diff --git a/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/metrics/test_final100_loop_true_reverse_h100_f399_v1_step00000000_metrics.json b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/metrics/test_final100_loop_true_reverse_h100_f399_v1_step00000000_metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a05aac20493d1930f7f8fd768103aafb5a4ae17e --- /dev/null +++ b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/metrics/test_final100_loop_true_reverse_h100_f399_v1_step00000000_metrics.json @@ -0,0 +1,130 @@ +{ + "artifact_type": "re10k_benchmark_metrics", + "benchmark": "loop_true_reverse_h100_f399_v1", + "case_count": 100, + "generated_frames": 30000, + "global_step": 0, + "memory_mode": "correct", + "method": "worldmem", + "metrics": { + "final/lpips": 0.876171886920929, + "final/mse": 0.14580927789211273, + "final/psnr": 8.9396390914917, + "final/ssim": 0.254367858171463, + "frames_000_025/lpips": 0.2293580025434494, + "frames_000_025/mse": 0.025142336264252663, + "frames_000_025/psnr": 18.989437103271484, + "frames_000_025/rfid": 56.5, + "frames_000_025/ssim": 0.5077254772186279, + "frames_000_050/lpips": 0.42186474800109863, + "frames_000_050/mse": 0.059256792068481445, + "frames_000_050/psnr": 15.043106079101562, + "frames_000_050/rfid": 81.5, + "frames_000_050/ssim": 0.396808385848999, + "frames_000_100/lpips": 0.6100778579711914, + "frames_000_100/mse": 0.0906115397810936, + "frames_000_100/psnr": 12.310774803161621, + "frames_000_100/rfid": 116.0, + "frames_000_100/ssim": 0.32401224970817566, + "frames_025_050/lpips": 0.6143699288368225, + "frames_025_050/mse": 0.09337124973535538, + "frames_025_050/psnr": 11.09677505493164, + "frames_025_050/rfid": 138.0, + "frames_025_050/ssim": 0.2858913540840149, + "frames_050_075/lpips": 0.7753156423568726, + "frames_050_075/mse": 0.120847687125206, + "frames_050_075/psnr": 9.657604217529297, + "frames_050_075/rfid": 193.0, + "frames_050_075/ssim": 0.25408196449279785, + "frames_050_100/lpips": 0.798291802406311, + "frames_050_100/mse": 0.12196629494428635, + "frames_050_100/psnr": 9.57844352722168, + "frames_050_100/rfid": 191.0, + "frames_050_100/ssim": 0.2512160837650299, + "frames_075_100/lpips": 0.8212679624557495, + "frames_075_100/mse": 0.12308488041162491, + "frames_075_100/psnr": 9.499282836914062, + "frames_075_100/rfid": 215.0, + "frames_075_100/ssim": 0.24835021793842316, + "frames_100_125/lpips": 0.8471609354019165, + "frames_100_125/mse": 0.12503862380981445, + "frames_100_125/psnr": 9.496912002563477, + "frames_100_125/rfid": 221.0, + "frames_100_125/ssim": 0.2582925856113434, + "frames_100_150/lpips": 0.8543499708175659, + "frames_100_150/mse": 0.12679652869701385, + "frames_100_150/psnr": 9.460912704467773, + "frames_100_150/rfid": 210.0, + "frames_100_150/ssim": 0.2583332657814026, + "frames_100_200/lpips": 0.863664448261261, + "frames_100_200/mse": 0.12420104444026947, + "frames_100_200/psnr": 9.533553123474121, + "frames_100_200/rfid": 201.0, + "frames_100_200/ssim": 0.26197361946105957, + "frames_100_300/lpips": 0.8692519664764404, + "frames_100_300/mse": 0.13436025381088257, + "frames_100_300/psnr": 9.201690673828125, + "frames_100_300/rfid": 186.0, + "frames_100_300/ssim": 0.2551032304763794, + "frames_125_150/lpips": 0.861537516117096, + "frames_125_150/mse": 0.12855443358421326, + "frames_125_150/psnr": 9.424912452697754, + "frames_125_150/rfid": 223.0, + "frames_125_150/ssim": 0.2583739459514618, + "frames_150_175/lpips": 0.8728874921798706, + "frames_150_175/mse": 0.12412379682064056, + "frames_150_175/psnr": 9.528512954711914, + "frames_150_175/rfid": 230.0, + "frames_150_175/ssim": 0.2662564218044281, + "frames_150_200/lpips": 0.8729812502861023, + "frames_150_200/mse": 0.12160556763410568, + "frames_150_200/psnr": 9.606195449829102, + "frames_150_200/rfid": 217.0, + "frames_150_200/ssim": 0.2656139135360718, + "frames_175_200/lpips": 0.8730718493461609, + "frames_175_200/mse": 0.1190873384475708, + "frames_175_200/psnr": 9.683878898620605, + "frames_175_200/rfid": 228.0, + "frames_175_200/ssim": 0.26497146487236023, + "frames_200_225/lpips": 0.8735374808311462, + "frames_200_225/mse": 0.13129284977912903, + "frames_200_225/psnr": 9.233220100402832, + "frames_200_225/rfid": 227.0, + "frames_200_225/ssim": 0.2500914931297302, + "frames_200_250/lpips": 0.8743476271629333, + "frames_200_250/mse": 0.14001718163490295, + "frames_200_250/psnr": 8.964224815368652, + "frames_200_250/rfid": 213.0, + "frames_200_250/ssim": 0.24350155889987946, + "frames_225_250/lpips": 0.8751609325408936, + "frames_225_250/mse": 0.14874149858951569, + "frames_225_250/psnr": 8.695231437683105, + "frames_225_250/rfid": 222.0, + "frames_225_250/ssim": 0.2369115948677063, + "frames_250_275/lpips": 0.8750859498977661, + "frames_250_275/mse": 0.1510658860206604, + "frames_250_275/psnr": 8.698421478271484, + "frames_250_275/rfid": 216.0, + "frames_250_275/ssim": 0.24556545913219452, + "frames_250_300/lpips": 0.8753296732902527, + "frames_250_300/mse": 0.14902180433273315, + "frames_250_300/psnr": 8.775429725646973, + "frames_250_300/rfid": 205.0, + "frames_250_300/ssim": 0.2529641091823578, + "frames_275_300/lpips": 0.875573456287384, + "frames_275_300/mse": 0.14697770774364471, + "frames_275_300/psnr": 8.852438926696777, + "frames_275_300/rfid": 217.0, + "frames_275_300/ssim": 0.26036280393600464, + "overall/lpips": 0.782860517501831, + "overall/mse": 0.11977734416723251, + "overall/psnr": 10.238052368164062, + "overall/rfid": 149.0, + "overall/ssim": 0.27807289361953735 + }, + "namespace": "test", + "sampling_steps": 50, + "scene_count": 100, + "schema_version": 1, + "scope": "final100" +} diff --git a/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/metrics/test_history.csv b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/metrics/test_history.csv new file mode 100644 index 0000000000000000000000000000000000000000..465d6577f6c85eb0e7312aaed041d25542a00c16 --- /dev/null +++ b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/metrics/test_history.csv @@ -0,0 +1,2 @@ +method,namespace,global_step,sampling_steps,selected_scenes,generated_frames,overall/mse,overall/psnr,overall/ssim,overall/lpips,overall/rfid,final/mse,final/psnr,final/ssim,final/lpips,frames_100_200/mse,frames_100_200/psnr,frames_100_200/ssim,frames_100_200/lpips,frames_100_200/rfid,frames_000_025/mse,frames_000_025/psnr,frames_000_025/ssim,frames_000_025/lpips,frames_000_025/rfid,frames_025_050/mse,frames_025_050/psnr,frames_025_050/ssim,frames_025_050/lpips,frames_025_050/rfid,frames_050_075/mse,frames_050_075/psnr,frames_050_075/ssim,frames_050_075/lpips,frames_050_075/rfid,frames_075_100/mse,frames_075_100/psnr,frames_075_100/ssim,frames_075_100/lpips,frames_075_100/rfid,frames_100_125/mse,frames_100_125/psnr,frames_100_125/ssim,frames_100_125/lpips,frames_100_125/rfid,frames_125_150/mse,frames_125_150/psnr,frames_125_150/ssim,frames_125_150/lpips,frames_125_150/rfid,frames_150_175/mse,frames_150_175/psnr,frames_150_175/ssim,frames_150_175/lpips,frames_150_175/rfid,frames_175_200/mse,frames_175_200/psnr,frames_175_200/ssim,frames_175_200/lpips,frames_175_200/rfid,frames_200_225/mse,frames_200_225/psnr,frames_200_225/ssim,frames_200_225/lpips,frames_200_225/rfid,frames_225_250/mse,frames_225_250/psnr,frames_225_250/ssim,frames_225_250/lpips,frames_225_250/rfid,frames_250_275/mse,frames_250_275/psnr,frames_250_275/ssim,frames_250_275/lpips,frames_250_275/rfid,frames_275_300/mse,frames_275_300/psnr,frames_275_300/ssim,frames_275_300/lpips,frames_275_300/rfid,frames_000_050/mse,frames_000_050/psnr,frames_000_050/ssim,frames_000_050/lpips,frames_000_050/rfid,frames_050_100/mse,frames_050_100/psnr,frames_050_100/ssim,frames_050_100/lpips,frames_050_100/rfid,frames_100_150/mse,frames_100_150/psnr,frames_100_150/ssim,frames_100_150/lpips,frames_100_150/rfid,frames_150_200/mse,frames_150_200/psnr,frames_150_200/ssim,frames_150_200/lpips,frames_150_200/rfid,frames_200_250/mse,frames_200_250/psnr,frames_200_250/ssim,frames_200_250/lpips,frames_200_250/rfid,frames_250_300/mse,frames_250_300/psnr,frames_250_300/ssim,frames_250_300/lpips,frames_250_300/rfid,frames_000_100/mse,frames_000_100/psnr,frames_000_100/ssim,frames_000_100/lpips,frames_000_100/rfid,frames_100_300/mse,frames_100_300/psnr,frames_100_300/ssim,frames_100_300/lpips,frames_100_300/rfid,frame_bin_size,frame_bin_interval +worldmem,test,0,50,100,30000,0.11977734416723251,10.238052368164062,0.27807289361953735,0.782860517501831,149.0,0.14580927789211273,8.9396390914917,0.254367858171463,0.876171886920929,0.12420104444026947,9.533553123474121,0.26197361946105957,0.863664448261261,201.0,0.025142336264252663,18.989437103271484,0.5077254772186279,0.2293580025434494,56.5,0.09337124973535538,11.09677505493164,0.2858913540840149,0.6143699288368225,138.0,0.120847687125206,9.657604217529297,0.25408196449279785,0.7753156423568726,193.0,0.12308488041162491,9.499282836914062,0.24835021793842316,0.8212679624557495,215.0,0.12503862380981445,9.496912002563477,0.2582925856113434,0.8471609354019165,221.0,0.12855443358421326,9.424912452697754,0.2583739459514618,0.861537516117096,223.0,0.12412379682064056,9.528512954711914,0.2662564218044281,0.8728874921798706,230.0,0.1190873384475708,9.683878898620605,0.26497146487236023,0.8730718493461609,228.0,0.13129284977912903,9.233220100402832,0.2500914931297302,0.8735374808311462,227.0,0.14874149858951569,8.695231437683105,0.2369115948677063,0.8751609325408936,222.0,0.1510658860206604,8.698421478271484,0.24556545913219452,0.8750859498977661,216.0,0.14697770774364471,8.852438926696777,0.26036280393600464,0.875573456287384,217.0,0.059256792068481445,15.043106079101562,0.396808385848999,0.42186474800109863,81.5,0.12196629494428635,9.57844352722168,0.2512160837650299,0.798291802406311,191.0,0.12679652869701385,9.460912704467773,0.2583332657814026,0.8543499708175659,210.0,0.12160556763410568,9.606195449829102,0.2656139135360718,0.8729812502861023,217.0,0.14001718163490295,8.964224815368652,0.24350155889987946,0.8743476271629333,213.0,0.14902180433273315,8.775429725646973,0.2529641091823578,0.8753296732902527,205.0,0.0906115397810936,12.310774803161621,0.32401224970817566,0.6100778579711914,116.0,0.13436025381088257,9.201690673828125,0.2551032304763794,0.8692519664764404,186.0,25,"[start,end) generated-frame offset" diff --git a/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/metrics/test_step00000000.json b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/metrics/test_step00000000.json new file mode 100644 index 0000000000000000000000000000000000000000..492abd6c36917c5416d555ef4e4f5546396cba2e --- /dev/null +++ b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/metrics/test_step00000000.json @@ -0,0 +1,124 @@ +{ + "final/lpips": 0.876171886920929, + "final/mse": 0.14580927789211273, + "final/psnr": 8.9396390914917, + "final/ssim": 0.254367858171463, + "frame_bin_interval": "[start,end) generated-frame offset", + "frame_bin_size": 25, + "frames_000_025/lpips": 0.2293580025434494, + "frames_000_025/mse": 0.025142336264252663, + "frames_000_025/psnr": 18.989437103271484, + "frames_000_025/rfid": 56.5, + "frames_000_025/ssim": 0.5077254772186279, + "frames_000_050/lpips": 0.42186474800109863, + "frames_000_050/mse": 0.059256792068481445, + "frames_000_050/psnr": 15.043106079101562, + "frames_000_050/rfid": 81.5, + "frames_000_050/ssim": 0.396808385848999, + "frames_000_100/lpips": 0.6100778579711914, + "frames_000_100/mse": 0.0906115397810936, + "frames_000_100/psnr": 12.310774803161621, + "frames_000_100/rfid": 116.0, + "frames_000_100/ssim": 0.32401224970817566, + "frames_025_050/lpips": 0.6143699288368225, + "frames_025_050/mse": 0.09337124973535538, + "frames_025_050/psnr": 11.09677505493164, + "frames_025_050/rfid": 138.0, + "frames_025_050/ssim": 0.2858913540840149, + "frames_050_075/lpips": 0.7753156423568726, + "frames_050_075/mse": 0.120847687125206, + "frames_050_075/psnr": 9.657604217529297, + "frames_050_075/rfid": 193.0, + "frames_050_075/ssim": 0.25408196449279785, + "frames_050_100/lpips": 0.798291802406311, + "frames_050_100/mse": 0.12196629494428635, + "frames_050_100/psnr": 9.57844352722168, + "frames_050_100/rfid": 191.0, + "frames_050_100/ssim": 0.2512160837650299, + "frames_075_100/lpips": 0.8212679624557495, + "frames_075_100/mse": 0.12308488041162491, + "frames_075_100/psnr": 9.499282836914062, + "frames_075_100/rfid": 215.0, + "frames_075_100/ssim": 0.24835021793842316, + "frames_100_125/lpips": 0.8471609354019165, + "frames_100_125/mse": 0.12503862380981445, + "frames_100_125/psnr": 9.496912002563477, + "frames_100_125/rfid": 221.0, + "frames_100_125/ssim": 0.2582925856113434, + "frames_100_150/lpips": 0.8543499708175659, + "frames_100_150/mse": 0.12679652869701385, + "frames_100_150/psnr": 9.460912704467773, + "frames_100_150/rfid": 210.0, + "frames_100_150/ssim": 0.2583332657814026, + "frames_100_200/lpips": 0.863664448261261, + "frames_100_200/mse": 0.12420104444026947, + "frames_100_200/psnr": 9.533553123474121, + "frames_100_200/rfid": 201.0, + "frames_100_200/ssim": 0.26197361946105957, + "frames_100_300/lpips": 0.8692519664764404, + "frames_100_300/mse": 0.13436025381088257, + "frames_100_300/psnr": 9.201690673828125, + "frames_100_300/rfid": 186.0, + "frames_100_300/ssim": 0.2551032304763794, + "frames_125_150/lpips": 0.861537516117096, + "frames_125_150/mse": 0.12855443358421326, + "frames_125_150/psnr": 9.424912452697754, + "frames_125_150/rfid": 223.0, + "frames_125_150/ssim": 0.2583739459514618, + "frames_150_175/lpips": 0.8728874921798706, + "frames_150_175/mse": 0.12412379682064056, + "frames_150_175/psnr": 9.528512954711914, + "frames_150_175/rfid": 230.0, + "frames_150_175/ssim": 0.2662564218044281, + "frames_150_200/lpips": 0.8729812502861023, + "frames_150_200/mse": 0.12160556763410568, + "frames_150_200/psnr": 9.606195449829102, + "frames_150_200/rfid": 217.0, + "frames_150_200/ssim": 0.2656139135360718, + "frames_175_200/lpips": 0.8730718493461609, + "frames_175_200/mse": 0.1190873384475708, + "frames_175_200/psnr": 9.683878898620605, + "frames_175_200/rfid": 228.0, + "frames_175_200/ssim": 0.26497146487236023, + "frames_200_225/lpips": 0.8735374808311462, + "frames_200_225/mse": 0.13129284977912903, + "frames_200_225/psnr": 9.233220100402832, + "frames_200_225/rfid": 227.0, + "frames_200_225/ssim": 0.2500914931297302, + "frames_200_250/lpips": 0.8743476271629333, + "frames_200_250/mse": 0.14001718163490295, + "frames_200_250/psnr": 8.964224815368652, + "frames_200_250/rfid": 213.0, + "frames_200_250/ssim": 0.24350155889987946, + "frames_225_250/lpips": 0.8751609325408936, + "frames_225_250/mse": 0.14874149858951569, + "frames_225_250/psnr": 8.695231437683105, + "frames_225_250/rfid": 222.0, + "frames_225_250/ssim": 0.2369115948677063, + "frames_250_275/lpips": 0.8750859498977661, + "frames_250_275/mse": 0.1510658860206604, + "frames_250_275/psnr": 8.698421478271484, + "frames_250_275/rfid": 216.0, + "frames_250_275/ssim": 0.24556545913219452, + "frames_250_300/lpips": 0.8753296732902527, + "frames_250_300/mse": 0.14902180433273315, + "frames_250_300/psnr": 8.775429725646973, + "frames_250_300/rfid": 205.0, + "frames_250_300/ssim": 0.2529641091823578, + "frames_275_300/lpips": 0.875573456287384, + "frames_275_300/mse": 0.14697770774364471, + "frames_275_300/psnr": 8.852438926696777, + "frames_275_300/rfid": 217.0, + "frames_275_300/ssim": 0.26036280393600464, + "generated_frames": 30000, + "global_step": 0, + "method": "worldmem", + "namespace": "test", + "overall/lpips": 0.782860517501831, + "overall/mse": 0.11977734416723251, + "overall/psnr": 10.238052368164062, + "overall/rfid": 149.0, + "overall/ssim": 0.27807289361953735, + "sampling_steps": 50, + "selected_scenes": 100 +} diff --git a/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/resolved_config.yaml b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/resolved_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d8f6cdb6127f2ce168c0d85be5c7a5a0f25811fb --- /dev/null +++ b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/resolved_config.yaml @@ -0,0 +1,403 @@ +experiment: + debug: false + tasks: + - training + num_nodes: 1 + training: + precision: bf16-mixed + compile: false + lr: 4.0e-05 + batch_size: 4 + max_epochs: -1 + max_steps: 100000 + max_time: null + data: + num_workers: 4 + shuffle: true + optim: + accumulate_grad_batches: 1 + gradient_clip_val: 1.0 + checkpointing: + every_n_train_steps: 2500 + every_n_epochs: null + train_time_interval: null + enable_version_counter: false + validation: + precision: bf16-mixed + compile: false + batch_size: 1 + val_every_n_step: 2500 + val_every_n_epoch: null + num_sanity_val_steps: 0 + limit_batch: 2 + data: + num_workers: 4 + shuffle: false + inference_mode: false + test: + precision: 16-mixed + compile: false + batch_size: 1 + limit_batch: 1.0 + data: + num_workers: 4 + shuffle: false + inference_mode: true + find_unused_parameters: false + reload_dataloaders_every_n_epochs: 1 + ema: + enable: false + decay: 0.9999 + validate_original_weights: false + _name: video_generation +dataset: + debug: false + save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/re10k_dfot/real-estate-10k + data_mean: + - - - 0.577 + - - - 0.517 + - - - 0.461 + data_std: + - - - 0.249 + - - - 0.249 + - - - 0.268 + latent: + enable: false + type: pre_sample + suffix: null + downsampling_factor: + - 1 + - 8 + num_channels: 4 + resolution: 256 + observation_shape: + - 3 + - 256 + - 256 + max_frames: 12 + n_frames: 12 + context_length: 10 + frame_skip: 1 + filter_min_len: null + external_cond_dim: 16 + external_cond_stack: false + external_cond_processing: null + preload: false + subdataset_size: null + num_eval_videos: 100 + maximize_training_data: false + augmentation: + frame_skip_increase: 0 + horizontal_flip_prob: 0 + reverse_prob: 0 + back_and_forth_prob: 0 + local_window_frames: 8 + training_frame_skip: 1 + training_history_frames: 10 + eval_history_frames: 100 + eval_min_generated_frames: 300 + eval_max_generated_frames: 300 + video_fps: 10.0 + seed: 0 + memory_selection: + enabled: true + selector_version: re10k_faithful_trajectory_h100_f399_v1 + method: worldmem + training_selection_mode: schedule_cache_direct + selector_cache_dir: /proj/cvl/users/x_fahkh2/DeMemWM/.exp_artifact/re10k_faithful_trajectory_h100_f399_v1/cache_worldmem_clipped_731edd42758e + selector_cache_source_commit: 731edd42758e7aec3b0bfd03ed52b72d716b65df + causal: true + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 0 + training_dropout: 0.0 + local_context_exclusion_frames: 8 + match_position_scale: 2.0 + match_view_angle_degrees: 110.0 + evaluation_manifest: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/re10k_real_trajectory_h100_f399_scopes_v1.csv + evaluation_manifest_scope: final100 + evaluation_benchmark: loop_true_reverse_h100_f399_v1 + evaluation_benchmarks: + - loop_true_reverse_h100_f399_v1 + evaluation_memory_mode: correct + evaluation_case_counts: + loop_true_reverse_h100_f399_v1: 100 + evaluation_case_order: scene_manifest_then_path_ordinal + evaluation_case_seed: 1000003*metadata_index+path_ordinal + evaluation_trajectory: G0..G199,G199..G0 + evaluation_output_frames: 400 + evaluation_observed_output_interval: + - 0 + - 100 + evaluation_generated_output_interval: + - 100 + - 400 + evaluation_generated_segments: + forward: + - 0 + - 100 + return: + - 100 + - 300 + return_chunks: + - - 100 + - 150 + - - 150 + - 200 + - - 200 + - 250 + - - 250 + - 300 + evaluation_frame_bin_size: 25 + evaluation_turnaround_duplication: true + _name: video_realestate10k +algorithm: + debug: false + lr: 4.0e-05 + backbone: + name: u_vit3d_pose + channels: + - 128 + - 256 + - 576 + - 1152 + emb_channels: 1024 + patch_size: 2 + block_types: + - ResBlock + - ResBlock + - TransformerBlock + - TransformerBlock + block_dropouts: + - 0.0 + - 0.0 + - 0.1 + - 0.1 + num_updown_blocks: + - 3 + - 3 + - 6 + num_mid_blocks: 20 + num_heads: 9 + pos_emb_type: rope + use_checkpointing: + - false + - false + - false + - false + conditioning: + dim: null + external_cond_dropout: 0.1 + use_fourier_noise_embedding: true + x_shape: + - 3 + - 256 + - 256 + max_frames: 12 + n_frames: 12 + frame_skip: 1 + context_frames: 10 + latent: + enable: false + type: pre_sample + suffix: null + downsampling_factor: + - 1 + - 8 + num_channels: 4 + data_mean: + - - - 0.577 + - - - 0.517 + - - - 0.461 + data_std: + - - - 0.249 + - - - 0.249 + - - - 0.268 + external_cond_dim: 16 + external_cond_stack: false + external_cond_processing: null + compile: false + weight_decay: 0.002 + optimizer_beta: + - 0.9 + - 0.99 + lr_scheduler: + name: constant_with_warmup + num_warmup_steps: 1000 + num_training_steps: 100000 + noise_level: random_independent + uniform_future: + enabled: false + fixed_context: + enabled: false + indices: null + dropout: 0 + variable_context: + enabled: false + prob: 0 + dropout: 0 + chunk_size: -1 + scheduling_matrix: full_sequence + replacement: noisy_scale + diffusion: + is_continuous: true + timesteps: 1000 + beta_schedule: cosine_simple_diffusion + schedule_fn_kwargs: + shift: 1.0 + shifted: 0.125 + interpolated: false + use_causal_mask: false + clip_noise: 20.0 + objective: pred_v + loss_weighting: + strategy: sigmoid + snr_clip: 5.0 + cum_snr_decay: 0.9 + sigmoid_bias: -1.0 + sampling_timesteps: 50 + ddim_sampling_eta: 0.0 + reconstruction_guidance: 0.0 + training_schedule: + name: cosine + shift: 0.125 + precond_scale: 0.125 + vae: + pretrained_path: null + pretrained_kwargs: {} + use_fp16: true + batch_size: 2 + checkpoint: + reset_optimizer: false + strict: true + tasks: + prediction: + enabled: false + history_guidance: + name: conditional + keyframe_density: null + sliding_context_len: null + interpolation: + enabled: false + history_guidance: + name: conditional + max_batch_size: null + logging: + deterministic: 0 + loss_freq: 1 + grad_norm_freq: 0 + max_num_videos: 0 + n_metrics_frames: null + metrics: [] + metrics_batch_size: 16 + sanity_generation: false + raw_dir: null + camera_pose_conditioning: + normalize_by: first + bound: null + type: ray_encoding + memory: + local_window_frames: 8 + memory_selection: + enabled: true + selector_version: re10k_faithful_trajectory_h100_f399_v1 + method: worldmem + training_selection_mode: schedule_cache_direct + selector_cache_dir: /proj/cvl/users/x_fahkh2/DeMemWM/.exp_artifact/re10k_faithful_trajectory_h100_f399_v1/cache_worldmem_clipped_731edd42758e + selector_cache_source_commit: 731edd42758e7aec3b0bfd03ed52b72d716b65df + causal: true + max_anchor_frames: 0 + max_dynamic_frames: 4 + max_revisit_frames: 0 + training_dropout: 0.0 + local_context_exclusion_frames: 8 + match_position_scale: 2.0 + match_view_angle_degrees: 110.0 + memory_projection_mode: worldmem_reference_v1 + memory_ffn_ratio: 2 + memory_conditioning_input_dim: 1024 + memory_conditioning_hidden_dim: 128 + memory_role_embedding: false + memory_source_embedding: false + memory_stream_count_correction: false + memory_lora_rank: 8 + memory_lora_alpha: 8 + memory_output_gate_init: 0.0 + memory_camera_hidden_dim: 128 + memory_anchor_zero_geometry: false + memory_age_conditioning: false + reference_timestamp_conditioning: true + memory_attention_key_only_geometry: false + memory_noise: + enabled: false + training_seed: 0 + training: + max_steps: 100000 + num_gpus: 8 + batch_size_per_gpu: 4 + accumulate_grad_batches: 1 + learning_rate: 4.0e-05 + memory_learning_rate_stage_a: 4.0e-05 + memory_learning_rate_stage_b: 2.0e-05 + lora_learning_rate_stage_b: 5.0e-06 + stage_b_start_step: 8000 + stage_c_start_step: 30000 + min_lr_ratio: 0.1 + optimizer_betas: + - 0.9 + - 0.99 + weight_decay: 0.002 + gradient_clip_val: 1.0 + warmup_steps: 1000 + ema_decay: 0.9999 + frame_skip: 1 + checkpoint_every_n_steps: 2500 + validation_every_n_steps: 2500 + monitoring_steps: + - 2500 + - 5000 + - 7500 + - 10000 + - 12500 + - 15000 + - 20000 + - 25000 + - 30000 + - 35000 + - 40000 + - 50000 + - 60000 + - 70000 + - 80000 + - 90000 + - 100000 + evaluation: + selector_mode: online + benchmarks: + - loop_true_reverse_h100_f399_v1 + benchmark: loop_true_reverse_h100_f399_v1 + history_frames: 100 + min_generated_frames: 300 + max_generated_frames: 300 + frame_skip: 1 + sampling_steps: 50 + deterministic_seed: 0 + history_guidance: + name: conditional + reverse_loop_probe: false + turnaround_duplication: true + frame_bin_size: 25 + num_videos: 100 + memory_mode: correct + manifest_scope: final100 + active_benchmarks: + - loop_true_reverse_h100_f399_v1 + _name: dfot_worldmem_realestate10k +debug: false +wandb: + entity: turlin + project: worldmem + mode: offline +resume: null +load: null +name: dfot_worldmem_realestate10k diff --git a/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/run_manifest.json b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/run_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..45c2ecd49c721b629d0c901883ebdea6d72a18bf --- /dev/null +++ b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/run_manifest.json @@ -0,0 +1,189 @@ +{ + "method": "worldmem", + "dataset_root": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/re10k_dfot/real-estate-10k", + "dfot_root": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/DFoT-code", + "dfot_commit": "530f8bf4db91a21964993f214fa4050dbd529591", + "checkpoint": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/DFoT_RE10K.ckpt", + "local_frames": 8, + "memory_frames": { + "anchor": 0, + "dynamic": 4, + "revisit": 0 + }, + "training_selector_version": "re10k_faithful_trajectory_h100_f399_v1", + "training_selector_method": "worldmem", + "training_selector_cache_source_commit": "731edd42758e7aec3b0bfd03ed52b72d716b65df", + "memory_output_gate_init": 0.0, + "batch_size_per_gpu": 4, + "accumulate_grad_batches": 1, + "effective_global_batch_size": 32, + "optimizer_lr_schedule": { + "name": "stagewise_warmup_stable_cosine_floor", + "warmup_steps": 1000, + "stage_b_start_step": 8000, + "decay_start_step": 30000, + "terminal_step": 100000, + "min_lr_ratio": 0.1 + }, + "runtime": { + "gradient_checkpointing": [ + false, + false, + false, + false + ], + "split_frozen_memory_stream": false, + "structural_memory_autograd": true + }, + "wandb": { + "mode": "offline", + "entity": "turlin", + "project": "worldmem", + "run_id": null + }, + "evaluation_history_frames": 100, + "evaluation_max_generated_frames": 300, + "evaluation_num_videos": 100, + "evaluation_sampling_steps": 50, + "evaluation_metrics": [ + "mse", + "psnr", + "ssim", + "lpips", + "rfid" + ], + "training_time_eval_batches_per_rank": 2, + "post_training_evaluation": true, + "faithful_profile": { + "method": "worldmem", + "capacity": "clipped", + "selector_version": "re10k_faithful_trajectory_h100_f399_v1", + "lifecycle": "group1_59306102" + }, + "trainable_parameter_manifest": "/proj/cvl/users/x_fahkh2/DeMemWM/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/trainable_parameter_manifest.json", + "training_batch_composition": { + "global_natural": 8, + "global_loop": 24, + "rank_order": [ + "natural", + "loop", + "loop", + "loop" + ] + }, + "training_trajectory": { + "source_mapping": "physical_start+[0..199,199..0]", + "virtual_positions": 400, + "query_frames": 8, + "phase_bounds_half_open": [ + [ + 100, + 159 + ], + [ + 159, + 218 + ], + [ + 218, + 277 + ], + [ + 277, + 335 + ], + [ + 335, + 393 + ] + ] + }, + "evaluation_weight_source": "ema", + "post_training_checkpoint_policy": null, + "worldmem_git": { + "branch": "re10k-dfot-integration", + "head": "73c1cc4f67879627494f7f714c9e97ceb3c91c55", + "status": [ + "?? results/" + ] + }, + "profile_stage_b": false, + "schedule_step_offset": 0, + "algorithm_config": "/proj/cvl/users/x_fahkh2/DeMemWM/configurations/algorithm/dfot_worldmem_faithful_f200b200_clipped_realestate10k.yaml", + "training_seed": 0, + "evaluation_seed": 0, + "training_seed_label": "seed0", + "checkpoint_kind": "ema-v2", + "evaluation_checkpoint": "/proj/cvl/users/x_fahkh2/DeMemWM/results/dfot_worldmem_clipped_f200b200_seed0_g32_100k_731edd42758e/checkpoints/step=30000.ckpt", + "evaluation_manifest": "/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/re10k_real_trajectory_h100_f399_scopes_v1.csv", + "evaluation_manifest_scope": "final100", + "evaluation_benchmark": "loop_true_reverse_h100_f399_v1", + "evaluation_benchmarks": [ + "loop_true_reverse_h100_f399_v1" + ], + "evaluation_memory_mode": "correct", + "evaluation_case_counts": { + "loop_true_reverse_h100_f399_v1": 100 + }, + "evaluation_case_order": "scene_manifest_then_path_ordinal", + "evaluation_case_seed": "1000003*metadata_index+path_ordinal", + "evaluation_trajectory": "G0..G199,G199..G0", + "evaluation_output_frames": 400, + "evaluation_observed_output_interval": [ + 0, + 100 + ], + "evaluation_generated_output_interval": [ + 100, + 400 + ], + "evaluation_generated_segments": { + "forward": [ + 0, + 100 + ], + "return": [ + 100, + 300 + ], + "return_chunks": [ + [ + 100, + 150 + ], + [ + 150, + 200 + ], + [ + 200, + 250 + ], + [ + 250, + 300 + ] + ] + }, + "evaluation_frame_bin_size": 25, + "evaluation_turnaround_duplication": true, + "monitoring_steps": [ + 2500, + 5000, + 7500, + 10000, + 12500, + 15000, + 20000, + 25000, + 30000, + 35000, + 40000, + 50000, + 60000, + 70000, + 80000, + 90000, + 100000 + ] +} diff --git a/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/test_loop_true_reverse_h100_f399_v1_case_manifest.json b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/test_loop_true_reverse_h100_f399_v1_case_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..64360cadd45961db8a55f27e48dbe3d4950c0ad7 --- /dev/null +++ b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/test_loop_true_reverse_h100_f399_v1_case_manifest.json @@ -0,0 +1,1316 @@ +{ + "split": "test", + "evaluation_seed": 0, + "manifest": "/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/re10k_real_trajectory_h100_f399_scopes_v1.csv", + "manifest_scope": "final100", + "history_frames": 100, + "benchmark": "loop_true_reverse_h100_f399_v1", + "scene_count": 100, + "scene_order": [ + { + "metadata_index": 2225, + "stem": "516faacb7687d0c7" + }, + { + "metadata_index": 1894, + "stem": "45536907ffef7585" + }, + { + "metadata_index": 720, + "stem": "1a55c677aabd1a44" + }, + { + "metadata_index": 1618, + "stem": "3b122e1becb5fcb7" + }, + { + "metadata_index": 4564, + "stem": "a4e9bdb6d746eb0a" + }, + { + "metadata_index": 77, + "stem": "02b618a34bc12ff9" + }, + { + "metadata_index": 6655, + "stem": "efa64a0035040381" + }, + { + "metadata_index": 120, + "stem": "0425df3e42ba0de3" + }, + { + "metadata_index": 6594, + "stem": "ed9cff668e7377be" + }, + { + "metadata_index": 3149, + "stem": "72c8c1a0b4d0a5da" + }, + { + "metadata_index": 1089, + "stem": "28760a14d0a5ff3a" + }, + { + "metadata_index": 5667, + "stem": "cc014d7be0dd9a2f" + }, + { + "metadata_index": 6069, + "stem": "dae13e63d6107502" + }, + { + "metadata_index": 2351, + "stem": "564659475de45eab" + }, + { + "metadata_index": 5981, + "stem": "d84343c25315d7e9" + }, + { + "metadata_index": 4327, + "stem": "9d377a5dcd18edad" + }, + { + "metadata_index": 2287, + "stem": "537b6025597bfa35" + }, + { + "metadata_index": 4713, + "stem": "aa4a4768c8635fde" + }, + { + "metadata_index": 4103, + "stem": "94d5e6280f699e08" + }, + { + "metadata_index": 156, + "stem": "05a6149f1fcee38d" + }, + { + "metadata_index": 6912, + "stem": "f815dfc42d162fea" + }, + { + "metadata_index": 311, + "stem": "0bcde26e5a802638" + }, + { + "metadata_index": 5730, + "stem": "ce5cb13f40a259e3" + }, + { + "metadata_index": 834, + "stem": "1e1e13de4ebea05a" + }, + { + "metadata_index": 6236, + "stem": "e16e616763c90f55" + }, + { + "metadata_index": 5484, + "stem": "c551b85ce38402a8" + }, + { + "metadata_index": 5767, + "stem": "cf845955d579f84c" + }, + { + "metadata_index": 3992, + "stem": "904a0f9fa6651f9f" + }, + { + "metadata_index": 2120, + "stem": "4dd400119f6032b7" + }, + { + "metadata_index": 3355, + "stem": "7a18ec7dc0bdf907" + }, + { + "metadata_index": 1839, + "stem": "431e6542fde13130" + }, + { + "metadata_index": 2394, + "stem": "579983b1a3881a69" + }, + { + "metadata_index": 6869, + "stem": "f6b24019e753e366" + }, + { + "metadata_index": 523, + "stem": "127884736471b631" + }, + { + "metadata_index": 2336, + "stem": "55ca776696247564" + }, + { + "metadata_index": 1496, + "stem": "36cb04da872d3bce" + }, + { + "metadata_index": 5335, + "stem": "c0b4b9d69f8694e1" + }, + { + "metadata_index": 3903, + "stem": "8c8ed2adba3dad7a" + }, + { + "metadata_index": 5406, + "stem": "c30ce1a9890c8bdb" + }, + { + "metadata_index": 199, + "stem": "072a60bc7e0b0dfd" + }, + { + "metadata_index": 6908, + "stem": "f7f7028b9db4277e" + }, + { + "metadata_index": 4668, + "stem": "a8e53509b42b91e1" + }, + { + "metadata_index": 5380, + "stem": "c211e68f8e67eccf" + }, + { + "metadata_index": 7061, + "stem": "fcfa188fdd8e4cdc" + }, + { + "metadata_index": 6956, + "stem": "f9e4896193305279" + }, + { + "metadata_index": 2533, + "stem": "5c30afe4faa45992" + }, + { + "metadata_index": 6675, + "stem": "f088c85322065722" + }, + { + "metadata_index": 1770, + "stem": "4091c41c6909da3b" + }, + { + "metadata_index": 1250, + "stem": "2deacec5c281fbac" + }, + { + "metadata_index": 4812, + "stem": "ade3e283780efbd7" + }, + { + "metadata_index": 4678, + "stem": "a9228c6ca6619266" + }, + { + "metadata_index": 3973, + "stem": "8f71898bc3662fa6" + }, + { + "metadata_index": 4057, + "stem": "929ed71c97317c55" + }, + { + "metadata_index": 1995, + "stem": "49235d402cbb8895" + }, + { + "metadata_index": 2405, + "stem": "57ed3518c548f1d9" + }, + { + "metadata_index": 3770, + "stem": "885c1a66b4025d94" + }, + { + "metadata_index": 2563, + "stem": "5d3888256b80fb8d" + }, + { + "metadata_index": 5105, + "stem": "b8605d7a56bada1c" + }, + { + "metadata_index": 2470, + "stem": "599619a790d2b77d" + }, + { + "metadata_index": 4115, + "stem": "952f6cb94695d32d" + }, + { + "metadata_index": 6785, + "stem": "f3de74db05a6b558" + }, + { + "metadata_index": 5387, + "stem": "c251d6e68d6d89eb" + }, + { + "metadata_index": 4030, + "stem": "91ca67e0e542a035" + }, + { + "metadata_index": 3848, + "stem": "8ad61ecb52a9122a" + }, + { + "metadata_index": 6586, + "stem": "ed7a4d97bab47c6d" + }, + { + "metadata_index": 498, + "stem": "118ffef2ad3950f5" + }, + { + "metadata_index": 26, + "stem": "0122933cf8ab3317" + }, + { + "metadata_index": 5169, + "stem": "bb0d9ff59505b17e" + }, + { + "metadata_index": 4993, + "stem": "b4c2c2cb9d279a99" + }, + { + "metadata_index": 6381, + "stem": "e6f8497a860c873c" + }, + { + "metadata_index": 2308, + "stem": "54a1bceb5d97272d" + }, + { + "metadata_index": 2987, + "stem": "6c376c523531fd19" + }, + { + "metadata_index": 2320, + "stem": "5535e78a16298072" + }, + { + "metadata_index": 3128, + "stem": "7215c47789bfed66" + }, + { + "metadata_index": 6845, + "stem": "f5a110214805f3f1" + }, + { + "metadata_index": 1026, + "stem": "25d8e374a2dafc8a" + }, + { + "metadata_index": 4441, + "stem": "a0bc9b509471e1dc" + }, + { + "metadata_index": 5905, + "stem": "d5409c93b6f88256" + }, + { + "metadata_index": 2215, + "stem": "512810bd5785cad6" + }, + { + "metadata_index": 3875, + "stem": "8ba67e66fe6cf9d8" + }, + { + "metadata_index": 677, + "stem": "186f18684ed4b516" + }, + { + "metadata_index": 6477, + "stem": "ea11822d0ab31414" + }, + { + "metadata_index": 913, + "stem": "20fba1d53c349851" + }, + { + "metadata_index": 2124, + "stem": "4df04a3ab680ac1d" + }, + { + "metadata_index": 6719, + "stem": "f20a49fc09b9ee6e" + }, + { + "metadata_index": 2483, + "stem": "5a2c2af8cf1765b2" + }, + { + "metadata_index": 6513, + "stem": "eb5b73ca4a18bbf9" + }, + { + "metadata_index": 4300, + "stem": "9c1398cb9ad44a17" + }, + { + "metadata_index": 2135, + "stem": "4e50d5610be16ce1" + }, + { + "metadata_index": 1585, + "stem": "39c662fa32a3b5c6" + }, + { + "metadata_index": 3894, + "stem": "8c3cf8cab7fcc03e" + }, + { + "metadata_index": 2564, + "stem": "5d3c7dd9db5e495a" + }, + { + "metadata_index": 5610, + "stem": "ca19b5ba41e05630" + }, + { + "metadata_index": 3230, + "stem": "756c2bdfad5cdd55" + }, + { + "metadata_index": 1128, + "stem": "2a1769dddc1dbf8d" + }, + { + "metadata_index": 6354, + "stem": "e60589d40eda6049" + }, + { + "metadata_index": 4634, + "stem": "a790bd601037ea91" + }, + { + "metadata_index": 794, + "stem": "1c9772d765e0679e" + }, + { + "metadata_index": 6418, + "stem": "e8312620692723b3" + }, + { + "metadata_index": 1459, + "stem": "3558c5283169e6ca" + } + ], + "case_count": 100, + "case_order": "scene_manifest_then_path_ordinal", + "total_generated_frames": 30000, + "cases": [ + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2225, + "scene_id": "516faacb7687d0c7", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2225006682 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1894, + "scene_id": "45536907ffef7585", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1894005689 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 720, + "scene_id": "1a55c677aabd1a44", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 720002167 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1618, + "scene_id": "3b122e1becb5fcb7", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1618004861 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4564, + "scene_id": "a4e9bdb6d746eb0a", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4564013699 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 77, + "scene_id": "02b618a34bc12ff9", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 77000238 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6655, + "scene_id": "efa64a0035040381", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6655019972 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 120, + "scene_id": "0425df3e42ba0de3", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 120000367 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6594, + "scene_id": "ed9cff668e7377be", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6594019789 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3149, + "scene_id": "72c8c1a0b4d0a5da", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3149009454 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1089, + "scene_id": "28760a14d0a5ff3a", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1089003274 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5667, + "scene_id": "cc014d7be0dd9a2f", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5667017008 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6069, + "scene_id": "dae13e63d6107502", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6069018214 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2351, + "scene_id": "564659475de45eab", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2351007060 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5981, + "scene_id": "d84343c25315d7e9", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5981017950 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4327, + "scene_id": "9d377a5dcd18edad", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4327012988 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2287, + "scene_id": "537b6025597bfa35", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2287006868 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4713, + "scene_id": "aa4a4768c8635fde", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4713014146 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4103, + "scene_id": "94d5e6280f699e08", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4103012316 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 156, + "scene_id": "05a6149f1fcee38d", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 156000475 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6912, + "scene_id": "f815dfc42d162fea", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6912020743 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 311, + "scene_id": "0bcde26e5a802638", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 311000940 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5730, + "scene_id": "ce5cb13f40a259e3", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5730017197 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 834, + "scene_id": "1e1e13de4ebea05a", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 834002509 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6236, + "scene_id": "e16e616763c90f55", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6236018715 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5484, + "scene_id": "c551b85ce38402a8", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5484016459 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5767, + "scene_id": "cf845955d579f84c", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5767017308 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3992, + "scene_id": "904a0f9fa6651f9f", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3992011983 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2120, + "scene_id": "4dd400119f6032b7", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2120006367 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3355, + "scene_id": "7a18ec7dc0bdf907", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3355010072 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1839, + "scene_id": "431e6542fde13130", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1839005524 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2394, + "scene_id": "579983b1a3881a69", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2394007189 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6869, + "scene_id": "f6b24019e753e366", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6869020614 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 523, + "scene_id": "127884736471b631", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 523001576 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2336, + "scene_id": "55ca776696247564", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2336007015 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1496, + "scene_id": "36cb04da872d3bce", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1496004495 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5335, + "scene_id": "c0b4b9d69f8694e1", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5335016012 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3903, + "scene_id": "8c8ed2adba3dad7a", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3903011716 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5406, + "scene_id": "c30ce1a9890c8bdb", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5406016225 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 199, + "scene_id": "072a60bc7e0b0dfd", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 199000604 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6908, + "scene_id": "f7f7028b9db4277e", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6908020731 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4668, + "scene_id": "a8e53509b42b91e1", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4668014011 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5380, + "scene_id": "c211e68f8e67eccf", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5380016147 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 7061, + "scene_id": "fcfa188fdd8e4cdc", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 7061021190 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6956, + "scene_id": "f9e4896193305279", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6956020875 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2533, + "scene_id": "5c30afe4faa45992", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2533007606 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6675, + "scene_id": "f088c85322065722", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6675020032 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1770, + "scene_id": "4091c41c6909da3b", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1770005317 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1250, + "scene_id": "2deacec5c281fbac", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1250003757 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4812, + "scene_id": "ade3e283780efbd7", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4812014443 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4678, + "scene_id": "a9228c6ca6619266", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4678014041 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3973, + "scene_id": "8f71898bc3662fa6", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3973011926 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4057, + "scene_id": "929ed71c97317c55", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4057012178 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1995, + "scene_id": "49235d402cbb8895", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1995005992 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2405, + "scene_id": "57ed3518c548f1d9", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2405007222 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3770, + "scene_id": "885c1a66b4025d94", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3770011317 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2563, + "scene_id": "5d3888256b80fb8d", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2563007696 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5105, + "scene_id": "b8605d7a56bada1c", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5105015322 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2470, + "scene_id": "599619a790d2b77d", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2470007417 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4115, + "scene_id": "952f6cb94695d32d", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4115012352 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6785, + "scene_id": "f3de74db05a6b558", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6785020362 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5387, + "scene_id": "c251d6e68d6d89eb", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5387016168 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4030, + "scene_id": "91ca67e0e542a035", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4030012097 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3848, + "scene_id": "8ad61ecb52a9122a", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3848011551 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6586, + "scene_id": "ed7a4d97bab47c6d", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6586019765 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 498, + "scene_id": "118ffef2ad3950f5", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 498001501 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 26, + "scene_id": "0122933cf8ab3317", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 26000085 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5169, + "scene_id": "bb0d9ff59505b17e", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5169015514 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4993, + "scene_id": "b4c2c2cb9d279a99", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4993014986 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6381, + "scene_id": "e6f8497a860c873c", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6381019150 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2308, + "scene_id": "54a1bceb5d97272d", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2308006931 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2987, + "scene_id": "6c376c523531fd19", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2987008968 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2320, + "scene_id": "5535e78a16298072", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2320006967 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3128, + "scene_id": "7215c47789bfed66", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3128009391 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6845, + "scene_id": "f5a110214805f3f1", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6845020542 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1026, + "scene_id": "25d8e374a2dafc8a", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1026003085 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4441, + "scene_id": "a0bc9b509471e1dc", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4441013330 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5905, + "scene_id": "d5409c93b6f88256", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5905017722 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2215, + "scene_id": "512810bd5785cad6", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2215006652 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3875, + "scene_id": "8ba67e66fe6cf9d8", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3875011632 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 677, + "scene_id": "186f18684ed4b516", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 677002038 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6477, + "scene_id": "ea11822d0ab31414", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6477019438 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 913, + "scene_id": "20fba1d53c349851", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 913002746 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2124, + "scene_id": "4df04a3ab680ac1d", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2124006379 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6719, + "scene_id": "f20a49fc09b9ee6e", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6719020164 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2483, + "scene_id": "5a2c2af8cf1765b2", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2483007456 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6513, + "scene_id": "eb5b73ca4a18bbf9", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6513019546 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4300, + "scene_id": "9c1398cb9ad44a17", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4300012907 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2135, + "scene_id": "4e50d5610be16ce1", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2135006412 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1585, + "scene_id": "39c662fa32a3b5c6", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1585004762 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3894, + "scene_id": "8c3cf8cab7fcc03e", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3894011689 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 2564, + "scene_id": "5d3c7dd9db5e495a", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 2564007699 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 5610, + "scene_id": "ca19b5ba41e05630", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 5610016837 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 3230, + "scene_id": "756c2bdfad5cdd55", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 3230009697 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1128, + "scene_id": "2a1769dddc1dbf8d", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1128003391 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6354, + "scene_id": "e60589d40eda6049", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6354019069 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 4634, + "scene_id": "a790bd601037ea91", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 4634013909 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 794, + "scene_id": "1c9772d765e0679e", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 794002389 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 6418, + "scene_id": "e8312620692723b3", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 6418019261 + }, + { + "benchmark": "loop_true_reverse_h100_f399_v1", + "metadata_index": 1459, + "scene_id": "3558c5283169e6ca", + "path_name": "true_trajectory_g0_g199_g199_g0", + "path_ordinal": 7, + "generated_frames": 300, + "case_seed": 1459004384 + } + ] +} diff --git a/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/test_scene_manifest.json b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/test_scene_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..6a90c488693e461a50ee682ecc6375e2209078f7 --- /dev/null +++ b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/test_scene_manifest.json @@ -0,0 +1,410 @@ +{ + "split": "test", + "evaluation_seed": 0, + "manifest": "/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/re10k_real_trajectory_h100_f399_scopes_v1.csv", + "manifest_scope": "final100", + "history_frames": 100, + "scene_count": 100, + "scenes": [ + { + "metadata_index": 2225, + "stem": "516faacb7687d0c7" + }, + { + "metadata_index": 1894, + "stem": "45536907ffef7585" + }, + { + "metadata_index": 720, + "stem": "1a55c677aabd1a44" + }, + { + "metadata_index": 1618, + "stem": "3b122e1becb5fcb7" + }, + { + "metadata_index": 4564, + "stem": "a4e9bdb6d746eb0a" + }, + { + "metadata_index": 77, + "stem": "02b618a34bc12ff9" + }, + { + "metadata_index": 6655, + "stem": "efa64a0035040381" + }, + { + "metadata_index": 120, + "stem": "0425df3e42ba0de3" + }, + { + "metadata_index": 6594, + "stem": "ed9cff668e7377be" + }, + { + "metadata_index": 3149, + "stem": "72c8c1a0b4d0a5da" + }, + { + "metadata_index": 1089, + "stem": "28760a14d0a5ff3a" + }, + { + "metadata_index": 5667, + "stem": "cc014d7be0dd9a2f" + }, + { + "metadata_index": 6069, + "stem": "dae13e63d6107502" + }, + { + "metadata_index": 2351, + "stem": "564659475de45eab" + }, + { + "metadata_index": 5981, + "stem": "d84343c25315d7e9" + }, + { + "metadata_index": 4327, + "stem": "9d377a5dcd18edad" + }, + { + "metadata_index": 2287, + "stem": "537b6025597bfa35" + }, + { + "metadata_index": 4713, + "stem": "aa4a4768c8635fde" + }, + { + "metadata_index": 4103, + "stem": "94d5e6280f699e08" + }, + { + "metadata_index": 156, + "stem": "05a6149f1fcee38d" + }, + { + "metadata_index": 6912, + "stem": "f815dfc42d162fea" + }, + { + "metadata_index": 311, + "stem": "0bcde26e5a802638" + }, + { + "metadata_index": 5730, + "stem": "ce5cb13f40a259e3" + }, + { + "metadata_index": 834, + "stem": "1e1e13de4ebea05a" + }, + { + "metadata_index": 6236, + "stem": "e16e616763c90f55" + }, + { + "metadata_index": 5484, + "stem": "c551b85ce38402a8" + }, + { + "metadata_index": 5767, + "stem": "cf845955d579f84c" + }, + { + "metadata_index": 3992, + "stem": "904a0f9fa6651f9f" + }, + { + "metadata_index": 2120, + "stem": "4dd400119f6032b7" + }, + { + "metadata_index": 3355, + "stem": "7a18ec7dc0bdf907" + }, + { + "metadata_index": 1839, + "stem": "431e6542fde13130" + }, + { + "metadata_index": 2394, + "stem": "579983b1a3881a69" + }, + { + "metadata_index": 6869, + "stem": "f6b24019e753e366" + }, + { + "metadata_index": 523, + "stem": "127884736471b631" + }, + { + "metadata_index": 2336, + "stem": "55ca776696247564" + }, + { + "metadata_index": 1496, + "stem": "36cb04da872d3bce" + }, + { + "metadata_index": 5335, + "stem": "c0b4b9d69f8694e1" + }, + { + "metadata_index": 3903, + "stem": "8c8ed2adba3dad7a" + }, + { + "metadata_index": 5406, + "stem": "c30ce1a9890c8bdb" + }, + { + "metadata_index": 199, + "stem": "072a60bc7e0b0dfd" + }, + { + "metadata_index": 6908, + "stem": "f7f7028b9db4277e" + }, + { + "metadata_index": 4668, + "stem": "a8e53509b42b91e1" + }, + { + "metadata_index": 5380, + "stem": "c211e68f8e67eccf" + }, + { + "metadata_index": 7061, + "stem": "fcfa188fdd8e4cdc" + }, + { + "metadata_index": 6956, + "stem": "f9e4896193305279" + }, + { + "metadata_index": 2533, + "stem": "5c30afe4faa45992" + }, + { + "metadata_index": 6675, + "stem": "f088c85322065722" + }, + { + "metadata_index": 1770, + "stem": "4091c41c6909da3b" + }, + { + "metadata_index": 1250, + "stem": "2deacec5c281fbac" + }, + { + "metadata_index": 4812, + "stem": "ade3e283780efbd7" + }, + { + "metadata_index": 4678, + "stem": "a9228c6ca6619266" + }, + { + "metadata_index": 3973, + "stem": "8f71898bc3662fa6" + }, + { + "metadata_index": 4057, + "stem": "929ed71c97317c55" + }, + { + "metadata_index": 1995, + "stem": "49235d402cbb8895" + }, + { + "metadata_index": 2405, + "stem": "57ed3518c548f1d9" + }, + { + "metadata_index": 3770, + "stem": "885c1a66b4025d94" + }, + { + "metadata_index": 2563, + "stem": "5d3888256b80fb8d" + }, + { + "metadata_index": 5105, + "stem": "b8605d7a56bada1c" + }, + { + "metadata_index": 2470, + "stem": "599619a790d2b77d" + }, + { + "metadata_index": 4115, + "stem": "952f6cb94695d32d" + }, + { + "metadata_index": 6785, + "stem": "f3de74db05a6b558" + }, + { + "metadata_index": 5387, + "stem": "c251d6e68d6d89eb" + }, + { + "metadata_index": 4030, + "stem": "91ca67e0e542a035" + }, + { + "metadata_index": 3848, + "stem": "8ad61ecb52a9122a" + }, + { + "metadata_index": 6586, + "stem": "ed7a4d97bab47c6d" + }, + { + "metadata_index": 498, + "stem": "118ffef2ad3950f5" + }, + { + "metadata_index": 26, + "stem": "0122933cf8ab3317" + }, + { + "metadata_index": 5169, + "stem": "bb0d9ff59505b17e" + }, + { + "metadata_index": 4993, + "stem": "b4c2c2cb9d279a99" + }, + { + "metadata_index": 6381, + "stem": "e6f8497a860c873c" + }, + { + "metadata_index": 2308, + "stem": "54a1bceb5d97272d" + }, + { + "metadata_index": 2987, + "stem": "6c376c523531fd19" + }, + { + "metadata_index": 2320, + "stem": "5535e78a16298072" + }, + { + "metadata_index": 3128, + "stem": "7215c47789bfed66" + }, + { + "metadata_index": 6845, + "stem": "f5a110214805f3f1" + }, + { + "metadata_index": 1026, + "stem": "25d8e374a2dafc8a" + }, + { + "metadata_index": 4441, + "stem": "a0bc9b509471e1dc" + }, + { + "metadata_index": 5905, + "stem": "d5409c93b6f88256" + }, + { + "metadata_index": 2215, + "stem": "512810bd5785cad6" + }, + { + "metadata_index": 3875, + "stem": "8ba67e66fe6cf9d8" + }, + { + "metadata_index": 677, + "stem": "186f18684ed4b516" + }, + { + "metadata_index": 6477, + "stem": "ea11822d0ab31414" + }, + { + "metadata_index": 913, + "stem": "20fba1d53c349851" + }, + { + "metadata_index": 2124, + "stem": "4df04a3ab680ac1d" + }, + { + "metadata_index": 6719, + "stem": "f20a49fc09b9ee6e" + }, + { + "metadata_index": 2483, + "stem": "5a2c2af8cf1765b2" + }, + { + "metadata_index": 6513, + "stem": "eb5b73ca4a18bbf9" + }, + { + "metadata_index": 4300, + "stem": "9c1398cb9ad44a17" + }, + { + "metadata_index": 2135, + "stem": "4e50d5610be16ce1" + }, + { + "metadata_index": 1585, + "stem": "39c662fa32a3b5c6" + }, + { + "metadata_index": 3894, + "stem": "8c3cf8cab7fcc03e" + }, + { + "metadata_index": 2564, + "stem": "5d3c7dd9db5e495a" + }, + { + "metadata_index": 5610, + "stem": "ca19b5ba41e05630" + }, + { + "metadata_index": 3230, + "stem": "756c2bdfad5cdd55" + }, + { + "metadata_index": 1128, + "stem": "2a1769dddc1dbf8d" + }, + { + "metadata_index": 6354, + "stem": "e60589d40eda6049" + }, + { + "metadata_index": 4634, + "stem": "a790bd601037ea91" + }, + { + "metadata_index": 794, + "stem": "1c9772d765e0679e" + }, + { + "metadata_index": 6418, + "stem": "e8312620692723b3" + }, + { + "metadata_index": 1459, + "stem": "3558c5283169e6ca" + } + ] +} diff --git a/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/trainable_parameter_manifest.json b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/trainable_parameter_manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..5512b64fd775153a9cce5c815400ad4450ad567d --- /dev/null +++ b/results/re10k_faithful_worldmem_clipped_final100_train731edd42758e_eval73c1cc4f6787_step30000_job80887/trainable_parameter_manifest.json @@ -0,0 +1,2049 @@ +{ + "base_lora_token_scope": "all_tokens", + "base_names": [ + "diffusion_model.model.noise_level_pos_embedding.embedding.linear_1.weight", + "diffusion_model.model.noise_level_pos_embedding.embedding.linear_1.bias", + "diffusion_model.model.noise_level_pos_embedding.embedding.linear_2.weight", + "diffusion_model.model.noise_level_pos_embedding.embedding.linear_2.bias", + "diffusion_model.model.external_cond_embedding.patch_embedder.proj.weight", + "diffusion_model.model.external_cond_embedding.patch_embedder.proj.bias", + "diffusion_model.model.embed_input.proj.weight", + "diffusion_model.model.embed_input.proj.bias", + "diffusion_model.model.project_output.proj.weight", + "diffusion_model.model.project_output.proj.bias", + "diffusion_model.model.down_blocks.0.0.emb_layer.weight", + "diffusion_model.model.down_blocks.0.0.emb_layer.bias", + "diffusion_model.model.down_blocks.0.0.in_layers.0.weight", + "diffusion_model.model.down_blocks.0.0.in_layers.0.bias", + "diffusion_model.model.down_blocks.0.0.in_layers.2.weight", + "diffusion_model.model.down_blocks.0.0.in_layers.2.bias", + "diffusion_model.model.down_blocks.0.0.out_norm.weight", + "diffusion_model.model.down_blocks.0.0.out_norm.bias", + "diffusion_model.model.down_blocks.0.0.out_rest.1.weight", + "diffusion_model.model.down_blocks.0.0.out_rest.1.bias", + "diffusion_model.model.down_blocks.0.1.emb_layer.weight", + "diffusion_model.model.down_blocks.0.1.emb_layer.bias", + "diffusion_model.model.down_blocks.0.1.in_layers.0.weight", + "diffusion_model.model.down_blocks.0.1.in_layers.0.bias", + "diffusion_model.model.down_blocks.0.1.in_layers.2.weight", + "diffusion_model.model.down_blocks.0.1.in_layers.2.bias", + "diffusion_model.model.down_blocks.0.1.out_norm.weight", + "diffusion_model.model.down_blocks.0.1.out_norm.bias", + "diffusion_model.model.down_blocks.0.1.out_rest.1.weight", + "diffusion_model.model.down_blocks.0.1.out_rest.1.bias", + "diffusion_model.model.down_blocks.0.2.emb_layer.weight", + "diffusion_model.model.down_blocks.0.2.emb_layer.bias", + "diffusion_model.model.down_blocks.0.2.in_layers.0.weight", + "diffusion_model.model.down_blocks.0.2.in_layers.0.bias", + "diffusion_model.model.down_blocks.0.2.in_layers.2.weight", + "diffusion_model.model.down_blocks.0.2.in_layers.2.bias", + "diffusion_model.model.down_blocks.0.2.out_norm.weight", + "diffusion_model.model.down_blocks.0.2.out_norm.bias", + "diffusion_model.model.down_blocks.0.2.out_rest.1.weight", + "diffusion_model.model.down_blocks.0.2.out_rest.1.bias", + "diffusion_model.model.down_blocks.0.3.conv.weight", + "diffusion_model.model.down_blocks.0.3.conv.bias", + "diffusion_model.model.down_blocks.1.0.emb_layer.weight", + "diffusion_model.model.down_blocks.1.0.emb_layer.bias", + "diffusion_model.model.down_blocks.1.0.in_layers.0.weight", + "diffusion_model.model.down_blocks.1.0.in_layers.0.bias", + "diffusion_model.model.down_blocks.1.0.in_layers.2.weight", + "diffusion_model.model.down_blocks.1.0.in_layers.2.bias", + "diffusion_model.model.down_blocks.1.0.out_norm.weight", + "diffusion_model.model.down_blocks.1.0.out_norm.bias", + "diffusion_model.model.down_blocks.1.0.out_rest.1.weight", + "diffusion_model.model.down_blocks.1.0.out_rest.1.bias", + "diffusion_model.model.down_blocks.1.1.emb_layer.weight", + "diffusion_model.model.down_blocks.1.1.emb_layer.bias", + "diffusion_model.model.down_blocks.1.1.in_layers.0.weight", + "diffusion_model.model.down_blocks.1.1.in_layers.0.bias", + "diffusion_model.model.down_blocks.1.1.in_layers.2.weight", + "diffusion_model.model.down_blocks.1.1.in_layers.2.bias", + "diffusion_model.model.down_blocks.1.1.out_norm.weight", + "diffusion_model.model.down_blocks.1.1.out_norm.bias", + "diffusion_model.model.down_blocks.1.1.out_rest.1.weight", + "diffusion_model.model.down_blocks.1.1.out_rest.1.bias", + "diffusion_model.model.down_blocks.1.2.emb_layer.weight", + "diffusion_model.model.down_blocks.1.2.emb_layer.bias", + "diffusion_model.model.down_blocks.1.2.in_layers.0.weight", + "diffusion_model.model.down_blocks.1.2.in_layers.0.bias", + "diffusion_model.model.down_blocks.1.2.in_layers.2.weight", + "diffusion_model.model.down_blocks.1.2.in_layers.2.bias", + "diffusion_model.model.down_blocks.1.2.out_norm.weight", + "diffusion_model.model.down_blocks.1.2.out_norm.bias", + "diffusion_model.model.down_blocks.1.2.out_rest.1.weight", + "diffusion_model.model.down_blocks.1.2.out_rest.1.bias", + "diffusion_model.model.down_blocks.1.3.conv.weight", + "diffusion_model.model.down_blocks.1.3.conv.bias", + "diffusion_model.model.down_blocks.2.0.norm.emb_layer.weight", + "diffusion_model.model.down_blocks.2.0.norm.emb_layer.bias", + "diffusion_model.model.down_blocks.2.0.norm.norm.weight", + "diffusion_model.model.down_blocks.2.0.fused_attn_mlp_proj.weight", + "diffusion_model.model.down_blocks.2.0.fused_attn_mlp_proj.bias", + "diffusion_model.model.down_blocks.2.0.q_norm.weight", + "diffusion_model.model.down_blocks.2.0.k_norm.weight", + "diffusion_model.model.down_blocks.2.0.attn_out.weight", + "diffusion_model.model.down_blocks.2.0.attn_out.bias", + "diffusion_model.model.down_blocks.2.0.mlp_out.2.weight", + "diffusion_model.model.down_blocks.2.0.mlp_out.2.bias", + "diffusion_model.model.down_blocks.2.1.norm.emb_layer.weight", + "diffusion_model.model.down_blocks.2.1.norm.emb_layer.bias", + "diffusion_model.model.down_blocks.2.1.norm.norm.weight", + "diffusion_model.model.down_blocks.2.1.fused_attn_mlp_proj.weight", + "diffusion_model.model.down_blocks.2.1.fused_attn_mlp_proj.bias", + "diffusion_model.model.down_blocks.2.1.q_norm.weight", + "diffusion_model.model.down_blocks.2.1.k_norm.weight", + "diffusion_model.model.down_blocks.2.1.attn_out.weight", + "diffusion_model.model.down_blocks.2.1.attn_out.bias", + "diffusion_model.model.down_blocks.2.1.mlp_out.2.weight", + "diffusion_model.model.down_blocks.2.1.mlp_out.2.bias", + "diffusion_model.model.down_blocks.2.2.norm.emb_layer.weight", + "diffusion_model.model.down_blocks.2.2.norm.emb_layer.bias", + "diffusion_model.model.down_blocks.2.2.norm.norm.weight", + "diffusion_model.model.down_blocks.2.2.fused_attn_mlp_proj.weight", + "diffusion_model.model.down_blocks.2.2.fused_attn_mlp_proj.bias", + "diffusion_model.model.down_blocks.2.2.q_norm.weight", + "diffusion_model.model.down_blocks.2.2.k_norm.weight", + "diffusion_model.model.down_blocks.2.2.attn_out.weight", + "diffusion_model.model.down_blocks.2.2.attn_out.bias", + "diffusion_model.model.down_blocks.2.2.mlp_out.2.weight", + "diffusion_model.model.down_blocks.2.2.mlp_out.2.bias", + "diffusion_model.model.down_blocks.2.3.norm.emb_layer.weight", + "diffusion_model.model.down_blocks.2.3.norm.emb_layer.bias", + "diffusion_model.model.down_blocks.2.3.norm.norm.weight", + "diffusion_model.model.down_blocks.2.3.fused_attn_mlp_proj.weight", + "diffusion_model.model.down_blocks.2.3.fused_attn_mlp_proj.bias", + "diffusion_model.model.down_blocks.2.3.q_norm.weight", + "diffusion_model.model.down_blocks.2.3.k_norm.weight", + "diffusion_model.model.down_blocks.2.3.attn_out.weight", + "diffusion_model.model.down_blocks.2.3.attn_out.bias", + "diffusion_model.model.down_blocks.2.3.mlp_out.2.weight", + "diffusion_model.model.down_blocks.2.3.mlp_out.2.bias", + "diffusion_model.model.down_blocks.2.4.norm.emb_layer.weight", + "diffusion_model.model.down_blocks.2.4.norm.emb_layer.bias", + "diffusion_model.model.down_blocks.2.4.norm.norm.weight", + "diffusion_model.model.down_blocks.2.4.fused_attn_mlp_proj.weight", + "diffusion_model.model.down_blocks.2.4.fused_attn_mlp_proj.bias", + "diffusion_model.model.down_blocks.2.4.q_norm.weight", + "diffusion_model.model.down_blocks.2.4.k_norm.weight", + "diffusion_model.model.down_blocks.2.4.attn_out.weight", + "diffusion_model.model.down_blocks.2.4.attn_out.bias", + "diffusion_model.model.down_blocks.2.4.mlp_out.2.weight", + "diffusion_model.model.down_blocks.2.4.mlp_out.2.bias", + "diffusion_model.model.down_blocks.2.5.norm.emb_layer.weight", + "diffusion_model.model.down_blocks.2.5.norm.emb_layer.bias", + "diffusion_model.model.down_blocks.2.5.norm.norm.weight", + "diffusion_model.model.down_blocks.2.5.fused_attn_mlp_proj.weight", + "diffusion_model.model.down_blocks.2.5.fused_attn_mlp_proj.bias", + "diffusion_model.model.down_blocks.2.5.q_norm.weight", + "diffusion_model.model.down_blocks.2.5.k_norm.weight", + "diffusion_model.model.down_blocks.2.5.attn_out.weight", + "diffusion_model.model.down_blocks.2.5.attn_out.bias", + "diffusion_model.model.down_blocks.2.5.mlp_out.2.weight", + "diffusion_model.model.down_blocks.2.5.mlp_out.2.bias", + "diffusion_model.model.down_blocks.2.6.conv.weight", + "diffusion_model.model.down_blocks.2.6.conv.bias", + "diffusion_model.model.up_blocks.0.0.conv.weight", + "diffusion_model.model.up_blocks.0.0.conv.bias", + "diffusion_model.model.up_blocks.0.1.norm.emb_layer.weight", + "diffusion_model.model.up_blocks.0.1.norm.emb_layer.bias", + "diffusion_model.model.up_blocks.0.1.norm.norm.weight", + "diffusion_model.model.up_blocks.0.1.fused_attn_mlp_proj.weight", + "diffusion_model.model.up_blocks.0.1.fused_attn_mlp_proj.bias", + "diffusion_model.model.up_blocks.0.1.q_norm.weight", + "diffusion_model.model.up_blocks.0.1.k_norm.weight", + "diffusion_model.model.up_blocks.0.1.attn_out.weight", + "diffusion_model.model.up_blocks.0.1.attn_out.bias", + "diffusion_model.model.up_blocks.0.1.mlp_out.2.weight", + "diffusion_model.model.up_blocks.0.1.mlp_out.2.bias", + "diffusion_model.model.up_blocks.0.2.norm.emb_layer.weight", + "diffusion_model.model.up_blocks.0.2.norm.emb_layer.bias", + "diffusion_model.model.up_blocks.0.2.norm.norm.weight", + "diffusion_model.model.up_blocks.0.2.fused_attn_mlp_proj.weight", + "diffusion_model.model.up_blocks.0.2.fused_attn_mlp_proj.bias", + "diffusion_model.model.up_blocks.0.2.q_norm.weight", + "diffusion_model.model.up_blocks.0.2.k_norm.weight", + "diffusion_model.model.up_blocks.0.2.attn_out.weight", + "diffusion_model.model.up_blocks.0.2.attn_out.bias", + "diffusion_model.model.up_blocks.0.2.mlp_out.2.weight", + "diffusion_model.model.up_blocks.0.2.mlp_out.2.bias", + "diffusion_model.model.up_blocks.0.3.norm.emb_layer.weight", + "diffusion_model.model.up_blocks.0.3.norm.emb_layer.bias", + "diffusion_model.model.up_blocks.0.3.norm.norm.weight", + "diffusion_model.model.up_blocks.0.3.fused_attn_mlp_proj.weight", + "diffusion_model.model.up_blocks.0.3.fused_attn_mlp_proj.bias", + "diffusion_model.model.up_blocks.0.3.q_norm.weight", + "diffusion_model.model.up_blocks.0.3.k_norm.weight", + "diffusion_model.model.up_blocks.0.3.attn_out.weight", + "diffusion_model.model.up_blocks.0.3.attn_out.bias", + "diffusion_model.model.up_blocks.0.3.mlp_out.2.weight", + "diffusion_model.model.up_blocks.0.3.mlp_out.2.bias", + "diffusion_model.model.up_blocks.0.4.norm.emb_layer.weight", + "diffusion_model.model.up_blocks.0.4.norm.emb_layer.bias", + "diffusion_model.model.up_blocks.0.4.norm.norm.weight", + "diffusion_model.model.up_blocks.0.4.fused_attn_mlp_proj.weight", + "diffusion_model.model.up_blocks.0.4.fused_attn_mlp_proj.bias", + "diffusion_model.model.up_blocks.0.4.q_norm.weight", + "diffusion_model.model.up_blocks.0.4.k_norm.weight", + "diffusion_model.model.up_blocks.0.4.attn_out.weight", + "diffusion_model.model.up_blocks.0.4.attn_out.bias", + "diffusion_model.model.up_blocks.0.4.mlp_out.2.weight", + "diffusion_model.model.up_blocks.0.4.mlp_out.2.bias", + "diffusion_model.model.up_blocks.0.5.norm.emb_layer.weight", + "diffusion_model.model.up_blocks.0.5.norm.emb_layer.bias", + "diffusion_model.model.up_blocks.0.5.norm.norm.weight", + "diffusion_model.model.up_blocks.0.5.fused_attn_mlp_proj.weight", + "diffusion_model.model.up_blocks.0.5.fused_attn_mlp_proj.bias", + "diffusion_model.model.up_blocks.0.5.q_norm.weight", + "diffusion_model.model.up_blocks.0.5.k_norm.weight", + "diffusion_model.model.up_blocks.0.5.attn_out.weight", + "diffusion_model.model.up_blocks.0.5.attn_out.bias", + "diffusion_model.model.up_blocks.0.5.mlp_out.2.weight", + "diffusion_model.model.up_blocks.0.5.mlp_out.2.bias", + "diffusion_model.model.up_blocks.0.6.norm.emb_layer.weight", + "diffusion_model.model.up_blocks.0.6.norm.emb_layer.bias", + "diffusion_model.model.up_blocks.0.6.norm.norm.weight", + "diffusion_model.model.up_blocks.0.6.fused_attn_mlp_proj.weight", + "diffusion_model.model.up_blocks.0.6.fused_attn_mlp_proj.bias", + "diffusion_model.model.up_blocks.0.6.q_norm.weight", + "diffusion_model.model.up_blocks.0.6.k_norm.weight", + "diffusion_model.model.up_blocks.0.6.attn_out.weight", + "diffusion_model.model.up_blocks.0.6.attn_out.bias", + "diffusion_model.model.up_blocks.0.6.mlp_out.2.weight", + "diffusion_model.model.up_blocks.0.6.mlp_out.2.bias", + "diffusion_model.model.up_blocks.1.0.conv.weight", + "diffusion_model.model.up_blocks.1.0.conv.bias", + "diffusion_model.model.up_blocks.1.1.emb_layer.weight", + "diffusion_model.model.up_blocks.1.1.emb_layer.bias", + "diffusion_model.model.up_blocks.1.1.in_layers.0.weight", + "diffusion_model.model.up_blocks.1.1.in_layers.0.bias", + "diffusion_model.model.up_blocks.1.1.in_layers.2.weight", + "diffusion_model.model.up_blocks.1.1.in_layers.2.bias", + "diffusion_model.model.up_blocks.1.1.out_norm.weight", + "diffusion_model.model.up_blocks.1.1.out_norm.bias", + "diffusion_model.model.up_blocks.1.1.out_rest.1.weight", + "diffusion_model.model.up_blocks.1.1.out_rest.1.bias", + "diffusion_model.model.up_blocks.1.2.emb_layer.weight", + "diffusion_model.model.up_blocks.1.2.emb_layer.bias", + "diffusion_model.model.up_blocks.1.2.in_layers.0.weight", + "diffusion_model.model.up_blocks.1.2.in_layers.0.bias", + "diffusion_model.model.up_blocks.1.2.in_layers.2.weight", + "diffusion_model.model.up_blocks.1.2.in_layers.2.bias", + "diffusion_model.model.up_blocks.1.2.out_norm.weight", + "diffusion_model.model.up_blocks.1.2.out_norm.bias", + "diffusion_model.model.up_blocks.1.2.out_rest.1.weight", + "diffusion_model.model.up_blocks.1.2.out_rest.1.bias", + "diffusion_model.model.up_blocks.1.3.emb_layer.weight", + "diffusion_model.model.up_blocks.1.3.emb_layer.bias", + "diffusion_model.model.up_blocks.1.3.in_layers.0.weight", + "diffusion_model.model.up_blocks.1.3.in_layers.0.bias", + "diffusion_model.model.up_blocks.1.3.in_layers.2.weight", + "diffusion_model.model.up_blocks.1.3.in_layers.2.bias", + "diffusion_model.model.up_blocks.1.3.out_norm.weight", + "diffusion_model.model.up_blocks.1.3.out_norm.bias", + "diffusion_model.model.up_blocks.1.3.out_rest.1.weight", + "diffusion_model.model.up_blocks.1.3.out_rest.1.bias", + "diffusion_model.model.up_blocks.2.0.conv.weight", + "diffusion_model.model.up_blocks.2.0.conv.bias", + "diffusion_model.model.up_blocks.2.1.emb_layer.weight", + "diffusion_model.model.up_blocks.2.1.emb_layer.bias", + "diffusion_model.model.up_blocks.2.1.in_layers.0.weight", + "diffusion_model.model.up_blocks.2.1.in_layers.0.bias", + "diffusion_model.model.up_blocks.2.1.in_layers.2.weight", + "diffusion_model.model.up_blocks.2.1.in_layers.2.bias", + "diffusion_model.model.up_blocks.2.1.out_norm.weight", + "diffusion_model.model.up_blocks.2.1.out_norm.bias", + "diffusion_model.model.up_blocks.2.1.out_rest.1.weight", + "diffusion_model.model.up_blocks.2.1.out_rest.1.bias", + "diffusion_model.model.up_blocks.2.2.emb_layer.weight", + "diffusion_model.model.up_blocks.2.2.emb_layer.bias", + "diffusion_model.model.up_blocks.2.2.in_layers.0.weight", + "diffusion_model.model.up_blocks.2.2.in_layers.0.bias", + "diffusion_model.model.up_blocks.2.2.in_layers.2.weight", + "diffusion_model.model.up_blocks.2.2.in_layers.2.bias", + "diffusion_model.model.up_blocks.2.2.out_norm.weight", + "diffusion_model.model.up_blocks.2.2.out_norm.bias", + "diffusion_model.model.up_blocks.2.2.out_rest.1.weight", + "diffusion_model.model.up_blocks.2.2.out_rest.1.bias", + "diffusion_model.model.up_blocks.2.3.emb_layer.weight", + "diffusion_model.model.up_blocks.2.3.emb_layer.bias", + "diffusion_model.model.up_blocks.2.3.in_layers.0.weight", + "diffusion_model.model.up_blocks.2.3.in_layers.0.bias", + "diffusion_model.model.up_blocks.2.3.in_layers.2.weight", + "diffusion_model.model.up_blocks.2.3.in_layers.2.bias", + "diffusion_model.model.up_blocks.2.3.out_norm.weight", + "diffusion_model.model.up_blocks.2.3.out_norm.bias", + "diffusion_model.model.up_blocks.2.3.out_rest.1.weight", + "diffusion_model.model.up_blocks.2.3.out_rest.1.bias", + "diffusion_model.model.mid_blocks.0.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.0.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.0.norm.norm.weight", + "diffusion_model.model.mid_blocks.0.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.0.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.0.q_norm.weight", + "diffusion_model.model.mid_blocks.0.k_norm.weight", + "diffusion_model.model.mid_blocks.0.attn_out.weight", + "diffusion_model.model.mid_blocks.0.attn_out.bias", + "diffusion_model.model.mid_blocks.0.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.0.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.1.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.1.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.1.norm.norm.weight", + "diffusion_model.model.mid_blocks.1.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.1.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.1.q_norm.weight", + "diffusion_model.model.mid_blocks.1.k_norm.weight", + "diffusion_model.model.mid_blocks.1.attn_out.weight", + "diffusion_model.model.mid_blocks.1.attn_out.bias", + "diffusion_model.model.mid_blocks.1.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.1.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.2.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.2.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.2.norm.norm.weight", + "diffusion_model.model.mid_blocks.2.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.2.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.2.q_norm.weight", + "diffusion_model.model.mid_blocks.2.k_norm.weight", + "diffusion_model.model.mid_blocks.2.attn_out.weight", + "diffusion_model.model.mid_blocks.2.attn_out.bias", + "diffusion_model.model.mid_blocks.2.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.2.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.3.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.3.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.3.norm.norm.weight", + "diffusion_model.model.mid_blocks.3.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.3.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.3.q_norm.weight", + "diffusion_model.model.mid_blocks.3.k_norm.weight", + "diffusion_model.model.mid_blocks.3.attn_out.weight", + "diffusion_model.model.mid_blocks.3.attn_out.bias", + "diffusion_model.model.mid_blocks.3.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.3.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.4.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.4.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.4.norm.norm.weight", + "diffusion_model.model.mid_blocks.4.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.4.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.4.q_norm.weight", + "diffusion_model.model.mid_blocks.4.k_norm.weight", + "diffusion_model.model.mid_blocks.4.attn_out.weight", + "diffusion_model.model.mid_blocks.4.attn_out.bias", + "diffusion_model.model.mid_blocks.4.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.4.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.5.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.5.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.5.norm.norm.weight", + "diffusion_model.model.mid_blocks.5.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.5.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.5.q_norm.weight", + "diffusion_model.model.mid_blocks.5.k_norm.weight", + "diffusion_model.model.mid_blocks.5.attn_out.weight", + "diffusion_model.model.mid_blocks.5.attn_out.bias", + "diffusion_model.model.mid_blocks.5.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.5.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.6.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.6.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.6.norm.norm.weight", + "diffusion_model.model.mid_blocks.6.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.6.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.6.q_norm.weight", + "diffusion_model.model.mid_blocks.6.k_norm.weight", + "diffusion_model.model.mid_blocks.6.attn_out.weight", + "diffusion_model.model.mid_blocks.6.attn_out.bias", + "diffusion_model.model.mid_blocks.6.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.6.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.7.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.7.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.7.norm.norm.weight", + "diffusion_model.model.mid_blocks.7.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.7.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.7.q_norm.weight", + "diffusion_model.model.mid_blocks.7.k_norm.weight", + "diffusion_model.model.mid_blocks.7.attn_out.weight", + "diffusion_model.model.mid_blocks.7.attn_out.bias", + "diffusion_model.model.mid_blocks.7.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.7.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.8.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.8.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.8.norm.norm.weight", + "diffusion_model.model.mid_blocks.8.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.8.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.8.q_norm.weight", + "diffusion_model.model.mid_blocks.8.k_norm.weight", + "diffusion_model.model.mid_blocks.8.attn_out.weight", + "diffusion_model.model.mid_blocks.8.attn_out.bias", + "diffusion_model.model.mid_blocks.8.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.8.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.9.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.9.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.9.norm.norm.weight", + "diffusion_model.model.mid_blocks.9.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.9.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.9.q_norm.weight", + "diffusion_model.model.mid_blocks.9.k_norm.weight", + "diffusion_model.model.mid_blocks.9.attn_out.weight", + "diffusion_model.model.mid_blocks.9.attn_out.bias", + "diffusion_model.model.mid_blocks.9.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.9.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.10.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.10.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.10.norm.norm.weight", + "diffusion_model.model.mid_blocks.10.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.10.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.10.q_norm.weight", + "diffusion_model.model.mid_blocks.10.k_norm.weight", + "diffusion_model.model.mid_blocks.10.attn_out.weight", + "diffusion_model.model.mid_blocks.10.attn_out.bias", + "diffusion_model.model.mid_blocks.10.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.10.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.11.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.11.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.11.norm.norm.weight", + "diffusion_model.model.mid_blocks.11.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.11.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.11.q_norm.weight", + "diffusion_model.model.mid_blocks.11.k_norm.weight", + "diffusion_model.model.mid_blocks.11.attn_out.weight", + "diffusion_model.model.mid_blocks.11.attn_out.bias", + "diffusion_model.model.mid_blocks.11.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.11.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.12.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.12.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.12.norm.norm.weight", + "diffusion_model.model.mid_blocks.12.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.12.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.12.q_norm.weight", + "diffusion_model.model.mid_blocks.12.k_norm.weight", + "diffusion_model.model.mid_blocks.12.attn_out.weight", + "diffusion_model.model.mid_blocks.12.attn_out.bias", + "diffusion_model.model.mid_blocks.12.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.12.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.13.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.13.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.13.norm.norm.weight", + "diffusion_model.model.mid_blocks.13.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.13.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.13.q_norm.weight", + "diffusion_model.model.mid_blocks.13.k_norm.weight", + "diffusion_model.model.mid_blocks.13.attn_out.weight", + "diffusion_model.model.mid_blocks.13.attn_out.bias", + "diffusion_model.model.mid_blocks.13.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.13.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.14.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.14.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.14.norm.norm.weight", + "diffusion_model.model.mid_blocks.14.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.14.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.14.q_norm.weight", + "diffusion_model.model.mid_blocks.14.k_norm.weight", + "diffusion_model.model.mid_blocks.14.attn_out.weight", + "diffusion_model.model.mid_blocks.14.attn_out.bias", + "diffusion_model.model.mid_blocks.14.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.14.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.15.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.15.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.15.norm.norm.weight", + "diffusion_model.model.mid_blocks.15.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.15.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.15.q_norm.weight", + "diffusion_model.model.mid_blocks.15.k_norm.weight", + "diffusion_model.model.mid_blocks.15.attn_out.weight", + "diffusion_model.model.mid_blocks.15.attn_out.bias", + "diffusion_model.model.mid_blocks.15.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.15.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.16.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.16.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.16.norm.norm.weight", + "diffusion_model.model.mid_blocks.16.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.16.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.16.q_norm.weight", + "diffusion_model.model.mid_blocks.16.k_norm.weight", + "diffusion_model.model.mid_blocks.16.attn_out.weight", + "diffusion_model.model.mid_blocks.16.attn_out.bias", + "diffusion_model.model.mid_blocks.16.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.16.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.17.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.17.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.17.norm.norm.weight", + "diffusion_model.model.mid_blocks.17.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.17.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.17.q_norm.weight", + "diffusion_model.model.mid_blocks.17.k_norm.weight", + "diffusion_model.model.mid_blocks.17.attn_out.weight", + "diffusion_model.model.mid_blocks.17.attn_out.bias", + "diffusion_model.model.mid_blocks.17.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.17.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.18.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.18.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.18.norm.norm.weight", + "diffusion_model.model.mid_blocks.18.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.18.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.18.q_norm.weight", + "diffusion_model.model.mid_blocks.18.k_norm.weight", + "diffusion_model.model.mid_blocks.18.attn_out.weight", + "diffusion_model.model.mid_blocks.18.attn_out.bias", + "diffusion_model.model.mid_blocks.18.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.18.mlp_out.2.bias", + "diffusion_model.model.mid_blocks.19.norm.emb_layer.weight", + "diffusion_model.model.mid_blocks.19.norm.emb_layer.bias", + "diffusion_model.model.mid_blocks.19.norm.norm.weight", + "diffusion_model.model.mid_blocks.19.fused_attn_mlp_proj.weight", + "diffusion_model.model.mid_blocks.19.fused_attn_mlp_proj.bias", + "diffusion_model.model.mid_blocks.19.q_norm.weight", + "diffusion_model.model.mid_blocks.19.k_norm.weight", + "diffusion_model.model.mid_blocks.19.attn_out.weight", + "diffusion_model.model.mid_blocks.19.attn_out.bias", + "diffusion_model.model.mid_blocks.19.mlp_out.2.weight", + "diffusion_model.model.mid_blocks.19.mlp_out.2.bias" + ], + "base_numel": 458818051, + "base_tensor_count": 494, + "dfot_update": "frozen_lora_stage8k", + "dynamic_selection_enabled": true, + "evaluation_weight_source": "ema", + "families": { + "ffn_input": 61106688, + "ffn_output": 61076736, + "pose_cond_mlp": 30553344, + "pose_encoder": 224704, + "reference_adaln": 23182848, + "reference_output": 30553344, + "reference_qkv": 91570176, + "timestamp_encoder": 2104704, + "timestep_encoder": 131200 + }, + "frozen_numel": 458818051, + "full_model_numel": 762916035, + "layers": [ + { + "block": "down_blocks.2.0", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "down_blocks.2.1", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "down_blocks.2.2", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "down_blocks.2.3", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "down_blocks.2.4", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "down_blocks.2.5", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "up_blocks.0.1", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "up_blocks.0.2", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "up_blocks.0.3", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "up_blocks.0.4", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "up_blocks.0.5", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "up_blocks.0.6", + "dim": 576, + "ffn_input": 664704, + "ffn_output": 664128, + "pose_cond_mlp": 332352, + "reference_adaln": 445824, + "reference_output": 332352, + "reference_qkv": 995328, + "tensor_count": 11, + "total": 3434688 + }, + { + "block": "mid_blocks.0", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.1", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.2", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.3", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.4", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.5", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.6", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.7", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.8", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.9", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.10", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.11", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.12", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.13", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.14", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.15", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.16", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.17", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.18", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + }, + { + "block": "mid_blocks.19", + "dim": 1152, + "ffn_input": 2656512, + "ffn_output": 2655360, + "pose_cond_mlp": 1328256, + "reference_adaln": 891648, + "reference_output": 1328256, + "reference_qkv": 3981312, + "tensor_count": 11, + "total": 12841344 + } + ], + "lifecycle": "group1_59306102", + "lora_names": [ + "diffusion_model.model.down_blocks.2.0.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_mlp_output_lora.up.weight" + ], + "lora_numel": 3594240, + "lora_tensor_count": 192, + "memory_age_conditioning": false, + "memory_capacity": "clipped", + "method": "worldmem", + "profile_variant": "faithful_f200b200", + "reference_names": [ + "diffusion_model.model.down_blocks.2.0.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.0.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.0.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.0.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.0.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.0.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.0.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.0.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.0.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.0.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.0.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.0.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.1.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.1.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.1.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.1.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.1.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.1.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.1.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.1.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.1.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.1.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.1.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.2.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.2.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.2.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.2.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.2.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.2.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.2.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.2.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.2.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.2.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.2.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.3.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.3.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.3.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.3.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.3.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.3.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.3.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.3.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.3.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.3.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.3.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.4.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.4.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.4.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.4.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.4.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.4.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.4.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.4.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.4.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.4.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.4.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.5.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.5.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.5.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.5.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.5.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.5.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.5.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.5.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.5.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.5.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.5.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.6.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.6.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.6.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.6.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.6.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.6.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.6.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.6.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.6.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.6.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.6.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.7.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.7.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.7.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.7.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.7.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.7.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.7.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.7.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.7.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.7.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.7.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.8.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.8.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.8.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.8.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.8.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.8.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.8.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.8.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.8.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.8.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.8.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.9.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.9.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.9.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.9.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.9.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.9.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.9.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.9.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.9.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.9.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.9.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.10.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.10.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.10.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.10.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.10.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.10.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.10.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.10.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.10.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.10.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.10.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.11.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.11.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.11.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.11.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.11.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.11.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.11.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.11.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.11.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.11.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.11.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.12.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.12.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.12.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.12.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.12.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.12.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.12.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.12.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.12.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.12.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.12.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.13.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.13.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.13.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.13.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.13.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.13.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.13.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.13.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.13.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.13.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.13.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.14.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.14.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.14.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.14.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.14.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.14.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.14.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.14.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.14.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.14.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.14.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.15.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.15.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.15.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.15.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.15.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.15.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.15.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.15.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.15.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.15.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.15.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.16.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.16.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.16.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.16.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.16.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.16.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.16.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.16.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.16.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.16.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.16.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.17.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.17.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.17.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.17.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.17.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.17.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.17.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.17.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.17.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.17.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.17.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.18.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.18.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.18.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.18.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.18.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.18.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.18.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.18.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.18.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.18.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.18.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.19.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.19.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.19.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.19.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.19.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.19.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.19.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.19.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.19.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.19.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.19.memory_block.reference_adaln.1.bias", + "diffusion_model.model.memory_conditioner.pose_encoder.576.0.weight", + "diffusion_model.model.memory_conditioner.pose_encoder.576.0.bias", + "diffusion_model.model.memory_conditioner.pose_encoder.576.2.weight", + "diffusion_model.model.memory_conditioner.pose_encoder.576.2.bias", + "diffusion_model.model.memory_conditioner.pose_encoder.1152.0.weight", + "diffusion_model.model.memory_conditioner.pose_encoder.1152.0.bias", + "diffusion_model.model.memory_conditioner.pose_encoder.1152.2.weight", + "diffusion_model.model.memory_conditioner.pose_encoder.1152.2.bias", + "diffusion_model.model.memory_conditioner.timestamp_encoder.576.0.weight", + "diffusion_model.model.memory_conditioner.timestamp_encoder.576.0.bias", + "diffusion_model.model.memory_conditioner.timestamp_encoder.576.2.weight", + "diffusion_model.model.memory_conditioner.timestamp_encoder.576.2.bias", + "diffusion_model.model.memory_conditioner.timestamp_encoder.1152.0.weight", + "diffusion_model.model.memory_conditioner.timestamp_encoder.1152.0.bias", + "diffusion_model.model.memory_conditioner.timestamp_encoder.1152.2.weight", + "diffusion_model.model.memory_conditioner.timestamp_encoder.1152.2.bias", + "diffusion_model.model.memory_conditioner.timestep_encoder.0.weight", + "diffusion_model.model.memory_conditioner.timestep_encoder.0.bias" + ], + "reference_numel": 300503744, + "reference_projection_initialization": "independent_xavier_bias_free_qkv_v1", + "reference_tensor_count": 370, + "reference_timestamp_conditioning": true, + "schema_version": 1, + "selector_version": "re10k_faithful_trajectory_h100_f399_v1", + "trainable_names": [ + "diffusion_model.model.down_blocks.2.0.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.0.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.0.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.1.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.1.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.2.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.2.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.3.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.3.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.4.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.4.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_qkv.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_output.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_output.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_ffn.0.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_ffn.0.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_ffn.2.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_ffn.2.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_adaln.1.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.reference_adaln.1.bias", + "diffusion_model.model.down_blocks.2.5.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.down_blocks.2.5.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.1.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.1.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.2.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.2.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.3.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.3.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.4.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.4.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.5.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.5.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_qkv.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_output.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_output.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_ffn.0.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_ffn.0.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_ffn.2.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_ffn.2.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_adaln.1.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.reference_adaln.1.bias", + "diffusion_model.model.up_blocks.0.6.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.up_blocks.0.6.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.0.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.0.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.0.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.0.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.0.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.0.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.0.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.0.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.0.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.0.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.0.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.0.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.0.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.1.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.1.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.1.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.1.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.1.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.1.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.1.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.1.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.1.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.1.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.1.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.1.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.1.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.2.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.2.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.2.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.2.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.2.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.2.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.2.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.2.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.2.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.2.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.2.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.2.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.2.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.3.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.3.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.3.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.3.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.3.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.3.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.3.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.3.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.3.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.3.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.3.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.3.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.3.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.4.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.4.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.4.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.4.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.4.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.4.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.4.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.4.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.4.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.4.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.4.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.4.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.4.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.5.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.5.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.5.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.5.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.5.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.5.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.5.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.5.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.5.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.5.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.5.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.5.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.5.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.6.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.6.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.6.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.6.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.6.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.6.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.6.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.6.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.6.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.6.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.6.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.6.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.6.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.7.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.7.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.7.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.7.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.7.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.7.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.7.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.7.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.7.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.7.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.7.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.7.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.7.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.8.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.8.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.8.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.8.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.8.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.8.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.8.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.8.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.8.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.8.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.8.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.8.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.8.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.9.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.9.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.9.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.9.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.9.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.9.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.9.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.9.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.9.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.9.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.9.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.9.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.9.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.10.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.10.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.10.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.10.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.10.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.10.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.10.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.10.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.10.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.10.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.10.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.10.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.10.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.11.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.11.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.11.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.11.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.11.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.11.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.11.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.11.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.11.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.11.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.11.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.11.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.11.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.12.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.12.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.12.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.12.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.12.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.12.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.12.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.12.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.12.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.12.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.12.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.12.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.12.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.13.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.13.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.13.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.13.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.13.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.13.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.13.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.13.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.13.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.13.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.13.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.13.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.13.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.14.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.14.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.14.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.14.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.14.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.14.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.14.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.14.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.14.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.14.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.14.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.14.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.14.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.15.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.15.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.15.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.15.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.15.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.15.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.15.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.15.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.15.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.15.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.15.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.15.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.15.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.16.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.16.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.16.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.16.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.16.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.16.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.16.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.16.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.16.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.16.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.16.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.16.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.16.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.17.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.17.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.17.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.17.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.17.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.17.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.17.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.17.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.17.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.17.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.17.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.17.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.17.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.18.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.18.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.18.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.18.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.18.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.18.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.18.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.18.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.18.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.18.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.18.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.18.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.18.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.mid_blocks.19.memory_block.reference_qkv.weight", + "diffusion_model.model.mid_blocks.19.memory_block.pose_cond_mlp.weight", + "diffusion_model.model.mid_blocks.19.memory_block.pose_cond_mlp.bias", + "diffusion_model.model.mid_blocks.19.memory_block.reference_output.weight", + "diffusion_model.model.mid_blocks.19.memory_block.reference_output.bias", + "diffusion_model.model.mid_blocks.19.memory_block.reference_ffn.0.weight", + "diffusion_model.model.mid_blocks.19.memory_block.reference_ffn.0.bias", + "diffusion_model.model.mid_blocks.19.memory_block.reference_ffn.2.weight", + "diffusion_model.model.mid_blocks.19.memory_block.reference_ffn.2.bias", + "diffusion_model.model.mid_blocks.19.memory_block.reference_adaln.1.weight", + "diffusion_model.model.mid_blocks.19.memory_block.reference_adaln.1.bias", + "diffusion_model.model.mid_blocks.19.memory_block.base_fused_lora.down.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_fused_lora.up.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_attention_output_lora.down.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_attention_output_lora.up.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_mlp_output_lora.down.weight", + "diffusion_model.model.mid_blocks.19.memory_block.base_mlp_output_lora.up.weight", + "diffusion_model.model.memory_conditioner.pose_encoder.576.0.weight", + "diffusion_model.model.memory_conditioner.pose_encoder.576.0.bias", + "diffusion_model.model.memory_conditioner.pose_encoder.576.2.weight", + "diffusion_model.model.memory_conditioner.pose_encoder.576.2.bias", + "diffusion_model.model.memory_conditioner.pose_encoder.1152.0.weight", + "diffusion_model.model.memory_conditioner.pose_encoder.1152.0.bias", + "diffusion_model.model.memory_conditioner.pose_encoder.1152.2.weight", + "diffusion_model.model.memory_conditioner.pose_encoder.1152.2.bias", + "diffusion_model.model.memory_conditioner.timestamp_encoder.576.0.weight", + "diffusion_model.model.memory_conditioner.timestamp_encoder.576.0.bias", + "diffusion_model.model.memory_conditioner.timestamp_encoder.576.2.weight", + "diffusion_model.model.memory_conditioner.timestamp_encoder.576.2.bias", + "diffusion_model.model.memory_conditioner.timestamp_encoder.1152.0.weight", + "diffusion_model.model.memory_conditioner.timestamp_encoder.1152.0.bias", + "diffusion_model.model.memory_conditioner.timestamp_encoder.1152.2.weight", + "diffusion_model.model.memory_conditioner.timestamp_encoder.1152.2.bias", + "diffusion_model.model.memory_conditioner.timestep_encoder.0.weight", + "diffusion_model.model.memory_conditioner.timestep_encoder.0.bias" + ], + "trainable_numel": 304097984, + "trainable_tensor_count": 562, + "training_ema_decay": 0.9999 +}