hmkang commited on Sep 14

Commit

49daf3e

verified ·

1 Parent(s): ef6fe85

eval: raw output (no videos) — avg_L3 @ 60000

Browse files

Files changed (18) hide show

avg_L3/eval/checkpoint-60000/nas16/TurnOffMicrowave_PandaOmron_Env/provenance.json +13 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOffMicrowave_PandaOmron_Env/requeue_env.txt +37 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOffSinkFaucet_PandaOmron_Env/provenance.json +13 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOffSinkFaucet_PandaOmron_Env/requeue_env.txt +37 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOffStove_PandaOmron_Env/provenance.json +13 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOffStove_PandaOmron_Env/requeue_env.txt +37 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnMicrowave_PandaOmron_Env/provenance.json +13 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnMicrowave_PandaOmron_Env/requeue_env.txt +37 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/host_ram_watch.log +126 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/provenance.json +13 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/requeue_env.txt +37 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/rollout.log +21 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/server.log +117 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/vram_watch.log +236 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnStove_PandaOmron_Env/provenance.json +13 -0
avg_L3/eval/checkpoint-60000/nas16/TurnOnStove_PandaOmron_Env/requeue_env.txt +37 -0
avg_L3/eval/checkpoint-60000/nas16/TurnSinkSpout_PandaOmron_Env/provenance.json +13 -0
avg_L3/eval/checkpoint-60000/nas16/TurnSinkSpout_PandaOmron_Env/requeue_env.txt +37 -0

avg_L3/eval/checkpoint-60000/nas16/TurnOffMicrowave_PandaOmron_Env/provenance.json ADDED Viewed

	@@ -0,0 +1,13 @@

+{
+  "file_sha256": {
+    "gr00t/model/wam_dit4dit/backbone.py": "e7b6088e791a71a76dda9df9185005d19a51517004620774c747b158eaf562e5",
+    "gr00t/model/wam_dit4dit/wan22_diffusers_extractor.py": "d14f5468c0f1c0d708f70e4d83e9e55080ea988ed88e72170160e56561e8faad",
+    "gr00t/model/wam_dit4dit/wan22_extractor.py": "fc8716f82e635e6934591f7154a8a235023e78c736a87cd388a5e7b0cadc1d00",
+    "run_scripts/eval/robocasa/eval_wam_dit4dit_kitchen.sh": "71557eaaf61ca2e45cf57a5b9a51e481de97e376805444f1a92e3bbacfcad12f"
+  },
+  "git_branch": "unknown",
+  "git_commit": "unknown",
+  "git_dirty": false,
+  "hostname": "worker-node107",
+  "repo_root": "/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77"
+}

avg_L3/eval/checkpoint-60000/nas16/TurnOffMicrowave_PandaOmron_Env/requeue_env.txt ADDED Viewed

	@@ -0,0 +1,37 @@

+# generated 2026-09-14T17:23:13Z by eval_wam_dit4dit_kitchen.sh
+# job=184137 array_task=6 task=TurnOffMicrowave_PandaOmron_Env
+# consumed by scripts/wam_dit4dit/requeue_incomplete_eval.sh --run-dir <this dir's parent>
+BACKBONE_TYPE=wan22_diffusers
+CKPT_CONFIGS_OVERRIDE=/rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3:60000
+CKPT_STEP=60000
+EVAL_BENCH=robocasa_kitchen
+EVAL_OUTPUT_ROOT=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77/output
+EVAL_SUFFIX=bmethod_ctxpool_avg_L3_60k_seed42av_emaW
+EVAL_WEIGHTS=ema
+MODEL_OUTPUT_DIR=/rlwrld-unified-checkpoints/hyeonmo/wan2
+N_ENVS=1
+N_EPISODES=50
+VIDEO_DIT_OVERRIDE=
+VIDEO_DIT_OVERRIDE_HF=
+WAM_ACTION_EMA=warmup
+WAM_ALIGN_TRD=0
+WAM_ALIGN_TRD_PROBE=0
+WAM_COMPILE_BLOCKS=0
+WAM_DITACTION_EXCLUDE_CURRENT=1
+WAM_EVAL_ACTION_SEED=42
+WAM_EVAL_ENV_MODE=parallel
+WAM_EVAL_SERVER_TIMEOUT_S=1800
+WAM_EVAL_TASK_DIR_SUFFIX=
+WAM_EVAL_TREE=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_TREE_REEXEC=1
+WAM_EVAL_TREE_SHA=eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_USE_ACTION_EMA=1
+WAM_EVAL_USE_EMA=1
+WAM_EVAL_VIDEO_SEED=42
+WAM_EXTRACT_EARLY_EXIT=0
+WAM_TEXT_EMBED_CACHE=/sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt
+WAM_TEXT_EMBED_REQUIRE_HIT=1
+WAM_VIDEO_EMA=0
+WAM_VIDEO_EMA_TRAIN_SOURCE=live
+WAM_WAN_EVAL_TRUNCATE_ABOVE=0
+WAM_WAN_VAE_ENCODE_ENV_CHUNK=0

avg_L3/eval/checkpoint-60000/nas16/TurnOffSinkFaucet_PandaOmron_Env/provenance.json ADDED Viewed

	@@ -0,0 +1,13 @@

+{
+  "file_sha256": {
+    "gr00t/model/wam_dit4dit/backbone.py": "e7b6088e791a71a76dda9df9185005d19a51517004620774c747b158eaf562e5",
+    "gr00t/model/wam_dit4dit/wan22_diffusers_extractor.py": "d14f5468c0f1c0d708f70e4d83e9e55080ea988ed88e72170160e56561e8faad",
+    "gr00t/model/wam_dit4dit/wan22_extractor.py": "fc8716f82e635e6934591f7154a8a235023e78c736a87cd388a5e7b0cadc1d00",
+    "run_scripts/eval/robocasa/eval_wam_dit4dit_kitchen.sh": "71557eaaf61ca2e45cf57a5b9a51e481de97e376805444f1a92e3bbacfcad12f"
+  },
+  "git_branch": "unknown",
+  "git_commit": "unknown",
+  "git_dirty": false,
+  "hostname": "worker-node1001",
+  "repo_root": "/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77"
+}

avg_L3/eval/checkpoint-60000/nas16/TurnOffSinkFaucet_PandaOmron_Env/requeue_env.txt ADDED Viewed

	@@ -0,0 +1,37 @@

+# generated 2026-09-14T17:22:50Z by eval_wam_dit4dit_kitchen.sh
+# job=184136 array_task=5 task=TurnOffSinkFaucet_PandaOmron_Env
+# consumed by scripts/wam_dit4dit/requeue_incomplete_eval.sh --run-dir <this dir's parent>
+BACKBONE_TYPE=wan22_diffusers
+CKPT_CONFIGS_OVERRIDE=/rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3:60000
+CKPT_STEP=60000
+EVAL_BENCH=robocasa_kitchen
+EVAL_OUTPUT_ROOT=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77/output
+EVAL_SUFFIX=bmethod_ctxpool_avg_L3_60k_seed42av_emaW
+EVAL_WEIGHTS=ema
+MODEL_OUTPUT_DIR=/rlwrld-unified-checkpoints/hyeonmo/wan2
+N_ENVS=1
+N_EPISODES=50
+VIDEO_DIT_OVERRIDE=
+VIDEO_DIT_OVERRIDE_HF=
+WAM_ACTION_EMA=warmup
+WAM_ALIGN_TRD=0
+WAM_ALIGN_TRD_PROBE=0
+WAM_COMPILE_BLOCKS=0
+WAM_DITACTION_EXCLUDE_CURRENT=1
+WAM_EVAL_ACTION_SEED=42
+WAM_EVAL_ENV_MODE=parallel
+WAM_EVAL_SERVER_TIMEOUT_S=1800
+WAM_EVAL_TASK_DIR_SUFFIX=
+WAM_EVAL_TREE=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_TREE_REEXEC=1
+WAM_EVAL_TREE_SHA=eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_USE_ACTION_EMA=1
+WAM_EVAL_USE_EMA=1
+WAM_EVAL_VIDEO_SEED=42
+WAM_EXTRACT_EARLY_EXIT=0
+WAM_TEXT_EMBED_CACHE=/sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt
+WAM_TEXT_EMBED_REQUIRE_HIT=1
+WAM_VIDEO_EMA=0
+WAM_VIDEO_EMA_TRAIN_SOURCE=live
+WAM_WAN_EVAL_TRUNCATE_ABOVE=0
+WAM_WAN_VAE_ENCODE_ENV_CHUNK=0

avg_L3/eval/checkpoint-60000/nas16/TurnOffStove_PandaOmron_Env/provenance.json ADDED Viewed

	@@ -0,0 +1,13 @@

+{
+  "file_sha256": {
+    "gr00t/model/wam_dit4dit/backbone.py": "e7b6088e791a71a76dda9df9185005d19a51517004620774c747b158eaf562e5",
+    "gr00t/model/wam_dit4dit/wan22_diffusers_extractor.py": "d14f5468c0f1c0d708f70e4d83e9e55080ea988ed88e72170160e56561e8faad",
+    "gr00t/model/wam_dit4dit/wan22_extractor.py": "fc8716f82e635e6934591f7154a8a235023e78c736a87cd388a5e7b0cadc1d00",
+    "run_scripts/eval/robocasa/eval_wam_dit4dit_kitchen.sh": "71557eaaf61ca2e45cf57a5b9a51e481de97e376805444f1a92e3bbacfcad12f"
+  },
+  "git_branch": "unknown",
+  "git_commit": "unknown",
+  "git_dirty": false,
+  "hostname": "worker-node107",
+  "repo_root": "/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77"
+}

avg_L3/eval/checkpoint-60000/nas16/TurnOffStove_PandaOmron_Env/requeue_env.txt ADDED Viewed

	@@ -0,0 +1,37 @@

+# generated 2026-09-14T17:21:36Z by eval_wam_dit4dit_kitchen.sh
+# job=184135 array_task=4 task=TurnOffStove_PandaOmron_Env
+# consumed by scripts/wam_dit4dit/requeue_incomplete_eval.sh --run-dir <this dir's parent>
+BACKBONE_TYPE=wan22_diffusers
+CKPT_CONFIGS_OVERRIDE=/rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3:60000
+CKPT_STEP=60000
+EVAL_BENCH=robocasa_kitchen
+EVAL_OUTPUT_ROOT=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77/output
+EVAL_SUFFIX=bmethod_ctxpool_avg_L3_60k_seed42av_emaW
+EVAL_WEIGHTS=ema
+MODEL_OUTPUT_DIR=/rlwrld-unified-checkpoints/hyeonmo/wan2
+N_ENVS=1
+N_EPISODES=50
+VIDEO_DIT_OVERRIDE=
+VIDEO_DIT_OVERRIDE_HF=
+WAM_ACTION_EMA=warmup
+WAM_ALIGN_TRD=0
+WAM_ALIGN_TRD_PROBE=0
+WAM_COMPILE_BLOCKS=0
+WAM_DITACTION_EXCLUDE_CURRENT=1
+WAM_EVAL_ACTION_SEED=42
+WAM_EVAL_ENV_MODE=parallel
+WAM_EVAL_SERVER_TIMEOUT_S=1800
+WAM_EVAL_TASK_DIR_SUFFIX=
+WAM_EVAL_TREE=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_TREE_REEXEC=1
+WAM_EVAL_TREE_SHA=eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_USE_ACTION_EMA=1
+WAM_EVAL_USE_EMA=1
+WAM_EVAL_VIDEO_SEED=42
+WAM_EXTRACT_EARLY_EXIT=0
+WAM_TEXT_EMBED_CACHE=/sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt
+WAM_TEXT_EMBED_REQUIRE_HIT=1
+WAM_VIDEO_EMA=0
+WAM_VIDEO_EMA_TRAIN_SOURCE=live
+WAM_WAN_EVAL_TRUNCATE_ABOVE=0
+WAM_WAN_VAE_ENCODE_ENV_CHUNK=0

avg_L3/eval/checkpoint-60000/nas16/TurnOnMicrowave_PandaOmron_Env/provenance.json ADDED Viewed

	@@ -0,0 +1,13 @@

+{
+  "file_sha256": {
+    "gr00t/model/wam_dit4dit/backbone.py": "e7b6088e791a71a76dda9df9185005d19a51517004620774c747b158eaf562e5",
+    "gr00t/model/wam_dit4dit/wan22_diffusers_extractor.py": "d14f5468c0f1c0d708f70e4d83e9e55080ea988ed88e72170160e56561e8faad",
+    "gr00t/model/wam_dit4dit/wan22_extractor.py": "fc8716f82e635e6934591f7154a8a235023e78c736a87cd388a5e7b0cadc1d00",
+    "run_scripts/eval/robocasa/eval_wam_dit4dit_kitchen.sh": "71557eaaf61ca2e45cf57a5b9a51e481de97e376805444f1a92e3bbacfcad12f"
+  },
+  "git_branch": "unknown",
+  "git_commit": "unknown",
+  "git_dirty": false,
+  "hostname": "worker-node1001",
+  "repo_root": "/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77"
+}

avg_L3/eval/checkpoint-60000/nas16/TurnOnMicrowave_PandaOmron_Env/requeue_env.txt ADDED Viewed

	@@ -0,0 +1,37 @@

+# generated 2026-09-14T17:21:20Z by eval_wam_dit4dit_kitchen.sh
+# job=184134 array_task=3 task=TurnOnMicrowave_PandaOmron_Env
+# consumed by scripts/wam_dit4dit/requeue_incomplete_eval.sh --run-dir <this dir's parent>
+BACKBONE_TYPE=wan22_diffusers
+CKPT_CONFIGS_OVERRIDE=/rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3:60000
+CKPT_STEP=60000
+EVAL_BENCH=robocasa_kitchen
+EVAL_OUTPUT_ROOT=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77/output
+EVAL_SUFFIX=bmethod_ctxpool_avg_L3_60k_seed42av_emaW
+EVAL_WEIGHTS=ema
+MODEL_OUTPUT_DIR=/rlwrld-unified-checkpoints/hyeonmo/wan2
+N_ENVS=1
+N_EPISODES=50
+VIDEO_DIT_OVERRIDE=
+VIDEO_DIT_OVERRIDE_HF=
+WAM_ACTION_EMA=warmup
+WAM_ALIGN_TRD=0
+WAM_ALIGN_TRD_PROBE=0
+WAM_COMPILE_BLOCKS=0
+WAM_DITACTION_EXCLUDE_CURRENT=1
+WAM_EVAL_ACTION_SEED=42
+WAM_EVAL_ENV_MODE=parallel
+WAM_EVAL_SERVER_TIMEOUT_S=1800
+WAM_EVAL_TASK_DIR_SUFFIX=
+WAM_EVAL_TREE=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_TREE_REEXEC=1
+WAM_EVAL_TREE_SHA=eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_USE_ACTION_EMA=1
+WAM_EVAL_USE_EMA=1
+WAM_EVAL_VIDEO_SEED=42
+WAM_EXTRACT_EARLY_EXIT=0
+WAM_TEXT_EMBED_CACHE=/sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt
+WAM_TEXT_EMBED_REQUIRE_HIT=1
+WAM_VIDEO_EMA=0
+WAM_VIDEO_EMA_TRAIN_SOURCE=live
+WAM_WAN_EVAL_TRUNCATE_ABOVE=0
+WAM_WAN_VAE_ENCODE_ENV_CHUNK=0

avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/host_ram_watch.log ADDED Viewed

	@@ -0,0 +1,126 @@

+--- 2026-09-14T17:21:08Z
+VmHWM:	    7364 kB
+VmRSS:	    7364 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          95         160          70        1760        1840
+Swap:              0           0           0
+--- 2026-09-14T17:21:18Z
+VmHWM:	22337852 kB
+VmRSS:	22337852 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          99         157          67        1758        1838
+Swap:              0           0           0
+--- 2026-09-14T17:21:28Z
+VmHWM:	33817872 kB
+VmRSS:	 8923672 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          99         152          67        1763        1837
+Swap:              0           0           0
+--- 2026-09-14T17:21:38Z
+VmHWM:	33817872 kB
+VmRSS:	 6901448 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          98         143          70        1773        1835
+Swap:              0           0           0
+--- 2026-09-14T17:21:48Z
+VmHWM:	33817872 kB
+VmRSS:	 9583760 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          99         141          68        1774        1838
+Swap:              0           0           0
+--- 2026-09-14T17:21:58Z
+VmHWM:	33817872 kB
+VmRSS:	 8714532 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         102         136          67        1776        1835
+Swap:              0           0           0
+--- 2026-09-14T17:22:08Z
+VmHWM:	33817872 kB
+VmRSS:	10804236 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          99         138          67        1778        1838
+Swap:              0           0           0
+--- 2026-09-14T17:22:18Z
+VmHWM:	33817872 kB
+VmRSS:	 7922612 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          99         136          67        1780        1838
+Swap:              0           0           0
+--- 2026-09-14T17:22:28Z
+VmHWM:	33817872 kB
+VmRSS:	10125020 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         102         130          67        1782        1835
+Swap:              0           0           0
+--- 2026-09-14T17:22:38Z
+VmHWM:	33817872 kB
+VmRSS:	12166740 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          98         132          67        1784        1838
+Swap:              0           0           0
+--- 2026-09-14T17:22:48Z
+VmHWM:	33817872 kB
+VmRSS:	 9555660 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          98         130          67        1786        1838
+Swap:              0           0           0
+--- 2026-09-14T17:22:58Z
+VmHWM:	33817872 kB
+VmRSS:	11509236 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         103         123          67        1788        1834
+Swap:              0           0           0
+--- 2026-09-14T17:23:08Z
+VmHWM:	33817872 kB
+VmRSS:	 8905416 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          99         126          67        1790        1838
+Swap:              0           0           0
+--- 2026-09-14T17:23:19Z
+VmHWM:	33817872 kB
+VmRSS:	 7936144 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015          99         124          67        1791        1838
+Swap:              0           0           0
+--- 2026-09-14T17:23:29Z
+VmHWM:	33817872 kB
+VmRSS:	 7936144 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         105         118          67        1791        1832
+Swap:              0           0           0
+--- 2026-09-14T17:23:39Z
+VmHWM:	33817872 kB
+VmRSS:	 7936144 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         102         120          67        1791        1834
+Swap:              0           0           0
+--- 2026-09-14T17:23:49Z
+VmHWM:	33817872 kB
+VmRSS:	 7936144 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         104         119          67        1791        1832
+Swap:              0           0           0
+--- 2026-09-14T17:23:59Z
+VmHWM:	33817872 kB
+VmRSS:	 8256332 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         116         107          67        1791        1821
+Swap:              0           0           0
+--- 2026-09-14T17:24:09Z
+VmHWM:	33817872 kB
+VmRSS:	 8256424 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         108         115          67        1791        1829
+Swap:              0           0           0
+--- 2026-09-14T17:24:19Z
+VmHWM:	33817872 kB
+VmRSS:	 8256468 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         109         114          67        1791        1828
+Swap:              0           0           0
+--- 2026-09-14T17:24:29Z
+VmHWM:	33817872 kB
+VmRSS:	 8256308 kB
+               total        used        free      shared  buff/cache   available
+Mem:            2015         111         112          67        1791        1826
+Swap:              0           0           0

avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/provenance.json ADDED Viewed

	@@ -0,0 +1,13 @@

+{
+  "file_sha256": {
+    "gr00t/model/wam_dit4dit/backbone.py": "e7b6088e791a71a76dda9df9185005d19a51517004620774c747b158eaf562e5",
+    "gr00t/model/wam_dit4dit/wan22_diffusers_extractor.py": "d14f5468c0f1c0d708f70e4d83e9e55080ea988ed88e72170160e56561e8faad",
+    "gr00t/model/wam_dit4dit/wan22_extractor.py": "fc8716f82e635e6934591f7154a8a235023e78c736a87cd388a5e7b0cadc1d00",
+    "run_scripts/eval/robocasa/eval_wam_dit4dit_kitchen.sh": "71557eaaf61ca2e45cf57a5b9a51e481de97e376805444f1a92e3bbacfcad12f"
+  },
+  "git_branch": "unknown",
+  "git_commit": "unknown",
+  "git_dirty": false,
+  "hostname": "worker-node1001",
+  "repo_root": "/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77"
+}

avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/requeue_env.txt ADDED Viewed

	@@ -0,0 +1,37 @@

+# generated 2026-09-14T17:21:05Z by eval_wam_dit4dit_kitchen.sh
+# job=184133 array_task=2 task=TurnOnSinkFaucet_PandaOmron_Env
+# consumed by scripts/wam_dit4dit/requeue_incomplete_eval.sh --run-dir <this dir's parent>
+BACKBONE_TYPE=wan22_diffusers
+CKPT_CONFIGS_OVERRIDE=/rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3:60000
+CKPT_STEP=60000
+EVAL_BENCH=robocasa_kitchen
+EVAL_OUTPUT_ROOT=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77/output
+EVAL_SUFFIX=bmethod_ctxpool_avg_L3_60k_seed42av_emaW
+EVAL_WEIGHTS=ema
+MODEL_OUTPUT_DIR=/rlwrld-unified-checkpoints/hyeonmo/wan2
+N_ENVS=1
+N_EPISODES=50
+VIDEO_DIT_OVERRIDE=
+VIDEO_DIT_OVERRIDE_HF=
+WAM_ACTION_EMA=warmup
+WAM_ALIGN_TRD=0
+WAM_ALIGN_TRD_PROBE=0
+WAM_COMPILE_BLOCKS=0
+WAM_DITACTION_EXCLUDE_CURRENT=1
+WAM_EVAL_ACTION_SEED=42
+WAM_EVAL_ENV_MODE=parallel
+WAM_EVAL_SERVER_TIMEOUT_S=1800
+WAM_EVAL_TASK_DIR_SUFFIX=
+WAM_EVAL_TREE=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_TREE_REEXEC=1
+WAM_EVAL_TREE_SHA=eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_USE_ACTION_EMA=1
+WAM_EVAL_USE_EMA=1
+WAM_EVAL_VIDEO_SEED=42
+WAM_EXTRACT_EARLY_EXIT=0
+WAM_TEXT_EMBED_CACHE=/sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt
+WAM_TEXT_EMBED_REQUIRE_HIT=1
+WAM_VIDEO_EMA=0
+WAM_VIDEO_EMA_TRAIN_SOURCE=live
+WAM_WAN_EVAL_TRUNCATE_ABOVE=0
+WAM_WAN_VAE_ENCODE_ENV_CHUNK=0

avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/rollout.log ADDED Viewed

	@@ -0,0 +1,21 @@

+[robosuite WARNING] No private macro file found! (macros.py:57)
+[robosuite WARNING] It is recommended to use a private macro file (macros.py:58)
+[robosuite WARNING] To setup, run: python /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/gr00t/eval/sim/robocasa/robocasa_uv/.venv/lib/python3.10/site-packages/robosuite/scripts/setup_macros.py (macros.py:59)
+[robosuite WARNING] Could not import robosuite_models. Some robots may not be available. If you want to use these robots, please install robosuite_models from source (https://github.com/ARISE-Initiative/robosuite_models) or through pip install. (__init__.py:30)
+[robosuite WARNING] Could not load the mink-based whole-body IK. Make sure you install related import properly (e.g. pip install mink==0.0.5), otherwise you will not be able to use the default IK controller setting for GR1 robot. (__init__.py:40)
+[robosuite WARNING] No private macro file found! (macros.py:28)
+[robosuite WARNING] It is recommended to use a private macro file (macros.py:29)
+[robosuite WARNING] To setup, run: python /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/external_dependencies/robocasa/robocasa/scripts/setup_macros.py (macros.py:30)
+[robosuite INFO] Loading controller configuration from: /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/gr00t/eval/sim/robocasa/robocasa_uv/.venv/lib/python3.10/site-packages/robosuite/controllers/config/robots/default_pandaomron.json (composite_controller_factory.py:121)
+/sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/gr00t/eval/sim/robocasa/robocasa_uv/.venv/lib/python3.10/site-packages/gymnasium/utils/passive_env_checker.py:159: UserWarning: [33mWARN: The obs returned by the `reset()` method is not within the observation space.[0m
+  logger.warn(f"{pre} is not within the observation space.")
+Using video_delta_indices: [-24 -22 -20 -18 -16 -14 -12 -10  -8  -6  -4  -2   0]
+Using state_delta_indices: [0]
+Running collecting 50 episodes for robocasa_panda_omron/TurnOnSinkFaucet_PandaOmron_Env with 1 vec envs
+[i] Detected 0 recorded episode(s) across envs; skipping them.
+[i] Detected existing_successes: []
+[i] Creating environment robocasa_panda_omron/TurnOnSinkFaucet_PandaOmron_Env (env index: 0) with seed=42, start_episode_id=0
+WARNING: mimicgen environments not imported since mimicgen is not installed!
+[rollout_policy] Created 1 session IDs for parallel environments: ['robocasa_panda_omron/TurnOnSinkFaucet_PandaOmron_Env_env0_5363b759']
+  logger.warn(f"{pre} is not within the observation space.")

avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/server.log ADDED Viewed

	@@ -0,0 +1,117 @@

+Starting GR00T inference server...
+  Embodiment tag: EmbodimentTag.NEW_EMBODIMENT
+  Model path: /rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3/checkpoint-60000
+  Device: cuda
+  Host: 127.0.0.1
+  Port: 40903
+  Backbone type: wan22_diffusers
+  (wan22_diffusers) Cosmos video-dit-override SKIPPED; wiring umT5 text cache
+  WAM_TEXT_EMBED_CACHE: /sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt
+  WAM_TEXT_EMBED_REQUIRE_HIT: 1 (umT5 encoder freed; every prompt must hit the cache)
+[VIDEO-SEED] decided=ON seed=42 — video σ=1.0 노이즈 고정 (prepare_latents 의 전역 RNG — ★샘플별 독립 유지, c3_policy.py:1024 와 같은 모양)
+[RNG-DRIFT] after seeding, before model load: 2f0992352d445d54
+[ACTION-SEED] decided=ON seed=42 (실제 소비 배너는 첫 predict_action 에서 action_head 가 찍는다)
+[SEED-REGIME] controlled — seed42(a+v): action+video 노이즈 ★둘 다 고정 (통제)
+[DEBUG] model/__init__.py MODEL_CONFIG_TYPES={'Gr00tN1d6': <class 'gr00t.configs.model.gr00t_n1d6.Gr00tN1d6Config'>, 'EAGLE_DIT4DIT': <class 'gr00t.configs.model.eagle_dit4dit.EagleDiT4DiTConfig'>, 'AlinVLAv0': <class 'gr00t.configs.model.alinvla_v0.AlinVLAv0Config'>, 'WAM_DIT4DIT': <class 'gr00t.configs.model.wam_dit4dit.WAMDiT4DiTConfig'>}
+[DEBUG] registry.py: register_model model_cfg_cls=<class 'gr00t.configs.model.alinvla_v0.AlinVLAv0Config'>, pipeline_cls=<class 'gr00t.model.alinvla_v0.setup.AlinVLAv0Pipeline'>
+[DEBUG] registry.py: register_model model_cfg_cls=<class 'gr00t.configs.model.wam_dit4dit.WAMDiT4DiTConfig'>, pipeline_cls=<class 'gr00t.model.wam_dit4dit.setup.WAMDiT4DiTPipeline'>
+[DEBUG] registry.py: register_model model_cfg_cls=<class 'gr00t.configs.model.eagle_dit4dit.EagleDiT4DiTConfig'>, pipeline_cls=<class 'gr00t.model.eagle_dit4dit.setup.EagleDiT4DiTPipeline'>
+[Gr00tPolicy] ★WAM_EVAL_EMA_KEYSWAP=1 (default eval load path) — EMA containers are NOT built; requested EMA weights land by key-swap into the live modules after load (single backbone / single action head).
+`torch_dtype` is deprecated! Use `dtype` instead!
+[WanDiffusersBackbone] loading Wan2.2-TI2V-5B (diffusers) from 'Wan-AI/Wan2.2-TI2V-5B-Diffusers'; extract_layer=L18 (0-indexed block 18, 18/30 = 60% depth); backbone_embedding_dim=3072
+[WAN22-LOCAL] resolved to cached snapshot /sjw_alinlab3/home/hyeonmo/.cache/huggingface/hub/models--Wan-AI--Wan2.2-TI2V-5B-Diffusers/snapshots/b8fff7315c768468a5333511427288870b2e9635 (no huggingface.co API call from this process)
+The config attributes {'clip_output': False} were passed to AutoencoderKLWan, but are not expected and will be ignored. Please verify your config.json configuration file.
+[wan22-diffusers] LOAD-FINGERPRINT block param-L2(fp32): L0=276.6411(ref 276.6411), L15=327.7687(ref 327.7687), L29=284.2876(ref 284.2876) (rtol=0.005, skip=False)
+[WAM_CONTEXT_POOL] materialized 1 pooler parameter(s) off the meta device onto cuda:0 (from_pretrained init_empty_weights path)
+[WAM_CONTEXT_POOL] ENGAGED on WanTransformer3DModel: mode=avg before block L3 (n_cond_lat=4, xattn_heads=-, ffn_dim=-) | new params 3,072 | external token geometry preserved (output + block taps re-expanded)
+[wan22-diffusers] VAE env-axis micro-batch OFF (WAM_WAN_VAE_ENCODE_ENV_CHUNK unset/0 → full-batch, byte-identical)
+[Wan22DiffusersFeatureExtractor] [TRAIN-COMPUTE] armed: regime banner = ★UNCONDITIONAL (once per role at the L18 tap, every run, incl. the full per-leaf WEIGHT census); per-leaf OUTPUT sweep WAM_WAN_TRAIN_COMPUTE_PROBE='' resolved=0 (OFF — not one leaf hook registered; the regime banner still prints).
+[WAM_BLOCK_CAUSAL] resolved=0 (env='', config.wan_block_causal=False)
+[Wan22DiffusersFeatureExtractor] wan_v2v_teacher parsed: V2V history-privileged distill — student = un-sliced V2V student (backbone keeps the full history stack; forwards (a)/(b) over the same window as the teacher, per-token timestep map), teacher = _video_ema_model (WAM_VIDEO_EMA copy — reuse, no second EMA) run ONCE on the full history stack (history latent slots pinned per-token t=0, future slots σ=1.0); align scope = FUTURE-ONLY (doctrine), student tap L18 → teacher L18, γ=wan_fadistill_align_weight. `_sf_teacher`/selfflow_ema_decay/selfflow_mask_ratio are NEVER consumed on this track.
+[Wan22DiffusersFeatureExtractor] ★_sf_projector was built on the META device (low-memory loading context); materialising with to_empty + default re-init instead of failing the forward.
+[Wan22DiffusersFeatureExtractor] WAM_EXTRACT_EARLY_EXIT resolved=0 (raw='0'): OFF — every forward runs full depth (byte-identical to pre-M#0216-B).
+[LT] sequence-append resolved=0 (source=config, config.wan_learnable_token=False, env WAM_LEARNABLE_TOKEN='')
+[LAYER-TAPS] convention = 0-indexed blocks[i] (= self.blocks[i]); count-from-1 = i+1 (blocks[18] = 19th layer)
+    action head reads : blocks[18]
+[Wan22FeatureExtractor] loaded WAM_TEXT_EMBED_CACHE: 366 entries from /sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt (per-entry shape=(512, 4096), dtype=torch.bfloat16, has_masks=True, require_hit=True, ★content_fp=96dc982e488bdac8)
+[Wan22FeatureExtractor] WAM_TEXT_EMBED_REQUIRE_HIT=1 -> text_encoder + tokenizer were never loaded (build-time prescan); nothing to free.
+[TEXT-ENC] skipped (cache=/sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt, entries=366) — umT5 never loaded, never on device; every prompt must come from this cache [Wan22DiffusersFeatureExtractor]
+[wan22-diffusers] truncate-above OFF (WAM_WAN_EVAL_TRUNCATE_ABOVE unset/0 + config flag False → all transformer blocks kept, byte-identical)
+/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77/gr00t/model/modules/dit.py:205: FutureWarning: Accessing config attribute `compute_dtype` directly via 'DiT' object attribute is deprecated. Please access 'compute_dtype' over 'DiT's config object instead, e.g. 'unet.config.compute_dtype'.
+  embedding_dim=self.inner_dim, compute_dtype=self.compute_dtype
+/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77/gr00t/model/modules/dit.py:236: FutureWarning: Accessing config attribute `output_dim` directly via 'DiT' object attribute is deprecated. Please access 'output_dim' over 'DiT's config object instead, e.g. 'unet.config.output_dim'.
+  self.proj_out_2 = nn.Linear(self.inner_dim, self.output_dim)
+[CosmosBackbone] froze modules under extractor: ['vae']
+Total number of DiT parameters:  164505088
+[WAMDiT4DiT] ★WAM_EVAL_EMA_KEYSWAP: action-EMA container NOT built — EMA arrives by key-swap into the live action head at load (single head).
+Some weights of the model checkpoint at /rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3/checkpoint-60000 were not used when initializing WAMDiT4DiT: ['_dino_head.last_layer.parametrizations.weight.original0', '_dino_head.last_layer.parametrizations.weight.original1', '_dino_head.mlp.0.bias', '_dino_head.mlp.0.weight', '_dino_head.mlp.2.bias', '_dino_head.mlp.2.weight', '_dino_head.mlp.4.bias', '_dino_head.mlp.4.weight', '_dino_shadow.center', '_dino_shadow.center2', '_dino_shadow.ema.last_layer.parametrizations.weight.original0', '_dino_shadow.ema.last_layer.parametrizations.weight.original1', '_dino_shadow.ema.mlp.0.bias', '_dino_shadow.ema.mlp.0.weight', '_dino_shadow.ema.mlp.2.bias', '_dino_shadow.ema.mlp.2.weight', '_dino_shadow.ema.mlp.4.bias', '_dino_shadow.ema.mlp.4.weight', 'action_ema.model.action_decoder.layer1.bias', 'action_ema.model.action_decoder.layer1.weight', 'action_ema.model.action_decoder.layer2.bias', 'action_ema.model.action_decoder.layer2.weight', 'action_ema.model.action_encoder.layer1.bias', 'action_ema.model.action_encoder.layer1.weight', 'action_ema.model.action_encoder.layer2.bias', 'action_ema.model.action_encoder.layer2.weight', 'action_ema.model.action_encoder.layer3.bias', 'action_ema.model.action_encoder.layer3.weight', 'action_ema.model.model.proj_out_1.bias', 'action_ema.model.model.proj_out_1.weight', 'action_ema.model.model.proj_out_2.bias', 'action_ema.model.model.proj_out_2.weight', 'action_ema.model.model.timestep_encoder.timestep_embedder.linear_1.bias', 'action_ema.model.model.timestep_encoder.timestep_embedder.linear_1.weight', 'action_ema.model.model.timestep_encoder.timestep_embedder.linear_2.bias', 'action_ema.model.model.timestep_encoder.timestep_embedder.linear_2.weight', 'action_ema.model.model.transformer_blocks.0.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.0.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.0.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.0.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.0.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.0.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.0.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.0.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.0.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.0.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.0.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.0.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.0.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.0.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.1.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.1.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.1.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.1.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.1.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.1.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.1.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.1.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.1.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.1.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.1.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.1.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.1.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.1.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.10.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.10.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.10.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.10.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.10.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.10.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.10.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.10.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.10.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.10.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.10.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.10.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.10.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.10.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.11.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.11.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.11.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.11.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.11.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.11.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.11.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.11.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.11.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.11.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.11.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.11.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.11.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.11.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.12.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.12.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.12.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.12.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.12.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.12.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.12.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.12.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.12.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.12.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.12.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.12.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.12.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.12.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.13.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.13.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.13.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.13.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.13.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.13.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.13.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.13.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.13.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.13.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.13.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.13.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.13.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.13.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.14.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.14.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.14.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.14.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.14.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.14.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.14.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.14.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.14.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.14.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.14.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.14.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.14.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.14.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.15.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.15.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.15.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.15.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.15.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.15.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.15.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.15.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.15.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.15.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.15.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.15.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.15.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.15.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.2.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.2.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.2.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.2.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.2.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.2.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.2.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.2.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.2.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.2.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.2.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.2.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.2.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.2.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.3.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.3.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.3.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.3.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.3.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.3.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.3.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.3.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.3.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.3.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.3.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.3.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.3.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.3.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.4.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.4.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.4.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.4.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.4.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.4.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.4.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.4.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.4.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.4.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.4.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.4.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.4.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.4.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.5.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.5.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.5.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.5.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.5.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.5.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.5.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.5.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.5.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.5.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.5.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.5.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.5.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.5.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.6.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.6.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.6.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.6.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.6.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.6.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.6.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.6.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.6.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.6.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.6.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.6.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.6.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.6.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.7.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.7.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.7.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.7.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.7.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.7.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.7.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.7.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.7.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.7.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.7.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.7.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.7.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.7.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.8.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.8.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.8.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.8.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.8.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.8.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.8.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.8.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.8.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.8.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.8.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.8.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.8.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.8.norm1.linear.weight', 'action_ema.model.model.transformer_blocks.9.attn1.to_k.bias', 'action_ema.model.model.transformer_blocks.9.attn1.to_k.weight', 'action_ema.model.model.transformer_blocks.9.attn1.to_out.0.bias', 'action_ema.model.model.transformer_blocks.9.attn1.to_out.0.weight', 'action_ema.model.model.transformer_blocks.9.attn1.to_q.bias', 'action_ema.model.model.transformer_blocks.9.attn1.to_q.weight', 'action_ema.model.model.transformer_blocks.9.attn1.to_v.bias', 'action_ema.model.model.transformer_blocks.9.attn1.to_v.weight', 'action_ema.model.model.transformer_blocks.9.ff.net.0.proj.bias', 'action_ema.model.model.transformer_blocks.9.ff.net.0.proj.weight', 'action_ema.model.model.transformer_blocks.9.ff.net.2.bias', 'action_ema.model.model.transformer_blocks.9.ff.net.2.weight', 'action_ema.model.model.transformer_blocks.9.norm1.linear.bias', 'action_ema.model.model.transformer_blocks.9.norm1.linear.weight', 'action_ema.model.position_embedding.weight', 'action_ema.model.state_encoder.layer1.bias', 'action_ema.model.state_encoder.layer1.weight', 'action_ema.model.state_encoder.layer2.bias', 'action_ema.model.state_encoder.layer2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.0.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.0.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.1.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.1.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.10.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.10.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.11.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.11.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.12.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.12.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.13.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.13.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.14.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.14.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.15.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.15.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.16.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.16.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.17.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.17.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.18.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.18.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.19.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.19.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.2.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.2.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.20.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.20.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.21.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.21.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.22.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.22.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.23.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.23.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.24.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.24.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.25.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.25.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.26.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.26.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.27.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.27.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.28.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.28.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.29.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.29.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.3.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.3.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.4.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.4.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.5.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.5.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.6.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.6.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.7.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.7.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.8.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.8.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn1.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.norm_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.norm_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.to_k.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.to_k.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.to_out.0.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.to_out.0.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.to_q.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.to_q.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.to_v.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.attn2.to_v.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.ffn.net.0.proj.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.ffn.net.0.proj.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.ffn.net.2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.ffn.net.2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.norm2.bias', 'backbone.extractor._video_ema_model.model.dit.blocks.9.norm2.weight', 'backbone.extractor._video_ema_model.model.dit.blocks.9.scale_shift_table', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.text_embedder.linear_1.bias', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.text_embedder.linear_1.weight', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.text_embedder.linear_2.bias', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.text_embedder.linear_2.weight', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.time_embedder.linear_1.bias', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.time_embedder.linear_1.weight', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.time_embedder.linear_2.bias', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.time_embedder.linear_2.weight', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.time_proj.bias', 'backbone.extractor._video_ema_model.model.dit.condition_embedder.time_proj.weight', 'backbone.extractor._video_ema_model.model.dit.ctx_pool_time_pos_emb', 'backbone.extractor._video_ema_model.model.dit.patch_embedding.bias', 'backbone.extractor._video_ema_model.model.dit.patch_embedding.weight', 'backbone.extractor._video_ema_model.model.dit.proj_out.bias', 'backbone.extractor._video_ema_model.model.dit.proj_out.weight', 'backbone.extractor._video_ema_model.model.dit.scale_shift_table']
+- This IS expected if you are initializing WAMDiT4DiT from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
+- This IS NOT expected if you are initializing WAMDiT4DiT from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
+Some weights of WAMDiT4DiT were not initialized from the model checkpoint at /rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3/checkpoint-60000 and are newly initialized: ['backbone.extractor._sf_projector.linear1.bias', 'backbone.extractor._sf_projector.linear1.weight', 'backbone.extractor._sf_projector.linear2.bias', 'backbone.extractor._sf_projector.linear2.weight']
+You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
+[EMA-DTYPE] live dtype source=backbone.extractor.transformer -> torch.float32 (★1-tensor probe — see note above; compute dtype is what governs the regime)
+[EMA-KEYSWAP] video_ema: swap begin — 826 keys across 5 shard(s) | host_maxrss=32.3GB cuda_alloc=11.1GB cuda_resv=21.9GB
+[EMA-KEYSWAP] video_ema: shard 1/5 'model-00003-of-00007.safetensors' (147 keys) open | host_maxrss=32.3GB cuda_alloc=11.1GB cuda_resv=21.9GB
+[EMA-KEYSWAP] video_ema: shard 1/5 progress 128/147 | host_maxrss=32.3GB cuda_alloc=12.5GB cuda_resv=21.9GB
+[EMA-KEYSWAP] video_ema: shard 2/5 'model-00004-of-00007.safetensors' (206 keys) open | host_maxrss=32.3GB cuda_alloc=12.7GB cuda_resv=21.9GB
+[EMA-KEYSWAP] video_ema: shard 2/5 progress 128/206 | host_maxrss=32.3GB cuda_alloc=14.1GB cuda_resv=21.9GB
+[EMA-KEYSWAP] video_ema: shard 3/5 'model-00005-of-00007.safetensors' (199 keys) open | host_maxrss=32.3GB cuda_alloc=15.0GB cuda_resv=21.9GB
+[EMA-KEYSWAP] video_ema: shard 3/5 progress 128/199 | host_maxrss=32.3GB cuda_alloc=16.5GB cuda_resv=21.9GB
+[EMA-KEYSWAP] video_ema: shard 4/5 'model-00006-of-00007.safetensors' (206 keys) open | host_maxrss=32.3GB cuda_alloc=17.3GB cuda_resv=21.9GB
+[EMA-KEYSWAP] video_ema: shard 4/5 progress 128/206 | host_maxrss=32.3GB cuda_alloc=18.7GB cuda_resv=22.8GB
+[EMA-KEYSWAP] video_ema: shard 5/5 'model-00007-of-00007.safetensors' (68 keys) open | host_maxrss=32.3GB cuda_alloc=19.6GB cuda_resv=24.2GB
+[EMA-KEYSWAP] video_ema: 826 keys swapped, fp=d11a7e0a6d62b6f7 (src==dst VERIFIED) — 'backbone.extractor._video_ema_model.model.*' -> 'backbone.extractor.transformer.*', source dtype preserved, zero module moves.
+[EMA-KEYSWAP] action: swap begin — 247 keys across 1 shard(s) | host_maxrss=32.3GB cuda_alloc=20.4GB cuda_resv=25.5GB
+[EMA-KEYSWAP] action: shard 1/1 'model-00007-of-00007.safetensors' (247 keys) open | host_maxrss=32.3GB cuda_alloc=20.4GB cuda_resv=25.5GB
+[EMA-KEYSWAP] action: shard 1/1 progress 128/247 | host_maxrss=32.3GB cuda_alloc=20.6GB cuda_resv=25.5GB
+[EMA-KEYSWAP] action: 247 keys swapped, fp=2fbfdc80fbef87cc (src==dst VERIFIED) — 'action_ema.model.*' -> 'action_head.*', source dtype preserved, zero module moves.
+[Gr00tPolicy] EMA-EVAL GUARD (keyswap): video-DiT EMA (WAM_VIDEO_EMA) VERIFIED — 826 'backbone.extractor._video_ema_model.*' keys were key-swapped into the live module (fp=d11a7e0a6d62b6f7), so the eval runs the TRAINED EMA weights on a single copy.
+[Gr00tPolicy] EMA-EVAL GUARD (keyswap): action-head EMA (WAM_ACTION_EMA) VERIFIED — 247 'action_ema.*' keys were key-swapped into the live module (fp=2fbfdc80fbef87cc), so the eval runs the TRAINED EMA weights on a single copy.
+[WAM_IMAGE_AUG] cosmos-style image aug ENABLED (stronger=False, moderate=True -> rotate=on, allview=True -> geometry=every view); eval counterpart = center-crop sqrt(0.9)→resize (persisted per-ckpt).
+[WAM_TEXT_DROPOUT] train caption dropout p=0.1 (requires '' in text cache).
+[VRAM-RECLAIM] empty_cache after load: reserved 26144 -> 21682 MiB (reclaimed 4462 MiB; allocated=21281 MiB) — allocator cache only, weights untouched
+[server-oom-probe] device=cuda:0(NVIDIA H100 80GB HBM3) device_free_at_probe=58700 MiB total=80995 MiB not_free=22295 MiB | this_process reserved=21682 MiB allocated=21281 MiB cache_held=401 MiB — snapshot AFTER load+empty_cache, BEFORE first inference
+[RNG-DRIFT] after model load: 66aa85e083c1bf4a (before=2f0992352d445d54) — compare 'after' between two runs of the SAME cell
+Server is ready and listening on tcp://127.0.0.1:40903
+[EVAL-COMPUTE] backbone: autocast_enabled=True compute_dtype=torch.bfloat16 weights=[830xfloat32, 196xbfloat16] of 1026 floating params (dominant=torch.float32) => regime=(torch.float32 weights, torch.bfloat16 compute). mixed weight residency is EXPECTED (diffusers _keep_in_fp32_modules); the regime is the PAIR, and compute_dtype is the half that is measured here. action head runs under a separate fp32 autocast (see below).
+[GEOM-RUNTIME] mode=eval student_latent_slots=6 cond_slots=4 grid=6x24x16 tokens=2304 (num_frames_in=13 num_frames_out=21 tfac=4 v2v=False)
+[TRAIN-COMPUTE] rank0 student/extract(a) (Wan22DiffusersFeatureExtractor): autocast_enabled=True autocast_dtype=bfloat16 grad_enabled=False block_output_dtype=bfloat16 weights=[27xfloat32] over 27 floating tensors (dominant=float32) => regime=(float32 weights, bfloat16 compute). norm_leaves=[5xfloat32, 2xno-params] other_leaves=[10xfloat32, 3xno-params]. tap=blocks[18] (student L18); block_input_dtype=bfloat16; fadistill_active=False
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     NORM norm1                              FP32LayerNorm          weights=no-params
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn1.to_q                         Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn1.to_k                         Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn1.to_v                         Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn1.to_out.0                     Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn1.to_out.1                     Dropout                weights=no-params
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     NORM attn1.norm_q                       RMSNorm                weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     NORM attn1.norm_k                       RMSNorm                weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn2.to_q                         Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn2.to_k                         Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn2.to_v                         Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn2.to_out.0                     Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest attn2.to_out.1                     Dropout                weights=no-params
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     NORM attn2.norm_q                       RMSNorm                weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     NORM attn2.norm_k                       RMSNorm                weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     NORM norm2                              FP32LayerNorm          weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest ffn.net.0.proj                     Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest ffn.net.1                          Dropout                weights=no-params
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     rest ffn.net.2                          Linear                 weights=float32
+[TRAIN-COMPUTE-LEAF] rank0 student/extract(a)     NORM norm3                              FP32LayerNorm          weights=no-params
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[WAMDiT4DiT] ★EMA-KEYSWAP: get_action predicts through the LIVE head HOLDING the swapped EMA weights (247 keys, dtype=torch.float32) — single head, no container.
+[ACTION-KV] mode=eval vl_embeds=(1, 192, 3072) (B,S,D) tokens_S=192 dtype=torch.bfloat16
+[ACTION-SEED] eval action noise seeded: seed=42 device=cuda:0 (paired 비교용 — 기본 OFF, 이 셀은 마커로 구분할 것)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)
+[DITACTION_EXCLUDE_CURRENT] dropped 384 cond tokens (past+current), kept 192 future (tokens 576->192)

avg_L3/eval/checkpoint-60000/nas16/TurnOnSinkFaucet_PandaOmron_Env/vram_watch.log ADDED Viewed

	@@ -0,0 +1,236 @@

+--- 2026-09-14T17:21:08Z
+1 MiB, 81559 MiB
+    Processes                             : None
+--- 2026-09-14T17:21:19Z
+4 MiB, 81559 MiB
+    Processes                             : None
+--- 2026-09-14T17:21:29Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 23046 MiB
+23056 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 23046 MiB
+--- 2026-09-14T17:21:39Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 23046 MiB
+23056 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 23046 MiB
+--- 2026-09-14T17:21:49Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 23048 MiB
+23058 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 23048 MiB
+--- 2026-09-14T17:22:00Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 23048 MiB
+23058 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 23048 MiB
+--- 2026-09-14T17:22:10Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 23048 MiB
+23058 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 23048 MiB
+--- 2026-09-14T17:22:20Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 23048 MiB
+23058 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 23048 MiB
+--- 2026-09-14T17:22:30Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 23050 MiB
+23060 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 23050 MiB
+--- 2026-09-14T17:22:40Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 23050 MiB
+23060 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 23050 MiB
+--- 2026-09-14T17:22:51Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 23386 MiB
+23396 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 23386 MiB
+--- 2026-09-14T17:23:01Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 24826 MiB
+24836 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 24826 MiB
+--- 2026-09-14T17:23:11Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 26410 MiB
+26420 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 26578 MiB
+--- 2026-09-14T17:23:21Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 22286 MiB
+22296 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 22286 MiB
+--- 2026-09-14T17:23:32Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 22286 MiB
+22296 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 22286 MiB
+--- 2026-09-14T17:23:42Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 22286 MiB
+24771 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 74044
+            Type                          : G
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/gr00t/eval/sim/robocasa/robocasa_uv/.venv/bin/python
+            Used GPU Memory               : 2494 MiB
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 22286 MiB
+--- 2026-09-14T17:23:52Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 22286 MiB
+26829 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 74044
+            Type                          : G
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/gr00t/eval/sim/robocasa/robocasa_uv/.venv/bin/python
+            Used GPU Memory               : 4528 MiB
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 22286 MiB
+--- 2026-09-14T17:24:02Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 32160 MiB
+36897 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 74044
+            Type                          : G
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/gr00t/eval/sim/robocasa/robocasa_uv/.venv/bin/python
+            Used GPU Memory               : 4722 MiB
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 32160 MiB
+--- 2026-09-14T17:24:13Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 32160 MiB
+36897 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 74044
+            Type                          : G
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/gr00t/eval/sim/robocasa/robocasa_uv/.venv/bin/python
+            Used GPU Memory               : 4722 MiB
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 32160 MiB
+--- 2026-09-14T17:24:23Z
+3430343, /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python, 32160 MiB
+36897 MiB, 81559 MiB
+    Processes
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 74044
+            Type                          : G
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/gr00t/eval/sim/robocasa/robocasa_uv/.venv/bin/python
+            Used GPU Memory               : 4722 MiB
+        GPU instance ID                   : N/A
+        Compute instance ID               : N/A
+        Process ID                        : 3430343
+            Type                          : C
+            Name                          : /sjw_alinlab1/home/huiwon/RobotTransformers/Isaac-GR00T-AlinVLA-SH/.venv/bin/python
+            Used GPU Memory               : 32160 MiB

avg_L3/eval/checkpoint-60000/nas16/TurnOnStove_PandaOmron_Env/provenance.json ADDED Viewed

	@@ -0,0 +1,13 @@

+{
+  "file_sha256": {
+    "gr00t/model/wam_dit4dit/backbone.py": "e7b6088e791a71a76dda9df9185005d19a51517004620774c747b158eaf562e5",
+    "gr00t/model/wam_dit4dit/wan22_diffusers_extractor.py": "d14f5468c0f1c0d708f70e4d83e9e55080ea988ed88e72170160e56561e8faad",
+    "gr00t/model/wam_dit4dit/wan22_extractor.py": "fc8716f82e635e6934591f7154a8a235023e78c736a87cd388a5e7b0cadc1d00",
+    "run_scripts/eval/robocasa/eval_wam_dit4dit_kitchen.sh": "71557eaaf61ca2e45cf57a5b9a51e481de97e376805444f1a92e3bbacfcad12f"
+  },
+  "git_branch": "unknown",
+  "git_commit": "unknown",
+  "git_dirty": false,
+  "hostname": "worker-node1001",
+  "repo_root": "/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77"
+}

avg_L3/eval/checkpoint-60000/nas16/TurnOnStove_PandaOmron_Env/requeue_env.txt ADDED Viewed

	@@ -0,0 +1,37 @@

+# generated 2026-09-14T17:20:20Z by eval_wam_dit4dit_kitchen.sh
+# job=184132 array_task=1 task=TurnOnStove_PandaOmron_Env
+# consumed by scripts/wam_dit4dit/requeue_incomplete_eval.sh --run-dir <this dir's parent>
+BACKBONE_TYPE=wan22_diffusers
+CKPT_CONFIGS_OVERRIDE=/rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3:60000
+CKPT_STEP=60000
+EVAL_BENCH=robocasa_kitchen
+EVAL_OUTPUT_ROOT=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77/output
+EVAL_SUFFIX=bmethod_ctxpool_avg_L3_60k_seed42av_emaW
+EVAL_WEIGHTS=ema
+MODEL_OUTPUT_DIR=/rlwrld-unified-checkpoints/hyeonmo/wan2
+N_ENVS=1
+N_EPISODES=50
+VIDEO_DIT_OVERRIDE=
+VIDEO_DIT_OVERRIDE_HF=
+WAM_ACTION_EMA=warmup
+WAM_ALIGN_TRD=0
+WAM_ALIGN_TRD_PROBE=0
+WAM_COMPILE_BLOCKS=0
+WAM_DITACTION_EXCLUDE_CURRENT=1
+WAM_EVAL_ACTION_SEED=42
+WAM_EVAL_ENV_MODE=parallel
+WAM_EVAL_SERVER_TIMEOUT_S=1800
+WAM_EVAL_TASK_DIR_SUFFIX=
+WAM_EVAL_TREE=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_TREE_REEXEC=1
+WAM_EVAL_TREE_SHA=eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_USE_ACTION_EMA=1
+WAM_EVAL_USE_EMA=1
+WAM_EVAL_VIDEO_SEED=42
+WAM_EXTRACT_EARLY_EXIT=0
+WAM_TEXT_EMBED_CACHE=/sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt
+WAM_TEXT_EMBED_REQUIRE_HIT=1
+WAM_VIDEO_EMA=0
+WAM_VIDEO_EMA_TRAIN_SOURCE=live
+WAM_WAN_EVAL_TRUNCATE_ABOVE=0
+WAM_WAN_VAE_ENCODE_ENV_CHUNK=0

avg_L3/eval/checkpoint-60000/nas16/TurnSinkSpout_PandaOmron_Env/provenance.json ADDED Viewed

	@@ -0,0 +1,13 @@

+{
+  "file_sha256": {
+    "gr00t/model/wam_dit4dit/backbone.py": "e7b6088e791a71a76dda9df9185005d19a51517004620774c747b158eaf562e5",
+    "gr00t/model/wam_dit4dit/wan22_diffusers_extractor.py": "d14f5468c0f1c0d708f70e4d83e9e55080ea988ed88e72170160e56561e8faad",
+    "gr00t/model/wam_dit4dit/wan22_extractor.py": "fc8716f82e635e6934591f7154a8a235023e78c736a87cd388a5e7b0cadc1d00",
+    "run_scripts/eval/robocasa/eval_wam_dit4dit_kitchen.sh": "71557eaaf61ca2e45cf57a5b9a51e481de97e376805444f1a92e3bbacfcad12f"
+  },
+  "git_branch": "unknown",
+  "git_commit": "unknown",
+  "git_dirty": false,
+  "hostname": "worker-node1001",
+  "repo_root": "/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77"
+}

avg_L3/eval/checkpoint-60000/nas16/TurnSinkSpout_PandaOmron_Env/requeue_env.txt ADDED Viewed

	@@ -0,0 +1,37 @@

+# generated 2026-09-14T17:19:35Z by eval_wam_dit4dit_kitchen.sh
+# job=184129 array_task=0 task=TurnSinkSpout_PandaOmron_Env
+# consumed by scripts/wam_dit4dit/requeue_incomplete_eval.sh --run-dir <this dir's parent>
+BACKBONE_TYPE=wan22_diffusers
+CKPT_CONFIGS_OVERRIDE=/rlwrld-unified-checkpoints/hyeonmo/wan2/hf_downloads/wam_ctxpool_bmethod/avg_L3:60000
+CKPT_STEP=60000
+EVAL_BENCH=robocasa_kitchen
+EVAL_OUTPUT_ROOT=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77/output
+EVAL_SUFFIX=bmethod_ctxpool_avg_L3_60k_seed42av_emaW
+EVAL_WEIGHTS=ema
+MODEL_OUTPUT_DIR=/rlwrld-unified-checkpoints/hyeonmo/wan2
+N_ENVS=1
+N_EPISODES=50
+VIDEO_DIT_OVERRIDE=
+VIDEO_DIT_OVERRIDE_HF=
+WAM_ACTION_EMA=warmup
+WAM_ALIGN_TRD=0
+WAM_ALIGN_TRD_PROBE=0
+WAM_COMPILE_BLOCKS=0
+WAM_DITACTION_EXCLUDE_CURRENT=1
+WAM_EVAL_ACTION_SEED=42
+WAM_EVAL_ENV_MODE=parallel
+WAM_EVAL_SERVER_TIMEOUT_S=1800
+WAM_EVAL_TASK_DIR_SUFFIX=
+WAM_EVAL_TREE=/sjw_alinlab3/home/hyeonmo/.wam-eval-trees/eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_TREE_REEXEC=1
+WAM_EVAL_TREE_SHA=eb321c8b96c4a8d7f89b3a78652582db35f99e77
+WAM_EVAL_USE_ACTION_EMA=1
+WAM_EVAL_USE_EMA=1
+WAM_EVAL_VIDEO_SEED=42
+WAM_EXTRACT_EARLY_EXIT=0
+WAM_TEXT_EMBED_CACHE=/sjw_alinlab1/home/huiwon/.cache/wam_text_cache/robocasa300_wan22_umt5.pt
+WAM_TEXT_EMBED_REQUIRE_HIT=1
+WAM_VIDEO_EMA=0
+WAM_VIDEO_EMA_TRAIN_SOURCE=live
+WAM_WAN_EVAL_TRUNCATE_ABOVE=0
+WAM_WAN_VAE_ENCODE_ENV_CHUNK=0