Upload short-loop WorldMem-Clipped job 81245 results
Browse files- .gitattributes +8 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/metrics/test_history.csv +2 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/metrics/test_monitor8_loop_true_reverse_h100_f199_v1_step00000000_metrics.json +55 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/metrics/test_step00000000.json +49 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/resolved_config.yaml +400 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/run_manifest.json +185 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/test_loop_true_reverse_h100_f199_v1_case_manifest.json +120 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/test_scene_manifest.json +42 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/trainable_parameter_manifest.json +0 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/0ece034988793847.mp4 +3 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/1735d8d1b4015669.mp4 +3 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/1cdfd3abfcc3a64e.mp4 +3 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/6afff9609b65a604.mp4 +3 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/b009fbe3950ebeb2.mp4 +3 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/bcc76b58e2585320.mp4 +3 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/c49a14de01d7bd35.mp4 +3 -0
- results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/f287c679c250a841.mp4 +3 -0
.gitattributes
CHANGED
|
@@ -14295,3 +14295,11 @@ results/re10k_short_reverse_dememwm_full_monitor8_train731edd42758e_eval940b6fb7
|
|
| 14295 |
results/re10k_short_reverse_dememwm_full_monitor8_train731edd42758e_eval940b6fb72bb1_step70000_job81244/videos/test_loop_true_reverse_h100_f199_v1_step00000000/bcc76b58e2585320.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14296 |
results/re10k_short_reverse_dememwm_full_monitor8_train731edd42758e_eval940b6fb72bb1_step70000_job81244/videos/test_loop_true_reverse_h100_f199_v1_step00000000/c49a14de01d7bd35.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14297 |
results/re10k_short_reverse_dememwm_full_monitor8_train731edd42758e_eval940b6fb72bb1_step70000_job81244/videos/test_loop_true_reverse_h100_f199_v1_step00000000/f287c679c250a841.mp4 filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14295 |
results/re10k_short_reverse_dememwm_full_monitor8_train731edd42758e_eval940b6fb72bb1_step70000_job81244/videos/test_loop_true_reverse_h100_f199_v1_step00000000/bcc76b58e2585320.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14296 |
results/re10k_short_reverse_dememwm_full_monitor8_train731edd42758e_eval940b6fb72bb1_step70000_job81244/videos/test_loop_true_reverse_h100_f199_v1_step00000000/c49a14de01d7bd35.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14297 |
results/re10k_short_reverse_dememwm_full_monitor8_train731edd42758e_eval940b6fb72bb1_step70000_job81244/videos/test_loop_true_reverse_h100_f199_v1_step00000000/f287c679c250a841.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14298 |
+
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/0ece034988793847.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14299 |
+
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/1735d8d1b4015669.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14300 |
+
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/1cdfd3abfcc3a64e.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14301 |
+
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/6afff9609b65a604.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14302 |
+
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/b009fbe3950ebeb2.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14303 |
+
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/bcc76b58e2585320.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14304 |
+
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/c49a14de01d7bd35.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 14305 |
+
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/f287c679c250a841.mp4 filter=lfs diff=lfs merge=lfs -text
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/metrics/test_history.csv
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
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_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_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,frame_bin_size,frame_bin_interval
|
| 2 |
+
worldmem,test,0,50,8,800,0.08874138444662094,12.040312767028809,0.3002903461456299,0.6288661360740662,166.0,0.09402408450841904,10.716129302978516,0.30140751600265503,0.8330078125,0.03203858807682991,17.623973846435547,0.44365641474723816,0.275991827249527,76.0,0.09714064002037048,10.674241065979004,0.2576228082180023,0.6598437428474426,198.0,0.12046549469232559,9.604694366455078,0.22269946336746216,0.7605859637260437,249.0,0.10532081872224808,10.258338928222656,0.2771826982498169,0.819042980670929,272.0,0.06458961218595505,14.14910888671875,0.35063961148262024,0.467917799949646,123.0,0.11289317160844803,9.931517601013184,0.24994109570980072,0.7898144721984863,246.0,25,"[start,end) generated-frame offset"
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/metrics/test_monitor8_loop_true_reverse_h100_f199_v1_step00000000_metrics.json
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"artifact_type": "re10k_benchmark_metrics",
|
| 3 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 4 |
+
"case_count": 8,
|
| 5 |
+
"generated_frames": 800,
|
| 6 |
+
"global_step": 0,
|
| 7 |
+
"memory_mode": "correct",
|
| 8 |
+
"method": "worldmem",
|
| 9 |
+
"metrics": {
|
| 10 |
+
"final/lpips": 0.8330078125,
|
| 11 |
+
"final/mse": 0.09402408450841904,
|
| 12 |
+
"final/psnr": 10.716129302978516,
|
| 13 |
+
"final/ssim": 0.30140751600265503,
|
| 14 |
+
"frames_000_025/lpips": 0.275991827249527,
|
| 15 |
+
"frames_000_025/mse": 0.03203858807682991,
|
| 16 |
+
"frames_000_025/psnr": 17.623973846435547,
|
| 17 |
+
"frames_000_025/rfid": 76.0,
|
| 18 |
+
"frames_000_025/ssim": 0.44365641474723816,
|
| 19 |
+
"frames_000_050/lpips": 0.467917799949646,
|
| 20 |
+
"frames_000_050/mse": 0.06458961218595505,
|
| 21 |
+
"frames_000_050/psnr": 14.14910888671875,
|
| 22 |
+
"frames_000_050/rfid": 123.0,
|
| 23 |
+
"frames_000_050/ssim": 0.35063961148262024,
|
| 24 |
+
"frames_025_050/lpips": 0.6598437428474426,
|
| 25 |
+
"frames_025_050/mse": 0.09714064002037048,
|
| 26 |
+
"frames_025_050/psnr": 10.674241065979004,
|
| 27 |
+
"frames_025_050/rfid": 198.0,
|
| 28 |
+
"frames_025_050/ssim": 0.2576228082180023,
|
| 29 |
+
"frames_050_075/lpips": 0.7605859637260437,
|
| 30 |
+
"frames_050_075/mse": 0.12046549469232559,
|
| 31 |
+
"frames_050_075/psnr": 9.604694366455078,
|
| 32 |
+
"frames_050_075/rfid": 249.0,
|
| 33 |
+
"frames_050_075/ssim": 0.22269946336746216,
|
| 34 |
+
"frames_050_100/lpips": 0.7898144721984863,
|
| 35 |
+
"frames_050_100/mse": 0.11289317160844803,
|
| 36 |
+
"frames_050_100/psnr": 9.931517601013184,
|
| 37 |
+
"frames_050_100/rfid": 246.0,
|
| 38 |
+
"frames_050_100/ssim": 0.24994109570980072,
|
| 39 |
+
"frames_075_100/lpips": 0.819042980670929,
|
| 40 |
+
"frames_075_100/mse": 0.10532081872224808,
|
| 41 |
+
"frames_075_100/psnr": 10.258338928222656,
|
| 42 |
+
"frames_075_100/rfid": 272.0,
|
| 43 |
+
"frames_075_100/ssim": 0.2771826982498169,
|
| 44 |
+
"overall/lpips": 0.6288661360740662,
|
| 45 |
+
"overall/mse": 0.08874138444662094,
|
| 46 |
+
"overall/psnr": 12.040312767028809,
|
| 47 |
+
"overall/rfid": 166.0,
|
| 48 |
+
"overall/ssim": 0.3002903461456299
|
| 49 |
+
},
|
| 50 |
+
"namespace": "test",
|
| 51 |
+
"sampling_steps": 50,
|
| 52 |
+
"scene_count": 8,
|
| 53 |
+
"schema_version": 1,
|
| 54 |
+
"scope": "monitor8"
|
| 55 |
+
}
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/metrics/test_step00000000.json
ADDED
|
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"final/lpips": 0.8330078125,
|
| 3 |
+
"final/mse": 0.09402408450841904,
|
| 4 |
+
"final/psnr": 10.716129302978516,
|
| 5 |
+
"final/ssim": 0.30140751600265503,
|
| 6 |
+
"frame_bin_interval": "[start,end) generated-frame offset",
|
| 7 |
+
"frame_bin_size": 25,
|
| 8 |
+
"frames_000_025/lpips": 0.275991827249527,
|
| 9 |
+
"frames_000_025/mse": 0.03203858807682991,
|
| 10 |
+
"frames_000_025/psnr": 17.623973846435547,
|
| 11 |
+
"frames_000_025/rfid": 76.0,
|
| 12 |
+
"frames_000_025/ssim": 0.44365641474723816,
|
| 13 |
+
"frames_000_050/lpips": 0.467917799949646,
|
| 14 |
+
"frames_000_050/mse": 0.06458961218595505,
|
| 15 |
+
"frames_000_050/psnr": 14.14910888671875,
|
| 16 |
+
"frames_000_050/rfid": 123.0,
|
| 17 |
+
"frames_000_050/ssim": 0.35063961148262024,
|
| 18 |
+
"frames_025_050/lpips": 0.6598437428474426,
|
| 19 |
+
"frames_025_050/mse": 0.09714064002037048,
|
| 20 |
+
"frames_025_050/psnr": 10.674241065979004,
|
| 21 |
+
"frames_025_050/rfid": 198.0,
|
| 22 |
+
"frames_025_050/ssim": 0.2576228082180023,
|
| 23 |
+
"frames_050_075/lpips": 0.7605859637260437,
|
| 24 |
+
"frames_050_075/mse": 0.12046549469232559,
|
| 25 |
+
"frames_050_075/psnr": 9.604694366455078,
|
| 26 |
+
"frames_050_075/rfid": 249.0,
|
| 27 |
+
"frames_050_075/ssim": 0.22269946336746216,
|
| 28 |
+
"frames_050_100/lpips": 0.7898144721984863,
|
| 29 |
+
"frames_050_100/mse": 0.11289317160844803,
|
| 30 |
+
"frames_050_100/psnr": 9.931517601013184,
|
| 31 |
+
"frames_050_100/rfid": 246.0,
|
| 32 |
+
"frames_050_100/ssim": 0.24994109570980072,
|
| 33 |
+
"frames_075_100/lpips": 0.819042980670929,
|
| 34 |
+
"frames_075_100/mse": 0.10532081872224808,
|
| 35 |
+
"frames_075_100/psnr": 10.258338928222656,
|
| 36 |
+
"frames_075_100/rfid": 272.0,
|
| 37 |
+
"frames_075_100/ssim": 0.2771826982498169,
|
| 38 |
+
"generated_frames": 800,
|
| 39 |
+
"global_step": 0,
|
| 40 |
+
"method": "worldmem",
|
| 41 |
+
"namespace": "test",
|
| 42 |
+
"overall/lpips": 0.6288661360740662,
|
| 43 |
+
"overall/mse": 0.08874138444662094,
|
| 44 |
+
"overall/psnr": 12.040312767028809,
|
| 45 |
+
"overall/rfid": 166.0,
|
| 46 |
+
"overall/ssim": 0.3002903461456299,
|
| 47 |
+
"sampling_steps": 50,
|
| 48 |
+
"selected_scenes": 8
|
| 49 |
+
}
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/resolved_config.yaml
ADDED
|
@@ -0,0 +1,400 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment:
|
| 2 |
+
debug: false
|
| 3 |
+
tasks:
|
| 4 |
+
- training
|
| 5 |
+
num_nodes: 1
|
| 6 |
+
training:
|
| 7 |
+
precision: bf16-mixed
|
| 8 |
+
compile: false
|
| 9 |
+
lr: 4.0e-05
|
| 10 |
+
batch_size: 4
|
| 11 |
+
max_epochs: -1
|
| 12 |
+
max_steps: 100000
|
| 13 |
+
max_time: null
|
| 14 |
+
data:
|
| 15 |
+
num_workers: 4
|
| 16 |
+
shuffle: true
|
| 17 |
+
optim:
|
| 18 |
+
accumulate_grad_batches: 1
|
| 19 |
+
gradient_clip_val: 1.0
|
| 20 |
+
checkpointing:
|
| 21 |
+
every_n_train_steps: 2500
|
| 22 |
+
every_n_epochs: null
|
| 23 |
+
train_time_interval: null
|
| 24 |
+
enable_version_counter: false
|
| 25 |
+
validation:
|
| 26 |
+
precision: bf16-mixed
|
| 27 |
+
compile: false
|
| 28 |
+
batch_size: 1
|
| 29 |
+
val_every_n_step: 2500
|
| 30 |
+
val_every_n_epoch: null
|
| 31 |
+
num_sanity_val_steps: 0
|
| 32 |
+
limit_batch: 2
|
| 33 |
+
data:
|
| 34 |
+
num_workers: 4
|
| 35 |
+
shuffle: false
|
| 36 |
+
inference_mode: false
|
| 37 |
+
test:
|
| 38 |
+
precision: 16-mixed
|
| 39 |
+
compile: false
|
| 40 |
+
batch_size: 1
|
| 41 |
+
limit_batch: 1.0
|
| 42 |
+
data:
|
| 43 |
+
num_workers: 4
|
| 44 |
+
shuffle: false
|
| 45 |
+
inference_mode: true
|
| 46 |
+
find_unused_parameters: false
|
| 47 |
+
reload_dataloaders_every_n_epochs: 1
|
| 48 |
+
ema:
|
| 49 |
+
enable: false
|
| 50 |
+
decay: 0.9999
|
| 51 |
+
validate_original_weights: false
|
| 52 |
+
_name: video_generation
|
| 53 |
+
dataset:
|
| 54 |
+
debug: false
|
| 55 |
+
save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/re10k_dfot/real-estate-10k
|
| 56 |
+
data_mean:
|
| 57 |
+
- - - 0.577
|
| 58 |
+
- - - 0.517
|
| 59 |
+
- - - 0.461
|
| 60 |
+
data_std:
|
| 61 |
+
- - - 0.249
|
| 62 |
+
- - - 0.249
|
| 63 |
+
- - - 0.268
|
| 64 |
+
latent:
|
| 65 |
+
enable: false
|
| 66 |
+
type: pre_sample
|
| 67 |
+
suffix: null
|
| 68 |
+
downsampling_factor:
|
| 69 |
+
- 1
|
| 70 |
+
- 8
|
| 71 |
+
num_channels: 4
|
| 72 |
+
resolution: 256
|
| 73 |
+
observation_shape:
|
| 74 |
+
- 3
|
| 75 |
+
- 256
|
| 76 |
+
- 256
|
| 77 |
+
max_frames: 12
|
| 78 |
+
n_frames: 12
|
| 79 |
+
context_length: 10
|
| 80 |
+
frame_skip: 1
|
| 81 |
+
filter_min_len: null
|
| 82 |
+
external_cond_dim: 16
|
| 83 |
+
external_cond_stack: false
|
| 84 |
+
external_cond_processing: null
|
| 85 |
+
preload: false
|
| 86 |
+
subdataset_size: null
|
| 87 |
+
num_eval_videos: 8
|
| 88 |
+
maximize_training_data: false
|
| 89 |
+
augmentation:
|
| 90 |
+
frame_skip_increase: 0
|
| 91 |
+
horizontal_flip_prob: 0
|
| 92 |
+
reverse_prob: 0
|
| 93 |
+
back_and_forth_prob: 0
|
| 94 |
+
local_window_frames: 8
|
| 95 |
+
training_frame_skip: 1
|
| 96 |
+
training_history_frames: 10
|
| 97 |
+
eval_history_frames: 100
|
| 98 |
+
eval_min_generated_frames: 100
|
| 99 |
+
eval_max_generated_frames: 100
|
| 100 |
+
video_fps: 10.0
|
| 101 |
+
seed: 0
|
| 102 |
+
memory_selection:
|
| 103 |
+
enabled: true
|
| 104 |
+
selector_version: re10k_faithful_trajectory_h100_f399_v1
|
| 105 |
+
method: worldmem
|
| 106 |
+
training_selection_mode: schedule_cache_direct
|
| 107 |
+
selector_cache_dir: /proj/cvl/users/x_fahkh2/DeMemWM/.exp_artifact/re10k_faithful_trajectory_h100_f399_v1/cache_worldmem_clipped_731edd42758e
|
| 108 |
+
selector_cache_source_commit: 731edd42758e7aec3b0bfd03ed52b72d716b65df
|
| 109 |
+
causal: true
|
| 110 |
+
max_anchor_frames: 0
|
| 111 |
+
max_dynamic_frames: 4
|
| 112 |
+
max_revisit_frames: 0
|
| 113 |
+
training_dropout: 0.0
|
| 114 |
+
local_context_exclusion_frames: 8
|
| 115 |
+
match_position_scale: 2.0
|
| 116 |
+
match_view_angle_degrees: 110.0
|
| 117 |
+
evaluation_manifest: /proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/re10k_real_trajectory_h100_f399_scopes_v1.csv
|
| 118 |
+
evaluation_manifest_scope: monitor8
|
| 119 |
+
evaluation_benchmark: loop_true_reverse_h100_f199_v1
|
| 120 |
+
evaluation_benchmarks:
|
| 121 |
+
- loop_true_reverse_h100_f199_v1
|
| 122 |
+
evaluation_memory_mode: correct
|
| 123 |
+
evaluation_case_counts:
|
| 124 |
+
loop_true_reverse_h100_f199_v1: 8
|
| 125 |
+
evaluation_case_order: scene_manifest_then_path_ordinal
|
| 126 |
+
evaluation_case_seed: 1000003*metadata_index+path_ordinal
|
| 127 |
+
evaluation_trajectory: G0..G99,G99..G0
|
| 128 |
+
evaluation_output_frames: 200
|
| 129 |
+
evaluation_observed_output_interval:
|
| 130 |
+
- 0
|
| 131 |
+
- 100
|
| 132 |
+
evaluation_generated_output_interval:
|
| 133 |
+
- 100
|
| 134 |
+
- 200
|
| 135 |
+
evaluation_generated_segments:
|
| 136 |
+
return:
|
| 137 |
+
- 0
|
| 138 |
+
- 100
|
| 139 |
+
return_chunks:
|
| 140 |
+
- - 0
|
| 141 |
+
- 25
|
| 142 |
+
- - 25
|
| 143 |
+
- 50
|
| 144 |
+
- - 50
|
| 145 |
+
- 75
|
| 146 |
+
- - 75
|
| 147 |
+
- 100
|
| 148 |
+
evaluation_frame_bin_size: 25
|
| 149 |
+
evaluation_turnaround_duplication: true
|
| 150 |
+
_name: video_realestate10k
|
| 151 |
+
algorithm:
|
| 152 |
+
debug: false
|
| 153 |
+
lr: 4.0e-05
|
| 154 |
+
backbone:
|
| 155 |
+
name: u_vit3d_pose
|
| 156 |
+
channels:
|
| 157 |
+
- 128
|
| 158 |
+
- 256
|
| 159 |
+
- 576
|
| 160 |
+
- 1152
|
| 161 |
+
emb_channels: 1024
|
| 162 |
+
patch_size: 2
|
| 163 |
+
block_types:
|
| 164 |
+
- ResBlock
|
| 165 |
+
- ResBlock
|
| 166 |
+
- TransformerBlock
|
| 167 |
+
- TransformerBlock
|
| 168 |
+
block_dropouts:
|
| 169 |
+
- 0.0
|
| 170 |
+
- 0.0
|
| 171 |
+
- 0.1
|
| 172 |
+
- 0.1
|
| 173 |
+
num_updown_blocks:
|
| 174 |
+
- 3
|
| 175 |
+
- 3
|
| 176 |
+
- 6
|
| 177 |
+
num_mid_blocks: 20
|
| 178 |
+
num_heads: 9
|
| 179 |
+
pos_emb_type: rope
|
| 180 |
+
use_checkpointing:
|
| 181 |
+
- false
|
| 182 |
+
- false
|
| 183 |
+
- false
|
| 184 |
+
- false
|
| 185 |
+
conditioning:
|
| 186 |
+
dim: null
|
| 187 |
+
external_cond_dropout: 0.1
|
| 188 |
+
use_fourier_noise_embedding: true
|
| 189 |
+
x_shape:
|
| 190 |
+
- 3
|
| 191 |
+
- 256
|
| 192 |
+
- 256
|
| 193 |
+
max_frames: 12
|
| 194 |
+
n_frames: 12
|
| 195 |
+
frame_skip: 1
|
| 196 |
+
context_frames: 10
|
| 197 |
+
latent:
|
| 198 |
+
enable: false
|
| 199 |
+
type: pre_sample
|
| 200 |
+
suffix: null
|
| 201 |
+
downsampling_factor:
|
| 202 |
+
- 1
|
| 203 |
+
- 8
|
| 204 |
+
num_channels: 4
|
| 205 |
+
data_mean:
|
| 206 |
+
- - - 0.577
|
| 207 |
+
- - - 0.517
|
| 208 |
+
- - - 0.461
|
| 209 |
+
data_std:
|
| 210 |
+
- - - 0.249
|
| 211 |
+
- - - 0.249
|
| 212 |
+
- - - 0.268
|
| 213 |
+
external_cond_dim: 16
|
| 214 |
+
external_cond_stack: false
|
| 215 |
+
external_cond_processing: null
|
| 216 |
+
compile: false
|
| 217 |
+
weight_decay: 0.002
|
| 218 |
+
optimizer_beta:
|
| 219 |
+
- 0.9
|
| 220 |
+
- 0.99
|
| 221 |
+
lr_scheduler:
|
| 222 |
+
name: constant_with_warmup
|
| 223 |
+
num_warmup_steps: 1000
|
| 224 |
+
num_training_steps: 100000
|
| 225 |
+
noise_level: random_independent
|
| 226 |
+
uniform_future:
|
| 227 |
+
enabled: false
|
| 228 |
+
fixed_context:
|
| 229 |
+
enabled: false
|
| 230 |
+
indices: null
|
| 231 |
+
dropout: 0
|
| 232 |
+
variable_context:
|
| 233 |
+
enabled: false
|
| 234 |
+
prob: 0
|
| 235 |
+
dropout: 0
|
| 236 |
+
chunk_size: -1
|
| 237 |
+
scheduling_matrix: full_sequence
|
| 238 |
+
replacement: noisy_scale
|
| 239 |
+
diffusion:
|
| 240 |
+
is_continuous: true
|
| 241 |
+
timesteps: 1000
|
| 242 |
+
beta_schedule: cosine_simple_diffusion
|
| 243 |
+
schedule_fn_kwargs:
|
| 244 |
+
shift: 1.0
|
| 245 |
+
shifted: 0.125
|
| 246 |
+
interpolated: false
|
| 247 |
+
use_causal_mask: false
|
| 248 |
+
clip_noise: 20.0
|
| 249 |
+
objective: pred_v
|
| 250 |
+
loss_weighting:
|
| 251 |
+
strategy: sigmoid
|
| 252 |
+
snr_clip: 5.0
|
| 253 |
+
cum_snr_decay: 0.9
|
| 254 |
+
sigmoid_bias: -1.0
|
| 255 |
+
sampling_timesteps: 50
|
| 256 |
+
ddim_sampling_eta: 0.0
|
| 257 |
+
reconstruction_guidance: 0.0
|
| 258 |
+
training_schedule:
|
| 259 |
+
name: cosine
|
| 260 |
+
shift: 0.125
|
| 261 |
+
precond_scale: 0.125
|
| 262 |
+
vae:
|
| 263 |
+
pretrained_path: null
|
| 264 |
+
pretrained_kwargs: {}
|
| 265 |
+
use_fp16: true
|
| 266 |
+
batch_size: 2
|
| 267 |
+
checkpoint:
|
| 268 |
+
reset_optimizer: false
|
| 269 |
+
strict: true
|
| 270 |
+
tasks:
|
| 271 |
+
prediction:
|
| 272 |
+
enabled: false
|
| 273 |
+
history_guidance:
|
| 274 |
+
name: conditional
|
| 275 |
+
keyframe_density: null
|
| 276 |
+
sliding_context_len: null
|
| 277 |
+
interpolation:
|
| 278 |
+
enabled: false
|
| 279 |
+
history_guidance:
|
| 280 |
+
name: conditional
|
| 281 |
+
max_batch_size: null
|
| 282 |
+
logging:
|
| 283 |
+
deterministic: 0
|
| 284 |
+
loss_freq: 1
|
| 285 |
+
grad_norm_freq: 0
|
| 286 |
+
max_num_videos: 0
|
| 287 |
+
n_metrics_frames: null
|
| 288 |
+
metrics: []
|
| 289 |
+
metrics_batch_size: 16
|
| 290 |
+
sanity_generation: false
|
| 291 |
+
raw_dir: null
|
| 292 |
+
camera_pose_conditioning:
|
| 293 |
+
normalize_by: first
|
| 294 |
+
bound: null
|
| 295 |
+
type: ray_encoding
|
| 296 |
+
memory:
|
| 297 |
+
local_window_frames: 8
|
| 298 |
+
memory_selection:
|
| 299 |
+
enabled: true
|
| 300 |
+
selector_version: re10k_faithful_trajectory_h100_f399_v1
|
| 301 |
+
method: worldmem
|
| 302 |
+
training_selection_mode: schedule_cache_direct
|
| 303 |
+
selector_cache_dir: /proj/cvl/users/x_fahkh2/DeMemWM/.exp_artifact/re10k_faithful_trajectory_h100_f399_v1/cache_worldmem_clipped_731edd42758e
|
| 304 |
+
selector_cache_source_commit: 731edd42758e7aec3b0bfd03ed52b72d716b65df
|
| 305 |
+
causal: true
|
| 306 |
+
max_anchor_frames: 0
|
| 307 |
+
max_dynamic_frames: 4
|
| 308 |
+
max_revisit_frames: 0
|
| 309 |
+
training_dropout: 0.0
|
| 310 |
+
local_context_exclusion_frames: 8
|
| 311 |
+
match_position_scale: 2.0
|
| 312 |
+
match_view_angle_degrees: 110.0
|
| 313 |
+
memory_projection_mode: worldmem_reference_v1
|
| 314 |
+
memory_ffn_ratio: 2
|
| 315 |
+
memory_conditioning_input_dim: 1024
|
| 316 |
+
memory_conditioning_hidden_dim: 128
|
| 317 |
+
memory_role_embedding: false
|
| 318 |
+
memory_source_embedding: false
|
| 319 |
+
memory_stream_count_correction: false
|
| 320 |
+
memory_lora_rank: 8
|
| 321 |
+
memory_lora_alpha: 8
|
| 322 |
+
memory_output_gate_init: 0.0
|
| 323 |
+
memory_camera_hidden_dim: 128
|
| 324 |
+
memory_anchor_zero_geometry: false
|
| 325 |
+
memory_age_conditioning: false
|
| 326 |
+
reference_timestamp_conditioning: true
|
| 327 |
+
memory_attention_key_only_geometry: false
|
| 328 |
+
memory_noise:
|
| 329 |
+
enabled: false
|
| 330 |
+
training_seed: 0
|
| 331 |
+
training:
|
| 332 |
+
max_steps: 100000
|
| 333 |
+
num_gpus: 8
|
| 334 |
+
batch_size_per_gpu: 4
|
| 335 |
+
accumulate_grad_batches: 1
|
| 336 |
+
learning_rate: 4.0e-05
|
| 337 |
+
memory_learning_rate_stage_a: 4.0e-05
|
| 338 |
+
memory_learning_rate_stage_b: 2.0e-05
|
| 339 |
+
lora_learning_rate_stage_b: 5.0e-06
|
| 340 |
+
stage_b_start_step: 8000
|
| 341 |
+
stage_c_start_step: 30000
|
| 342 |
+
min_lr_ratio: 0.1
|
| 343 |
+
optimizer_betas:
|
| 344 |
+
- 0.9
|
| 345 |
+
- 0.99
|
| 346 |
+
weight_decay: 0.002
|
| 347 |
+
gradient_clip_val: 1.0
|
| 348 |
+
warmup_steps: 1000
|
| 349 |
+
ema_decay: 0.9999
|
| 350 |
+
frame_skip: 1
|
| 351 |
+
checkpoint_every_n_steps: 2500
|
| 352 |
+
validation_every_n_steps: 2500
|
| 353 |
+
monitoring_steps:
|
| 354 |
+
- 2500
|
| 355 |
+
- 5000
|
| 356 |
+
- 7500
|
| 357 |
+
- 10000
|
| 358 |
+
- 12500
|
| 359 |
+
- 15000
|
| 360 |
+
- 20000
|
| 361 |
+
- 25000
|
| 362 |
+
- 30000
|
| 363 |
+
- 35000
|
| 364 |
+
- 40000
|
| 365 |
+
- 50000
|
| 366 |
+
- 60000
|
| 367 |
+
- 70000
|
| 368 |
+
- 80000
|
| 369 |
+
- 90000
|
| 370 |
+
- 100000
|
| 371 |
+
evaluation:
|
| 372 |
+
selector_mode: online
|
| 373 |
+
benchmarks:
|
| 374 |
+
- loop_true_reverse_h100_f199_v1
|
| 375 |
+
benchmark: loop_true_reverse_h100_f199_v1
|
| 376 |
+
history_frames: 100
|
| 377 |
+
min_generated_frames: 100
|
| 378 |
+
max_generated_frames: 100
|
| 379 |
+
frame_skip: 1
|
| 380 |
+
sampling_steps: 50
|
| 381 |
+
deterministic_seed: 0
|
| 382 |
+
history_guidance:
|
| 383 |
+
name: conditional
|
| 384 |
+
reverse_loop_probe: false
|
| 385 |
+
turnaround_duplication: true
|
| 386 |
+
frame_bin_size: 25
|
| 387 |
+
num_videos: 8
|
| 388 |
+
memory_mode: correct
|
| 389 |
+
manifest_scope: monitor8
|
| 390 |
+
active_benchmarks:
|
| 391 |
+
- loop_true_reverse_h100_f199_v1
|
| 392 |
+
_name: dfot_worldmem_realestate10k
|
| 393 |
+
debug: false
|
| 394 |
+
wandb:
|
| 395 |
+
entity: turlin
|
| 396 |
+
project: worldmem
|
| 397 |
+
mode: offline
|
| 398 |
+
resume: null
|
| 399 |
+
load: null
|
| 400 |
+
name: dfot_worldmem_realestate10k
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/run_manifest.json
ADDED
|
@@ -0,0 +1,185 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"method": "worldmem",
|
| 3 |
+
"dataset_root": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/re10k_dfot/real-estate-10k",
|
| 4 |
+
"dfot_root": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/DFoT-code",
|
| 5 |
+
"dfot_commit": "530f8bf4db91a21964993f214fa4050dbd529591",
|
| 6 |
+
"checkpoint": "/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/DFoT_RE10K.ckpt",
|
| 7 |
+
"local_frames": 8,
|
| 8 |
+
"memory_frames": {
|
| 9 |
+
"anchor": 0,
|
| 10 |
+
"dynamic": 4,
|
| 11 |
+
"revisit": 0
|
| 12 |
+
},
|
| 13 |
+
"training_selector_version": "re10k_faithful_trajectory_h100_f399_v1",
|
| 14 |
+
"training_selector_method": "worldmem",
|
| 15 |
+
"training_selector_cache_source_commit": "731edd42758e7aec3b0bfd03ed52b72d716b65df",
|
| 16 |
+
"memory_output_gate_init": 0.0,
|
| 17 |
+
"batch_size_per_gpu": 4,
|
| 18 |
+
"accumulate_grad_batches": 1,
|
| 19 |
+
"effective_global_batch_size": 32,
|
| 20 |
+
"optimizer_lr_schedule": {
|
| 21 |
+
"name": "stagewise_warmup_stable_cosine_floor",
|
| 22 |
+
"warmup_steps": 1000,
|
| 23 |
+
"stage_b_start_step": 8000,
|
| 24 |
+
"decay_start_step": 30000,
|
| 25 |
+
"terminal_step": 100000,
|
| 26 |
+
"min_lr_ratio": 0.1
|
| 27 |
+
},
|
| 28 |
+
"runtime": {
|
| 29 |
+
"gradient_checkpointing": [
|
| 30 |
+
false,
|
| 31 |
+
false,
|
| 32 |
+
false,
|
| 33 |
+
false
|
| 34 |
+
],
|
| 35 |
+
"split_frozen_memory_stream": false,
|
| 36 |
+
"structural_memory_autograd": true
|
| 37 |
+
},
|
| 38 |
+
"wandb": {
|
| 39 |
+
"mode": "offline",
|
| 40 |
+
"entity": "turlin",
|
| 41 |
+
"project": "worldmem",
|
| 42 |
+
"run_id": null
|
| 43 |
+
},
|
| 44 |
+
"evaluation_history_frames": 100,
|
| 45 |
+
"evaluation_max_generated_frames": 100,
|
| 46 |
+
"evaluation_num_videos": 8,
|
| 47 |
+
"evaluation_sampling_steps": 50,
|
| 48 |
+
"evaluation_metrics": [
|
| 49 |
+
"mse",
|
| 50 |
+
"psnr",
|
| 51 |
+
"ssim",
|
| 52 |
+
"lpips",
|
| 53 |
+
"rfid"
|
| 54 |
+
],
|
| 55 |
+
"training_time_eval_batches_per_rank": 2,
|
| 56 |
+
"post_training_evaluation": true,
|
| 57 |
+
"faithful_profile": {
|
| 58 |
+
"method": "worldmem",
|
| 59 |
+
"capacity": "clipped",
|
| 60 |
+
"selector_version": "re10k_faithful_trajectory_h100_f399_v1",
|
| 61 |
+
"lifecycle": "group1_59306102"
|
| 62 |
+
},
|
| 63 |
+
"trainable_parameter_manifest": "/proj/cvl/users/x_fahkh2/DeMemWM/results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/trainable_parameter_manifest.json",
|
| 64 |
+
"training_batch_composition": {
|
| 65 |
+
"global_natural": 8,
|
| 66 |
+
"global_loop": 24,
|
| 67 |
+
"rank_order": [
|
| 68 |
+
"natural",
|
| 69 |
+
"loop",
|
| 70 |
+
"loop",
|
| 71 |
+
"loop"
|
| 72 |
+
]
|
| 73 |
+
},
|
| 74 |
+
"training_trajectory": {
|
| 75 |
+
"source_mapping": "physical_start+[0..199,199..0]",
|
| 76 |
+
"virtual_positions": 400,
|
| 77 |
+
"query_frames": 8,
|
| 78 |
+
"phase_bounds_half_open": [
|
| 79 |
+
[
|
| 80 |
+
100,
|
| 81 |
+
159
|
| 82 |
+
],
|
| 83 |
+
[
|
| 84 |
+
159,
|
| 85 |
+
218
|
| 86 |
+
],
|
| 87 |
+
[
|
| 88 |
+
218,
|
| 89 |
+
277
|
| 90 |
+
],
|
| 91 |
+
[
|
| 92 |
+
277,
|
| 93 |
+
335
|
| 94 |
+
],
|
| 95 |
+
[
|
| 96 |
+
335,
|
| 97 |
+
393
|
| 98 |
+
]
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"evaluation_weight_source": "ema",
|
| 102 |
+
"post_training_checkpoint_policy": null,
|
| 103 |
+
"worldmem_git": {
|
| 104 |
+
"branch": "re10k-dfot-integration",
|
| 105 |
+
"head": "940b6fb72bb1543b1e5bfad72baf837140a603f0",
|
| 106 |
+
"status": [
|
| 107 |
+
"?? results/"
|
| 108 |
+
]
|
| 109 |
+
},
|
| 110 |
+
"profile_stage_b": false,
|
| 111 |
+
"schedule_step_offset": 0,
|
| 112 |
+
"algorithm_config": "/proj/cvl/users/x_fahkh2/DeMemWM/configurations/algorithm/dfot_worldmem_faithful_f200b200_clipped_realestate10k.yaml",
|
| 113 |
+
"training_seed": 0,
|
| 114 |
+
"evaluation_seed": 0,
|
| 115 |
+
"training_seed_label": "seed0",
|
| 116 |
+
"checkpoint_kind": "ema-v2",
|
| 117 |
+
"evaluation_checkpoint": "/proj/cvl/users/x_fahkh2/DeMemWM/results/dfot_worldmem_clipped_f200b200_seed0_g32_100k_731edd42758e/checkpoints/step=30000.ckpt",
|
| 118 |
+
"evaluation_manifest": "/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/re10k_real_trajectory_h100_f399_scopes_v1.csv",
|
| 119 |
+
"evaluation_manifest_scope": "monitor8",
|
| 120 |
+
"evaluation_benchmark": "loop_true_reverse_h100_f199_v1",
|
| 121 |
+
"evaluation_benchmarks": [
|
| 122 |
+
"loop_true_reverse_h100_f199_v1"
|
| 123 |
+
],
|
| 124 |
+
"evaluation_memory_mode": "correct",
|
| 125 |
+
"evaluation_case_counts": {
|
| 126 |
+
"loop_true_reverse_h100_f199_v1": 8
|
| 127 |
+
},
|
| 128 |
+
"evaluation_case_order": "scene_manifest_then_path_ordinal",
|
| 129 |
+
"evaluation_case_seed": "1000003*metadata_index+path_ordinal",
|
| 130 |
+
"evaluation_trajectory": "G0..G99,G99..G0",
|
| 131 |
+
"evaluation_output_frames": 200,
|
| 132 |
+
"evaluation_observed_output_interval": [
|
| 133 |
+
0,
|
| 134 |
+
100
|
| 135 |
+
],
|
| 136 |
+
"evaluation_generated_output_interval": [
|
| 137 |
+
100,
|
| 138 |
+
200
|
| 139 |
+
],
|
| 140 |
+
"evaluation_generated_segments": {
|
| 141 |
+
"return": [
|
| 142 |
+
0,
|
| 143 |
+
100
|
| 144 |
+
],
|
| 145 |
+
"return_chunks": [
|
| 146 |
+
[
|
| 147 |
+
0,
|
| 148 |
+
25
|
| 149 |
+
],
|
| 150 |
+
[
|
| 151 |
+
25,
|
| 152 |
+
50
|
| 153 |
+
],
|
| 154 |
+
[
|
| 155 |
+
50,
|
| 156 |
+
75
|
| 157 |
+
],
|
| 158 |
+
[
|
| 159 |
+
75,
|
| 160 |
+
100
|
| 161 |
+
]
|
| 162 |
+
]
|
| 163 |
+
},
|
| 164 |
+
"evaluation_frame_bin_size": 25,
|
| 165 |
+
"evaluation_turnaround_duplication": true,
|
| 166 |
+
"monitoring_steps": [
|
| 167 |
+
2500,
|
| 168 |
+
5000,
|
| 169 |
+
7500,
|
| 170 |
+
10000,
|
| 171 |
+
12500,
|
| 172 |
+
15000,
|
| 173 |
+
20000,
|
| 174 |
+
25000,
|
| 175 |
+
30000,
|
| 176 |
+
35000,
|
| 177 |
+
40000,
|
| 178 |
+
50000,
|
| 179 |
+
60000,
|
| 180 |
+
70000,
|
| 181 |
+
80000,
|
| 182 |
+
90000,
|
| 183 |
+
100000
|
| 184 |
+
]
|
| 185 |
+
}
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/test_loop_true_reverse_h100_f199_v1_case_manifest.json
ADDED
|
@@ -0,0 +1,120 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"split": "test",
|
| 3 |
+
"evaluation_seed": 0,
|
| 4 |
+
"manifest": "/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/re10k_real_trajectory_h100_f399_scopes_v1.csv",
|
| 5 |
+
"manifest_scope": "monitor8",
|
| 6 |
+
"history_frames": 100,
|
| 7 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 8 |
+
"scene_count": 8,
|
| 9 |
+
"scene_order": [
|
| 10 |
+
{
|
| 11 |
+
"metadata_index": 4865,
|
| 12 |
+
"stem": "b009fbe3950ebeb2"
|
| 13 |
+
},
|
| 14 |
+
{
|
| 15 |
+
"metadata_index": 418,
|
| 16 |
+
"stem": "0ece034988793847"
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"metadata_index": 5224,
|
| 20 |
+
"stem": "bcc76b58e2585320"
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"metadata_index": 640,
|
| 24 |
+
"stem": "1735d8d1b4015669"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"metadata_index": 2952,
|
| 28 |
+
"stem": "6afff9609b65a604"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"metadata_index": 5459,
|
| 32 |
+
"stem": "c49a14de01d7bd35"
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"metadata_index": 6743,
|
| 36 |
+
"stem": "f287c679c250a841"
|
| 37 |
+
},
|
| 38 |
+
{
|
| 39 |
+
"metadata_index": 806,
|
| 40 |
+
"stem": "1cdfd3abfcc3a64e"
|
| 41 |
+
}
|
| 42 |
+
],
|
| 43 |
+
"case_count": 8,
|
| 44 |
+
"case_order": "scene_manifest_then_path_ordinal",
|
| 45 |
+
"total_generated_frames": 800,
|
| 46 |
+
"cases": [
|
| 47 |
+
{
|
| 48 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 49 |
+
"metadata_index": 4865,
|
| 50 |
+
"scene_id": "b009fbe3950ebeb2",
|
| 51 |
+
"path_name": "true_trajectory_g0_g99_g99_g0",
|
| 52 |
+
"path_ordinal": 8,
|
| 53 |
+
"generated_frames": 100,
|
| 54 |
+
"case_seed": 4865014603
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 58 |
+
"metadata_index": 418,
|
| 59 |
+
"scene_id": "0ece034988793847",
|
| 60 |
+
"path_name": "true_trajectory_g0_g99_g99_g0",
|
| 61 |
+
"path_ordinal": 8,
|
| 62 |
+
"generated_frames": 100,
|
| 63 |
+
"case_seed": 418001262
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 67 |
+
"metadata_index": 5224,
|
| 68 |
+
"scene_id": "bcc76b58e2585320",
|
| 69 |
+
"path_name": "true_trajectory_g0_g99_g99_g0",
|
| 70 |
+
"path_ordinal": 8,
|
| 71 |
+
"generated_frames": 100,
|
| 72 |
+
"case_seed": 5224015680
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 76 |
+
"metadata_index": 640,
|
| 77 |
+
"scene_id": "1735d8d1b4015669",
|
| 78 |
+
"path_name": "true_trajectory_g0_g99_g99_g0",
|
| 79 |
+
"path_ordinal": 8,
|
| 80 |
+
"generated_frames": 100,
|
| 81 |
+
"case_seed": 640001928
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 85 |
+
"metadata_index": 2952,
|
| 86 |
+
"scene_id": "6afff9609b65a604",
|
| 87 |
+
"path_name": "true_trajectory_g0_g99_g99_g0",
|
| 88 |
+
"path_ordinal": 8,
|
| 89 |
+
"generated_frames": 100,
|
| 90 |
+
"case_seed": 2952008864
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 94 |
+
"metadata_index": 5459,
|
| 95 |
+
"scene_id": "c49a14de01d7bd35",
|
| 96 |
+
"path_name": "true_trajectory_g0_g99_g99_g0",
|
| 97 |
+
"path_ordinal": 8,
|
| 98 |
+
"generated_frames": 100,
|
| 99 |
+
"case_seed": 5459016385
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 103 |
+
"metadata_index": 6743,
|
| 104 |
+
"scene_id": "f287c679c250a841",
|
| 105 |
+
"path_name": "true_trajectory_g0_g99_g99_g0",
|
| 106 |
+
"path_ordinal": 8,
|
| 107 |
+
"generated_frames": 100,
|
| 108 |
+
"case_seed": 6743020237
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"benchmark": "loop_true_reverse_h100_f199_v1",
|
| 112 |
+
"metadata_index": 806,
|
| 113 |
+
"scene_id": "1cdfd3abfcc3a64e",
|
| 114 |
+
"path_name": "true_trajectory_g0_g99_g99_g0",
|
| 115 |
+
"path_ordinal": 8,
|
| 116 |
+
"generated_frames": 100,
|
| 117 |
+
"case_seed": 806002426
|
| 118 |
+
}
|
| 119 |
+
]
|
| 120 |
+
}
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/test_scene_manifest.json
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"split": "test",
|
| 3 |
+
"evaluation_seed": 0,
|
| 4 |
+
"manifest": "/proj/cvl/users/x_fahkh2/DeMemWM/evaluation/manifests/re10k_real_trajectory_h100_f399_scopes_v1.csv",
|
| 5 |
+
"manifest_scope": "monitor8",
|
| 6 |
+
"history_frames": 100,
|
| 7 |
+
"scene_count": 8,
|
| 8 |
+
"scenes": [
|
| 9 |
+
{
|
| 10 |
+
"metadata_index": 4865,
|
| 11 |
+
"stem": "b009fbe3950ebeb2"
|
| 12 |
+
},
|
| 13 |
+
{
|
| 14 |
+
"metadata_index": 418,
|
| 15 |
+
"stem": "0ece034988793847"
|
| 16 |
+
},
|
| 17 |
+
{
|
| 18 |
+
"metadata_index": 5224,
|
| 19 |
+
"stem": "bcc76b58e2585320"
|
| 20 |
+
},
|
| 21 |
+
{
|
| 22 |
+
"metadata_index": 640,
|
| 23 |
+
"stem": "1735d8d1b4015669"
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"metadata_index": 2952,
|
| 27 |
+
"stem": "6afff9609b65a604"
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"metadata_index": 5459,
|
| 31 |
+
"stem": "c49a14de01d7bd35"
|
| 32 |
+
},
|
| 33 |
+
{
|
| 34 |
+
"metadata_index": 6743,
|
| 35 |
+
"stem": "f287c679c250a841"
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"metadata_index": 806,
|
| 39 |
+
"stem": "1cdfd3abfcc3a64e"
|
| 40 |
+
}
|
| 41 |
+
]
|
| 42 |
+
}
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/trainable_parameter_manifest.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/0ece034988793847.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8c30cc49297bb760504ed4df2f6d3e9fa66358444d262e00c5318c4cbb0fdbe8
|
| 3 |
+
size 1242094
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/1735d8d1b4015669.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bb9f0b0b55b4520adbdf265f2ee8082654cd9f4254df185c9ef028ba2731181d
|
| 3 |
+
size 1492655
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/1cdfd3abfcc3a64e.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:676b339a74b6655b970d43ac64a46e2f843ed3be2f8990803299f3a566c9c9d5
|
| 3 |
+
size 2546246
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/6afff9609b65a604.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:116944e329855fa9964d50515303eee3336da054413845063021a83d74b1798b
|
| 3 |
+
size 1569548
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/b009fbe3950ebeb2.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:297d0ef380a093091e820caab4ac38b8bce23f6f9c2aec87be28b5c29c0be31a
|
| 3 |
+
size 1759351
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/bcc76b58e2585320.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f155fa6d1e071beac028fe98bd0aea4852d60c4fce5cbd4d09e177b4c0c7dc67
|
| 3 |
+
size 1709270
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/c49a14de01d7bd35.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d505890e93585e21a80f0c158cd24446fa71b448d9123f26d52bd0dd43943b4b
|
| 3 |
+
size 1588155
|
results/re10k_short_reverse_worldmem_clipped_monitor8_train731edd42758e_eval940b6fb72bb1_step30000_job81245/videos/test_loop_true_reverse_h100_f199_v1_step00000000/f287c679c250a841.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:daf6e6199974b620994829985628ad3192f253b4aa777c90e6af24e307f2b6ef
|
| 3 |
+
size 1473196
|