Add files using upload-large-folder tool

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/latest_checkpoint.txt

@@ -0,0 +1 @@
+iter_000002500.pt

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/model/iter_000000300.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8daa2dee5e3623fa8cb553f9d02685cf99e93b183c94a9f526a175d8f552fc4e
+size 4005911812

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000000300.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1376e97b6216936fe773a9cb0c92fd6c5a18d0a7cf38bd3c9a19e6e2926d7921
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000000600.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d78eba575b91c250389b7cf5075660897e28d8ef63e941e7a62e9cca171a0ab1
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000000900.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8069dfd6e2acf47dfd6e282bdbde519ffded23834129d550d0cb6fc5df0ec0ef
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000001000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea9430a8478cf5e24fa989cdeb74142f870bac689a63a577c3ca93c641af0076
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000001100.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a6acdec2dea772446522e4c39e418ae22fe7bfb52ee98ba817cc1fde4af0398
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000001200.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5b9068ac6142edae0d816c38578e0f0b19397ab7455e613ecf5890cab48bdc3
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000001400.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:357fd01f9a1ebe49ead3314bf333b7f793badc53881009a9a2069d8822fb962b
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000001500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dec3285aa6519fef112cfffd1167355acffb4a10c5174f86e201c307dbb92bb5
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000001700.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7880925cfa6a16877c128ad95301186ba0a1fed4437bdc53b9a8d9457ddd9e07
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000001900.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8254ad73ccb1fb97fe3099743caa0c56f4ae676b2024c2ceec823a4d1c87ccc0
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000002300.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1c2f30d8288c9bc5475d7af5daa09558ac7dc135ecb74dcad1cbd194d47f9f8
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/optim/iter_000002500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3891adc45e663c3cb4e2ae282d815bf686c7678cbcefb56f82830217e379b079
+size 1792982648

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000000300.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:335332347d209098e04f222c4ea965c135e93a1c5cfbb0174c8e05ed4a60c74e
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000000600.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fc1c2b6bab1e64e23227ed9d41949ec424142fc15a1e21a0fa2b6816b7824b9
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000000900.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06f91ee0a2a43af5cdaca380bea2c0235f9cc413c4c190635dd7d1d056f55890
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d348f12f8b56c5328f69671dd6140dd00f108663f2aa60f2aa9b6253325bd354
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001100.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16bde946535bc4cb97efa0dce41d8e4ef0053b17e354cdf4a22725522ce40c95
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001200.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab4c97895672ce35d6c5e7ae32c92b320808359009a943044fc9cce6b2bb0cc5
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001300.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:819b51f2f3685a6da038be87077e7e4c133626d2c7b56a532478ec1b8d133822
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001400.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d46f7c5071e707fe3be3aa0a003e8fdf498a24302159b792f4e611a78148726f
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4928ba932e574be34bf0aade8476a07b2ad09732b5fc9b0a380c5c80c3c7ff23
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001600.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:200f9f4c7033ca20ccb9f7813a2e911f782586d82ea77567aca0113ff0bc0e65
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001700.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c5998235d9fc1d64308302593a314aa5a7e0deccea225f8f1fbbeb37afab68d
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001800.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8e80282061c2fdfd91a2460954604492285a8b5705bcf5bbfc18bbbfd99dee1
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000001900.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e362aa27cac87936de9e5cb902297aac41b8c921c1946e50862c9e5db8ba37f
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000002000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85f9e4bedf80049a696cd58e057fbaed153d897e1bfd14634a51d393077f0000
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000002100.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f4e2b0c21af495503606b2334b267df35d1c7307559ae41e67dc3ae1f363b3a
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000002200.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ffedde310d21ae69a4dcbaebc422d199c781401f03a89e8e12d2bf9eba38190
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000002300.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f738aa2d63a50bd7f0635118655db595cf8c0ee3a62dcdb1aeeb16cfc6f60036
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000002400.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1e1e6f1adaaf95a438b9ed9f4a10e26b3922ebc23dda99b44d3bbfa941d9006
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/scheduler/iter_000002500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb9c3d332e23f21533d1c01ce42d9903189529c59ccb81e964ba0de9a0653b3d
+size 1602

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000000300.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0f362aa11769c1a4c5261f596d15c1d4dd509a0eb98fbded2add580ad97b56c
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000000600.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d4241c8657960fc56d56e303b2f10176953d10a3c4c475826e14070d4f889bd
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000000900.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf3af3f1095f92752b6396497ccccfef3f15a363cebb395bb892d595d7442321
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000001000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42a8eb54f575f232d4f03d7c9594005cb06ae9ffb4bc249977d1154493cfbf75
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000001100.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72f4dc142705e94743fbb00be363e3befd9b4642adcf1511bcdfa5493786293b
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000001300.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6d19ade99b7fdec1a42c107b3df22a6fe8945fc407843972b83551262df4306
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000001500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7053c4ac2524bdbf97217455f074cf8d8f747463838eda2b9ef2ca7541c455dd
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000001600.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:064e3e842533ed6f4fbbbf831111751c87bfe938b153b7c635cdb2b7c9a59540
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000001700.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d90aea88abdb41139f0b9594b7d450840341d8915f6895f9052883942b2242bd
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000001800.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:190ee185eb7fdff804e5fe1ddf9ce00ff9920e5e2bb5d4755cb96932ff1bed82
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000002000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2b3774816544e78e539e7d51e64b43a30c795a2947d1fe5a7dbd70c48e434f7
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000002100.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1149581fc64ad3bb27b55c1f262129c99510dc49e609d8394a04cfad5cb2479d
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000002200.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66a6708b0dedfc90c1d322f3b5ce321cf9c4ea23e9deaf2216846b9bbcdae305
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/checkpoints/trainer/iter_000002400.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebd3ca4055c7d21f21f4ac41393ba855e7099df1f555bc39de59329362209266
+size 892

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/config.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9979e0349cb9378211c0b26d885b2eb0ae983134375a769b432b2d0ffdb27638
+size 77101

posttraining/video2world/v2w_push_combined_lora_rank32_lr1.778e-04_bsz32/config.yaml

@@ -0,0 +1,541 @@
+checkpoint:
+  broadcast_via_filesystem: 'False'
+  dcp_allow_mismatched_size: 'False'
+  dcp_async_mode_enabled: 'False'
+  jit:
+    device: cuda
+    dtype: bfloat16
+    enabled: 'False'
+    input_shape: null
+    strict: 'True'
+  keys_not_to_resume: []
+  load_ema_to_reg: 'False'
+  load_path: ''
+  load_training_state: 'False'
+  only_load_scheduler_state: 'False'
+  save_iter: '100'
+  strict_resume: 'True'
+  type:
+    _target_: <class 'cosmos_predict2.checkpointer.Checkpointer'>
+    callbacks: null
+  verbose: 'True'
+data_config: null
+dataloader_train:
+  _target_: <class 'torch.utils.data.dataloader.DataLoader'>
+  batch_sampler: null
+  batch_size: '4'
+  collate_fn: null
+  dataset:
+    _target_: <class 'cosmos_predict2.data.dataset_video.MultiDataset'>
+    push_task1:
+      _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+      data_fps: '30.0'
+      dataset_dir: /home/ubuntu/data/push-mimic-task1
+      exclude_with_substring: null
+      include_only_with_substrings: null
+      is_multi_img: 'False'
+      is_val: 'False'
+      num_frames: '61'
+      obs_history: '5'
+      val_ratio: '0.0'
+      video_size:
+      - '480'
+      - '640'
+    push_task2:
+      _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+      data_fps: '30.0'
+      dataset_dir: /home/ubuntu/data/push-mimic-task2
+      exclude_with_substring: null
+      include_only_with_substrings: null
+      is_multi_img: 'False'
+      is_val: 'False'
+      num_frames: '61'
+      obs_history: '5'
+      val_ratio: '0.0'
+      video_size:
+      - '480'
+      - '640'
+  drop_last: 'True'
+  generator: null
+  in_order: 'False'
+  multiprocessing_context: null
+  num_workers: '12'
+  persistent_workers: 'True'
+  pin_memory: 'True'
+  pin_memory_device: ''
+  prefetch_factor: '8'
+  sampler:
+    _target_: <function get_sampler at 0x7bd0b99e0e50>
+    dataset:
+      _target_: <class 'cosmos_predict2.data.dataset_video.MultiDataset'>
+      push_task1:
+        _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+        data_fps: '30.0'
+        dataset_dir: /home/ubuntu/data/push-mimic-task1
+        exclude_with_substring: null
+        include_only_with_substrings: null
+        is_multi_img: 'False'
+        is_val: 'False'
+        num_frames: '61'
+        obs_history: '5'
+        val_ratio: '0.0'
+        video_size:
+        - '480'
+        - '640'
+      push_task2:
+        _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+        data_fps: '30.0'
+        dataset_dir: /home/ubuntu/data/push-mimic-task2
+        exclude_with_substring: null
+        include_only_with_substrings: null
+        is_multi_img: 'False'
+        is_val: 'False'
+        num_frames: '61'
+        obs_history: '5'
+        val_ratio: '0.0'
+        video_size:
+        - '480'
+        - '640'
+  shuffle: null
+  timeout: '0'
+  worker_init_fn: null
+dataloader_val:
+  _target_: <class 'torch.utils.data.dataloader.DataLoader'>
+  batch_sampler: null
+  batch_size: '1'
+  collate_fn: null
+  dataset:
+    _target_: <class 'cosmos_predict2.data.dataset_video.MultiDataset'>
+    push_task1:
+      _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+      data_fps: '30.0'
+      dataset_dir: /home/ubuntu/data/push-mimic-task1
+      exclude_with_substring: null
+      include_only_with_substrings: null
+      is_multi_img: 'False'
+      is_val: 'True'
+      num_frames: '61'
+      obs_history: '5'
+      val_ratio: '0.0'
+      video_size:
+      - '480'
+      - '640'
+    push_task2:
+      _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+      data_fps: '30.0'
+      dataset_dir: /home/ubuntu/data/push-mimic-task2
+      exclude_with_substring: null
+      include_only_with_substrings: null
+      is_multi_img: 'False'
+      is_val: 'True'
+      num_frames: '61'
+      obs_history: '5'
+      val_ratio: '0.0'
+      video_size:
+      - '480'
+      - '640'
+  drop_last: 'False'
+  generator: null
+  in_order: 'False'
+  multiprocessing_context: null
+  num_workers: '0'
+  persistent_workers: 'False'
+  pin_memory: 'False'
+  pin_memory_device: ''
+  prefetch_factor: null
+  sampler:
+    _target_: <function get_sampler at 0x7bd0b99e0e50>
+    dataset:
+      _target_: <class 'cosmos_predict2.data.dataset_video.MultiDataset'>
+      push_task1:
+        _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+        data_fps: '30.0'
+        dataset_dir: /home/ubuntu/data/push-mimic-task1
+        exclude_with_substring: null
+        include_only_with_substrings: null
+        is_multi_img: 'False'
+        is_val: 'True'
+        num_frames: '61'
+        obs_history: '5'
+        val_ratio: '0.0'
+        video_size:
+        - '480'
+        - '640'
+      push_task2:
+        _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+        data_fps: '30.0'
+        dataset_dir: /home/ubuntu/data/push-mimic-task2
+        exclude_with_substring: null
+        include_only_with_substrings: null
+        is_multi_img: 'False'
+        is_val: 'True'
+        num_frames: '61'
+        obs_history: '5'
+        val_ratio: '0.0'
+        video_size:
+        - '480'
+        - '640'
+  shuffle: null
+  timeout: '0'
+  worker_init_fn: null
+defaults:
+- _self_
+- data_config: null
+- video_dataset_train: null
+- video_dataset_val: null
+- dataloader_train: null
+- dataloader_val: null
+- world2action_pipe: null
+- optimizer: fusedadamw
+- scheduler: constant
+- model: null
+- callbacks:
+  - basic
+- net: null
+- ema: null
+- checkpoint: null
+- ckpt_type: null
+- experiment: null
+job:
+  group: video2world
+  name: v2w_push_combined_lora_rank32_lr1.778e-04_bsz32
+  project: posttraining
+model:
+  _recursive_: 'False'
+  _target_: <class 'cosmos_predict2.models.video2world_model.Predict2Video2WorldModel'>
+  config:
+    adjust_video_noise: true
+    debug_without_randomness: false
+    fsdp_shard_size: 0
+    high_sigma_ratio: 0.05
+    init_lora_weights: true
+    input_image_key: images
+    input_video_key: video
+    lora_alpha: 32
+    lora_rank: 32
+    lora_target_modules: q_proj,k_proj,v_proj,output_proj,x_embedder.proj.1,linear_1,linear_2,mlp.layer1,mlp.layer2
+    loss_reduce: mean
+    loss_scale: 100.0
+    model_manager_config:
+      _target_: cosmos_predict2.models.video2world_model.Predict2ModelManagerConfig
+      dit_path: /home/ubuntu/pdt-mimic/mimic-video/model/checkpoints/video_backbone//v2w_pretrained_cosmos.pt
+      text_encoder_path: ''
+    pipe_config:
+      adjust_video_noise: true
+      conditioner:
+        _target_: <class 'cosmos_predict2.conditioner.VideoConditioner'>
+        fps:
+          _target_: <class 'cosmos_predict2.conditioner.ReMapkey'>
+          dropout_rate: '0.0'
+          dtype: null
+          input_key: fps
+          output_key: fps
+        padding_mask:
+          _target_: <class 'cosmos_predict2.conditioner.ReMapkey'>
+          dropout_rate: '0.0'
+          dtype: null
+          input_key: padding_mask
+          output_key: padding_mask
+        text:
+          _target_: <class 'cosmos_predict2.conditioner.TextAttr'>
+          dropout_rate: '0.0'
+          input_key:
+          - obs/language_embedding
+        use_video_condition:
+          _target_: <class 'cosmos_predict2.conditioner.BooleanFlag'>
+          dropout_rate: '0.0'
+          input_key: fps
+          output_key: use_video_condition
+      conditioning_strategy: frame_replace
+      ema:
+        _target_: cosmos_predict2.configs.defaults.ema.EMAConfig
+        enabled: 'False'
+        iteration_shift: '0'
+        rate: '0.1'
+      guardrail_config:
+        checkpoint_dir: /home/ubuntu/pdt-mimic/mimic-video/model/checkpoints
+        enabled: false
+        offload_model_to_cpu: true
+      input_image_key: images
+      input_video_key: video
+      max_num_conditional_frames: 2
+      min_num_conditional_frames: 1
+      net:
+        _target_: <class 'cosmos_predict2.models.video2world_dit.MinimalV1LVGDiT'>
+        adaln_lora_dim: '256'
+        atten_backend: minimal_a2a
+        concat_padding_mask: 'True'
+        extra_per_block_abs_pos_emb: 'False'
+        in_channels: '16'
+        max_frames: '128'
+        max_img_h: '240'
+        max_img_w: '240'
+        model_channels: '2048'
+        num_blocks: '28'
+        num_heads: '16'
+        out_channels: '16'
+        patch_spatial: '2'
+        patch_temporal: '1'
+        pos_emb_cls: rope3d
+        pos_emb_interpolation: crop
+        pos_emb_learnable: 'True'
+        rope_enable_fps_modulation: 'False'
+        rope_h_extrapolation_ratio: '3.0'
+        rope_t_extrapolation_ratio: '1.0'
+        rope_w_extrapolation_ratio: '3.0'
+        sac_config:
+          _target_: cosmos_predict2.models.text2image_dit.SACConfig
+          every_n_blocks: '1'
+          mode: predict2_2b_720
+        use_adaln_lora: 'True'
+      precision: bfloat16
+      rectified_flow_loss_weight_uniform: true
+      rectified_flow_t_scaling_factor: 1.0
+      resize_online: false
+      resolution: '480'
+      sigma_conditional: 0.0001
+      sigma_data: 1.0
+      state_ch: 16
+      state_t: 16
+      text_encoder:
+        cls: !!python/object/apply:imaginaire.constants.TextEncoderClass
+        - t5
+        t5:
+          ckpt_path: /home/ubuntu/pdt-mimic/mimic-video/model/checkpoints/text_encoder/t5-11b
+          embed_dim: 1024
+          num_tokens: 512
+      timestamps:
+        is_forward: false
+        nfe: 35
+        order: 7.0
+        t_max: 80.0
+        t_min: 0.002
+      tokenizer:
+        _target_: <class 'cosmos_predict2.tokenizers.tokenizer.TokenizerInterface'>
+        chunk_duration: '81'
+        load_mean_std: 'False'
+        name: tokenizer
+        temporal_window: '16'
+        vae_pth: /home/ubuntu/pdt-mimic/mimic-video/model/checkpoints/video_backbone//tokenizer/tokenizer.pth
+    precision: bfloat16
+    train_architecture: lora
+model_parallel:
+  _cpu_offloading_context: null
+  async_tensor_model_parallel_allreduce: false
+  autocast_dtype: torch.float32
+  barrier_with_L1_time: true
+  batch_p2p_comm: true
+  batch_p2p_sync: true
+  bf16: false
+  context_parallel_size: 1
+  cpu_offloading: false
+  cpu_offloading_activations: false
+  cpu_offloading_num_layers: 0
+  cpu_offloading_weights: false
+  cross_entropy_fusion_impl: native
+  cross_entropy_loss_fusion: false
+  deallocate_pipeline_outputs: false
+  defer_embedding_wgrad_compute: false
+  deterministic_mode: false
+  enable_autocast: false
+  expert_model_parallel_size: 1
+  expert_tensor_parallel_size: 1
+  finalize_model_grads_func: null
+  fp16: false
+  grad_scale_func: null
+  grad_sync_func: null
+  gradient_accumulation_fusion: false
+  hierarchical_context_parallel_sizes: null
+  microbatch_group_size_per_vp_stage: 1
+  moe_extended_tp: false
+  no_sync_func: null
+  num_microbatches_with_partial_activation_checkpoints: null
+  overlap_p2p_comm: false
+  overlap_p2p_comm_warmup_flush: false
+  param_sync_func: null
+  params_dtype: torch.float32
+  perform_initialization: true
+  pipeline_dtype: null
+  pipeline_model_parallel_comm_backend: null
+  pipeline_model_parallel_size: 1
+  pipeline_model_parallel_split_rank: null
+  sequence_parallel: false
+  tensor_model_parallel_size: 1
+  timers: null
+  tp_comm_atomic_ag: false
+  tp_comm_atomic_rs: false
+  tp_comm_bootstrap_backend: nccl
+  tp_comm_bulk_dgrad: true
+  tp_comm_bulk_wgrad: true
+  tp_comm_overlap: false
+  tp_comm_overlap_ag: true
+  tp_comm_overlap_disable_fc1: false
+  tp_comm_overlap_disable_qkv: false
+  tp_comm_overlap_rs: true
+  tp_comm_overlap_rs_dgrad: false
+  tp_comm_split_ag: true
+  tp_comm_split_rs: true
+  use_cpu_initialization: false
+  use_ring_exchange_p2p: false
+  use_te_rng_tracker: false
+  variable_seq_lengths: false
+  virtual_pipeline_model_parallel_size: null
+  wgrad_deferral_limit: 0
+optimizer:
+  _target_: <function get_base_optimizer at 0x7bd0aab87a30>
+  betas:
+  - '0.9'
+  - '0.99'
+  capturable: 'True'
+  eps: 1e-08
+  lr: '0.0001778279410038923'
+  master_weights: 'True'
+  model: null
+  optim_type: fusedadam
+  weight_decay: '0.1'
+scheduler:
+  _target_: <class 'cosmos_predict2.configs.defaults.scheduler.ConstantScheduler'>
+trainer:
+  callbacks:
+    device_monitor:
+      _target_: <class 'cosmos_predict2.callbacks.device_monitor.DeviceMonitor'>
+      every_n: '10'
+      log_memory_detail: 'True'
+      step_size: '1'
+    ema:
+      _target_: <class 'imaginaire.utils.callback.EMAModelCallback'>
+      config: null
+      trainer: null
+    grad_clip:
+      _target_: <class 'cosmos_predict2.callbacks.grad_clip.GradClip'>
+      clip_norm: '10.0'
+      force_finite: 'True'
+      log_wandb: 'True'
+    iter_speed:
+      _target_: <class 'cosmos_predict2.callbacks.iter_speed.IterSpeed'>
+      every_n: '10'
+      hit_thres: '5'
+    low_prec:
+      _target_: <class 'imaginaire.utils.callback.LowPrecisionCallback'>
+      config: null
+      trainer: null
+      update_iter: '1'
+    manual_gc:
+      _target_: <class 'imaginaire.callbacks.manual_gc.ManualGarbageCollection'>
+      every_n: '5'
+      warm_up: '5'
+    progress_bar:
+      _target_: <class 'imaginaire.utils.callback.ProgressBarCallback'>
+      config: null
+      trainer: null
+    rollout_logger:
+      _target_: <class 'cosmos_predict2.callbacks.rollout_logger.RolloutLogger'>
+      guidance: '5.0'
+      num_latent_conditional_frames: '2'
+      num_sampling_step: '25'
+      num_video_frames: '61'
+      rollout_iter: '100'
+      seed: '0'
+      test_clips:
+      - dataset_dir: /home/ubuntu/data/push-mimic-task1
+        ep_idx: '0'
+        name: task1
+      - dataset_dir: /home/ubuntu/data/push-mimic-task2
+        ep_idx: '0'
+        name: task2
+    wandb:
+      _target_: <class 'cosmos_predict2.callbacks.wandb_logger.WandbLogger'>
+      entity: theo-lefur
+      every_n: '10'
+      group: null
+      project: cosmos-predict-push
+  cudnn:
+    benchmark: 'True'
+    deterministic: 'False'
+  ddp:
+    broadcast_buffers: 'True'
+    find_unused_parameters: 'False'
+    static_graph: 'True'
+  distributed_parallelism: ddp
+  grad_accum_iter: '8'
+  grad_scaler_args:
+    enabled: 'False'
+  logging_iter: '10'
+  max_iter: '3000'
+  max_val_iter: null
+  memory_format: torch.preserve_format
+  profiling:
+    enable_memory_snapshot: 'False'
+    enable_profiling: 'False'
+    first_n_rank: '4'
+    profile_freq: '1'
+    profile_memory: 'True'
+    record_shape: 'True'
+    with_modules: 'True'
+    with_stack: 'True'
+  run_validation: 'False'
+  seed: '0'
+  timeout_period: '999999999'
+  type: <class 'imaginaire.trainer.ImaginaireTrainer'>
+  validation_iter: '999999999'
+video_dataset_train:
+  _target_: <class 'cosmos_predict2.data.dataset_video.MultiDataset'>
+  push_task1:
+    _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+    data_fps: '30.0'
+    dataset_dir: /home/ubuntu/data/push-mimic-task1
+    exclude_with_substring: null
+    include_only_with_substrings: null
+    is_multi_img: 'False'
+    is_val: 'False'
+    num_frames: '61'
+    obs_history: '5'
+    val_ratio: '0.0'
+    video_size:
+    - '480'
+    - '640'
+  push_task2:
+    _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+    data_fps: '30.0'
+    dataset_dir: /home/ubuntu/data/push-mimic-task2
+    exclude_with_substring: null
+    include_only_with_substrings: null
+    is_multi_img: 'False'
+    is_val: 'False'
+    num_frames: '61'
+    obs_history: '5'
+    val_ratio: '0.0'
+    video_size:
+    - '480'
+    - '640'
+video_dataset_val:
+  _target_: <class 'cosmos_predict2.data.dataset_video.MultiDataset'>
+  push_task1:
+    _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+    data_fps: '30.0'
+    dataset_dir: /home/ubuntu/data/push-mimic-task1
+    exclude_with_substring: null
+    include_only_with_substrings: null
+    is_multi_img: 'False'
+    is_val: 'True'
+    num_frames: '61'
+    obs_history: '5'
+    val_ratio: '0.0'
+    video_size:
+    - '480'
+    - '640'
+  push_task2:
+    _target_: <class 'cosmos_predict2.data.dataset_video.Dataset'>
+    data_fps: '30.0'
+    dataset_dir: /home/ubuntu/data/push-mimic-task2
+    exclude_with_substring: null
+    include_only_with_substrings: null
+    is_multi_img: 'False'
+    is_val: 'True'
+    num_frames: '61'
+    obs_history: '5'
+    val_ratio: '0.0'
+    video_size:
+    - '480'
+    - '640'
+world2action_pipe: null