_composite_mse_mean": 0.11642114937837635, + "w2_raw_to_w4_mse_ratio": 7.407508270044756, + "w2_composite_to_w4_mse_ratio": 7.407508270044756, + "old_rank128_improvement_vs_raw": 0.9999999999914104, + "phase1b_improvement_vs_raw": 0.9999999999914104, + "phase2_improvement_vs_raw": 0.9999999999914104, + "composite_improvement_vs_raw": 0.9999999999914104, + "remaining_w2_specific_gap_score": 0.7459694774574699, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 6, + "module_name": "up_blocks.1.attentions.1.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01650732191878238, + "w2_raw_mse_mean": 0.11842691459293876, + "w2_old_rank128_mse_mean": 0.11842691459293876, + "w2_phase1b_mse_mean": 0.11842691459293876, + "w2_phase2_mse_mean": 0.11842691459293876, + "w2_composite_mse_mean": 0.11842691459293876, + "w2_raw_to_w4_mse_ratio": 7.174205190183874, + "w2_composite_to_w4_mse_ratio": 7.174205190183874, + "old_rank128_improvement_vs_raw": 0.999999999991556, + "phase1b_improvement_vs_raw": 0.999999999991556, + "phase2_improvement_vs_raw": 0.999999999991556, + "composite_improvement_vs_raw": 0.999999999991556, + "remaining_w2_specific_gap_score": 0.7311920707443591, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 7, + "module_name": "mid_block.attentions.0.proj_out", + "ldm_key": "model.diffusion_model.middle_block.1.proj_out.weight", + "suffix": "proj_out", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": true, + "has_phase1b": true, + "has_phase2": true, + "composite_source": "phase2", + "w4_mse_mean": 0.18414856466863835, + "w2_raw_mse_mean": 1.3472472195114409, + "w2_old_rank128_mse_mean": 0.85862555035523, + "w2_phase1b_mse_mean": 0.4570868414427553, + "w2_phase2_mse_mean": 0.4683431349694729, + "w2_composite_mse_mean": 0.4683431349694729, + "w2_raw_to_w4_mse_ratio": 7.316088626204586, + "w2_composite_to_w4_mse_ratio": 2.543289630357305, + "old_rank128_improvement_vs_raw": 1.5690742244421791, + "phase1b_improvement_vs_raw": 2.9474644582986103, + "phase2_improvement_vs_raw": 2.8766242502868744, + "composite_improvement_vs_raw": 2.8766242502868744, + "remaining_w2_specific_gap_score": 0.7227891036499625, + "recommended_action": "train_block_level_or_rank256_plus_real_latents", + "recommended_rank": 256 + }, + { + "training_priority_rank": 8, + "module_name": "up_blocks.0.attentions.1.proj_out", + "ldm_key": "model.diffusion_model.output_blocks.1.1.proj_out.weight", + "suffix": "proj_out", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": true, + "has_phase1b": true, + "has_phase2": true, + "composite_source": "phase2", + "w4_mse_mean": 0.1777853944471904, + "w2_raw_mse_mean": 1.3340980570231165, + "w2_old_rank128_mse_mean": 0.7680485237921987, + "w2_phase1b_mse_mean": 0.6045318022370338, + "w2_phase2_mse_mean": 0.4564306310244969, + "w2_composite_mse_mean": 0.4564306310244969, + "w2_raw_to_w4_mse_ratio": 7.503980071950705, + "w2_composite_to_w4_mse_ratio": 2.5673123061720813, + "old_rank128_improvement_vs_raw": 1.7369970980926328, + "phase1b_improvement_vs_raw": 2.2068285772297824, + "phase2_improvement_vs_raw": 2.9228933518894173, + "composite_improvement_vs_raw": 2.9228933518894173, + "remaining_w2_specific_gap_score": 0.7153693449211498, + "recommended_action": "train_block_level_or_rank256_plus_real_latents", + "recommended_rank": 256 + }, + { + "training_priority_rank": 9, + "module_name": "up_blocks.1.attentions.0.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01555688923690468, + "w2_raw_mse_mean": 0.11352902118648801, + "w2_old_rank128_mse_mean": 0.11352902118648801, + "w2_phase1b_mse_mean": 0.11352902118648801, + "w2_phase2_mse_mean": 0.11352902118648801, + "w2_composite_mse_mean": 0.11352902118648801, + "w2_raw_to_w4_mse_ratio": 7.29766854094919, + "w2_composite_to_w4_mse_ratio": 7.29766854094919, + "old_rank128_improvement_vs_raw": 0.9999999999911916, + "phase1b_improvement_vs_raw": 0.9999999999911916, + "phase2_improvement_vs_raw": 0.9999999999911916, + "composite_improvement_vs_raw": 0.9999999999911916, + "remaining_w2_specific_gap_score": 0.7149681452181973, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 10, + "module_name": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.015393487393989094, + "w2_raw_mse_mean": 0.11155489552766085, + "w2_old_rank128_mse_mean": 0.11155489552766085, + "w2_phase1b_mse_mean": 0.11155489552766085, + "w2_phase2_mse_mean": 0.11155489552766085, + "w2_composite_mse_mean": 0.11155489552766085, + "w2_raw_to_w4_mse_ratio": 7.246889068423464, + "w2_composite_to_w4_mse_ratio": 7.246889068423464, + "old_rank128_improvement_vs_raw": 0.9999999999910357, + "phase1b_improvement_vs_raw": 0.9999999999910357, + "phase2_improvement_vs_raw": 0.9999999999910357, + "composite_improvement_vs_raw": 0.9999999999910357, + "remaining_w2_specific_gap_score": 0.6968710574081131, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 11, + "module_name": "down_blocks.1.attentions.1.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.015574711058954043, + "w2_raw_mse_mean": 0.11173469506736312, + "w2_old_rank128_mse_mean": 0.11173469506736312, + "w2_phase1b_mse_mean": 0.11173469506736312, + "w2_phase2_mse_mean": 0.11173469506736312, + "w2_composite_mse_mean": 0.11173469506736312, + "w2_raw_to_w4_mse_ratio": 7.174110302094608, + "w2_composite_to_w4_mse_ratio": 7.174110302094608, + "old_rank128_improvement_vs_raw": 0.9999999999910502, + "phase1b_improvement_vs_raw": 0.9999999999910502, + "phase2_improvement_vs_raw": 0.9999999999910502, + "composite_improvement_vs_raw": 0.9999999999910502, + "remaining_w2_specific_gap_score": 0.6898623319239804, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 12, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.9.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006283481646927872, + "w2_raw_mse_mean": 0.06878376293129154, + "w2_old_rank128_mse_mean": 0.06878376293129154, + "w2_phase1b_mse_mean": 0.06878376293129154, + "w2_phase2_mse_mean": 0.06878376293129154, + "w2_composite_mse_mean": 0.06878376293129154, + "w2_raw_to_w4_mse_ratio": 10.94675958096808, + "w2_composite_to_w4_mse_ratio": 10.94675958096808, + "old_rank128_improvement_vs_raw": 0.9999999999854616, + "phase1b_improvement_vs_raw": 0.9999999999854616, + "phase2_improvement_vs_raw": 0.9999999999854616, + "composite_improvement_vs_raw": 0.9999999999854616, + "remaining_w2_specific_gap_score": 0.6841755529628081, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 13, + "module_name": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.012902505862127458, + "w2_raw_mse_mean": 0.0985980766958424, + "w2_old_rank128_mse_mean": 0.0985980766958424, + "w2_phase1b_mse_mean": 0.0985980766958424, + "w2_phase2_mse_mean": 0.0985980766958424, + "w2_composite_mse_mean": 0.0985980766958424, + "w2_raw_to_w4_mse_ratio": 7.641777321536753, + "w2_composite_to_w4_mse_ratio": 7.641777321536753, + "old_rank128_improvement_vs_raw": 0.9999999999898578, + "phase1b_improvement_vs_raw": 0.9999999999898578, + "phase2_improvement_vs_raw": 0.9999999999898578, + "composite_improvement_vs_raw": 0.9999999999898578, + "remaining_w2_specific_gap_score": 0.6548664697532293, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 14, + "module_name": "up_blocks.1.attentions.1.transformer_blocks.1.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.1.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.014045619937990392, + "w2_raw_mse_mean": 0.10288021394184657, + "w2_old_rank128_mse_mean": 0.10288021394184657, + "w2_phase1b_mse_mean": 0.10288021394184657, + "w2_phase2_mse_mean": 0.10288021394184657, + "w2_composite_mse_mean": 0.10288021394184657, + "w2_raw_to_w4_mse_ratio": 7.3247186232238075, + "w2_composite_to_w4_mse_ratio": 7.3247186232238075, + "old_rank128_improvement_vs_raw": 0.9999999999902799, + "phase1b_improvement_vs_raw": 0.9999999999902799, + "phase2_improvement_vs_raw": 0.9999999999902799, + "composite_improvement_vs_raw": 0.9999999999902799, + "remaining_w2_specific_gap_score": 0.6506884050865713, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 15, + "module_name": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.015184641383322222, + "w2_raw_mse_mean": 0.10630956571549177, + "w2_old_rank128_mse_mean": 0.10630956571549177, + "w2_phase1b_mse_mean": 0.10630956571549177, + "w2_phase2_mse_mean": 0.10630956571549177, + "w2_composite_mse_mean": 0.10630956571549177, + "w2_raw_to_w4_mse_ratio": 7.0011245590069615, + "w2_composite_to_w4_mse_ratio": 7.0011245590069615, + "old_rank128_improvement_vs_raw": 0.9999999999905934, + "phase1b_improvement_vs_raw": 0.9999999999905934, + "phase2_improvement_vs_raw": 0.9999999999905934, + "composite_improvement_vs_raw": 0.9999999999905934, + "remaining_w2_specific_gap_score": 0.6379769456796033, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 16, + "module_name": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.015266332110123975, + "w2_raw_mse_mean": 0.10559878458401986, + "w2_old_rank128_mse_mean": 0.10559878458401986, + "w2_phase1b_mse_mean": 0.10559878458401986, + "w2_phase2_mse_mean": 0.10559878458401986, + "w2_composite_mse_mean": 0.10559878458401986, + "w2_raw_to_w4_mse_ratio": 6.917102537489944, + "w2_composite_to_w4_mse_ratio": 6.917102537489944, + "old_rank128_improvement_vs_raw": 0.9999999999905301, + "phase1b_improvement_vs_raw": 0.9999999999905301, + "phase2_improvement_vs_raw": 0.9999999999905301, + "composite_improvement_vs_raw": 0.9999999999905301, + "remaining_w2_specific_gap_score": 0.624838836224875, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 17, + "module_name": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0125871665908822, + "w2_raw_mse_mean": 0.09344679423208747, + "w2_old_rank128_mse_mean": 0.09344679423208747, + "w2_phase1b_mse_mean": 0.09344679423208747, + "w2_phase2_mse_mean": 0.09344679423208747, + "w2_composite_mse_mean": 0.09344679423208747, + "w2_raw_to_w4_mse_ratio": 7.423973739439805, + "w2_composite_to_w4_mse_ratio": 7.423973739439805, + "old_rank128_improvement_vs_raw": 0.9999999999892987, + "phase1b_improvement_vs_raw": 0.9999999999892987, + "phase2_improvement_vs_raw": 0.9999999999892987, + "composite_improvement_vs_raw": 0.9999999999892987, + "remaining_w2_specific_gap_score": 0.6002997521891889, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 18, + "module_name": "down_blocks.1.attentions.1.transformer_blocks.1.attn1.to_v", + "ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.1.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.011522587821153658, + "w2_raw_mse_mean": 0.08661618735641241, + "w2_old_rank128_mse_mean": 0.08661618735641241, + "w2_phase1b_mse_mean": 0.08661618735641241, + "w2_phase2_mse_mean": 0.08661618735641241, + "w2_composite_mse_mean": 0.08661618735641241, + "w2_raw_to_w4_mse_ratio": 7.517077647252264, + "w2_composite_to_w4_mse_ratio": 7.517077647252264, + "old_rank128_improvement_vs_raw": 0.9999999999884548, + "phase1b_improvement_vs_raw": 0.9999999999884548, + "phase2_improvement_vs_raw": 0.9999999999884548, + "composite_improvement_vs_raw": 0.9999999999884548, + "remaining_w2_specific_gap_score": 0.5644844185182065, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 19, + "module_name": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.011824695875735156, + "w2_raw_mse_mean": 0.08770120383373328, + "w2_old_rank128_mse_mean": 0.08770120383373328, + "w2_phase1b_mse_mean": 0.08770120383373328, + "w2_phase2_mse_mean": 0.08770120383373328, + "w2_composite_mse_mean": 0.08770120383373328, + "w2_raw_to_w4_mse_ratio": 7.416783040169649, + "w2_composite_to_w4_mse_ratio": 7.416783040169649, + "old_rank128_improvement_vs_raw": 0.9999999999885976, + "phase1b_improvement_vs_raw": 0.9999999999885976, + "phase2_improvement_vs_raw": 0.9999999999885976, + "composite_improvement_vs_raw": 0.9999999999885976, + "remaining_w2_specific_gap_score": 0.5627595973701779, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 20, + "module_name": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.013221082560318922, + "w2_raw_mse_mean": 0.09296748986733812, + "w2_old_rank128_mse_mean": 0.09296748986733812, + "w2_phase1b_mse_mean": 0.09296748986733812, + "w2_phase2_mse_mean": 0.09296748986733812, + "w2_composite_mse_mean": 0.09296748986733812, + "w2_raw_to_w4_mse_ratio": 7.031760783291242, + "w2_composite_to_w4_mse_ratio": 7.031760783291242, + "old_rank128_improvement_vs_raw": 0.9999999999892435, + "phase1b_improvement_vs_raw": 0.9999999999892435, + "phase2_improvement_vs_raw": 0.9999999999892435, + "composite_improvement_vs_raw": 0.9999999999892435, + "remaining_w2_specific_gap_score": 0.5607576595098677, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 21, + "module_name": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.011692194467676538, + "w2_raw_mse_mean": 0.08697623320456062, + "w2_old_rank128_mse_mean": 0.08697623320456062, + "w2_phase1b_mse_mean": 0.08697623320456062, + "w2_phase2_mse_mean": 0.08697623320456062, + "w2_composite_mse_mean": 0.08697623320456062, + "w2_raw_to_w4_mse_ratio": 7.438828821875183, + "w2_composite_to_w4_mse_ratio": 7.438828821875183, + "old_rank128_improvement_vs_raw": 0.9999999999885025, + "phase1b_improvement_vs_raw": 0.9999999999885025, + "phase2_improvement_vs_raw": 0.9999999999885025, + "composite_improvement_vs_raw": 0.9999999999885025, + "remaining_w2_specific_gap_score": 0.560025077183101, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 22, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.8.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.8.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01207385812553444, + "w2_raw_mse_mean": 0.08799946933452572, + "w2_old_rank128_mse_mean": 0.08799946933452572, + "w2_phase1b_mse_mean": 0.08799946933452572, + "w2_phase2_mse_mean": 0.08799946933452572, + "w2_composite_mse_mean": 0.08799946933452572, + "w2_raw_to_w4_mse_ratio": 7.288429962675419, + "w2_composite_to_w4_mse_ratio": 7.288429962675419, + "old_rank128_improvement_vs_raw": 0.9999999999886362, + "phase1b_improvement_vs_raw": 0.9999999999886362, + "phase2_improvement_vs_raw": 0.9999999999886362, + "composite_improvement_vs_raw": 0.9999999999886362, + "remaining_w2_specific_gap_score": 0.5533784996700567, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 23, + "module_name": "down_blocks.1.attentions.0.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.012419876642525196, + "w2_raw_mse_mean": 0.08907664686973606, + "w2_old_rank128_mse_mean": 0.08907664686973606, + "w2_phase1b_mse_mean": 0.08907664686973606, + "w2_phase2_mse_mean": 0.08907664686973606, + "w2_composite_mse_mean": 0.08907664686973606, + "w2_raw_to_w4_mse_ratio": 7.1721039931724295, + "w2_composite_to_w4_mse_ratio": 7.1721039931724295, + "old_rank128_improvement_vs_raw": 0.9999999999887736, + "phase1b_improvement_vs_raw": 0.9999999999887736, + "phase2_improvement_vs_raw": 0.9999999999887736, + "composite_improvement_vs_raw": 0.9999999999887736, + "remaining_w2_specific_gap_score": 0.5497903278502804, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 24, + "module_name": "up_blocks.1.attentions.2.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.012714513893505293, + "w2_raw_mse_mean": 0.0901942936969655, + "w2_old_rank128_mse_mean": 0.0901942936969655, + "w2_phase1b_mse_mean": 0.0901942936969655, + "w2_phase2_mse_mean": 0.0901942936969655, + "w2_composite_mse_mean": 0.0901942936969655, + "w2_raw_to_w4_mse_ratio": 7.09380590129709, + "w2_composite_to_w4_mse_ratio": 7.09380590129709, + "old_rank128_improvement_vs_raw": 0.9999999999889128, + "phase1b_improvement_vs_raw": 0.9999999999889128, + "phase2_improvement_vs_raw": 0.9999999999889128, + "composite_improvement_vs_raw": 0.9999999999889128, + "remaining_w2_specific_gap_score": 0.549626519200985, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 25, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.7.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.7.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.012551478842007262, + "w2_raw_mse_mean": 0.08876089937984943, + "w2_old_rank128_mse_mean": 0.08876089937984943, + "w2_phase1b_mse_mean": 0.08876089937984943, + "w2_phase2_mse_mean": 0.08876089937984943, + "w2_composite_mse_mean": 0.08876089937984943, + "w2_raw_to_w4_mse_ratio": 7.071748316677466, + "w2_composite_to_w4_mse_ratio": 7.071748316677466, + "old_rank128_improvement_vs_raw": 0.9999999999887337, + "phase1b_improvement_vs_raw": 0.9999999999887337, + "phase2_improvement_vs_raw": 0.9999999999887337, + "composite_improvement_vs_raw": 0.9999999999887337, + "remaining_w2_specific_gap_score": 0.5389338414034505, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 26, + "module_name": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.011307834963580328, + "w2_raw_mse_mean": 0.08250298537313938, + "w2_old_rank128_mse_mean": 0.08250298537313938, + "w2_phase1b_mse_mean": 0.08250298537313938, + "w2_phase2_mse_mean": 0.08250298537313938, + "w2_composite_mse_mean": 0.08250298537313938, + "w2_raw_to_w4_mse_ratio": 7.296090333080073, + "w2_composite_to_w4_mse_ratio": 7.296090333080073, + "old_rank128_improvement_vs_raw": 0.9999999999878791, + "phase1b_improvement_vs_raw": 0.9999999999878791, + "phase2_improvement_vs_raw": 0.9999999999878791, + "composite_improvement_vs_raw": 0.9999999999878791, + "remaining_w2_specific_gap_score": 0.5194462486653656, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 27, + "module_name": "down_blocks.1.attentions.0.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.012032134258853537, + "w2_raw_mse_mean": 0.08493821855102267, + "w2_old_rank128_mse_mean": 0.08493821855102267, + "w2_phase1b_mse_mean": 0.08493821855102267, + "w2_phase2_mse_mean": 0.08493821855102267, + "w2_composite_mse_mean": 0.08493821855102267, + "w2_raw_to_w4_mse_ratio": 7.059281148019419, + "w2_composite_to_w4_mse_ratio": 7.059281148019419, + "old_rank128_improvement_vs_raw": 0.9999999999882266, + "phase1b_improvement_vs_raw": 0.9999999999882266, + "phase2_improvement_vs_raw": 0.9999999999882266, + "composite_improvement_vs_raw": 0.9999999999882266, + "remaining_w2_specific_gap_score": 0.5146645464196242, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 28, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.8.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.8.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01177227936152901, + "w2_raw_mse_mean": 0.08393190681402173, + "w2_old_rank128_mse_mean": 0.08393190681402173, + "w2_phase1b_mse_mean": 0.08393190681402173, + "w2_phase2_mse_mean": 0.08393190681402173, + "w2_composite_mse_mean": 0.08393190681402173, + "w2_raw_to_w4_mse_ratio": 7.129622414600161, + "w2_composite_to_w4_mse_ratio": 7.129622414600161, + "old_rank128_improvement_vs_raw": 0.9999999999880855, + "phase1b_improvement_vs_raw": 0.9999999999880855, + "phase2_improvement_vs_raw": 0.9999999999880855, + "composite_improvement_vs_raw": 0.9999999999880855, + "remaining_w2_specific_gap_score": 0.5144708973144891, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 29, + "module_name": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.011715796842638935, + "w2_raw_mse_mean": 0.0831726472824812, + "w2_old_rank128_mse_mean": 0.0831726472824812, + "w2_phase1b_mse_mean": 0.0831726472824812, + "w2_phase2_mse_mean": 0.0831726472824812, + "w2_composite_mse_mean": 0.0831726472824812, + "w2_raw_to_w4_mse_ratio": 7.09918824920898, + "w2_composite_to_w4_mse_ratio": 7.09918824920898, + "old_rank128_improvement_vs_raw": 0.9999999999879767, + "phase1b_improvement_vs_raw": 0.9999999999879767, + "phase2_improvement_vs_raw": 0.9999999999879767, + "composite_improvement_vs_raw": 0.9999999999879767, + "remaining_w2_specific_gap_score": 0.5072856329680118, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 30, + "module_name": "up_blocks.1.attentions.2.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.011052693789159613, + "w2_raw_mse_mean": 0.0802028871008328, + "w2_old_rank128_mse_mean": 0.0802028871008328, + "w2_phase1b_mse_mean": 0.0802028871008328, + "w2_phase2_mse_mean": 0.0802028871008328, + "w2_composite_mse_mean": 0.0802028871008328, + "w2_raw_to_w4_mse_ratio": 7.25641084639825, + "w2_composite_to_w4_mse_ratio": 7.25641084639825, + "old_rank128_improvement_vs_raw": 0.9999999999875315, + "phase1b_improvement_vs_raw": 0.9999999999875315, + "phase2_improvement_vs_raw": 0.9999999999875315, + "composite_improvement_vs_raw": 0.9999999999875315, + "remaining_w2_specific_gap_score": 0.501782212777361, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 31, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.6.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.6.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010030013425940914, + "w2_raw_mse_mean": 0.076130835339427, + "w2_old_rank128_mse_mean": 0.076130835339427, + "w2_phase1b_mse_mean": 0.076130835339427, + "w2_phase2_mse_mean": 0.076130835339427, + "w2_composite_mse_mean": 0.076130835339427, + "w2_raw_to_w4_mse_ratio": 7.590302435182919, + "w2_composite_to_w4_mse_ratio": 7.590302435182919, + "old_rank128_improvement_vs_raw": 0.9999999999868646, + "phase1b_improvement_vs_raw": 0.9999999999868646, + "phase2_improvement_vs_raw": 0.9999999999868646, + "composite_improvement_vs_raw": 0.9999999999868646, + "remaining_w2_specific_gap_score": 0.5017252295375259, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 32, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.7.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.7.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.012142672136958157, + "w2_raw_mse_mean": 0.08435673732310534, + "w2_old_rank128_mse_mean": 0.08435673732310534, + "w2_phase1b_mse_mean": 0.08435673732310534, + "w2_phase2_mse_mean": 0.08435673732310534, + "w2_composite_mse_mean": 0.08435673732310534, + "w2_raw_to_w4_mse_ratio": 6.9471312709996536, + "w2_composite_to_w4_mse_ratio": 6.9471312709996536, + "old_rank128_improvement_vs_raw": 0.9999999999881455, + "phase1b_improvement_vs_raw": 0.9999999999881455, + "phase2_improvement_vs_raw": 0.9999999999881455, + "composite_improvement_vs_raw": 0.9999999999881455, + "remaining_w2_specific_gap_score": 0.5016805904606905, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 33, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.4.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.4.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010702501915927445, + "w2_raw_mse_mean": 0.07881842686661653, + "w2_old_rank128_mse_mean": 0.07881842686661653, + "w2_phase1b_mse_mean": 0.07881842686661653, + "w2_phase2_mse_mean": 0.07881842686661653, + "w2_composite_mse_mean": 0.07881842686661653, + "w2_raw_to_w4_mse_ratio": 7.364486124684041, + "w2_composite_to_w4_mse_ratio": 7.364486124684041, + "old_rank128_improvement_vs_raw": 0.9999999999873126, + "phase1b_improvement_vs_raw": 0.9999999999873126, + "phase2_improvement_vs_raw": 0.9999999999873126, + "composite_improvement_vs_raw": 0.9999999999873126, + "remaining_w2_specific_gap_score": 0.5016387841693692, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 34, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.4.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.4.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010684565009017075, + "w2_raw_mse_mean": 0.0787019906565547, + "w2_old_rank128_mse_mean": 0.0787019906565547, + "w2_phase1b_mse_mean": 0.0787019906565547, + "w2_phase2_mse_mean": 0.0787019906565547, + "w2_composite_mse_mean": 0.0787019906565547, + "w2_raw_to_w4_mse_ratio": 7.365951780233393, + "w2_composite_to_w4_mse_ratio": 7.365951780233393, + "old_rank128_improvement_vs_raw": 0.9999999999872937, + "phase1b_improvement_vs_raw": 0.9999999999872937, + "phase2_improvement_vs_raw": 0.9999999999872937, + "composite_improvement_vs_raw": 0.9999999999872937, + "remaining_w2_specific_gap_score": 0.5010130775353723, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 35, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.9.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.9.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.004660304186732641, + "w2_raw_mse_mean": 0.050641766084092, + "w2_old_rank128_mse_mean": 0.050641766084092, + "w2_phase1b_mse_mean": 0.050641766084092, + "w2_phase2_mse_mean": 0.050641766084092, + "w2_composite_mse_mean": 0.050641766084092, + "w2_raw_to_w4_mse_ratio": 10.866622444388234, + "w2_composite_to_w4_mse_ratio": 10.866622444388234, + "old_rank128_improvement_vs_raw": 0.9999999999802535, + "phase1b_improvement_vs_raw": 0.9999999999802535, + "phase2_improvement_vs_raw": 0.9999999999802535, + "composite_improvement_vs_raw": 0.9999999999802535, + "remaining_w2_specific_gap_score": 0.49966318587962766, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 36, + "module_name": "mid_block.attentions.0.transformer_blocks.4.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.4.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010012863469975335, + "w2_raw_mse_mean": 0.07585773031626429, + "w2_old_rank128_mse_mean": 0.07585773031626429, + "w2_phase1b_mse_mean": 0.07585773031626429, + "w2_phase2_mse_mean": 0.07585773031626429, + "w2_composite_mse_mean": 0.07585773031626429, + "w2_raw_to_w4_mse_ratio": 7.576027630473136, + "w2_composite_to_w4_mse_ratio": 7.576027630473136, + "old_rank128_improvement_vs_raw": 0.9999999999868173, + "phase1b_improvement_vs_raw": 0.9999999999868173, + "phase2_improvement_vs_raw": 0.9999999999868173, + "composite_improvement_vs_raw": 0.9999999999868173, + "remaining_w2_specific_gap_score": 0.49884253055230965, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 37, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006486877957026341, + "w2_raw_mse_mean": 0.060104410363627334, + "w2_old_rank128_mse_mean": 0.060104410363627334, + "w2_phase1b_mse_mean": 0.060104410363627334, + "w2_phase2_mse_mean": 0.060104410363627334, + "w2_composite_mse_mean": 0.060104410363627334, + "w2_raw_to_w4_mse_ratio": 9.265537405287388, + "w2_composite_to_w4_mse_ratio": 9.265537405287388, + "old_rank128_improvement_vs_raw": 0.9999999999833623, + "phase1b_improvement_vs_raw": 0.9999999999833623, + "phase2_improvement_vs_raw": 0.9999999999833623, + "composite_improvement_vs_raw": 0.9999999999833623, + "remaining_w2_specific_gap_score": 0.4967952520925702, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 38, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.7.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.7.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.012249513346302723, + "w2_raw_mse_mean": 0.08413445058145694, + "w2_old_rank128_mse_mean": 0.08413445058145694, + "w2_phase1b_mse_mean": 0.08413445058145694, + "w2_phase2_mse_mean": 0.08413445058145694, + "w2_composite_mse_mean": 0.08413445058145694, + "w2_raw_to_w4_mse_ratio": 6.868391273681325, + "w2_composite_to_w4_mse_ratio": 6.868391273681325, + "old_rank128_improvement_vs_raw": 0.9999999999881142, + "phase1b_improvement_vs_raw": 0.9999999999881142, + "phase2_improvement_vs_raw": 0.9999999999881142, + "composite_improvement_vs_raw": 0.9999999999881142, + "remaining_w2_specific_gap_score": 0.49373387561506293, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 39, + "module_name": "mid_block.attentions.0.transformer_blocks.4.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.4.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008935813060296434, + "w2_raw_mse_mean": 0.07091100793331861, + "w2_old_rank128_mse_mean": 0.07091100793331861, + "w2_phase1b_mse_mean": 0.07091100793331861, + "w2_phase2_mse_mean": 0.07091100793331861, + "w2_composite_mse_mean": 0.07091100793331861, + "w2_raw_to_w4_mse_ratio": 7.935596620799343, + "w2_composite_to_w4_mse_ratio": 7.935596620799343, + "old_rank128_improvement_vs_raw": 0.9999999999858977, + "phase1b_improvement_vs_raw": 0.9999999999858977, + "phase2_improvement_vs_raw": 0.9999999999858977, + "composite_improvement_vs_raw": 0.9999999999858977, + "remaining_w2_specific_gap_score": 0.49181014700773557, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 40, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007506641941810293, + "w2_raw_mse_mean": 0.06461270240002445, + "w2_old_rank128_mse_mean": 0.06461270240002445, + "w2_phase1b_mse_mean": 0.06461270240002445, + "w2_phase2_mse_mean": 0.06461270240002445, + "w2_composite_mse_mean": 0.06461270240002445, + "w2_raw_to_w4_mse_ratio": 8.60740433502482, + "w2_composite_to_w4_mse_ratio": 8.60740433502482, + "old_rank128_improvement_vs_raw": 0.999999999984523, + "phase1b_improvement_vs_raw": 0.999999999984523, + "phase2_improvement_vs_raw": 0.999999999984523, + "composite_improvement_vs_raw": 0.999999999984523, + "remaining_w2_specific_gap_score": 0.491534952344222, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 41, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.6.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.6.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.011722295345472438, + "w2_raw_mse_mean": 0.08174368473035949, + "w2_old_rank128_mse_mean": 0.08174368473035949, + "w2_phase1b_mse_mean": 0.08174368473035949, + "w2_phase2_mse_mean": 0.08174368473035949, + "w2_composite_mse_mean": 0.08174368473035949, + "w2_raw_to_w4_mse_ratio": 6.973351405529841, + "w2_composite_to_w4_mse_ratio": 6.973351405529841, + "old_rank128_improvement_vs_raw": 0.9999999999877666, + "phase1b_improvement_vs_raw": 0.9999999999877666, + "phase2_improvement_vs_raw": 0.9999999999877666, + "composite_improvement_vs_raw": 0.9999999999877666, + "remaining_w2_specific_gap_score": 0.48828375408425445, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 42, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.6.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.6.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010069620113686792, + "w2_raw_mse_mean": 0.07518453284033708, + "w2_old_rank128_mse_mean": 0.07518453284033708, + "w2_phase1b_mse_mean": 0.07518453284033708, + "w2_phase2_mse_mean": 0.07518453284033708, + "w2_composite_mse_mean": 0.07518453284033708, + "w2_raw_to_w4_mse_ratio": 7.466471622964064, + "w2_composite_to_w4_mse_ratio": 7.466471622964064, + "old_rank128_improvement_vs_raw": 0.9999999999866993, + "phase1b_improvement_vs_raw": 0.9999999999866993, + "phase2_improvement_vs_raw": 0.9999999999866993, + "composite_improvement_vs_raw": 0.9999999999866993, + "remaining_w2_specific_gap_score": 0.48617864810531586, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 43, + "module_name": "up_blocks.1.attentions.2.transformer_blocks.1.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.1.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009663990116678178, + "w2_raw_mse_mean": 0.07352370609130178, + "w2_old_rank128_mse_mean": 0.07352370609130178, + "w2_phase1b_mse_mean": 0.07352370609130178, + "w2_phase2_mse_mean": 0.07352370609130178, + "w2_composite_mse_mean": 0.07352370609130178, + "w2_raw_to_w4_mse_ratio": 7.608007168468236, + "w2_composite_to_w4_mse_ratio": 7.608007168468236, + "old_rank128_improvement_vs_raw": 0.9999999999863989, + "phase1b_improvement_vs_raw": 0.9999999999863989, + "phase2_improvement_vs_raw": 0.9999999999863989, + "composite_improvement_vs_raw": 0.9999999999863989, + "remaining_w2_specific_gap_score": 0.48584517691128193, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 44, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.7.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.7.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00983373707692538, + "w2_raw_mse_mean": 0.07414118839161736, + "w2_old_rank128_mse_mean": 0.07414118839161736, + "w2_phase1b_mse_mean": 0.07414118839161736, + "w2_phase2_mse_mean": 0.07414118839161736, + "w2_composite_mse_mean": 0.07414118839161736, + "w2_raw_to_w4_mse_ratio": 7.539472308858892, + "w2_composite_to_w4_mse_ratio": 7.539472308858892, + "old_rank128_improvement_vs_raw": 0.9999999999865121, + "phase1b_improvement_vs_raw": 0.9999999999865121, + "phase2_improvement_vs_raw": 0.9999999999865121, + "composite_improvement_vs_raw": 0.9999999999865121, + "remaining_w2_specific_gap_score": 0.4848442484404116, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 45, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.9.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.9.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0113746911686446, + "w2_raw_mse_mean": 0.07972693496516772, + "w2_old_rank128_mse_mean": 0.07972693496516772, + "w2_phase1b_mse_mean": 0.07972693496516772, + "w2_phase2_mse_mean": 0.07972693496516772, + "w2_composite_mse_mean": 0.07972693496516772, + "w2_raw_to_w4_mse_ratio": 7.009151613534205, + "w2_composite_to_w4_mse_ratio": 7.009151613534205, + "old_rank128_improvement_vs_raw": 0.9999999999874571, + "phase1b_improvement_vs_raw": 0.9999999999874571, + "phase2_improvement_vs_raw": 0.9999999999874571, + "composite_improvement_vs_raw": 0.9999999999874571, + "remaining_w2_specific_gap_score": 0.47909123989508334, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 46, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.3.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.3.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010654779566851045, + "w2_raw_mse_mean": 0.07692748667406184, + "w2_old_rank128_mse_mean": 0.07692748667406184, + "w2_phase1b_mse_mean": 0.07692748667406184, + "w2_phase2_mse_mean": 0.07692748667406184, + "w2_composite_mse_mean": 0.07692748667406184, + "w2_raw_to_w4_mse_ratio": 7.219997953423384, + "w2_composite_to_w4_mse_ratio": 7.219997953423384, + "old_rank128_improvement_vs_raw": 0.9999999999870006, + "phase1b_improvement_vs_raw": 0.9999999999870006, + "phase2_improvement_vs_raw": 0.9999999999870006, + "composite_improvement_vs_raw": 0.9999999999870006, + "remaining_w2_specific_gap_score": 0.4784888096818893, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 47, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.5.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.5.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010159684727633638, + "w2_raw_mse_mean": 0.07482210207464439, + "w2_old_rank128_mse_mean": 0.07482210207464439, + "w2_phase1b_mse_mean": 0.07482210207464439, + "w2_phase2_mse_mean": 0.07482210207464439, + "w2_composite_mse_mean": 0.07482210207464439, + "w2_raw_to_w4_mse_ratio": 7.364608654023372, + "w2_composite_to_w4_mse_ratio": 7.364608654023372, + "old_rank128_improvement_vs_raw": 0.9999999999866349, + "phase1b_improvement_vs_raw": 0.9999999999866349, + "phase2_improvement_vs_raw": 0.9999999999866349, + "composite_improvement_vs_raw": 0.9999999999866349, + "remaining_w2_specific_gap_score": 0.4762133983838664, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 48, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.3.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.3.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01047705065658582, + "w2_raw_mse_mean": 0.07595028488763741, + "w2_old_rank128_mse_mean": 0.07595028488763741, + "w2_phase1b_mse_mean": 0.07595028488763741, + "w2_phase2_mse_mean": 0.07595028488763741, + "w2_composite_mse_mean": 0.07595028488763741, + "w2_raw_to_w4_mse_ratio": 7.249204701768455, + "w2_composite_to_w4_mse_ratio": 7.249204701768455, + "old_rank128_improvement_vs_raw": 0.9999999999868334, + "phase1b_improvement_vs_raw": 0.9999999999868334, + "phase2_improvement_vs_raw": 0.9999999999868334, + "composite_improvement_vs_raw": 0.9999999999868334, + "remaining_w2_specific_gap_score": 0.4746288774277266, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 49, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.6.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.6.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.011311921789976103, + "w2_raw_mse_mean": 0.07889604688222919, + "w2_old_rank128_mse_mean": 0.07889604688222919, + "w2_phase1b_mse_mean": 0.07889604688222919, + "w2_phase2_mse_mean": 0.07889604688222919, + "w2_composite_mse_mean": 0.07889604688222919, + "w2_raw_to_w4_mse_ratio": 6.974592676654394, + "w2_composite_to_w4_mse_ratio": 6.974592676654394, + "old_rank128_improvement_vs_raw": 0.999999999987325, + "phase1b_improvement_vs_raw": 0.999999999987325, + "phase2_improvement_vs_raw": 0.999999999987325, + "composite_improvement_vs_raw": 0.999999999987325, + "remaining_w2_specific_gap_score": 0.47137174392652287, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 50, + "module_name": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010193767342051225, + "w2_raw_mse_mean": 0.07446694174515349, + "w2_old_rank128_mse_mean": 0.07446694174515349, + "w2_phase1b_mse_mean": 0.07446694174515349, + "w2_phase2_mse_mean": 0.07446694174515349, + "w2_composite_mse_mean": 0.07446694174515349, + "w2_raw_to_w4_mse_ratio": 7.305144333700659, + "w2_composite_to_w4_mse_ratio": 7.305144333700659, + "old_rank128_improvement_vs_raw": 0.9999999999865712, + "phase1b_improvement_vs_raw": 0.9999999999865712, + "phase2_improvement_vs_raw": 0.9999999999865712, + "composite_improvement_vs_raw": 0.9999999999865712, + "remaining_w2_specific_gap_score": 0.46952481579977673, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 51, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.5.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.5.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010397774782697005, + "w2_raw_mse_mean": 0.07525793596037797, + "w2_old_rank128_mse_mean": 0.07525793596037797, + "w2_phase1b_mse_mean": 0.07525793596037797, + "w2_phase2_mse_mean": 0.07525793596037797, + "w2_composite_mse_mean": 0.07525793596037797, + "w2_raw_to_w4_mse_ratio": 7.237888637324327, + "w2_composite_to_w4_mse_ratio": 7.237888637324327, + "old_rank128_improvement_vs_raw": 0.9999999999867123, + "phase1b_improvement_vs_raw": 0.9999999999867123, + "phase2_improvement_vs_raw": 0.9999999999867123, + "composite_improvement_vs_raw": 0.9999999999867123, + "remaining_w2_specific_gap_score": 0.4694506236029615, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 52, + "module_name": "mid_block.attentions.0.transformer_blocks.5.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.5.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008858994713851384, + "w2_raw_mse_mean": 0.06868407822081021, + "w2_old_rank128_mse_mean": 0.06868407822081021, + "w2_phase1b_mse_mean": 0.06868407822081021, + "w2_phase2_mse_mean": 0.06868407822081021, + "w2_composite_mse_mean": 0.06868407822081021, + "w2_raw_to_w4_mse_ratio": 7.753032983038914, + "w2_composite_to_w4_mse_ratio": 7.753032983038914, + "old_rank128_improvement_vs_raw": 0.9999999999854405, + "phase1b_improvement_vs_raw": 0.9999999999854405, + "phase2_improvement_vs_raw": 0.9999999999854405, + "composite_improvement_vs_raw": 0.9999999999854405, + "remaining_w2_specific_gap_score": 0.46382584564250917, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 53, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.8.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.8.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01147363827164684, + "w2_raw_mse_mean": 0.0788167794900281, + "w2_old_rank128_mse_mean": 0.0788167794900281, + "w2_phase1b_mse_mean": 0.0788167794900281, + "w2_phase2_mse_mean": 0.0788167794900281, + "w2_composite_mse_mean": 0.0788167794900281, + "w2_raw_to_w4_mse_ratio": 6.869379844223201, + "w2_composite_to_w4_mse_ratio": 6.869379844223201, + "old_rank128_improvement_vs_raw": 0.9999999999873123, + "phase1b_improvement_vs_raw": 0.9999999999873123, + "phase2_improvement_vs_raw": 0.9999999999873123, + "composite_improvement_vs_raw": 0.9999999999873123, + "remaining_w2_specific_gap_score": 0.46260561693222496, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 54, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.8.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.8.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009421276161447167, + "w2_raw_mse_mean": 0.07069274237645524, + "w2_old_rank128_mse_mean": 0.07069274237645524, + "w2_phase1b_mse_mean": 0.07069274237645524, + "w2_phase2_mse_mean": 0.07069274237645524, + "w2_composite_mse_mean": 0.07069274237645524, + "w2_raw_to_w4_mse_ratio": 7.503520877376857, + "w2_composite_to_w4_mse_ratio": 7.503520877376857, + "old_rank128_improvement_vs_raw": 0.9999999999858542, + "phase1b_improvement_vs_raw": 0.9999999999858542, + "phase2_improvement_vs_raw": 0.9999999999858542, + "composite_improvement_vs_raw": 0.9999999999858542, + "remaining_w2_specific_gap_score": 0.45975172593180386, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 55, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.9.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.9.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.011102498625405133, + "w2_raw_mse_mean": 0.07708972040563822, + "w2_old_rank128_mse_mean": 0.07708972040563822, + "w2_phase1b_mse_mean": 0.07708972040563822, + "w2_phase2_mse_mean": 0.07708972040563822, + "w2_composite_mse_mean": 0.07708972040563822, + "w2_raw_to_w4_mse_ratio": 6.943456873959463, + "w2_composite_to_w4_mse_ratio": 6.943456873959463, + "old_rank128_improvement_vs_raw": 0.999999999987028, + "phase1b_improvement_vs_raw": 0.999999999987028, + "phase2_improvement_vs_raw": 0.999999999987028, + "composite_improvement_vs_raw": 0.999999999987028, + "remaining_w2_specific_gap_score": 0.458179428663447, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 56, + "module_name": "mid_block.attentions.0.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00733477233110794, + "w2_raw_mse_mean": 0.06174906090434108, + "w2_old_rank128_mse_mean": 0.06174906090434108, + "w2_phase1b_mse_mean": 0.06174906090434108, + "w2_phase2_mse_mean": 0.06174906090434108, + "w2_composite_mse_mean": 0.06174906090434108, + "w2_raw_to_w4_mse_ratio": 8.418674514822332, + "w2_composite_to_w4_mse_ratio": 8.418674514822332, + "old_rank128_improvement_vs_raw": 0.9999999999838054, + "phase1b_improvement_vs_raw": 0.9999999999838054, + "phase2_improvement_vs_raw": 0.9999999999838054, + "composite_improvement_vs_raw": 0.9999999999838054, + "remaining_w2_specific_gap_score": 0.45809618445366584, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 57, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00640476293795343, + "w2_raw_mse_mean": 0.05743642417447908, + "w2_old_rank128_mse_mean": 0.05743642417447908, + "w2_phase1b_mse_mean": 0.05743642417447908, + "w2_phase2_mse_mean": 0.05743642417447908, + "w2_composite_mse_mean": 0.05743642417447908, + "w2_raw_to_w4_mse_ratio": 8.967767382170193, + "w2_composite_to_w4_mse_ratio": 8.967767382170193, + "old_rank128_improvement_vs_raw": 0.9999999999825895, + "phase1b_improvement_vs_raw": 0.9999999999825895, + "phase2_improvement_vs_raw": 0.9999999999825895, + "composite_improvement_vs_raw": 0.9999999999825895, + "remaining_w2_specific_gap_score": 0.4576400670948737, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 58, + "module_name": "mid_block.attentions.0.transformer_blocks.5.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.5.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009446067120214658, + "w2_raw_mse_mean": 0.070597412862948, + "w2_old_rank128_mse_mean": 0.070597412862948, + "w2_phase1b_mse_mean": 0.070597412862948, + "w2_phase2_mse_mean": 0.070597412862948, + "w2_composite_mse_mean": 0.070597412862948, + "w2_raw_to_w4_mse_ratio": 7.4737361017047235, + "w2_composite_to_w4_mse_ratio": 7.4737361017047235, + "old_rank128_improvement_vs_raw": 0.9999999999858351, + "phase1b_improvement_vs_raw": 0.9999999999858351, + "phase2_improvement_vs_raw": 0.9999999999858351, + "composite_improvement_vs_raw": 0.9999999999858351, + "remaining_w2_specific_gap_score": 0.45702902034529364, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 59, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.9.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.9.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010663017148285039, + "w2_raw_mse_mean": 0.07515409948038203, + "w2_old_rank128_mse_mean": 0.07515409948038203, + "w2_phase1b_mse_mean": 0.07515409948038203, + "w2_phase2_mse_mean": 0.07515409948038203, + "w2_composite_mse_mean": 0.07515409948038203, + "w2_raw_to_w4_mse_ratio": 7.048108281943554, + "w2_composite_to_w4_mse_ratio": 7.048108281943554, + "old_rank128_improvement_vs_raw": 0.999999999986694, + "phase1b_improvement_vs_raw": 0.999999999986694, + "phase2_improvement_vs_raw": 0.999999999986694, + "composite_improvement_vs_raw": 0.999999999986694, + "remaining_w2_specific_gap_score": 0.4545401314963564, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 60, + "module_name": "mid_block.attentions.0.transformer_blocks.4.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.4.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008841160368839545, + "w2_raw_mse_mean": 0.06781472851123128, + "w2_old_rank128_mse_mean": 0.06781472851123128, + "w2_phase1b_mse_mean": 0.06781472851123128, + "w2_phase2_mse_mean": 0.06781472851123128, + "w2_composite_mse_mean": 0.06781472851123128, + "w2_raw_to_w4_mse_ratio": 7.6703425426567655, + "w2_composite_to_w4_mse_ratio": 7.6703425426567655, + "old_rank128_improvement_vs_raw": 0.9999999999852539, + "phase1b_improvement_vs_raw": 0.9999999999852539, + "phase2_improvement_vs_raw": 0.9999999999852539, + "composite_improvement_vs_raw": 0.9999999999852539, + "remaining_w2_specific_gap_score": 0.4523474686148551, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 61, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.4.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.4.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010702945524826646, + "w2_raw_mse_mean": 0.0750300454507981, + "w2_old_rank128_mse_mean": 0.0750300454507981, + "w2_phase1b_mse_mean": 0.0750300454507981, + "w2_phase2_mse_mean": 0.0750300454507981, + "w2_composite_mse_mean": 0.0750300454507981, + "w2_raw_to_w4_mse_ratio": 7.010223986448172, + "w2_composite_to_w4_mse_ratio": 7.010223986448172, + "old_rank128_improvement_vs_raw": 0.9999999999866719, + "phase1b_improvement_vs_raw": 0.9999999999866719, + "phase2_improvement_vs_raw": 0.9999999999866719, + "composite_improvement_vs_raw": 0.9999999999866719, + "remaining_w2_specific_gap_score": 0.4509473788796935, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 62, + "module_name": "mid_block.attentions.0.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007322497865451234, + "w2_raw_mse_mean": 0.061190334148705006, + "w2_old_rank128_mse_mean": 0.061190334148705006, + "w2_phase1b_mse_mean": 0.061190334148705006, + "w2_phase2_mse_mean": 0.061190334148705006, + "w2_composite_mse_mean": 0.061190334148705006, + "w2_raw_to_w4_mse_ratio": 8.35648371152892, + "w2_composite_to_w4_mse_ratio": 8.35648371152892, + "old_rank128_improvement_vs_raw": 0.9999999999836575, + "phase1b_improvement_vs_raw": 0.9999999999836575, + "phase2_improvement_vs_raw": 0.9999999999836575, + "composite_improvement_vs_raw": 0.9999999999836575, + "remaining_w2_specific_gap_score": 0.4501456964763167, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 63, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.4.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009656090896377074, + "w2_raw_mse_mean": 0.07083165266418032, + "w2_old_rank128_mse_mean": 0.07083165266418032, + "w2_phase1b_mse_mean": 0.07083165266418032, + "w2_phase2_mse_mean": 0.07083165266418032, + "w2_composite_mse_mean": 0.07083165266418032, + "w2_raw_to_w4_mse_ratio": 7.335437644173443, + "w2_composite_to_w4_mse_ratio": 7.335437644173443, + "old_rank128_improvement_vs_raw": 0.999999999985882, + "phase1b_improvement_vs_raw": 0.999999999985882, + "phase2_improvement_vs_raw": 0.999999999985882, + "composite_improvement_vs_raw": 0.999999999985882, + "remaining_w2_specific_gap_score": 0.44874951869500157, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 64, + "module_name": "mid_block.attentions.0.transformer_blocks.5.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.5.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008727636853499072, + "w2_raw_mse_mean": 0.0668703146012766, + "w2_old_rank128_mse_mean": 0.0668703146012766, + "w2_phase1b_mse_mean": 0.0668703146012766, + "w2_phase2_mse_mean": 0.0668703146012766, + "w2_composite_mse_mean": 0.0668703146012766, + "w2_raw_to_w4_mse_ratio": 7.66190387112694, + "w2_composite_to_w4_mse_ratio": 7.66190387112694, + "old_rank128_improvement_vs_raw": 0.9999999999850456, + "phase1b_improvement_vs_raw": 0.9999999999850456, + "phase2_improvement_vs_raw": 0.9999999999850456, + "composite_improvement_vs_raw": 0.9999999999850456, + "remaining_w2_specific_gap_score": 0.4454836077133828, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 65, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.8.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009430031724540251, + "w2_raw_mse_mean": 0.06955591861956886, + "w2_old_rank128_mse_mean": 0.06955591861956886, + "w2_phase1b_mse_mean": 0.06955591861956886, + "w2_phase2_mse_mean": 0.06955591861956886, + "w2_composite_mse_mean": 0.06955591861956886, + "w2_raw_to_w4_mse_ratio": 7.376000489074067, + "w2_composite_to_w4_mse_ratio": 7.376000489074067, + "old_rank128_improvement_vs_raw": 0.999999999985623, + "phase1b_improvement_vs_raw": 0.999999999985623, + "phase2_improvement_vs_raw": 0.999999999985623, + "composite_improvement_vs_raw": 0.999999999985623, + "remaining_w2_specific_gap_score": 0.44348857114374307, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 66, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.7.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.7.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009708761470392346, + "w2_raw_mse_mean": 0.07058029176135149, + "w2_old_rank128_mse_mean": 0.07058029176135149, + "w2_phase1b_mse_mean": 0.07058029176135149, + "w2_phase2_mse_mean": 0.07058029176135149, + "w2_composite_mse_mean": 0.07058029176135149, + "w2_raw_to_w4_mse_ratio": 7.269752374628014, + "w2_composite_to_w4_mse_ratio": 7.269752374628014, + "old_rank128_improvement_vs_raw": 0.9999999999858317, + "phase1b_improvement_vs_raw": 0.9999999999858317, + "phase2_improvement_vs_raw": 0.9999999999858317, + "composite_improvement_vs_raw": 0.9999999999858317, + "remaining_w2_specific_gap_score": 0.4425209518799413, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 67, + "module_name": "mid_block.attentions.0.transformer_blocks.8.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.8.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010308127151802182, + "w2_raw_mse_mean": 0.07286726390676838, + "w2_old_rank128_mse_mean": 0.07286726390676838, + "w2_phase1b_mse_mean": 0.07286726390676838, + "w2_phase2_mse_mean": 0.07286726390676838, + "w2_composite_mse_mean": 0.07286726390676838, + "w2_raw_to_w4_mse_ratio": 7.068913957562116, + "w2_composite_to_w4_mse_ratio": 7.068913957562116, + "old_rank128_improvement_vs_raw": 0.9999999999862763, + "phase1b_improvement_vs_raw": 0.9999999999862763, + "phase2_improvement_vs_raw": 0.9999999999862763, + "composite_improvement_vs_raw": 0.9999999999862763, + "remaining_w2_specific_gap_score": 0.44222515498021775, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 68, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.9.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.9.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009393613278267108, + "w2_raw_mse_mean": 0.06922137384702053, + "w2_old_rank128_mse_mean": 0.06922137384702053, + "w2_phase1b_mse_mean": 0.06922137384702053, + "w2_phase2_mse_mean": 0.06922137384702053, + "w2_composite_mse_mean": 0.06922137384702053, + "w2_raw_to_w4_mse_ratio": 7.368982710817024, + "w2_composite_to_w4_mse_ratio": 7.368982710817024, + "old_rank128_improvement_vs_raw": 0.9999999999855536, + "phase1b_improvement_vs_raw": 0.9999999999855536, + "phase2_improvement_vs_raw": 0.9999999999855536, + "composite_improvement_vs_raw": 0.9999999999855536, + "remaining_w2_specific_gap_score": 0.4408697332580444, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 69, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.6.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.6.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01106060981484396, + "w2_raw_mse_mean": 0.07544799401823964, + "w2_old_rank128_mse_mean": 0.07544799401823964, + "w2_phase1b_mse_mean": 0.07544799401823964, + "w2_phase2_mse_mean": 0.07544799401823964, + "w2_composite_mse_mean": 0.07544799401823964, + "w2_raw_to_w4_mse_ratio": 6.821323170641357, + "w2_composite_to_w4_mse_ratio": 6.821323170641357, + "old_rank128_improvement_vs_raw": 0.9999999999867457, + "phase1b_improvement_vs_raw": 0.9999999999867457, + "phase2_improvement_vs_raw": 0.9999999999867457, + "composite_improvement_vs_raw": 0.9999999999867457, + "remaining_w2_specific_gap_score": 0.43920715576361025, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 70, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.7.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00981543879190992, + "w2_raw_mse_mean": 0.07071997697598167, + "w2_old_rank128_mse_mean": 0.07071997697598167, + "w2_phase1b_mse_mean": 0.07071997697598167, + "w2_phase2_mse_mean": 0.07071997697598167, + "w2_composite_mse_mean": 0.07071997697598167, + "w2_raw_to_w4_mse_ratio": 7.2049735593140785, + "w2_composite_to_w4_mse_ratio": 7.2049735593140785, + "old_rank128_improvement_vs_raw": 0.9999999999858596, + "phase1b_improvement_vs_raw": 0.9999999999858596, + "phase2_improvement_vs_raw": 0.9999999999858596, + "composite_improvement_vs_raw": 0.9999999999858596, + "remaining_w2_specific_gap_score": 0.43881558725847164, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 71, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.5.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.5.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010823137609155051, + "w2_raw_mse_mean": 0.07453868751014982, + "w2_old_rank128_mse_mean": 0.07453868751014982, + "w2_phase1b_mse_mean": 0.07453868751014982, + "w2_phase2_mse_mean": 0.07453868751014982, + "w2_composite_mse_mean": 0.07453868751014982, + "w2_raw_to_w4_mse_ratio": 6.88697586550246, + "w2_composite_to_w4_mse_ratio": 6.88697586550246, + "old_rank128_improvement_vs_raw": 0.9999999999865841, + "phase1b_improvement_vs_raw": 0.9999999999865841, + "phase2_improvement_vs_raw": 0.9999999999865841, + "composite_improvement_vs_raw": 0.9999999999865841, + "remaining_w2_specific_gap_score": 0.4388074544253686, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 72, + "module_name": "mid_block.attentions.0.transformer_blocks.2.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.2.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007376158194217298, + "w2_raw_mse_mean": 0.06063237999166761, + "w2_old_rank128_mse_mean": 0.06063237999166761, + "w2_phase1b_mse_mean": 0.06063237999166761, + "w2_phase2_mse_mean": 0.06063237999166761, + "w2_composite_mse_mean": 0.06063237999166761, + "w2_raw_to_w4_mse_ratio": 8.220048755323829, + "w2_composite_to_w4_mse_ratio": 8.220048755323829, + "old_rank128_improvement_vs_raw": 0.9999999999835072, + "phase1b_improvement_vs_raw": 0.9999999999835072, + "phase2_improvement_vs_raw": 0.9999999999835072, + "composite_improvement_vs_raw": 0.9999999999835072, + "remaining_w2_specific_gap_score": 0.4377687396993812, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 73, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.3.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.3.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01018157139021371, + "w2_raw_mse_mean": 0.07186773046851158, + "w2_old_rank128_mse_mean": 0.07186773046851158, + "w2_phase1b_mse_mean": 0.07186773046851158, + "w2_phase2_mse_mean": 0.07186773046851158, + "w2_composite_mse_mean": 0.07186773046851158, + "w2_raw_to_w4_mse_ratio": 7.058608902996109, + "w2_composite_to_w4_mse_ratio": 7.058608902996109, + "old_rank128_improvement_vs_raw": 0.9999999999860855, + "phase1b_improvement_vs_raw": 0.9999999999860855, + "phase2_improvement_vs_raw": 0.9999999999860855, + "composite_improvement_vs_raw": 0.9999999999860855, + "remaining_w2_specific_gap_score": 0.43541847166170766, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 74, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.2.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.2.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008740628371015191, + "w2_raw_mse_mean": 0.06610725108268005, + "w2_old_rank128_mse_mean": 0.06610725108268005, + "w2_phase1b_mse_mean": 0.06610725108268005, + "w2_phase2_mse_mean": 0.06610725108268005, + "w2_composite_mse_mean": 0.06610725108268005, + "w2_raw_to_w4_mse_ratio": 7.563214939366965, + "w2_composite_to_w4_mse_ratio": 7.563214939366965, + "old_rank128_improvement_vs_raw": 0.999999999984873, + "phase1b_improvement_vs_raw": 0.999999999984873, + "phase2_improvement_vs_raw": 0.999999999984873, + "composite_improvement_vs_raw": 0.999999999984873, + "remaining_w2_specific_gap_score": 0.43387609791389187, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 75, + "module_name": "mid_block.attentions.0.transformer_blocks.3.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.3.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008083509248016136, + "w2_raw_mse_mean": 0.06333210512197443, + "w2_old_rank128_mse_mean": 0.06333210512197443, + "w2_phase1b_mse_mean": 0.06333210512197443, + "w2_phase2_mse_mean": 0.06333210512197443, + "w2_composite_mse_mean": 0.06333210512197443, + "w2_raw_to_w4_mse_ratio": 7.83472909735122, + "w2_composite_to_w4_mse_ratio": 7.83472909735122, + "old_rank128_improvement_vs_raw": 0.9999999999842102, + "phase1b_improvement_vs_raw": 0.9999999999842102, + "phase2_improvement_vs_raw": 0.9999999999842102, + "composite_improvement_vs_raw": 0.9999999999842102, + "remaining_w2_specific_gap_score": 0.4328577816814996, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 76, + "module_name": "down_blocks.1.attentions.0.transformer_blocks.1.attn1.to_v", + "ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.1.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009237130039504595, + "w2_raw_mse_mean": 0.06785488275012799, + "w2_old_rank128_mse_mean": 0.06785488275012799, + "w2_phase1b_mse_mean": 0.06785488275012799, + "w2_phase2_mse_mean": 0.06785488275012799, + "w2_composite_mse_mean": 0.06785488275012799, + "w2_raw_to_w4_mse_ratio": 7.345883673022459, + "w2_composite_to_w4_mse_ratio": 7.345883673022459, + "old_rank128_improvement_vs_raw": 0.9999999999852626, + "phase1b_improvement_vs_raw": 0.9999999999852626, + "phase2_improvement_vs_raw": 0.9999999999852626, + "composite_improvement_vs_raw": 0.9999999999852626, + "remaining_w2_specific_gap_score": 0.43059919258623636, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 77, + "module_name": "down_blocks.2.attentions.0.proj_out", + "ldm_key": "model.diffusion_model.input_blocks.7.1.proj_out.weight", + "suffix": "proj_out", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": true, + "has_phase1b": true, + "has_phase2": true, + "composite_source": "phase2", + "w4_mse_mean": 0.0490127966207053, + "w2_raw_mse_mean": 0.33593689410814215, + "w2_old_rank128_mse_mean": 0.2647447625973395, + "w2_phase1b_mse_mean": 0.16342671867460012, + "w2_phase2_mse_mean": 0.17179679710950171, + "w2_composite_mse_mean": 0.17179679710950171, + "w2_raw_to_w4_mse_ratio": 6.854065004717006, + "w2_composite_to_w4_mse_ratio": 3.5051416966772626, + "old_rank128_improvement_vs_raw": 1.2689085548325376, + "phase1b_improvement_vs_raw": 2.0555812221560443, + "phase2_improvement_vs_raw": 1.955431648077022, + "composite_improvement_vs_raw": 1.955431648077022, + "remaining_w2_specific_gap_score": 0.4303753197981217, + "recommended_action": "train_block_level_or_rank256_plus_real_latents", + "recommended_rank": 128 + }, + { + "training_priority_rank": 78, + "module_name": "mid_block.attentions.0.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006331111179731254, + "w2_raw_mse_mean": 0.0554331235720643, + "w2_old_rank128_mse_mean": 0.0554331235720643, + "w2_phase1b_mse_mean": 0.0554331235720643, + "w2_phase2_mse_mean": 0.0554331235720643, + "w2_composite_mse_mean": 0.0554331235720643, + "w2_raw_to_w4_mse_ratio": 8.755670527596338, + "w2_composite_to_w4_mse_ratio": 8.755670527596338, + "old_rank128_improvement_vs_raw": 0.9999999999819603, + "phase1b_improvement_vs_raw": 0.9999999999819603, + "phase2_improvement_vs_raw": 0.9999999999819603, + "composite_improvement_vs_raw": 0.9999999999819603, + "remaining_w2_specific_gap_score": 0.42992104274922066, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 79, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.5.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.5.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010671445318231625, + "w2_raw_mse_mean": 0.07325183586882693, + "w2_old_rank128_mse_mean": 0.07325183586882693, + "w2_phase1b_mse_mean": 0.07325183586882693, + "w2_phase2_mse_mean": 0.07325183586882693, + "w2_composite_mse_mean": 0.07325183586882693, + "w2_raw_to_w4_mse_ratio": 6.864284422356134, + "w2_composite_to_w4_mse_ratio": 6.864284422356134, + "old_rank128_improvement_vs_raw": 0.9999999999863484, + "phase1b_improvement_vs_raw": 0.9999999999863484, + "phase2_improvement_vs_raw": 0.9999999999863484, + "composite_improvement_vs_raw": 0.9999999999863484, + "remaining_w2_specific_gap_score": 0.42956960000141436, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 80, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.8.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.8.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009484410103011345, + "w2_raw_mse_mean": 0.06858773077172893, + "w2_old_rank128_mse_mean": 0.06858773077172893, + "w2_phase1b_mse_mean": 0.06858773077172893, + "w2_phase2_mse_mean": 0.06858773077172893, + "w2_composite_mse_mean": 0.06858773077172893, + "w2_raw_to_w4_mse_ratio": 7.231628537732713, + "w2_composite_to_w4_mse_ratio": 7.231628537732713, + "old_rank128_improvement_vs_raw": 0.99999999998542, + "phase1b_improvement_vs_raw": 0.99999999998542, + "phase2_improvement_vs_raw": 0.99999999998542, + "composite_improvement_vs_raw": 0.99999999998542, + "remaining_w2_specific_gap_score": 0.42741326042266575, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 81, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.2.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.2.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008612354535476438, + "w2_raw_mse_mean": 0.06495540016996008, + "w2_old_rank128_mse_mean": 0.06495540016996008, + "w2_phase1b_mse_mean": 0.06495540016996008, + "w2_phase2_mse_mean": 0.06495540016996008, + "w2_composite_mse_mean": 0.06495540016996008, + "w2_raw_to_w4_mse_ratio": 7.542118696443633, + "w2_composite_to_w4_mse_ratio": 7.542118696443633, + "old_rank128_improvement_vs_raw": 0.9999999999846048, + "phase1b_improvement_vs_raw": 0.9999999999846048, + "phase2_improvement_vs_raw": 0.9999999999846048, + "composite_improvement_vs_raw": 0.9999999999846048, + "remaining_w2_specific_gap_score": 0.4249459378944159, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 82, + "module_name": "mid_block.attentions.0.transformer_blocks.3.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.3.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008931877507295991, + "w2_raw_mse_mean": 0.06621908170304128, + "w2_old_rank128_mse_mean": 0.06621908170304128, + "w2_phase1b_mse_mean": 0.06621908170304128, + "w2_phase2_mse_mean": 0.06621908170304128, + "w2_composite_mse_mean": 0.06621908170304128, + "w2_raw_to_w4_mse_ratio": 7.413791964963304, + "w2_composite_to_w4_mse_ratio": 7.413791964963304, + "old_rank128_improvement_vs_raw": 0.9999999999848985, + "phase1b_improvement_vs_raw": 0.9999999999848985, + "phase2_improvement_vs_raw": 0.9999999999848985, + "composite_improvement_vs_raw": 0.9999999999848985, + "remaining_w2_specific_gap_score": 0.4247154141616285, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 83, + "module_name": "mid_block.attentions.0.transformer_blocks.7.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.7.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008478846617176064, + "w2_raw_mse_mean": 0.06428828635918242, + "w2_old_rank128_mse_mean": 0.06428828635918242, + "w2_phase1b_mse_mean": 0.06428828635918242, + "w2_phase2_mse_mean": 0.06428828635918242, + "w2_composite_mse_mean": 0.06428828635918242, + "w2_raw_to_w4_mse_ratio": 7.582197114094259, + "w2_composite_to_w4_mse_ratio": 7.582197114094259, + "old_rank128_improvement_vs_raw": 0.999999999984445, + "phase1b_improvement_vs_raw": 0.999999999984445, + "phase2_improvement_vs_raw": 0.999999999984445, + "composite_improvement_vs_raw": 0.999999999984445, + "remaining_w2_specific_gap_score": 0.42315817295105806, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 84, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.4.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.4.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010669421587538506, + "w2_raw_mse_mean": 0.07271738576569728, + "w2_old_rank128_mse_mean": 0.07271738576569728, + "w2_phase1b_mse_mean": 0.07271738576569728, + "w2_phase2_mse_mean": 0.07271738576569728, + "w2_composite_mse_mean": 0.07271738576569728, + "w2_raw_to_w4_mse_ratio": 6.815494650976491, + "w2_composite_to_w4_mse_ratio": 6.815494650976491, + "old_rank128_improvement_vs_raw": 0.999999999986248, + "phase1b_improvement_vs_raw": 0.999999999986248, + "phase2_improvement_vs_raw": 0.999999999986248, + "composite_improvement_vs_raw": 0.999999999986248, + "remaining_w2_specific_gap_score": 0.42288756796022203, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 85, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.6.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.6.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009471761977433093, + "w2_raw_mse_mean": 0.06814888286005173, + "w2_old_rank128_mse_mean": 0.06814888286005173, + "w2_phase1b_mse_mean": 0.06814888286005173, + "w2_phase2_mse_mean": 0.06814888286005173, + "w2_composite_mse_mean": 0.06814888286005173, + "w2_raw_to_w4_mse_ratio": 7.194953063139109, + "w2_composite_to_w4_mse_ratio": 7.194953063139109, + "old_rank128_improvement_vs_raw": 0.9999999999853262, + "phase1b_improvement_vs_raw": 0.9999999999853262, + "phase2_improvement_vs_raw": 0.9999999999853262, + "composite_improvement_vs_raw": 0.9999999999853262, + "remaining_w2_specific_gap_score": 0.42217913063058077, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 86, + "module_name": "mid_block.attentions.0.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006334711622912437, + "w2_raw_mse_mean": 0.05482997605577111, + "w2_old_rank128_mse_mean": 0.05482997605577111, + "w2_phase1b_mse_mean": 0.05482997605577111, + "w2_phase2_mse_mean": 0.05482997605577111, + "w2_composite_mse_mean": 0.05482997605577111, + "w2_raw_to_w4_mse_ratio": 8.655480992820173, + "w2_composite_to_w4_mse_ratio": 8.655480992820173, + "old_rank128_improvement_vs_raw": 0.9999999999817618, + "phase1b_improvement_vs_raw": 0.9999999999817618, + "phase2_improvement_vs_raw": 0.9999999999817618, + "composite_improvement_vs_raw": 0.9999999999817618, + "remaining_w2_specific_gap_score": 0.41974983954039646, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 87, + "module_name": "mid_block.attentions.0.transformer_blocks.7.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.7.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009647387296094425, + "w2_raw_mse_mean": 0.06863855331071786, + "w2_old_rank128_mse_mean": 0.06863855331071786, + "w2_phase1b_mse_mean": 0.06863855331071786, + "w2_phase2_mse_mean": 0.06863855331071786, + "w2_composite_mse_mean": 0.06863855331071786, + "w2_raw_to_w4_mse_ratio": 7.114729739459121, + "w2_composite_to_w4_mse_ratio": 7.114729739459121, + "old_rank128_improvement_vs_raw": 0.9999999999854309, + "phase1b_improvement_vs_raw": 0.9999999999854309, + "phase2_improvement_vs_raw": 0.9999999999854309, + "composite_improvement_vs_raw": 0.9999999999854309, + "remaining_w2_specific_gap_score": 0.41970620320961155, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 88, + "module_name": "mid_block.attentions.0.transformer_blocks.3.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.3.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008072188722768001, + "w2_raw_mse_mean": 0.06234529135482652, + "w2_old_rank128_mse_mean": 0.06234529135482652, + "w2_phase1b_mse_mean": 0.06234529135482652, + "w2_phase2_mse_mean": 0.06234529135482652, + "w2_composite_mse_mean": 0.06234529135482652, + "w2_raw_to_w4_mse_ratio": 7.723468007041897, + "w2_composite_to_w4_mse_ratio": 7.723468007041897, + "old_rank128_improvement_vs_raw": 0.9999999999839603, + "phase1b_improvement_vs_raw": 0.9999999999839603, + "phase2_improvement_vs_raw": 0.9999999999839603, + "composite_improvement_vs_raw": 0.9999999999839603, + "remaining_w2_specific_gap_score": 0.4191765718216054, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 89, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.9.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008552260780041772, + "w2_raw_mse_mean": 0.0642980641153242, + "w2_old_rank128_mse_mean": 0.0642980641153242, + "w2_phase1b_mse_mean": 0.0642980641153242, + "w2_phase2_mse_mean": 0.0642980641153242, + "w2_composite_mse_mean": 0.0642980641153242, + "w2_raw_to_w4_mse_ratio": 7.518253449176498, + "w2_composite_to_w4_mse_ratio": 7.518253449176498, + "old_rank128_improvement_vs_raw": 0.9999999999844473, + "phase1b_improvement_vs_raw": 0.9999999999844473, + "phase2_improvement_vs_raw": 0.9999999999844473, + "composite_improvement_vs_raw": 0.9999999999844473, + "remaining_w2_specific_gap_score": 0.4191110782026018, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 90, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.5.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.5.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00906950087353055, + "w2_raw_mse_mean": 0.06626612433631505, + "w2_old_rank128_mse_mean": 0.06626612433631505, + "w2_phase1b_mse_mean": 0.06626612433631505, + "w2_phase2_mse_mean": 0.06626612433631505, + "w2_composite_mse_mean": 0.06626612433631505, + "w2_raw_to_w4_mse_ratio": 7.306479733896611, + "w2_composite_to_w4_mse_ratio": 7.306479733896611, + "old_rank128_improvement_vs_raw": 0.9999999999849093, + "phase1b_improvement_vs_raw": 0.9999999999849093, + "phase2_improvement_vs_raw": 0.9999999999849093, + "composite_improvement_vs_raw": 0.9999999999849093, + "remaining_w2_specific_gap_score": 0.41790597017815045, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 91, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.8.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009394964303023048, + "w2_raw_mse_mean": 0.06752665407423462, + "w2_old_rank128_mse_mean": 0.06752665407423462, + "w2_phase1b_mse_mean": 0.06752665407423462, + "w2_phase2_mse_mean": 0.06752665407423462, + "w2_composite_mse_mean": 0.06752665407423462, + "w2_raw_to_w4_mse_ratio": 7.187537055922481, + "w2_composite_to_w4_mse_ratio": 7.187537055922481, + "old_rank128_improvement_vs_raw": 0.999999999985191, + "phase1b_improvement_vs_raw": 0.999999999985191, + "phase2_improvement_vs_raw": 0.999999999985191, + "composite_improvement_vs_raw": 0.999999999985191, + "remaining_w2_specific_gap_score": 0.417823674353973, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 92, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.7.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009576536565353828, + "w2_raw_mse_mean": 0.06821094998823744, + "w2_old_rank128_mse_mean": 0.06821094998823744, + "w2_phase1b_mse_mean": 0.06821094998823744, + "w2_phase2_mse_mean": 0.06821094998823744, + "w2_composite_mse_mean": 0.06821094998823744, + "w2_raw_to_w4_mse_ratio": 7.122715975198128, + "w2_composite_to_w4_mse_ratio": 7.122715975198128, + "old_rank128_improvement_vs_raw": 0.9999999999853395, + "phase1b_improvement_vs_raw": 0.9999999999853395, + "phase2_improvement_vs_raw": 0.9999999999853395, + "composite_improvement_vs_raw": 0.9999999999853395, + "remaining_w2_specific_gap_score": 0.4176362731835447, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 93, + "module_name": "mid_block.attentions.0.transformer_blocks.8.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.8.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008950482122600079, + "w2_raw_mse_mean": 0.06546537298709154, + "w2_old_rank128_mse_mean": 0.06546537298709154, + "w2_phase1b_mse_mean": 0.06546537298709154, + "w2_phase2_mse_mean": 0.06546537298709154, + "w2_composite_mse_mean": 0.06546537298709154, + "w2_raw_to_w4_mse_ratio": 7.314172810253035, + "w2_composite_to_w4_mse_ratio": 7.314172810253035, + "old_rank128_improvement_vs_raw": 0.9999999999847247, + "phase1b_improvement_vs_raw": 0.9999999999847247, + "phase2_improvement_vs_raw": 0.9999999999847247, + "composite_improvement_vs_raw": 0.9999999999847247, + "remaining_w2_specific_gap_score": 0.4133596781354811, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 94, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.3.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.3.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009778586681932211, + "w2_raw_mse_mean": 0.06852757345352854, + "w2_old_rank128_mse_mean": 0.06852757345352854, + "w2_phase1b_mse_mean": 0.06852757345352854, + "w2_phase2_mse_mean": 0.06852757345352854, + "w2_composite_mse_mean": 0.06852757345352854, + "w2_raw_to_w4_mse_ratio": 7.007922072536134, + "w2_composite_to_w4_mse_ratio": 7.007922072536134, + "old_rank128_improvement_vs_raw": 0.9999999999854072, + "phase1b_improvement_vs_raw": 0.9999999999854072, + "phase2_improvement_vs_raw": 0.9999999999854072, + "composite_improvement_vs_raw": 0.9999999999854072, + "remaining_w2_specific_gap_score": 0.4117083211358033, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 95, + "module_name": "mid_block.attentions.0.transformer_blocks.2.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.2.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007539212853381676, + "w2_raw_mse_mean": 0.05957258951717189, + "w2_old_rank128_mse_mean": 0.05957258951717189, + "w2_phase1b_mse_mean": 0.05957258951717189, + "w2_phase2_mse_mean": 0.05957258951717189, + "w2_composite_mse_mean": 0.05957258951717189, + "w2_raw_to_w4_mse_ratio": 7.901698846789981, + "w2_composite_to_w4_mse_ratio": 7.901698846789981, + "old_rank128_improvement_vs_raw": 0.9999999999832138, + "phase1b_improvement_vs_raw": 0.9999999999832138, + "phase2_improvement_vs_raw": 0.9999999999832138, + "composite_improvement_vs_raw": 0.9999999999832138, + "remaining_w2_specific_gap_score": 0.41115207237885987, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 96, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007336104862458471, + "w2_raw_mse_mean": 0.058705355772482495, + "w2_old_rank128_mse_mean": 0.058705355772482495, + "w2_phase1b_mse_mean": 0.058705355772482495, + "w2_phase2_mse_mean": 0.058705355772482495, + "w2_composite_mse_mean": 0.058705355772482495, + "w2_raw_to_w4_mse_ratio": 8.002251448844058, + "w2_composite_to_w4_mse_ratio": 8.002251448844058, + "old_rank128_improvement_vs_raw": 0.9999999999829658, + "phase1b_improvement_vs_raw": 0.9999999999829658, + "phase2_improvement_vs_raw": 0.9999999999829658, + "composite_improvement_vs_raw": 0.9999999999829658, + "remaining_w2_specific_gap_score": 0.4110696625207737, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 97, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.9.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.9.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009438967057836376, + "w2_raw_mse_mean": 0.06710386921518616, + "w2_old_rank128_mse_mean": 0.06710386921518616, + "w2_phase1b_mse_mean": 0.06710386921518616, + "w2_phase2_mse_mean": 0.06710386921518616, + "w2_composite_mse_mean": 0.06710386921518616, + "w2_raw_to_w4_mse_ratio": 7.109238627161671, + "w2_composite_to_w4_mse_ratio": 7.109238627161671, + "old_rank128_improvement_vs_raw": 0.9999999999850977, + "phase1b_improvement_vs_raw": 0.9999999999850977, + "phase2_improvement_vs_raw": 0.9999999999850977, + "composite_improvement_vs_raw": 0.9999999999850977, + "remaining_w2_specific_gap_score": 0.4099535498485295, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 98, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.9.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.9.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009401705149295074, + "w2_raw_mse_mean": 0.06691283280295986, + "w2_old_rank128_mse_mean": 0.06691283280295986, + "w2_phase1b_mse_mean": 0.06691283280295986, + "w2_phase2_mse_mean": 0.06691283280295986, + "w2_composite_mse_mean": 0.06691283280295986, + "w2_raw_to_w4_mse_ratio": 7.1170954346361075, + "w2_composite_to_w4_mse_ratio": 7.1170954346361075, + "old_rank128_improvement_vs_raw": 0.9999999999850551, + "phase1b_improvement_vs_raw": 0.9999999999850551, + "phase2_improvement_vs_raw": 0.9999999999850551, + "composite_improvement_vs_raw": 0.9999999999850551, + "remaining_w2_specific_gap_score": 0.409312184064672, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 99, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.3.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.3.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00876774950717975, + "w2_raw_mse_mean": 0.06425148574635386, + "w2_old_rank128_mse_mean": 0.06425148574635386, + "w2_phase1b_mse_mean": 0.06425148574635386, + "w2_phase2_mse_mean": 0.06425148574635386, + "w2_composite_mse_mean": 0.06425148574635386, + "w2_raw_to_w4_mse_ratio": 7.328161654984707, + "w2_composite_to_w4_mse_ratio": 7.328161654984707, + "old_rank128_improvement_vs_raw": 0.9999999999844361, + "phase1b_improvement_vs_raw": 0.9999999999844361, + "phase2_improvement_vs_raw": 0.9999999999844361, + "composite_improvement_vs_raw": 0.9999999999844361, + "remaining_w2_specific_gap_score": 0.40659378838320115, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 100, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.4.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.4.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.010289091700022774, + "w2_raw_mse_mean": 0.06991122450147357, + "w2_old_rank128_mse_mean": 0.06991122450147357, + "w2_phase1b_mse_mean": 0.06991122450147357, + "w2_phase2_mse_mean": 0.06991122450147357, + "w2_composite_mse_mean": 0.06991122450147357, + "w2_raw_to_w4_mse_ratio": 6.7946934999640565, + "w2_composite_to_w4_mse_ratio": 6.7946934999640565, + "old_rank128_improvement_vs_raw": 0.9999999999856961, + "phase1b_improvement_vs_raw": 0.9999999999856961, + "phase2_improvement_vs_raw": 0.9999999999856961, + "composite_improvement_vs_raw": 0.9999999999856961, + "remaining_w2_specific_gap_score": 0.4051141182000114, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 101, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.5.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.5.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01074123103171587, + "w2_raw_mse_mean": 0.07145838692252125, + "w2_old_rank128_mse_mean": 0.07145838692252125, + "w2_phase1b_mse_mean": 0.07145838692252125, + "w2_phase2_mse_mean": 0.07145838692252125, + "w2_composite_mse_mean": 0.07145838692252125, + "w2_raw_to_w4_mse_ratio": 6.652718548262464, + "w2_composite_to_w4_mse_ratio": 6.652718548262464, + "old_rank128_improvement_vs_raw": 0.9999999999860057, + "phase1b_improvement_vs_raw": 0.9999999999860057, + "phase2_improvement_vs_raw": 0.9999999999860057, + "composite_improvement_vs_raw": 0.9999999999860057, + "remaining_w2_specific_gap_score": 0.4039341491925045, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 102, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.9.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008575785090215504, + "w2_raw_mse_mean": 0.06329776878867831, + "w2_old_rank128_mse_mean": 0.06329776878867831, + "w2_phase1b_mse_mean": 0.06329776878867831, + "w2_phase2_mse_mean": 0.06329776878867831, + "w2_composite_mse_mean": 0.06329776878867831, + "w2_raw_to_w4_mse_ratio": 7.380988226199437, + "w2_composite_to_w4_mse_ratio": 7.380988226199437, + "old_rank128_improvement_vs_raw": 0.9999999999842015, + "phase1b_improvement_vs_raw": 0.9999999999842015, + "phase2_improvement_vs_raw": 0.9999999999842015, + "composite_improvement_vs_raw": 0.9999999999842015, + "remaining_w2_specific_gap_score": 0.4039023173926315, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 103, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.2.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.2.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008927146588186068, + "w2_raw_mse_mean": 0.0646245514840952, + "w2_old_rank128_mse_mean": 0.0646245514840952, + "w2_phase1b_mse_mean": 0.0646245514840952, + "w2_phase2_mse_mean": 0.0646245514840952, + "w2_composite_mse_mean": 0.0646245514840952, + "w2_raw_to_w4_mse_ratio": 7.239104997153109, + "w2_composite_to_w4_mse_ratio": 7.239104997153109, + "old_rank128_improvement_vs_raw": 0.9999999999845259, + "phase1b_improvement_vs_raw": 0.9999999999845259, + "phase2_improvement_vs_raw": 0.9999999999845259, + "composite_improvement_vs_raw": 0.9999999999845259, + "remaining_w2_specific_gap_score": 0.40319936211043583, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 104, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.8.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.8.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00837516610044986, + "w2_raw_mse_mean": 0.062348778026976755, + "w2_old_rank128_mse_mean": 0.062348778026976755, + "w2_phase1b_mse_mean": 0.062348778026976755, + "w2_phase2_mse_mean": 0.062348778026976755, + "w2_composite_mse_mean": 0.062348778026976755, + "w2_raw_to_w4_mse_ratio": 7.444482565687061, + "w2_composite_to_w4_mse_ratio": 7.444482565687061, + "old_rank128_improvement_vs_raw": 0.9999999999839612, + "phase1b_improvement_vs_raw": 0.9999999999839612, + "phase2_improvement_vs_raw": 0.9999999999839612, + "composite_improvement_vs_raw": 0.9999999999839612, + "remaining_w2_specific_gap_score": 0.40180561299418865, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 105, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.9.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.9.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009161271420972688, + "w2_raw_mse_mean": 0.06527550385466643, + "w2_old_rank128_mse_mean": 0.06527550385466643, + "w2_phase1b_mse_mean": 0.06527550385466643, + "w2_phase2_mse_mean": 0.06527550385466643, + "w2_composite_mse_mean": 0.06527550385466643, + "w2_raw_to_w4_mse_ratio": 7.125157726263581, + "w2_composite_to_w4_mse_ratio": 7.125157726263581, + "old_rank128_improvement_vs_raw": 0.9999999999846803, + "phase1b_improvement_vs_raw": 0.9999999999846803, + "phase2_improvement_vs_raw": 0.9999999999846803, + "composite_improvement_vs_raw": 0.9999999999846803, + "remaining_w2_specific_gap_score": 0.3998227567782834, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 106, + "module_name": "mid_block.attentions.0.transformer_blocks.2.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.2.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007752585233122643, + "w2_raw_mse_mean": 0.05962593228157077, + "w2_old_rank128_mse_mean": 0.05962593228157077, + "w2_phase1b_mse_mean": 0.05962593228157077, + "w2_phase2_mse_mean": 0.05962593228157077, + "w2_composite_mse_mean": 0.05962593228157077, + "w2_raw_to_w4_mse_ratio": 7.69110309411756, + "w2_composite_to_w4_mse_ratio": 7.69110309411756, + "old_rank128_improvement_vs_raw": 0.9999999999832287, + "phase1b_improvement_vs_raw": 0.9999999999832287, + "phase2_improvement_vs_raw": 0.9999999999832287, + "composite_improvement_vs_raw": 0.9999999999832287, + "remaining_w2_specific_gap_score": 0.3989632599865534, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 107, + "module_name": "mid_block.attentions.0.transformer_blocks.9.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.9.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008580211732935692, + "w2_raw_mse_mean": 0.06277605265911136, + "w2_old_rank128_mse_mean": 0.06277605265911136, + "w2_phase1b_mse_mean": 0.06277605265911136, + "w2_phase2_mse_mean": 0.06277605265911136, + "w2_composite_mse_mean": 0.06277605265911136, + "w2_raw_to_w4_mse_ratio": 7.316375703273741, + "w2_composite_to_w4_mse_ratio": 7.316375703273741, + "old_rank128_improvement_vs_raw": 0.9999999999840703, + "phase1b_improvement_vs_raw": 0.9999999999840703, + "phase2_improvement_vs_raw": 0.9999999999840703, + "composite_improvement_vs_raw": 0.9999999999840703, + "remaining_w2_specific_gap_score": 0.39651713377076037, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 108, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.3.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.3.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00978369581779199, + "w2_raw_mse_mean": 0.06725552212446928, + "w2_old_rank128_mse_mean": 0.06725552212446928, + "w2_phase1b_mse_mean": 0.06725552212446928, + "w2_phase2_mse_mean": 0.06725552212446928, + "w2_composite_mse_mean": 0.06725552212446928, + "w2_raw_to_w4_mse_ratio": 6.87424500619577, + "w2_composite_to_w4_mse_ratio": 6.87424500619577, + "old_rank128_improvement_vs_raw": 0.9999999999851312, + "phase1b_improvement_vs_raw": 0.9999999999851312, + "phase2_improvement_vs_raw": 0.9999999999851312, + "composite_improvement_vs_raw": 0.9999999999851312, + "remaining_w2_specific_gap_score": 0.3950754149856271, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 109, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.3.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.3.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009806850825303368, + "w2_raw_mse_mean": 0.0672571255958506, + "w2_old_rank128_mse_mean": 0.0672571255958506, + "w2_phase1b_mse_mean": 0.0672571255958506, + "w2_phase2_mse_mean": 0.0672571255958506, + "w2_composite_mse_mean": 0.0672571255958506, + "w2_raw_to_w4_mse_ratio": 6.8581776950718405, + "w2_composite_to_w4_mse_ratio": 6.8581776950718405, + "old_rank128_improvement_vs_raw": 0.9999999999851316, + "phase1b_improvement_vs_raw": 0.9999999999851316, + "phase2_improvement_vs_raw": 0.9999999999851316, + "composite_improvement_vs_raw": 0.9999999999851316, + "remaining_w2_specific_gap_score": 0.3940041930071156, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 110, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.6.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.6.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0090646689391828, + "w2_raw_mse_mean": 0.06439808809331485, + "w2_old_rank128_mse_mean": 0.06439808809331485, + "w2_phase1b_mse_mean": 0.06439808809331485, + "w2_phase2_mse_mean": 0.06439808809331485, + "w2_composite_mse_mean": 0.06439808809331485, + "w2_raw_to_w4_mse_ratio": 7.1042956470086125, + "w2_composite_to_w4_mse_ratio": 7.1042956470086125, + "old_rank128_improvement_vs_raw": 0.9999999999844715, + "phase1b_improvement_vs_raw": 0.9999999999844715, + "phase2_improvement_vs_raw": 0.9999999999844715, + "composite_improvement_vs_raw": 0.9999999999844715, + "remaining_w2_specific_gap_score": 0.3931049688308033, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 111, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.2.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.2.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008240919302417231, + "w2_raw_mse_mean": 0.06116146434630666, + "w2_old_rank128_mse_mean": 0.06116146434630666, + "w2_phase1b_mse_mean": 0.06116146434630666, + "w2_phase2_mse_mean": 0.06116146434630666, + "w2_composite_mse_mean": 0.06116146434630666, + "w2_raw_to_w4_mse_ratio": 7.421679802269762, + "w2_composite_to_w4_mse_ratio": 7.421679802269762, + "old_rank128_improvement_vs_raw": 0.9999999999836499, + "phase1b_improvement_vs_raw": 0.9999999999836499, + "phase2_improvement_vs_raw": 0.9999999999836499, + "composite_improvement_vs_raw": 0.9999999999836499, + "remaining_w2_specific_gap_score": 0.39275934027734133, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 112, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.4.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.4.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.01052490490422185, + "w2_raw_mse_mean": 0.06971109685088907, + "w2_old_rank128_mse_mean": 0.06971109685088907, + "w2_phase1b_mse_mean": 0.06971109685088907, + "w2_phase2_mse_mean": 0.06971109685088907, + "w2_composite_mse_mean": 0.06971109685088907, + "w2_raw_to_w4_mse_ratio": 6.623441967281096, + "w2_composite_to_w4_mse_ratio": 6.623441967281096, + "old_rank128_improvement_vs_raw": 0.999999999985655, + "phase1b_improvement_vs_raw": 0.999999999985655, + "phase2_improvement_vs_raw": 0.999999999985655, + "composite_improvement_vs_raw": 0.999999999985655, + "remaining_w2_specific_gap_score": 0.39201630762311007, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 113, + "module_name": "mid_block.attentions.0.transformer_blocks.7.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.7.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008351019034827394, + "w2_raw_mse_mean": 0.06149342489827957, + "w2_old_rank128_mse_mean": 0.06149342489827957, + "w2_phase1b_mse_mean": 0.06149342489827957, + "w2_phase2_mse_mean": 0.06149342489827957, + "w2_composite_mse_mean": 0.06149342489827957, + "w2_raw_to_w4_mse_ratio": 7.363583370420012, + "w2_composite_to_w4_mse_ratio": 7.363583370420012, + "old_rank128_improvement_vs_raw": 0.9999999999837381, + "phase1b_improvement_vs_raw": 0.9999999999837381, + "phase2_improvement_vs_raw": 0.9999999999837381, + "composite_improvement_vs_raw": 0.9999999999837381, + "remaining_w2_specific_gap_score": 0.3913185360802274, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 114, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.8.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00517206741330613, + "w2_raw_mse_mean": 0.04740510654768774, + "w2_old_rank128_mse_mean": 0.04740510654768774, + "w2_phase1b_mse_mean": 0.04740510654768774, + "w2_phase2_mse_mean": 0.04740510654768774, + "w2_composite_mse_mean": 0.04740510654768774, + "w2_raw_to_w4_mse_ratio": 9.165601054727837, + "w2_composite_to_w4_mse_ratio": 9.165601054727837, + "old_rank128_improvement_vs_raw": 0.9999999999789052, + "phase1b_improvement_vs_raw": 0.9999999999789052, + "phase2_improvement_vs_raw": 0.9999999999789052, + "composite_improvement_vs_raw": 0.9999999999789052, + "remaining_w2_specific_gap_score": 0.3870911880344501, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 115, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.5.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.5.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00996589002066425, + "w2_raw_mse_mean": 0.06728293154654759, + "w2_old_rank128_mse_mean": 0.06728293154654759, + "w2_phase1b_mse_mean": 0.06728293154654759, + "w2_phase2_mse_mean": 0.06728293154654759, + "w2_composite_mse_mean": 0.06728293154654759, + "w2_raw_to_w4_mse_ratio": 6.751321899026104, + "w2_composite_to_w4_mse_ratio": 6.751321899026104, + "old_rank128_improvement_vs_raw": 0.9999999999851373, + "phase1b_improvement_vs_raw": 0.9999999999851373, + "phase2_improvement_vs_raw": 0.9999999999851373, + "composite_improvement_vs_raw": 0.9999999999851373, + "remaining_w2_specific_gap_score": 0.38696579764108474, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 116, + "module_name": "mid_block.attentions.0.transformer_blocks.6.attn1.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.6.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007750729389954358, + "w2_raw_mse_mean": 0.05873665180323379, + "w2_old_rank128_mse_mean": 0.05873665180323379, + "w2_phase1b_mse_mean": 0.05873665180323379, + "w2_phase2_mse_mean": 0.05873665180323379, + "w2_composite_mse_mean": 0.05873665180323379, + "w2_raw_to_w4_mse_ratio": 7.578209590413976, + "w2_composite_to_w4_mse_ratio": 7.578209590413976, + "old_rank128_improvement_vs_raw": 0.9999999999829748, + "phase1b_improvement_vs_raw": 0.9999999999829748, + "phase2_improvement_vs_raw": 0.9999999999829748, + "composite_improvement_vs_raw": 0.9999999999829748, + "remaining_w2_specific_gap_score": 0.38638200620841706, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 117, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.4.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009648218917261277, + "w2_raw_mse_mean": 0.06605836822252188, + "w2_old_rank128_mse_mean": 0.06605836822252188, + "w2_phase1b_mse_mean": 0.06605836822252188, + "w2_phase2_mse_mean": 0.06605836822252188, + "w2_composite_mse_mean": 0.06605836822252188, + "w2_raw_to_w4_mse_ratio": 6.846690439153756, + "w2_composite_to_w4_mse_ratio": 6.846690439153756, + "old_rank128_improvement_vs_raw": 0.9999999999848618, + "phase1b_improvement_vs_raw": 0.9999999999848618, + "phase2_improvement_vs_raw": 0.9999999999848618, + "composite_improvement_vs_raw": 0.9999999999848618, + "remaining_w2_specific_gap_score": 0.38622282991956364, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 118, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.3.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.3.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008834491708382432, + "w2_raw_mse_mean": 0.06294071009116513, + "w2_old_rank128_mse_mean": 0.06294071009116513, + "w2_phase1b_mse_mean": 0.06294071009116513, + "w2_phase2_mse_mean": 0.06294071009116513, + "w2_composite_mse_mean": 0.06294071009116513, + "w2_raw_to_w4_mse_ratio": 7.12442913091657, + "w2_composite_to_w4_mse_ratio": 7.12442913091657, + "old_rank128_improvement_vs_raw": 0.9999999999841119, + "phase1b_improvement_vs_raw": 0.9999999999841119, + "phase2_improvement_vs_raw": 0.9999999999841119, + "composite_improvement_vs_raw": 0.9999999999841119, + "remaining_w2_specific_gap_score": 0.38547591841003065, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 119, + "module_name": "mid_block.attentions.0.transformer_blocks.8.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.8.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008935831803163248, + "w2_raw_mse_mean": 0.06324338680133224, + "w2_old_rank128_mse_mean": 0.06324338680133224, + "w2_phase1b_mse_mean": 0.06324338680133224, + "w2_phase2_mse_mean": 0.06324338680133224, + "w2_composite_mse_mean": 0.06324338680133224, + "w2_raw_to_w4_mse_ratio": 7.077504163839211, + "w2_composite_to_w4_mse_ratio": 7.077504163839211, + "old_rank128_improvement_vs_raw": 0.999999999984188, + "phase1b_improvement_vs_raw": 0.999999999984188, + "phase2_improvement_vs_raw": 0.999999999984188, + "composite_improvement_vs_raw": 0.999999999984188, + "remaining_w2_specific_gap_score": 0.3843619466274679, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 120, + "module_name": "mid_block.attentions.0.transformer_blocks.1.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.1.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007110940658354333, + "w2_raw_mse_mean": 0.055922536711607664, + "w2_old_rank128_mse_mean": 0.055922536711607664, + "w2_phase1b_mse_mean": 0.055922536711607664, + "w2_phase2_mse_mean": 0.055922536711607664, + "w2_composite_mse_mean": 0.055922536711607664, + "w2_raw_to_w4_mse_ratio": 7.864295230482963, + "w2_composite_to_w4_mse_ratio": 7.864295230482963, + "old_rank128_improvement_vs_raw": 0.9999999999821182, + "phase1b_improvement_vs_raw": 0.9999999999821182, + "phase2_improvement_vs_raw": 0.9999999999821182, + "composite_improvement_vs_raw": 0.9999999999821182, + "remaining_w2_specific_gap_score": 0.3838688020338612, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 121, + "module_name": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007073104805645666, + "w2_raw_mse_mean": 0.05567965490211334, + "w2_old_rank128_mse_mean": 0.05567965490211334, + "w2_phase1b_mse_mean": 0.05567965490211334, + "w2_phase2_mse_mean": 0.05567965490211334, + "w2_composite_mse_mean": 0.05567965490211334, + "w2_raw_to_w4_mse_ratio": 7.8720245810296055, + "w2_composite_to_w4_mse_ratio": 7.8720245810296055, + "old_rank128_improvement_vs_raw": 0.9999999999820401, + "phase1b_improvement_vs_raw": 0.9999999999820401, + "phase2_improvement_vs_raw": 0.9999999999820401, + "composite_improvement_vs_raw": 0.9999999999820401, + "remaining_w2_specific_gap_score": 0.38263195715844045, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 122, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.3.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.3.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009941752151852208, + "w2_raw_mse_mean": 0.06679752289450594, + "w2_old_rank128_mse_mean": 0.06679752289450594, + "w2_phase1b_mse_mean": 0.06679752289450594, + "w2_phase2_mse_mean": 0.06679752289450594, + "w2_composite_mse_mean": 0.06679752289450594, + "w2_raw_to_w4_mse_ratio": 6.718888367715169, + "w2_composite_to_w4_mse_ratio": 6.718888367715169, + "old_rank128_improvement_vs_raw": 0.9999999999850293, + "phase1b_improvement_vs_raw": 0.9999999999850293, + "phase2_improvement_vs_raw": 0.9999999999850293, + "composite_improvement_vs_raw": 0.9999999999850293, + "remaining_w2_specific_gap_score": 0.3820075766802966, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 123, + "module_name": "mid_block.attentions.0.transformer_blocks.9.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.9.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008835031384868281, + "w2_raw_mse_mean": 0.06262099097615906, + "w2_old_rank128_mse_mean": 0.06262099097615906, + "w2_phase1b_mse_mean": 0.06262099097615906, + "w2_phase2_mse_mean": 0.06262099097615906, + "w2_composite_mse_mean": 0.06262099097615906, + "w2_raw_to_w4_mse_ratio": 7.087806284007314, + "w2_composite_to_w4_mse_ratio": 7.087806284007314, + "old_rank128_improvement_vs_raw": 0.9999999999840308, + "phase1b_improvement_vs_raw": 0.9999999999840308, + "phase2_improvement_vs_raw": 0.9999999999840308, + "composite_improvement_vs_raw": 0.9999999999840308, + "remaining_w2_specific_gap_score": 0.38122446238251423, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 124, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.2.attn2.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.2.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007653157667456461, + "w2_raw_mse_mean": 0.05793328995683363, + "w2_old_rank128_mse_mean": 0.05793328995683363, + "w2_phase1b_mse_mean": 0.05793328995683363, + "w2_phase2_mse_mean": 0.05793328995683363, + "w2_composite_mse_mean": 0.05793328995683363, + "w2_raw_to_w4_mse_ratio": 7.569854492298471, + "w2_composite_to_w4_mse_ratio": 7.569854492298471, + "old_rank128_improvement_vs_raw": 0.9999999999827388, + "phase1b_improvement_vs_raw": 0.9999999999827388, + "phase2_improvement_vs_raw": 0.9999999999827388, + "composite_improvement_vs_raw": 0.9999999999827388, + "remaining_w2_specific_gap_score": 0.3806132852841032, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 125, + "module_name": "mid_block.attentions.0.transformer_blocks.6.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.6.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008609382851448442, + "w2_raw_mse_mean": 0.061615151552749534, + "w2_old_rank128_mse_mean": 0.061615151552749534, + "w2_phase1b_mse_mean": 0.061615151552749534, + "w2_phase2_mse_mean": 0.061615151552749534, + "w2_composite_mse_mean": 0.061615151552749534, + "w2_raw_to_w4_mse_ratio": 7.1567442880329875, + "w2_composite_to_w4_mse_ratio": 7.1567442880329875, + "old_rank128_improvement_vs_raw": 0.9999999999837702, + "phase1b_improvement_vs_raw": 0.9999999999837702, + "phase2_improvement_vs_raw": 0.9999999999837702, + "composite_improvement_vs_raw": 0.9999999999837702, + "remaining_w2_specific_gap_score": 0.37934873238583433, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 126, + "module_name": "up_blocks.0.attentions.0.proj_out", + "ldm_key": "model.diffusion_model.output_blocks.0.1.proj_out.weight", + "suffix": "proj_out", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": true, + "has_phase1b": true, + "has_phase2": true, + "composite_source": "phase2", + "w4_mse_mean": 0.15619481448084116, + "w2_raw_mse_mean": 1.0925663400973593, + "w2_old_rank128_mse_mean": 0.5609709428889411, + "w2_phase1b_mse_mean": 0.3312290856348617, + "w2_phase2_mse_mean": 0.3327453477042062, + "w2_composite_mse_mean": 0.3327453477042062, + "w2_raw_to_w4_mse_ratio": 6.994895084845332, + "w2_composite_to_w4_mse_ratio": 2.130322628238663, + "old_rank128_improvement_vs_raw": 1.947634461187616, + "phase1b_improvement_vs_raw": 3.2985217406253913, + "phase2_improvement_vs_raw": 3.2834909567700765, + "composite_improvement_vs_raw": 3.2834909567700765, + "remaining_w2_specific_gap_score": 0.37610959595333643, + "recommended_action": "train_block_level_or_rank256_plus_real_latents", + "recommended_rank": 128 + }, + { + "training_priority_rank": 127, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.2.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.2.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008190823411236383, + "w2_raw_mse_mean": 0.059717762017888684, + "w2_old_rank128_mse_mean": 0.059717762017888684, + "w2_phase1b_mse_mean": 0.059717762017888684, + "w2_phase2_mse_mean": 0.059717762017888684, + "w2_composite_mse_mean": 0.059717762017888684, + "w2_raw_to_w4_mse_ratio": 7.290813024814513, + "w2_composite_to_w4_mse_ratio": 7.290813024814513, + "old_rank128_improvement_vs_raw": 0.9999999999832546, + "phase1b_improvement_vs_raw": 0.9999999999832546, + "phase2_improvement_vs_raw": 0.9999999999832546, + "composite_improvement_vs_raw": 0.9999999999832546, + "remaining_w2_specific_gap_score": 0.37567327512219834, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 128, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006979057922892805, + "w2_raw_mse_mean": 0.054687963732119114, + "w2_old_rank128_mse_mean": 0.054687963732119114, + "w2_phase1b_mse_mean": 0.054687963732119114, + "w2_phase2_mse_mean": 0.054687963732119114, + "w2_composite_mse_mean": 0.054687963732119114, + "w2_raw_to_w4_mse_ratio": 7.836009434008975, + "w2_composite_to_w4_mse_ratio": 7.836009434008975, + "old_rank128_improvement_vs_raw": 0.9999999999817145, + "phase1b_improvement_vs_raw": 0.9999999999817145, + "phase2_improvement_vs_raw": 0.9999999999817145, + "composite_improvement_vs_raw": 0.9999999999817145, + "remaining_w2_specific_gap_score": 0.37384743600734294, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 129, + "module_name": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_v", + "ldm_key": "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0078036936881419805, + "w2_raw_mse_mean": 0.05800568398886493, + "w2_old_rank128_mse_mean": 0.05800568398886493, + "w2_phase1b_mse_mean": 0.05800568398886493, + "w2_phase2_mse_mean": 0.05800568398886493, + "w2_composite_mse_mean": 0.05800568398886493, + "w2_raw_to_w4_mse_ratio": 7.4331062057924875, + "w2_composite_to_w4_mse_ratio": 7.4331062057924875, + "old_rank128_improvement_vs_raw": 0.9999999999827603, + "phase1b_improvement_vs_raw": 0.9999999999827603, + "phase2_improvement_vs_raw": 0.9999999999827603, + "composite_improvement_vs_raw": 0.9999999999827603, + "remaining_w2_specific_gap_score": 0.373156725647438, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 130, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.3.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006935483318687018, + "w2_raw_mse_mean": 0.054112316707947423, + "w2_old_rank128_mse_mean": 0.054112316707947423, + "w2_phase1b_mse_mean": 0.054112316707947423, + "w2_phase2_mse_mean": 0.054112316707947423, + "w2_composite_mse_mean": 0.054112316707947423, + "w2_raw_to_w4_mse_ratio": 7.802241633880735, + "w2_composite_to_w4_mse_ratio": 7.802241633880735, + "old_rank128_improvement_vs_raw": 0.9999999999815199, + "phase1b_improvement_vs_raw": 0.9999999999815199, + "phase2_improvement_vs_raw": 0.9999999999815199, + "composite_improvement_vs_raw": 0.9999999999815199, + "remaining_w2_specific_gap_score": 0.3680850536243423, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 131, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006908955077441143, + "w2_raw_mse_mean": 0.053971642827881236, + "w2_old_rank128_mse_mean": 0.053971642827881236, + "w2_phase1b_mse_mean": 0.053971642827881236, + "w2_phase2_mse_mean": 0.053971642827881236, + "w2_composite_mse_mean": 0.053971642827881236, + "w2_raw_to_w4_mse_ratio": 7.811838724541073, + "w2_composite_to_w4_mse_ratio": 7.811838724541073, + "old_rank128_improvement_vs_raw": 0.9999999999814718, + "phase1b_improvement_vs_raw": 0.9999999999814718, + "phase2_improvement_vs_raw": 0.9999999999814718, + "composite_improvement_vs_raw": 0.9999999999814718, + "remaining_w2_specific_gap_score": 0.36764612664987273, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 132, + "module_name": "down_blocks.2.attentions.1.proj_out", + "ldm_key": "model.diffusion_model.input_blocks.8.1.proj_out.weight", + "suffix": "proj_out", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": true, + "has_phase1b": true, + "has_phase2": true, + "composite_source": "phase2", + "w4_mse_mean": 0.03394595310757203, + "w2_raw_mse_mean": 0.24734551991735185, + "w2_old_rank128_mse_mean": 0.19901637533413513, + "w2_phase1b_mse_mean": 0.13690019665019854, + "w2_phase2_mse_mean": 0.12977141780512674, + "w2_composite_mse_mean": 0.12977141780512674, + "w2_raw_to_w4_mse_ratio": 7.286450880499572, + "w2_composite_to_w4_mse_ratio": 3.822883316608273, + "old_rank128_improvement_vs_raw": 1.2428400401767568, + "phase1b_improvement_vs_raw": 1.80675795921281, + "phase2_improvement_vs_raw": 1.9060092283716592, + "composite_improvement_vs_raw": 1.9060092283716592, + "remaining_w2_specific_gap_score": 0.3663295702985169, + "recommended_action": "train_block_level_or_rank256_plus_real_latents", + "recommended_rank": 128 + }, + { + "training_priority_rank": 133, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.6.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.6.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.009443284178684865, + "w2_raw_mse_mean": 0.06367529243496912, + "w2_old_rank128_mse_mean": 0.06367529243496912, + "w2_phase1b_mse_mean": 0.06367529243496912, + "w2_phase2_mse_mean": 0.06367529243496912, + "w2_composite_mse_mean": 0.06367529243496912, + "w2_raw_to_w4_mse_ratio": 6.742918165266318, + "w2_composite_to_w4_mse_ratio": 6.742918165266318, + "old_rank128_improvement_vs_raw": 0.9999999999842952, + "phase1b_improvement_vs_raw": 0.9999999999842952, + "phase2_improvement_vs_raw": 0.9999999999842952, + "composite_improvement_vs_raw": 0.9999999999842952, + "remaining_w2_specific_gap_score": 0.36568199361017206, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 134, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.7.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.7.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007920356400843178, + "w2_raw_mse_mean": 0.057760783338121006, + "w2_old_rank128_mse_mean": 0.057760783338121006, + "w2_phase1b_mse_mean": 0.057760783338121006, + "w2_phase2_mse_mean": 0.057760783338121006, + "w2_composite_mse_mean": 0.057760783338121006, + "w2_raw_to_w4_mse_ratio": 7.2927000260593395, + "w2_composite_to_w4_mse_ratio": 7.2927000260593395, + "old_rank128_improvement_vs_raw": 0.9999999999826873, + "phase1b_improvement_vs_raw": 0.9999999999826873, + "phase2_improvement_vs_raw": 0.9999999999826873, + "composite_improvement_vs_raw": 0.9999999999826873, + "remaining_w2_specific_gap_score": 0.36347128282429464, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 135, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.5.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.5.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00891767070528918, + "w2_raw_mse_mean": 0.061484518593975475, + "w2_old_rank128_mse_mean": 0.061484518593975475, + "w2_phase1b_mse_mean": 0.061484518593975475, + "w2_phase2_mse_mean": 0.061484518593975475, + "w2_composite_mse_mean": 0.061484518593975475, + "w2_raw_to_w4_mse_ratio": 6.894683670099364, + "w2_composite_to_w4_mse_ratio": 6.894683670099364, + "old_rank128_improvement_vs_raw": 0.9999999999837358, + "phase1b_improvement_vs_raw": 0.9999999999837358, + "phase2_improvement_vs_raw": 0.9999999999837358, + "composite_improvement_vs_raw": 0.9999999999837358, + "remaining_w2_specific_gap_score": 0.36243178772672263, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 136, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.2.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.2.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008697727894676583, + "w2_raw_mse_mean": 0.06046478423689093, + "w2_old_rank128_mse_mean": 0.06046478423689093, + "w2_phase1b_mse_mean": 0.06046478423689093, + "w2_phase2_mse_mean": 0.06046478423689093, + "w2_composite_mse_mean": 0.06046478423689093, + "w2_raw_to_w4_mse_ratio": 6.951790739159179, + "w2_composite_to_w4_mse_ratio": 6.951790739159179, + "old_rank128_improvement_vs_raw": 0.9999999999834615, + "phase1b_improvement_vs_raw": 0.9999999999834615, + "phase2_improvement_vs_raw": 0.9999999999834615, + "composite_improvement_vs_raw": 0.9999999999834615, + "remaining_w2_specific_gap_score": 0.3598737428733371, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 137, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.3.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.3.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008602311285877866, + "w2_raw_mse_mean": 0.059842638804444244, + "w2_old_rank128_mse_mean": 0.059842638804444244, + "w2_phase1b_mse_mean": 0.059842638804444244, + "w2_phase2_mse_mean": 0.059842638804444244, + "w2_composite_mse_mean": 0.059842638804444244, + "w2_raw_to_w4_mse_ratio": 6.956576762774136, + "w2_composite_to_w4_mse_ratio": 6.956576762774136, + "old_rank128_improvement_vs_raw": 0.9999999999832895, + "phase1b_improvement_vs_raw": 0.9999999999832895, + "phase2_improvement_vs_raw": 0.9999999999832895, + "composite_improvement_vs_raw": 0.9999999999832895, + "remaining_w2_specific_gap_score": 0.356457271732595, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 138, + "module_name": "mid_block.attentions.0.transformer_blocks.9.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.9.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008544387362365211, + "w2_raw_mse_mean": 0.059585620649158955, + "w2_old_rank128_mse_mean": 0.059585620649158955, + "w2_phase1b_mse_mean": 0.059585620649158955, + "w2_phase2_mse_mean": 0.059585620649158955, + "w2_composite_mse_mean": 0.059585620649158955, + "w2_raw_to_w4_mse_ratio": 6.973656286305251, + "w2_composite_to_w4_mse_ratio": 6.973656286305251, + "old_rank128_improvement_vs_raw": 0.9999999999832174, + "phase1b_improvement_vs_raw": 0.9999999999832174, + "phase2_improvement_vs_raw": 0.9999999999832174, + "composite_improvement_vs_raw": 0.9999999999832174, + "remaining_w2_specific_gap_score": 0.355944017371222, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 139, + "module_name": "mid_block.attentions.0.transformer_blocks.6.attn1.to_k", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.6.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0076459818436498085, + "w2_raw_mse_mean": 0.05607844629724111, + "w2_old_rank128_mse_mean": 0.05607844629724111, + "w2_phase1b_mse_mean": 0.05607844629724111, + "w2_phase2_mse_mean": 0.05607844629724111, + "w2_composite_mse_mean": 0.05607844629724111, + "w2_raw_to_w4_mse_ratio": 7.334368225904353, + "w2_composite_to_w4_mse_ratio": 7.334368225904353, + "old_rank128_improvement_vs_raw": 0.9999999999821678, + "phase1b_improvement_vs_raw": 0.9999999999821678, + "phase2_improvement_vs_raw": 0.9999999999821678, + "composite_improvement_vs_raw": 0.9999999999821678, + "remaining_w2_specific_gap_score": 0.35522152839066207, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 140, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.6.attn2.to_q", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.002672525650788365, + "w2_raw_mse_mean": 0.03208718428920422, + "w2_old_rank128_mse_mean": 0.03208718428920422, + "w2_phase1b_mse_mean": 0.03208718428920422, + "w2_phase2_mse_mean": 0.03208718428920422, + "w2_composite_mse_mean": 0.03208718428920422, + "w2_raw_to_w4_mse_ratio": 12.006314801031957, + "w2_composite_to_w4_mse_ratio": 12.006314801031957, + "old_rank128_improvement_vs_raw": 0.9999999999688349, + "phase1b_improvement_vs_raw": 0.9999999999688349, + "phase2_improvement_vs_raw": 0.9999999999688349, + "composite_improvement_vs_raw": 0.9999999999688349, + "remaining_w2_specific_gap_score": 0.35316165137771477, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 141, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.8.attn2.to_q", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.8.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.002627550408111087, + "w2_raw_mse_mean": 0.03158448007889092, + "w2_old_rank128_mse_mean": 0.03158448007889092, + "w2_phase1b_mse_mean": 0.03158448007889092, + "w2_phase2_mse_mean": 0.03158448007889092, + "w2_composite_mse_mean": 0.03158448007889092, + "w2_raw_to_w4_mse_ratio": 12.020503952796135, + "w2_composite_to_w4_mse_ratio": 12.020503952796135, + "old_rank128_improvement_vs_raw": 0.9999999999683389, + "phase1b_improvement_vs_raw": 0.9999999999683389, + "phase2_improvement_vs_raw": 0.9999999999683389, + "composite_improvement_vs_raw": 0.9999999999683389, + "remaining_w2_specific_gap_score": 0.3480768875684487, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 142, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.1.attn2.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.1.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006467956317854779, + "w2_raw_mse_mean": 0.050779854878783226, + "w2_old_rank128_mse_mean": 0.050779854878783226, + "w2_phase1b_mse_mean": 0.050779854878783226, + "w2_phase2_mse_mean": 0.050779854878783226, + "w2_composite_mse_mean": 0.050779854878783226, + "w2_raw_to_w4_mse_ratio": 7.850989149502224, + "w2_composite_to_w4_mse_ratio": 7.850989149502224, + "old_rank128_improvement_vs_raw": 0.9999999999803072, + "phase1b_improvement_vs_raw": 0.9999999999803072, + "phase2_improvement_vs_raw": 0.9999999999803072, + "composite_improvement_vs_raw": 0.9999999999803072, + "remaining_w2_specific_gap_score": 0.3478922347956924, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 143, + "module_name": "mid_block.attentions.0.transformer_blocks.0.attn1.to_v", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.005460693259790007, + "w2_raw_mse_mean": 0.04639915882476738, + "w2_old_rank128_mse_mean": 0.04639915882476738, + "w2_phase1b_mse_mean": 0.04639915882476738, + "w2_phase2_mse_mean": 0.04639915882476738, + "w2_composite_mse_mean": 0.04639915882476738, + "w2_raw_to_w4_mse_ratio": 8.496935573717748, + "w2_composite_to_w4_mse_ratio": 8.496935573717748, + "old_rank128_improvement_vs_raw": 0.9999999999784479, + "phase1b_improvement_vs_raw": 0.9999999999784479, + "phase2_improvement_vs_raw": 0.9999999999784479, + "composite_improvement_vs_raw": 0.9999999999784479, + "remaining_w2_specific_gap_score": 0.3478515043924753, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 144, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.8.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.8.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008245563068028008, + "w2_raw_mse_mean": 0.05773270549252629, + "w2_old_rank128_mse_mean": 0.05773270549252629, + "w2_phase1b_mse_mean": 0.05773270549252629, + "w2_phase2_mse_mean": 0.05773270549252629, + "w2_composite_mse_mean": 0.05773270549252629, + "w2_raw_to_w4_mse_ratio": 7.001669262513064, + "w2_composite_to_w4_mse_ratio": 7.001669262513064, + "old_rank128_improvement_vs_raw": 0.9999999999826789, + "phase1b_improvement_vs_raw": 0.9999999999826789, + "phase2_improvement_vs_raw": 0.9999999999826789, + "composite_improvement_vs_raw": 0.9999999999826789, + "remaining_w2_specific_gap_score": 0.34649260400321585, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 145, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.7.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.7.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00917213159014604, + "w2_raw_mse_mean": 0.06113677312220846, + "w2_old_rank128_mse_mean": 0.06113677312220846, + "w2_phase1b_mse_mean": 0.06113677312220846, + "w2_phase2_mse_mean": 0.06113677312220846, + "w2_composite_mse_mean": 0.06113677312220846, + "w2_raw_to_w4_mse_ratio": 6.6654923683415594, + "w2_composite_to_w4_mse_ratio": 6.6654923683415594, + "old_rank128_improvement_vs_raw": 0.9999999999836433, + "phase1b_improvement_vs_raw": 0.9999999999836433, + "phase2_improvement_vs_raw": 0.9999999999836433, + "composite_improvement_vs_raw": 0.9999999999836433, + "remaining_w2_specific_gap_score": 0.34636992155556684, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 146, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.1.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.1.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0063619872671552, + "w2_raw_mse_mean": 0.05018920870497823, + "w2_old_rank128_mse_mean": 0.05018920870497823, + "w2_phase1b_mse_mean": 0.05018920870497823, + "w2_phase2_mse_mean": 0.05018920870497823, + "w2_composite_mse_mean": 0.05018920870497823, + "w2_raw_to_w4_mse_ratio": 7.888920016580244, + "w2_composite_to_w4_mse_ratio": 7.888920016580244, + "old_rank128_improvement_vs_raw": 0.9999999999800754, + "phase1b_improvement_vs_raw": 0.9999999999800754, + "phase2_improvement_vs_raw": 0.9999999999800754, + "composite_improvement_vs_raw": 0.9999999999800754, + "remaining_w2_specific_gap_score": 0.3457494444719369, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 147, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.2.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.2.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0081203036243096, + "w2_raw_mse_mean": 0.05718486782695566, + "w2_old_rank128_mse_mean": 0.05718486782695566, + "w2_phase1b_mse_mean": 0.05718486782695566, + "w2_phase2_mse_mean": 0.05718486782695566, + "w2_composite_mse_mean": 0.05718486782695566, + "w2_raw_to_w4_mse_ratio": 7.042208083047559, + "w2_composite_to_w4_mse_ratio": 7.042208083047559, + "old_rank128_improvement_vs_raw": 0.9999999999825129, + "phase1b_improvement_vs_raw": 0.9999999999825129, + "phase2_improvement_vs_raw": 0.9999999999825129, + "composite_improvement_vs_raw": 0.9999999999825129, + "remaining_w2_specific_gap_score": 0.34552287061907994, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 148, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.2.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.2.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008130993835428464, + "w2_raw_mse_mean": 0.0572148109120982, + "w2_old_rank128_mse_mean": 0.0572148109120982, + "w2_phase1b_mse_mean": 0.0572148109120982, + "w2_phase2_mse_mean": 0.0572148109120982, + "w2_composite_mse_mean": 0.0572148109120982, + "w2_raw_to_w4_mse_ratio": 7.036631937385625, + "w2_composite_to_w4_mse_ratio": 7.036631937385625, + "old_rank128_improvement_vs_raw": 0.999999999982522, + "phase1b_improvement_vs_raw": 0.999999999982522, + "phase2_improvement_vs_raw": 0.999999999982522, + "composite_improvement_vs_raw": 0.999999999982522, + "remaining_w2_specific_gap_score": 0.3453847548504882, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 149, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.1.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.1.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007230637977564973, + "w2_raw_mse_mean": 0.05363220362258809, + "w2_old_rank128_mse_mean": 0.05363220362258809, + "w2_phase1b_mse_mean": 0.05363220362258809, + "w2_phase2_mse_mean": 0.05363220362258809, + "w2_composite_mse_mean": 0.05363220362258809, + "w2_raw_to_w4_mse_ratio": 7.417354289010081, + "w2_composite_to_w4_mse_ratio": 7.417354289010081, + "old_rank128_improvement_vs_raw": 0.9999999999813545, + "phase1b_improvement_vs_raw": 0.9999999999813545, + "phase2_improvement_vs_raw": 0.9999999999813545, + "composite_improvement_vs_raw": 0.9999999999813545, + "remaining_w2_specific_gap_score": 0.3441768519538951, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 150, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006847605391937707, + "w2_raw_mse_mean": 0.052002187352627516, + "w2_old_rank128_mse_mean": 0.052002187352627516, + "w2_phase1b_mse_mean": 0.052002187352627516, + "w2_phase2_mse_mean": 0.052002187352627516, + "w2_composite_mse_mean": 0.052002187352627516, + "w2_raw_to_w4_mse_ratio": 7.594214965462245, + "w2_composite_to_w4_mse_ratio": 7.594214965462245, + "old_rank128_improvement_vs_raw": 0.99999999998077, + "phase1b_improvement_vs_raw": 0.99999999998077, + "phase2_improvement_vs_raw": 0.99999999998077, + "composite_improvement_vs_raw": 0.99999999998077, + "remaining_w2_specific_gap_score": 0.3429136020850621, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 151, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.3.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006766570490851466, + "w2_raw_mse_mean": 0.051655350519078116, + "w2_old_rank128_mse_mean": 0.051655350519078116, + "w2_phase1b_mse_mean": 0.051655350519078116, + "w2_phase2_mse_mean": 0.051655350519078116, + "w2_composite_mse_mean": 0.051655350519078116, + "w2_raw_to_w4_mse_ratio": 7.633904144098291, + "w2_composite_to_w4_mse_ratio": 7.633904144098291, + "old_rank128_improvement_vs_raw": 0.9999999999806409, + "phase1b_improvement_vs_raw": 0.9999999999806409, + "phase2_improvement_vs_raw": 0.9999999999806409, + "composite_improvement_vs_raw": 0.9999999999806409, + "remaining_w2_specific_gap_score": 0.34267664388099606, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 152, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.4.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006068530267969306, + "w2_raw_mse_mean": 0.04859945510647127, + "w2_old_rank128_mse_mean": 0.04859945510647127, + "w2_phase1b_mse_mean": 0.04859945510647127, + "w2_phase2_mse_mean": 0.04859945510647127, + "w2_composite_mse_mean": 0.04859945510647127, + "w2_raw_to_w4_mse_ratio": 8.008439103447945, + "w2_composite_to_w4_mse_ratio": 8.008439103447945, + "old_rank128_improvement_vs_raw": 0.9999999999794237, + "phase1b_improvement_vs_raw": 0.9999999999794237, + "phase2_improvement_vs_raw": 0.9999999999794237, + "composite_improvement_vs_raw": 0.9999999999794237, + "remaining_w2_specific_gap_score": 0.34060632158246457, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 153, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.9.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.9.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0039035734620743562, + "w2_raw_mse_mean": 0.03838608613503831, + "w2_old_rank128_mse_mean": 0.03838608613503831, + "w2_phase1b_mse_mean": 0.03838608613503831, + "w2_phase2_mse_mean": 0.03838608613503831, + "w2_composite_mse_mean": 0.03838608613503831, + "w2_raw_to_w4_mse_ratio": 9.833575952431644, + "w2_composite_to_w4_mse_ratio": 9.833575952431644, + "old_rank128_improvement_vs_raw": 0.999999999973949, + "phase1b_improvement_vs_raw": 0.999999999973949, + "phase2_improvement_vs_raw": 0.999999999973949, + "composite_improvement_vs_raw": 0.999999999973949, + "remaining_w2_specific_gap_score": 0.33908640740027773, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 154, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.5.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.005924319391072329, + "w2_raw_mse_mean": 0.047873335890471935, + "w2_old_rank128_mse_mean": 0.047873335890471935, + "w2_phase1b_mse_mean": 0.047873335890471935, + "w2_phase2_mse_mean": 0.047873335890471935, + "w2_composite_mse_mean": 0.047873335890471935, + "w2_raw_to_w4_mse_ratio": 8.080816161690064, + "w2_composite_to_w4_mse_ratio": 8.080816161690064, + "old_rank128_improvement_vs_raw": 0.9999999999791116, + "phase1b_improvement_vs_raw": 0.9999999999791116, + "phase2_improvement_vs_raw": 0.9999999999791116, + "composite_improvement_vs_raw": 0.9999999999791116, + "remaining_w2_specific_gap_score": 0.3389822904953515, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 155, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.2.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.2.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008133498660754412, + "w2_raw_mse_mean": 0.05651983479037881, + "w2_old_rank128_mse_mean": 0.05651983479037881, + "w2_phase1b_mse_mean": 0.05651983479037881, + "w2_phase2_mse_mean": 0.05651983479037881, + "w2_composite_mse_mean": 0.05651983479037881, + "w2_raw_to_w4_mse_ratio": 6.949018760665457, + "w2_composite_to_w4_mse_ratio": 6.949018760665457, + "old_rank128_improvement_vs_raw": 0.9999999999823072, + "phase1b_improvement_vs_raw": 0.9999999999823072, + "phase2_improvement_vs_raw": 0.9999999999823072, + "composite_improvement_vs_raw": 0.9999999999823072, + "remaining_w2_specific_gap_score": 0.33623755752462475, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 156, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.3.attn2.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.3.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007257051350149725, + "w2_raw_mse_mean": 0.05314673676288554, + "w2_old_rank128_mse_mean": 0.05314673676288554, + "w2_phase1b_mse_mean": 0.05314673676288554, + "w2_phase2_mse_mean": 0.05314673676288554, + "w2_composite_mse_mean": 0.05314673676288554, + "w2_raw_to_w4_mse_ratio": 7.323461581192418, + "w2_composite_to_w4_mse_ratio": 7.323461581192418, + "old_rank128_improvement_vs_raw": 0.9999999999811842, + "phase1b_improvement_vs_raw": 0.9999999999811842, + "phase2_improvement_vs_raw": 0.9999999999811842, + "composite_improvement_vs_raw": 0.9999999999811842, + "remaining_w2_specific_gap_score": 0.3360713480931769, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 157, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.6.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.6.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00695298974668341, + "w2_raw_mse_mean": 0.05193483091092536, + "w2_old_rank128_mse_mean": 0.05193483091092536, + "w2_phase1b_mse_mean": 0.05193483091092536, + "w2_phase2_mse_mean": 0.05193483091092536, + "w2_composite_mse_mean": 0.05193483091092536, + "w2_raw_to_w4_mse_ratio": 7.469424347738891, + "w2_composite_to_w4_mse_ratio": 7.469424347738891, + "old_rank128_improvement_vs_raw": 0.9999999999807452, + "phase1b_improvement_vs_raw": 0.9999999999807452, + "phase2_improvement_vs_raw": 0.9999999999807452, + "composite_improvement_vs_raw": 0.9999999999807452, + "remaining_w2_specific_gap_score": 0.3359884595983123, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 158, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.1.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.1.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007190611055453441, + "w2_raw_mse_mean": 0.05280047023136701, + "w2_old_rank128_mse_mean": 0.05280047023136701, + "w2_phase1b_mse_mean": 0.05280047023136701, + "w2_phase2_mse_mean": 0.05280047023136701, + "w2_composite_mse_mean": 0.05280047023136701, + "w2_raw_to_w4_mse_ratio": 7.342974027774671, + "w2_composite_to_w4_mse_ratio": 7.342974027774671, + "old_rank128_improvement_vs_raw": 0.9999999999810608, + "phase1b_improvement_vs_raw": 0.9999999999810608, + "phase2_improvement_vs_raw": 0.9999999999810608, + "composite_improvement_vs_raw": 0.9999999999810608, + "remaining_w2_specific_gap_score": 0.3349120113391936, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 159, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.8.attn1.to_v", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.8.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00820906734692731, + "w2_raw_mse_mean": 0.056493712056960375, + "w2_old_rank128_mse_mean": 0.056493712056960375, + "w2_phase1b_mse_mean": 0.056493712056960375, + "w2_phase2_mse_mean": 0.056493712056960375, + "w2_composite_mse_mean": 0.056493712056960375, + "w2_raw_to_w4_mse_ratio": 6.881867289251117, + "w2_composite_to_w4_mse_ratio": 6.881867289251117, + "old_rank128_improvement_vs_raw": 0.9999999999822989, + "phase1b_improvement_vs_raw": 0.9999999999822989, + "phase2_improvement_vs_raw": 0.9999999999822989, + "composite_improvement_vs_raw": 0.9999999999822989, + "remaining_w2_specific_gap_score": 0.33228851700308853, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 160, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.7.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.7.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007825615524780005, + "w2_raw_mse_mean": 0.05502519224371229, + "w2_old_rank128_mse_mean": 0.05502519224371229, + "w2_phase1b_mse_mean": 0.05502519224371229, + "w2_phase2_mse_mean": 0.05502519224371229, + "w2_composite_mse_mean": 0.05502519224371229, + "w2_raw_to_w4_mse_ratio": 7.031420348015083, + "w2_composite_to_w4_mse_ratio": 7.031420348015083, + "old_rank128_improvement_vs_raw": 0.9999999999818265, + "phase1b_improvement_vs_raw": 0.9999999999818265, + "phase2_improvement_vs_raw": 0.9999999999818265, + "composite_improvement_vs_raw": 0.9999999999818265, + "remaining_w2_specific_gap_score": 0.33188006415919946, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 161, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.8.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.8.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008694804265230362, + "w2_raw_mse_mean": 0.058207141634609015, + "w2_old_rank128_mse_mean": 0.058207141634609015, + "w2_phase1b_mse_mean": 0.058207141634609015, + "w2_phase2_mse_mean": 0.058207141634609015, + "w2_composite_mse_mean": 0.058207141634609015, + "w2_raw_to_w4_mse_ratio": 6.694474062018736, + "w2_composite_to_w4_mse_ratio": 6.694474062018736, + "old_rank128_improvement_vs_raw": 0.99999999998282, + "phase1b_improvement_vs_raw": 0.99999999998282, + "phase2_improvement_vs_raw": 0.99999999998282, + "composite_improvement_vs_raw": 0.99999999998282, + "remaining_w2_specific_gap_score": 0.3314590582692264, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 162, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007028926513157785, + "w2_raw_mse_mean": 0.05151086114346981, + "w2_old_rank128_mse_mean": 0.05151086114346981, + "w2_phase1b_mse_mean": 0.05151086114346981, + "w2_phase2_mse_mean": 0.05151086114346981, + "w2_composite_mse_mean": 0.05151086114346981, + "w2_raw_to_w4_mse_ratio": 7.328410823433101, + "w2_composite_to_w4_mse_ratio": 7.328410823433101, + "old_rank128_improvement_vs_raw": 0.9999999999805866, + "phase1b_improvement_vs_raw": 0.9999999999805866, + "phase2_improvement_vs_raw": 0.9999999999805866, + "composite_improvement_vs_raw": 0.9999999999805866, + "remaining_w2_specific_gap_score": 0.3259818911920223, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 163, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.9.attn1.to_v", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.9.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008140871650539339, + "w2_raw_mse_mean": 0.05567192918221865, + "w2_old_rank128_mse_mean": 0.05567192918221865, + "w2_phase1b_mse_mean": 0.05567192918221865, + "w2_phase2_mse_mean": 0.05567192918221865, + "w2_composite_mse_mean": 0.05567192918221865, + "w2_raw_to_w4_mse_ratio": 6.838571048063603, + "w2_composite_to_w4_mse_ratio": 6.838571048063603, + "old_rank128_improvement_vs_raw": 0.9999999999820377, + "phase1b_improvement_vs_raw": 0.9999999999820377, + "phase2_improvement_vs_raw": 0.9999999999820377, + "composite_improvement_vs_raw": 0.9999999999820377, + "remaining_w2_specific_gap_score": 0.3250445139199876, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 164, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.4.attn2.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.4.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006580516587876316, + "w2_raw_mse_mean": 0.04957554243238909, + "w2_old_rank128_mse_mean": 0.04957554243238909, + "w2_phase1b_mse_mean": 0.04957554243238909, + "w2_phase2_mse_mean": 0.04957554243238909, + "w2_composite_mse_mean": 0.04957554243238909, + "w2_raw_to_w4_mse_ratio": 7.533685503686963, + "w2_composite_to_w4_mse_ratio": 7.533685503686963, + "old_rank128_improvement_vs_raw": 0.9999999999798288, + "phase1b_improvement_vs_raw": 0.9999999999798288, + "phase2_improvement_vs_raw": 0.9999999999798288, + "composite_improvement_vs_raw": 0.9999999999798288, + "remaining_w2_specific_gap_score": 0.3239110029354522, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 165, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.4.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.4.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008210500086923795, + "w2_raw_mse_mean": 0.05567643678347979, + "w2_old_rank128_mse_mean": 0.05567643678347979, + "w2_phase1b_mse_mean": 0.05567643678347979, + "w2_phase2_mse_mean": 0.05567643678347979, + "w2_composite_mse_mean": 0.05567643678347979, + "w2_raw_to_w4_mse_ratio": 6.78112614180104, + "w2_composite_to_w4_mse_ratio": 6.78112614180104, + "old_rank128_improvement_vs_raw": 0.9999999999820391, + "phase1b_improvement_vs_raw": 0.9999999999820391, + "phase2_improvement_vs_raw": 0.9999999999820391, + "composite_improvement_vs_raw": 0.9999999999820391, + "remaining_w2_specific_gap_score": 0.32187250417808916, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 166, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.2.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.2.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008297459372053189, + "w2_raw_mse_mean": 0.05586122861132026, + "w2_old_rank128_mse_mean": 0.05586122861132026, + "w2_phase1b_mse_mean": 0.05586122861132026, + "w2_phase2_mse_mean": 0.05586122861132026, + "w2_composite_mse_mean": 0.05586122861132026, + "w2_raw_to_w4_mse_ratio": 6.732329270901291, + "w2_composite_to_w4_mse_ratio": 6.732329270901291, + "old_rank128_improvement_vs_raw": 0.9999999999820985, + "phase1b_improvement_vs_raw": 0.9999999999820985, + "phase2_improvement_vs_raw": 0.9999999999820985, + "composite_improvement_vs_raw": 0.9999999999820985, + "remaining_w2_specific_gap_score": 0.32021495588391213, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 167, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.1.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.1.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006873148327161159, + "w2_raw_mse_mean": 0.050424908487392325, + "w2_old_rank128_mse_mean": 0.050424908487392325, + "w2_phase1b_mse_mean": 0.050424908487392325, + "w2_phase2_mse_mean": 0.050424908487392325, + "w2_composite_mse_mean": 0.050424908487392325, + "w2_raw_to_w4_mse_ratio": 7.336508115325803, + "w2_composite_to_w4_mse_ratio": 7.336508115325803, + "old_rank128_improvement_vs_raw": 0.9999999999801685, + "phase1b_improvement_vs_raw": 0.9999999999801685, + "phase2_improvement_vs_raw": 0.9999999999801685, + "composite_improvement_vs_raw": 0.9999999999801685, + "remaining_w2_specific_gap_score": 0.31951784185225895, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 168, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.9.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.9.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.008353837449768824, + "w2_raw_mse_mean": 0.05582169263756701, + "w2_old_rank128_mse_mean": 0.05582169263756701, + "w2_phase1b_mse_mean": 0.05582169263756701, + "w2_phase2_mse_mean": 0.05582169263756701, + "w2_composite_mse_mean": 0.05582169263756701, + "w2_raw_to_w4_mse_ratio": 6.682161697128738, + "w2_composite_to_w4_mse_ratio": 6.682161697128738, + "old_rank128_improvement_vs_raw": 0.9999999999820859, + "phase1b_improvement_vs_raw": 0.9999999999820859, + "phase2_improvement_vs_raw": 0.9999999999820859, + "composite_improvement_vs_raw": 0.9999999999820859, + "remaining_w2_specific_gap_score": 0.3171878837807587, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 169, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.9.attn2.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.9.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0035961148553594413, + "w2_raw_mse_mean": 0.035519292877454846, + "w2_old_rank128_mse_mean": 0.035519292877454846, + "w2_phase1b_mse_mean": 0.035519292877454846, + "w2_phase2_mse_mean": 0.035519292877454846, + "w2_composite_mse_mean": 0.035519292877454846, + "w2_raw_to_w4_mse_ratio": 9.877129706978579, + "w2_composite_to_w4_mse_ratio": 9.877129706978579, + "old_rank128_improvement_vs_raw": 0.9999999999718463, + "phase1b_improvement_vs_raw": 0.9999999999718463, + "phase2_improvement_vs_raw": 0.9999999999718463, + "composite_improvement_vs_raw": 0.9999999999718463, + "remaining_w2_specific_gap_score": 0.3153093699832042, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 170, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.2.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.2.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006807816963243697, + "w2_raw_mse_mean": 0.04938648268580437, + "w2_old_rank128_mse_mean": 0.04938648268580437, + "w2_phase1b_mse_mean": 0.04938648268580437, + "w2_phase2_mse_mean": 0.04938648268580437, + "w2_composite_mse_mean": 0.04938648268580437, + "w2_raw_to_w4_mse_ratio": 7.254378745079976, + "w2_composite_to_w4_mse_ratio": 7.254378745079976, + "old_rank128_improvement_vs_raw": 0.9999999999797515, + "phase1b_improvement_vs_raw": 0.9999999999797515, + "phase2_improvement_vs_raw": 0.9999999999797515, + "composite_improvement_vs_raw": 0.9999999999797515, + "remaining_w2_specific_gap_score": 0.3088817676116095, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 171, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.003558679679242362, + "w2_raw_mse_mean": 0.034886259119957685, + "w2_old_rank128_mse_mean": 0.034886259119957685, + "w2_phase1b_mse_mean": 0.034886259119957685, + "w2_phase2_mse_mean": 0.034886259119957685, + "w2_composite_mse_mean": 0.034886259119957685, + "w2_raw_to_w4_mse_ratio": 9.803146743901877, + "w2_composite_to_w4_mse_ratio": 9.803146743901877, + "old_rank128_improvement_vs_raw": 0.9999999999713355, + "phase1b_improvement_vs_raw": 0.9999999999713355, + "phase2_improvement_vs_raw": 0.9999999999713355, + "composite_improvement_vs_raw": 0.9999999999713355, + "remaining_w2_specific_gap_score": 0.3071088583885758, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 172, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.4.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.4.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006838459247124514, + "w2_raw_mse_mean": 0.049264275402362855, + "w2_old_rank128_mse_mean": 0.049264275402362855, + "w2_phase1b_mse_mean": 0.049264275402362855, + "w2_phase2_mse_mean": 0.049264275402362855, + "w2_composite_mse_mean": 0.049264275402362855, + "w2_raw_to_w4_mse_ratio": 7.204002190387236, + "w2_composite_to_w4_mse_ratio": 7.204002190387236, + "old_rank128_improvement_vs_raw": 0.9999999999797013, + "phase1b_improvement_vs_raw": 0.9999999999797013, + "phase2_improvement_vs_raw": 0.9999999999797013, + "composite_improvement_vs_raw": 0.9999999999797013, + "remaining_w2_specific_gap_score": 0.3056356725113032, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 173, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.005026322115944433, + "w2_raw_mse_mean": 0.041745843299265416, + "w2_old_rank128_mse_mean": 0.041745843299265416, + "w2_phase1b_mse_mean": 0.041745843299265416, + "w2_phase2_mse_mean": 0.041745843299265416, + "w2_composite_mse_mean": 0.041745843299265416, + "w2_raw_to_w4_mse_ratio": 8.30544527946873, + "w2_composite_to_w4_mse_ratio": 8.30544527946873, + "old_rank128_improvement_vs_raw": 0.9999999999760456, + "phase1b_improvement_vs_raw": 0.9999999999760456, + "phase2_improvement_vs_raw": 0.9999999999760456, + "composite_improvement_vs_raw": 0.9999999999760456, + "remaining_w2_specific_gap_score": 0.30497197387636527, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 174, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.5.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.5.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.00711254177648308, + "w2_raw_mse_mean": 0.050212835360850604, + "w2_old_rank128_mse_mean": 0.050212835360850604, + "w2_phase1b_mse_mean": 0.050212835360850604, + "w2_phase2_mse_mean": 0.050212835360850604, + "w2_composite_mse_mean": 0.050212835360850604, + "w2_raw_to_w4_mse_ratio": 7.059759637520112, + "w2_composite_to_w4_mse_ratio": 7.059759637520112, + "old_rank128_improvement_vs_raw": 0.9999999999800848, + "phase1b_improvement_vs_raw": 0.9999999999800848, + "phase2_improvement_vs_raw": 0.9999999999800848, + "composite_improvement_vs_raw": 0.9999999999800848, + "remaining_w2_specific_gap_score": 0.30427771301218487, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 175, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0035091634781565517, + "w2_raw_mse_mean": 0.03445518994703889, + "w2_old_rank128_mse_mean": 0.03445518994703889, + "w2_phase1b_mse_mean": 0.03445518994703889, + "w2_phase2_mse_mean": 0.03445518994703889, + "w2_composite_mse_mean": 0.03445518994703889, + "w2_raw_to_w4_mse_ratio": 9.818633458285163, + "w2_composite_to_w4_mse_ratio": 9.818633458285163, + "old_rank128_improvement_vs_raw": 0.9999999999709769, + "phase1b_improvement_vs_raw": 0.9999999999709769, + "phase2_improvement_vs_raw": 0.9999999999709769, + "composite_improvement_vs_raw": 0.9999999999709769, + "remaining_w2_specific_gap_score": 0.30384769088834634, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 176, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.4.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007767786579539201, + "w2_raw_mse_mean": 0.052618608211300204, + "w2_old_rank128_mse_mean": 0.052618608211300204, + "w2_phase1b_mse_mean": 0.052618608211300204, + "w2_phase2_mse_mean": 0.052618608211300204, + "w2_composite_mse_mean": 0.052618608211300204, + "w2_raw_to_w4_mse_ratio": 6.773951326511301, + "w2_composite_to_w4_mse_ratio": 6.773951326511301, + "old_rank128_improvement_vs_raw": 0.9999999999809953, + "phase1b_improvement_vs_raw": 0.9999999999809953, + "phase2_improvement_vs_raw": 0.9999999999809953, + "composite_improvement_vs_raw": 0.9999999999809953, + "remaining_w2_specific_gap_score": 0.3038172826875892, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 177, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006552080889897687, + "w2_raw_mse_mean": 0.04790414770000747, + "w2_old_rank128_mse_mean": 0.04790414770000747, + "w2_phase1b_mse_mean": 0.04790414770000747, + "w2_phase2_mse_mean": 0.04790414770000747, + "w2_composite_mse_mean": 0.04790414770000747, + "w2_raw_to_w4_mse_ratio": 7.3112875890401625, + "w2_composite_to_w4_mse_ratio": 7.3112875890401625, + "old_rank128_improvement_vs_raw": 0.999999999979125, + "phase1b_improvement_vs_raw": 0.999999999979125, + "phase2_improvement_vs_raw": 0.999999999979125, + "composite_improvement_vs_raw": 0.999999999979125, + "remaining_w2_specific_gap_score": 0.3023368528499153, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 178, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.1.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.1.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0066317775552826265, + "w2_raw_mse_mean": 0.048116493238402266, + "w2_old_rank128_mse_mean": 0.048116493238402266, + "w2_phase1b_mse_mean": 0.048116493238402266, + "w2_phase2_mse_mean": 0.048116493238402266, + "w2_composite_mse_mean": 0.048116493238402266, + "w2_raw_to_w4_mse_ratio": 7.255444385769396, + "w2_composite_to_w4_mse_ratio": 7.255444385769396, + "old_rank128_improvement_vs_raw": 0.9999999999792172, + "phase1b_improvement_vs_raw": 0.9999999999792172, + "phase2_improvement_vs_raw": 0.9999999999792172, + "composite_improvement_vs_raw": 0.9999999999792172, + "remaining_w2_specific_gap_score": 0.30099004749833, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 179, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.6.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.6.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006942451756913215, + "w2_raw_mse_mean": 0.049280748170401366, + "w2_old_rank128_mse_mean": 0.049280748170401366, + "w2_phase1b_mse_mean": 0.049280748170401366, + "w2_phase2_mse_mean": 0.049280748170401366, + "w2_composite_mse_mean": 0.049280748170401366, + "w2_raw_to_w4_mse_ratio": 7.098464618674474, + "w2_composite_to_w4_mse_ratio": 7.098464618674474, + "old_rank128_improvement_vs_raw": 0.9999999999797081, + "phase1b_improvement_vs_raw": 0.9999999999797081, + "phase2_improvement_vs_raw": 0.9999999999797081, + "composite_improvement_vs_raw": 0.9999999999797081, + "remaining_w2_specific_gap_score": 0.30053689910609804, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 180, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.2.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.2.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.005556420173629054, + "w2_raw_mse_mean": 0.04372347878026111, + "w2_old_rank128_mse_mean": 0.04372347878026111, + "w2_phase1b_mse_mean": 0.04372347878026111, + "w2_phase2_mse_mean": 0.04372347878026111, + "w2_composite_mse_mean": 0.04372347878026111, + "w2_raw_to_w4_mse_ratio": 7.869001516462905, + "w2_composite_to_w4_mse_ratio": 7.869001516462905, + "old_rank128_improvement_vs_raw": 0.9999999999771291, + "phase1b_improvement_vs_raw": 0.9999999999771291, + "phase2_improvement_vs_raw": 0.9999999999771291, + "composite_improvement_vs_raw": 0.9999999999771291, + "remaining_w2_specific_gap_score": 0.3003366420545162, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 181, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.3.attn1.to_v", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.3.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006697809032630175, + "w2_raw_mse_mean": 0.048191075545868704, + "w2_old_rank128_mse_mean": 0.048191075545868704, + "w2_phase1b_mse_mean": 0.048191075545868704, + "w2_phase2_mse_mean": 0.048191075545868704, + "w2_composite_mse_mean": 0.048191075545868704, + "w2_raw_to_w4_mse_ratio": 7.195050695518174, + "w2_composite_to_w4_mse_ratio": 7.195050695518174, + "old_rank128_improvement_vs_raw": 0.9999999999792493, + "phase1b_improvement_vs_raw": 0.9999999999792493, + "phase2_improvement_vs_raw": 0.9999999999792493, + "composite_improvement_vs_raw": 0.9999999999792493, + "remaining_w2_specific_gap_score": 0.2985461560853978, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 182, + "module_name": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_v", + "ldm_key": "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 640, + 640 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006275745449654225, + "w2_raw_mse_mean": 0.04647703528670328, + "w2_old_rank128_mse_mean": 0.04647703528670328, + "w2_phase1b_mse_mean": 0.04647703528670328, + "w2_phase2_mse_mean": 0.04647703528670328, + "w2_composite_mse_mean": 0.04647703528670328, + "w2_raw_to_w4_mse_ratio": 7.405819061998157, + "w2_composite_to_w4_mse_ratio": 7.405819061998157, + "old_rank128_improvement_vs_raw": 0.999999999978484, + "phase1b_improvement_vs_raw": 0.999999999978484, + "phase2_improvement_vs_raw": 0.999999999978484, + "composite_improvement_vs_raw": 0.999999999978484, + "remaining_w2_specific_gap_score": 0.29772347859213066, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 183, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.0.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.0.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.005173523303320897, + "w2_raw_mse_mean": 0.0419008936067777, + "w2_old_rank128_mse_mean": 0.0419008936067777, + "w2_phase1b_mse_mean": 0.0419008936067777, + "w2_phase2_mse_mean": 0.0419008936067777, + "w2_composite_mse_mean": 0.0419008936067777, + "w2_raw_to_w4_mse_ratio": 8.099102128675502, + "w2_composite_to_w4_mse_ratio": 8.099102128675502, + "old_rank128_improvement_vs_raw": 0.9999999999761342, + "phase1b_improvement_vs_raw": 0.9999999999761342, + "phase2_improvement_vs_raw": 0.9999999999761342, + "composite_improvement_vs_raw": 0.9999999999761342, + "remaining_w2_specific_gap_score": 0.29745872300538045, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 184, + "module_name": "mid_block.attentions.0.transformer_blocks.9.attn2.to_q", + "ldm_key": "model.diffusion_model.middle_block.1.transformer_blocks.9.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0025735883973538876, + "w2_raw_mse_mean": 0.028946281809891974, + "w2_old_rank128_mse_mean": 0.028946281809891974, + "w2_phase1b_mse_mean": 0.028946281809891974, + "w2_phase2_mse_mean": 0.028946281809891974, + "w2_composite_mse_mean": 0.028946281809891974, + "w2_raw_to_w4_mse_ratio": 11.247440277709725, + "w2_composite_to_w4_mse_ratio": 11.247440277709725, + "old_rank128_improvement_vs_raw": 0.9999999999654533, + "phase1b_improvement_vs_raw": 0.9999999999654533, + "phase2_improvement_vs_raw": 0.9999999999654533, + "composite_improvement_vs_raw": 0.9999999999654533, + "remaining_w2_specific_gap_score": 0.2966252941198708, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 185, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.1.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.1.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006145016165516738, + "w2_raw_mse_mean": 0.0458448794670403, + "w2_old_rank128_mse_mean": 0.0458448794670403, + "w2_phase1b_mse_mean": 0.0458448794670403, + "w2_phase2_mse_mean": 0.0458448794670403, + "w2_composite_mse_mean": 0.0458448794670403, + "w2_raw_to_w4_mse_ratio": 7.460497779784877, + "w2_composite_to_w4_mse_ratio": 7.460497779784877, + "old_rank128_improvement_vs_raw": 0.9999999999781873, + "phase1b_improvement_vs_raw": 0.9999999999781873, + "phase2_improvement_vs_raw": 0.9999999999781873, + "composite_improvement_vs_raw": 0.9999999999781873, + "remaining_w2_specific_gap_score": 0.29618074201877964, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 186, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.4.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.005976258885181908, + "w2_raw_mse_mean": 0.04512940599982228, + "w2_old_rank128_mse_mean": 0.04512940599982228, + "w2_phase1b_mse_mean": 0.04512940599982228, + "w2_phase2_mse_mean": 0.04512940599982228, + "w2_composite_mse_mean": 0.04512940599982228, + "w2_raw_to_w4_mse_ratio": 7.551447629581254, + "w2_composite_to_w4_mse_ratio": 7.551447629581254, + "old_rank128_improvement_vs_raw": 0.9999999999778415, + "phase1b_improvement_vs_raw": 0.9999999999778415, + "phase2_improvement_vs_raw": 0.9999999999778415, + "composite_improvement_vs_raw": 0.9999999999778415, + "remaining_w2_specific_gap_score": 0.29566293996949716, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 187, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.0.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.0.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.004630002151576004, + "w2_raw_mse_mean": 0.039363279206944365, + "w2_old_rank128_mse_mean": 0.039363279206944365, + "w2_phase1b_mse_mean": 0.039363279206944365, + "w2_phase2_mse_mean": 0.039363279206944365, + "w2_composite_mse_mean": 0.039363279206944365, + "w2_raw_to_w4_mse_ratio": 8.501784213003818, + "w2_composite_to_w4_mse_ratio": 8.501784213003818, + "old_rank128_improvement_vs_raw": 0.9999999999745957, + "phase1b_improvement_vs_raw": 0.9999999999745957, + "phase2_improvement_vs_raw": 0.9999999999745957, + "composite_improvement_vs_raw": 0.9999999999745957, + "remaining_w2_specific_gap_score": 0.2952948265352185, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 188, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.3.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.3.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.005812832994186985, + "w2_raw_mse_mean": 0.044433628608073504, + "w2_old_rank128_mse_mean": 0.044433628608073504, + "w2_phase1b_mse_mean": 0.044433628608073504, + "w2_phase2_mse_mean": 0.044433628608073504, + "w2_composite_mse_mean": 0.044433628608073504, + "w2_raw_to_w4_mse_ratio": 7.6440573202196695, + "w2_composite_to_w4_mse_ratio": 7.6440573202196695, + "old_rank128_improvement_vs_raw": 0.9999999999774946, + "phase1b_improvement_vs_raw": 0.9999999999774946, + "phase2_improvement_vs_raw": 0.9999999999774946, + "composite_improvement_vs_raw": 0.9999999999774946, + "remaining_w2_specific_gap_score": 0.29521957542503696, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 189, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.2.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.2.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0056498262482429186, + "w2_raw_mse_mean": 0.04361926930557404, + "w2_old_rank128_mse_mean": 0.04361926930557404, + "w2_phase1b_mse_mean": 0.04361926930557404, + "w2_phase2_mse_mean": 0.04361926930557404, + "w2_composite_mse_mean": 0.04361926930557404, + "w2_raw_to_w4_mse_ratio": 7.720462078177901, + "w2_composite_to_w4_mse_ratio": 7.720462078177901, + "old_rank128_improvement_vs_raw": 0.9999999999770743, + "phase1b_improvement_vs_raw": 0.9999999999770743, + "phase2_improvement_vs_raw": 0.9999999999770743, + "composite_improvement_vs_raw": 0.9999999999770743, + "remaining_w2_specific_gap_score": 0.29314164525366015, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 190, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.5.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.5.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006440597511495331, + "w2_raw_mse_mean": 0.04677186221150415, + "w2_old_rank128_mse_mean": 0.04677186221150415, + "w2_phase1b_mse_mean": 0.04677186221150415, + "w2_phase2_mse_mean": 0.04677186221150415, + "w2_composite_mse_mean": 0.04677186221150415, + "w2_raw_to_w4_mse_ratio": 7.262037741182024, + "w2_composite_to_w4_mse_ratio": 7.262037741182024, + "old_rank128_improvement_vs_raw": 0.9999999999786197, + "phase1b_improvement_vs_raw": 0.9999999999786197, + "phase2_improvement_vs_raw": 0.9999999999786197, + "composite_improvement_vs_raw": 0.9999999999786197, + "remaining_w2_specific_gap_score": 0.2928871664010663, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 191, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.1.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.1.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.005045917907929314, + "w2_raw_mse_mean": 0.04100019957071969, + "w2_old_rank128_mse_mean": 0.04100019957071969, + "w2_phase1b_mse_mean": 0.04100019957071969, + "w2_phase2_mse_mean": 0.04100019957071969, + "w2_composite_mse_mean": 0.04100019957071969, + "w2_raw_to_w4_mse_ratio": 8.12541945998869, + "w2_composite_to_w4_mse_ratio": 8.12541945998869, + "old_rank128_improvement_vs_raw": 0.99999999997561, + "phase1b_improvement_vs_raw": 0.99999999997561, + "phase2_improvement_vs_raw": 0.99999999997561, + "composite_improvement_vs_raw": 0.99999999997561, + "remaining_w2_specific_gap_score": 0.29214361989275145, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 192, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.0.attn1.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.0.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.005402986251283437, + "w2_raw_mse_mean": 0.04236168560705015, + "w2_old_rank128_mse_mean": 0.04236168560705015, + "w2_phase1b_mse_mean": 0.04236168560705015, + "w2_phase2_mse_mean": 0.04236168560705015, + "w2_composite_mse_mean": 0.04236168560705015, + "w2_raw_to_w4_mse_ratio": 7.840420765303084, + "w2_composite_to_w4_mse_ratio": 7.840420765303084, + "old_rank128_improvement_vs_raw": 0.9999999999763938, + "phase1b_improvement_vs_raw": 0.9999999999763938, + "phase2_improvement_vs_raw": 0.9999999999763938, + "composite_improvement_vs_raw": 0.9999999999763938, + "remaining_w2_specific_gap_score": 0.28977175388754706, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 193, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.6.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.6.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.006458820184759263, + "w2_raw_mse_mean": 0.04656909898455654, + "w2_old_rank128_mse_mean": 0.04656909898455654, + "w2_phase1b_mse_mean": 0.04656909898455654, + "w2_phase2_mse_mean": 0.04656909898455654, + "w2_composite_mse_mean": 0.04656909898455654, + "w2_raw_to_w4_mse_ratio": 7.210155670107439, + "w2_composite_to_w4_mse_ratio": 7.210155670107439, + "old_rank128_improvement_vs_raw": 0.9999999999785265, + "phase1b_improvement_vs_raw": 0.9999999999785265, + "phase2_improvement_vs_raw": 0.9999999999785265, + "composite_improvement_vs_raw": 0.9999999999785265, + "remaining_w2_specific_gap_score": 0.28920135411794856, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 194, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.6.attn1.to_q", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_q.weight", + "suffix": "attn1.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.004988977502632354, + "w2_raw_mse_mean": 0.040489877840237956, + "w2_old_rank128_mse_mean": 0.040489877840237956, + "w2_phase1b_mse_mean": 0.040489877840237956, + "w2_phase2_mse_mean": 0.040489877840237956, + "w2_composite_mse_mean": 0.040489877840237956, + "w2_raw_to_w4_mse_ratio": 8.115866990933164, + "w2_composite_to_w4_mse_ratio": 8.115866990933164, + "old_rank128_improvement_vs_raw": 0.9999999999753025, + "phase1b_improvement_vs_raw": 0.9999999999753025, + "phase2_improvement_vs_raw": 0.9999999999753025, + "composite_improvement_vs_raw": 0.9999999999753025, + "remaining_w2_specific_gap_score": 0.28812058519838135, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 195, + "module_name": "up_blocks.0.attentions.1.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.output_blocks.1.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0053578837874478525, + "w2_raw_mse_mean": 0.0420537871707763, + "w2_old_rank128_mse_mean": 0.0420537871707763, + "w2_phase1b_mse_mean": 0.0420537871707763, + "w2_phase2_mse_mean": 0.0420537871707763, + "w2_composite_mse_mean": 0.0420537871707763, + "w2_raw_to_w4_mse_ratio": 7.848954705111107, + "w2_composite_to_w4_mse_ratio": 7.848954705111107, + "old_rank128_improvement_vs_raw": 0.9999999999762209, + "phase1b_improvement_vs_raw": 0.9999999999762209, + "phase2_improvement_vs_raw": 0.9999999999762209, + "composite_improvement_vs_raw": 0.9999999999762209, + "remaining_w2_specific_gap_score": 0.28802448351887844, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 196, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.7.attn2.to_q", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.7.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.002531217731302604, + "w2_raw_mse_mean": 0.028214162814297845, + "w2_old_rank128_mse_mean": 0.028214162814297845, + "w2_phase1b_mse_mean": 0.028214162814297845, + "w2_phase2_mse_mean": 0.028214162814297845, + "w2_composite_mse_mean": 0.028214162814297845, + "w2_raw_to_w4_mse_ratio": 11.14647801895411, + "w2_composite_to_w4_mse_ratio": 11.14647801895411, + "old_rank128_improvement_vs_raw": 0.9999999999645569, + "phase1b_improvement_vs_raw": 0.9999999999645569, + "phase2_improvement_vs_raw": 0.9999999999645569, + "composite_improvement_vs_raw": 0.9999999999645569, + "remaining_w2_specific_gap_score": 0.286274382829612, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 197, + "module_name": "up_blocks.0.attentions.2.transformer_blocks.7.attn1.to_v", + "ldm_key": "model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn1.to_v.weight", + "suffix": "attn1.to_v", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.007521395297122321, + "w2_raw_mse_mean": 0.05007568587149892, + "w2_old_rank128_mse_mean": 0.05007568587149892, + "w2_phase1b_mse_mean": 0.05007568587149892, + "w2_phase2_mse_mean": 0.05007568587149892, + "w2_composite_mse_mean": 0.05007568587149892, + "w2_raw_to_w4_mse_ratio": 6.657765465936895, + "w2_composite_to_w4_mse_ratio": 6.657765465936895, + "old_rank128_improvement_vs_raw": 0.9999999999800303, + "phase1b_improvement_vs_raw": 0.9999999999800303, + "phase2_improvement_vs_raw": 0.9999999999800303, + "composite_improvement_vs_raw": 0.9999999999800303, + "remaining_w2_specific_gap_score": 0.28331648621352845, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 198, + "module_name": "up_blocks.0.attentions.0.transformer_blocks.8.attn2.to_q", + "ldm_key": "model.diffusion_model.output_blocks.0.1.transformer_blocks.8.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.0038138889441532747, + "w2_raw_mse_mean": 0.03480188023032887, + "w2_old_rank128_mse_mean": 0.03480188023032887, + "w2_phase1b_mse_mean": 0.03480188023032887, + "w2_phase2_mse_mean": 0.03480188023032887, + "w2_composite_mse_mean": 0.03480188023032887, + "w2_raw_to_w4_mse_ratio": 9.125037653378822, + "w2_composite_to_w4_mse_ratio": 9.125037653378822, + "old_rank128_improvement_vs_raw": 0.999999999971266, + "phase1b_improvement_vs_raw": 0.999999999971266, + "phase2_improvement_vs_raw": 0.999999999971266, + "composite_improvement_vs_raw": 0.999999999971266, + "remaining_w2_specific_gap_score": 0.2827665872889271, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 199, + "module_name": "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_k", + "ldm_key": "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_k.weight", + "suffix": "attn1.to_k", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.004382886641126659, + "w2_raw_mse_mean": 0.03746225332309093, + "w2_old_rank128_mse_mean": 0.03746225332309093, + "w2_phase1b_mse_mean": 0.03746225332309093, + "w2_phase2_mse_mean": 0.03746225332309093, + "w2_composite_mse_mean": 0.03746225332309093, + "w2_raw_to_w4_mse_ratio": 8.547392707586328, + "w2_composite_to_w4_mse_ratio": 8.547392707586328, + "old_rank128_improvement_vs_raw": 0.9999999999733065, + "phase1b_improvement_vs_raw": 0.9999999999733065, + "phase2_improvement_vs_raw": 0.9999999999733065, + "composite_improvement_vs_raw": 0.9999999999733065, + "remaining_w2_specific_gap_score": 0.2827423375489955, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + }, + { + "training_priority_rank": 200, + "module_name": "down_blocks.2.attentions.1.transformer_blocks.8.attn2.to_q", + "ldm_key": "model.diffusion_model.input_blocks.8.1.transformer_blocks.8.attn2.to_q.weight", + "suffix": "attn2.to_q", + "shape": [ + 1280, + 1280 + ], + "has_old_rank128": false, + "has_phase1b": false, + "has_phase2": false, + "composite_source": "raw", + "w4_mse_mean": 0.004055892811655732, + "w2_raw_mse_mean": 0.03589118059192385, + "w2_old_rank128_mse_mean": 0.03589118059192385, + "w2_phase1b_mse_mean": 0.03589118059192385, + "w2_phase2_mse_mean": 0.03589118059192385, + "w2_composite_mse_mean": 0.03589118059192385, + "w2_raw_to_w4_mse_ratio": 8.849144257444735, + "w2_composite_to_w4_mse_ratio": 8.849144257444735, + "old_rank128_improvement_vs_raw": 0.9999999999721381, + "phase1b_improvement_vs_raw": 0.9999999999721381, + "phase2_improvement_vs_raw": 0.9999999999721381, + "composite_improvement_vs_raw": 0.9999999999721381, + "remaining_w2_specific_gap_score": 0.28171505404486014, + "recommended_action": "add_adapter_missing_layer", + "recommended_rank": 256 + } + ], + "suffix_summary": [ + { + "suffix": "attn1.to_q", + "count": 70, + "old_rank128_count": 0, + "phase1b_count": 0, + "phase2_count": 0, + "composite_count": 0, + "w4_mse_mean": 0.008834530145039174, + "w2_raw_mse_mean": 0.06604109280267541, + "w2_old_rank128_mse_mean": 0.06604109280267541, + "w2_phase1b_mse_mean": 0.06604109280267541, + "w2_phase2_mse_mean": 0.06604109280267541, + "w2_composite_mse_mean": 0.06604109280267541, + "w2_raw_to_w4_ratio_mean": 7.5747877444341, + "w2_composite_to_w4_ratio_mean": 7.5747877444341, + "composite_improvement_vs_raw_mean": 0.9999999999832079, + "phase1b_improvement_vs_raw_mean": 0.9999999999832079, + "phase2_improvement_vs_raw_mean": 0.9999999999832079, + "remaining_gap_score_mean": 0.42929220002638296, + "remaining_gap_score_max": 0.8156589509387167 + }, + { + "suffix": "attn1.to_k", + "count": 70, + "old_rank128_count": 0, + "phase1b_count": 0, + "phase2_count": 0, + "composite_count": 0, + "w4_mse_mean": 0.00869125885917444, + "w2_raw_mse_mean": 0.06352441829714772, + "w2_old_rank128_mse_mean": 0.06352441829714772, + "w2_phase1b_mse_mean": 0.06352441829714772, + "w2_phase2_mse_mean": 0.06352441829714772, + "w2_composite_mse_mean": 0.06352441829714772, + "w2_raw_to_w4_ratio_mean": 7.395005113804398, + "w2_composite_to_w4_ratio_mean": 7.395005113804398, + "composite_improvement_vs_raw_mean": 0.9999999999825412, + "phase1b_improvement_vs_raw_mean": 0.9999999999825412, + "phase2_improvement_vs_raw_mean": 0.9999999999825412, + "remaining_gap_score_mean": 0.4027360624185252, + "remaining_gap_score_max": 0.7311920707443591 + }, + { + "suffix": "attn1.to_v", + "count": 70, + "old_rank128_count": 0, + "phase1b_count": 0, + "phase2_count": 0, + "composite_count": 0, + "w4_mse_mean": 0.0078085727364238244, + "w2_raw_mse_mean": 0.055820106880796354, + "w2_old_rank128_mse_mean": 0.055820106880796354, + "w2_phase1b_mse_mean": 0.055820106880796354, + "w2_phase2_mse_mean": 0.055820106880796354, + "w2_composite_mse_mean": 0.055820106880796354, + "w2_raw_to_w4_ratio_mean": 7.1828383652128185, + "w2_composite_to_w4_ratio_mean": 7.1828383652128185, + "composite_improvement_vs_raw_mean": 0.999999999979725, + "phase1b_improvement_vs_raw_mean": 0.999999999979725, + "phase2_improvement_vs_raw_mean": 0.999999999979725, + "remaining_gap_score_mean": 0.34468865166305246, + "remaining_gap_score_max": 0.8469947038879494 + }, + { + "suffix": "proj_out", + "count": 11, + "old_rank128_count": 11, + "phase1b_count": 11, + "phase2_count": 8, + "composite_count": 11, + "w4_mse_mean": 0.08516350478106238, + "w2_raw_mse_mean": 0.5946647835556756, + "w2_old_rank128_mse_mean": 0.35843181815421715, + "w2_phase1b_mse_mean": 0.21572238788637635, + "w2_phase2_mse_mean": 0.2251953121226329, + "w2_composite_mse_mean": 0.20004504513866328, + "w2_raw_to_w4_ratio_mean": 6.913556153067888, + "w2_composite_to_w4_ratio_mean": 2.452973561882824, + "composite_improvement_vs_raw_mean": 2.969396590218453, + "phase1b_improvement_vs_raw_mean": 2.8728746957207307, + "phase2_improvement_vs_raw_mean": 2.388979858331337, + "remaining_gap_score_mean": 0.2912881384368135, + "remaining_gap_score_max": 0.7227891036499625 + }, + { + "suffix": "attn2.to_q", + "count": 70, + "old_rank128_count": 0, + "phase1b_count": 0, + "phase2_count": 0, + "composite_count": 0, + "w4_mse_mean": 0.004732320876079327, + "w2_raw_mse_mean": 0.037790954292618806, + "w2_old_rank128_mse_mean": 0.037790954292618806, + "w2_phase1b_mse_mean": 0.037790954292618806, + "w2_phase2_mse_mean": 0.037790954292618806, + "w2_composite_mse_mean": 0.037790954292618806, + "w2_raw_to_w4_ratio_mean": 8.31881567261658, + "w2_composite_to_w4_ratio_mean": 8.31881567261658, + "composite_improvement_vs_raw_mean": 0.9999999999712236, + "phase1b_improvement_vs_raw_mean": 0.9999999999712236, + "phase2_improvement_vs_raw_mean": 0.9999999999712236, + "remaining_gap_score_mean": 0.2711913227898678, + "remaining_gap_score_max": 0.6841755529628081 + }, + { + "suffix": "ff.net.2", + "count": 70, + "old_rank128_count": 70, + "phase1b_count": 70, + "phase2_count": 0, + "composite_count": 70, + "w4_mse_mean": 0.008616097393292431, + "w2_raw_mse_mean": 0.05920763128883756, + "w2_old_rank128_mse_mean": 0.04301157893223346, + "w2_phase1b_mse_mean": 0.02601208476651441, + "w2_phase2_mse_mean": 0.05920763128883756, + "w2_composite_mse_mean": 0.02601208476651441, + "w2_raw_to_w4_ratio_mean": 6.932097684619555, + "w2_composite_to_w4_ratio_mean": 3.2009256575082317, + "composite_improvement_vs_raw_mean": 2.3014962630105096, + "phase1b_improvement_vs_raw_mean": 2.3014962630105096, + "phase2_improvement_vs_raw_mean": 0.9999999999747406, + "remaining_gap_score_mean": 0.0572061937220024, + "remaining_gap_score_max": 0.20667232622681672 + }, + { + "suffix": "ff.net.0.proj", + "count": 70, + "old_rank128_count": 70, + "phase1b_count": 70, + "phase2_count": 0, + "composite_count": 70, + "w4_mse_mean": 0.006887027408722171, + "w2_raw_mse_mean": 0.04879796355012424, + "w2_old_rank128_mse_mean": 0.03164028028936164, + "w2_phase1b_mse_mean": 0.0176630408599336, + "w2_phase2_mse_mean": 0.04879796355012424, + "w2_composite_mse_mean": 0.0176630408599336, + "w2_raw_to_w4_ratio_mean": 7.171405831067578, + "w2_composite_to_w4_ratio_mean": 2.504632836281881, + "composite_improvement_vs_raw_mean": 2.998213235077801, + "phase1b_improvement_vs_raw_mean": 2.998213235077801, + "phase2_improvement_vs_raw_mean": 0.9999999999754532, + "remaining_gap_score_mean": 0.02951947878339631, + "remaining_gap_score_max": 0.1137982287591066 + }, + { + "suffix": "proj_in", + "count": 11, + "old_rank128_count": 11, + "phase1b_count": 11, + "phase2_count": 0, + "composite_count": 11, + "w4_mse_mean": 0.008933692221838206, + "w2_raw_mse_mean": 0.06406978043038157, + "w2_old_rank128_mse_mean": 0.03873854998329824, + "w2_phase1b_mse_mean": 0.018474196746806285, + "w2_phase2_mse_mean": 0.06406978043038157, + "w2_composite_mse_mean": 0.018474196746806285, + "w2_raw_to_w4_ratio_mean": 7.050454051901254, + "w2_composite_to_w4_ratio_mean": 1.9995457179267833, + "composite_improvement_vs_raw_mean": 3.6405517751576513, + "phase1b_improvement_vs_raw_mean": 3.6405517751576513, + "phase2_improvement_vs_raw_mean": 0.9999999999808292, + "remaining_gap_score_mean": 0.020491349940082455, + "remaining_gap_score_max": 0.05859008625009349 + }, + { + "suffix": "attn2.to_out.0", + "count": 70, + "old_rank128_count": 70, + "phase1b_count": 70, + "phase2_count": 0, + "composite_count": 70, + "w4_mse_mean": 0.0003375165737176238, + "w2_raw_mse_mean": 0.002381115644698648, + "w2_old_rank128_mse_mean": 0.002395683516781632, + "w2_phase1b_mse_mean": 0.00222385320224494, + "w2_phase2_mse_mean": 0.002381115644698648, + "w2_composite_mse_mean": 0.00222385320224494, + "w2_raw_to_w4_ratio_mean": 7.598942937360687, + "w2_composite_to_w4_ratio_mean": 6.7306297362361756, + "composite_improvement_vs_raw_mean": 1.8347576429611256, + "phase1b_improvement_vs_raw_mean": 1.8347576429611256, + "phase2_improvement_vs_raw_mean": 0.9999999981489446, + "remaining_gap_score_mean": 0.012840779013027317, + "remaining_gap_score_max": 0.05913571060692841 + }, + { + "suffix": "attn1.to_out.0", + "count": 70, + "old_rank128_count": 70, + "phase1b_count": 70, + "phase2_count": 0, + "composite_count": 70, + "w4_mse_mean": 0.0042316013468252445, + "w2_raw_mse_mean": 0.030931819595006883, + "w2_old_rank128_mse_mean": 0.01690724441420515, + "w2_phase1b_mse_mean": 0.005962002335374044, + "w2_phase2_mse_mean": 0.030931819595006883, + "w2_composite_mse_mean": 0.005962002335374044, + "w2_raw_to_w4_ratio_mean": 7.476497691796118, + "w2_composite_to_w4_ratio_mean": 1.4532056634829924, + "composite_improvement_vs_raw_mean": 5.661566033087392, + "phase1b_improvement_vs_raw_mean": 5.661566033087392, + "phase2_improvement_vs_raw_mean": 0.9999999999555428, + "remaining_gap_score_mean": 0.0034474704133703057, + "remaining_gap_score_max": 0.021837045500065322 + }, + { + "suffix": "attn2.to_v", + "count": 70, + "old_rank128_count": 0, + "phase1b_count": 0, + "phase2_count": 70, + "composite_count": 70, + "w4_mse_mean": 0.05053253279267145, + "w2_raw_mse_mean": 0.33406722963388474, + "w2_old_rank128_mse_mean": 0.33406722963388474, + "w2_phase1b_mse_mean": 0.33406722963388474, + "w2_phase2_mse_mean": 0.020737389160785824, + "w2_composite_mse_mean": 0.020737389160785824, + "w2_raw_to_w4_ratio_mean": 6.628456469971321, + "w2_composite_to_w4_ratio_mean": 0.4083913035125052, + "composite_improvement_vs_raw_mean": 16.325964238694546, + "phase1b_improvement_vs_raw_mean": 0.9999999999957944, + "phase2_improvement_vs_raw_mean": 16.325964238694546, + "remaining_gap_score_mean": 0.0, + "remaining_gap_score_max": 0.0 + }, + { + "suffix": "attn2.to_k", + "count": 70, + "old_rank128_count": 0, + "phase1b_count": 0, + "phase2_count": 70, + "composite_count": 70, + "w4_mse_mean": 0.0519350185824026, + "w2_raw_mse_mean": 0.42513413368059055, + "w2_old_rank128_mse_mean": 0.42513413368059055, + "w2_phase1b_mse_mean": 0.42513413368059055, + "w2_phase2_mse_mean": 0.024206796191200346, + "w2_composite_mse_mean": 0.024206796191200346, + "w2_raw_to_w4_ratio_mean": 8.66438299826568, + "w2_composite_to_w4_ratio_mean": 0.47265911793929566, + "composite_improvement_vs_raw_mean": 18.609512818870797, + "phase1b_improvement_vs_raw_mean": 0.9999999999969618, + "phase2_improvement_vs_raw_mean": 18.609512818870797, + "remaining_gap_score_mean": 0.0, + "remaining_gap_score_max": 0.0 + } + ] +} \ No newline at end of file