diff --git a/PUBLICATION_SCOPE.json b/PUBLICATION_SCOPE.json
deleted file mode 100644
index 3dfe3899e45dadfab4d8e7aa79bf2640dabaf13e..0000000000000000000000000000000000000000
--- a/PUBLICATION_SCOPE.json
+++ /dev/null
@@ -1,5 +0,0 @@
-{
-  "scope": "model weights, masks, architecture, normalization and provenance only",
-  "optimizer_resume_states": "deferred by explicit user instruction",
-  "original_files_modified": false
-}
diff --git a/README.md b/README.md
index 1959a9ec6aed18e75b6c7204ef42df84b35c0350..1907aa5e25fd4c68688e0ef3a065a5953dff5361 100644
--- a/README.md
+++ b/README.md
@@ -3,59 +3,63 @@ language: en
 tags:
 - robotics
 - vpp2
-- model-compression
 - structured-pruning
 - robodojo
 ---
 
 # AdaPrune — Bagel-Robotics-1 checkpoint 1000
 
-Complete: listed model artifacts verified against local receipts and remote checksums.
+Public inference release of the **exact checkpoint 1000** used in the Bagel-Robotics-1 evaluation campaign. Downloads require no account, token or access request. This is distinct from Bagel-Robotics-1-Preview.
 
-This is the **physically pruned policy checkpoint selected for the full-suite evaluation campaign**. The uploaded inference files match the SHA-256 fingerprints in all five suite protocols (Precision, Memory, Generalization, Long Horizon and Open). This establishes artifact identity, not completion of every evaluation episode.
+The video expert has **12,468,409,664 parameters**, 23.95% fewer than its recorded parent. Learned masks were converted into smaller FFN and cross-attention layers. Self-attention heads and the action expert are retained. No teacher or learned-mask module is required for inference. Parameter reduction is not a measured inference-speedup claim.
 
-Keep these four files together: `video.pt`, `action.pt`, `compact_architecture.json`, `dataset_stats.json`. No learned-mask module or full teacher is needed for deployed inference.
+## Fixed checkpoint identity
 
-### Evaluation settings
+- Original checkpoint revision: `2651631c5d08819088475e8778676f5cf6d9afcf`.
+- Identity: `790bffc9464fa7a026aa43b217aec2033d186895cad643166daa0bcc01914d5c`.
+- [artifact-manifest.json](artifact-manifest.json) pins all six bundle files and the separate public encoder/tokenizer assets.
+- [SHA256SUMS](SHA256SUMS) records the unchanged SHA-256 hashes of all six bundle files.
 
-- Action horizon: 32; execute 16 actions per query.
-- Action sampling: 10 Euler steps; shift 1; policy seed 1.
-- Temporal ensembling: 90% newly predicted actions, 10% previous overlapping predictions, using the recorded quaternion handling in the evaluation code.
+The release adds documentation and inference packaging without changing any checkpoint file. The original revision is preserved, so historical provenance remains accessible in Git history. Current release files focus on inference. `COMPLETE.json` is the original export receipt, including export-time paths and `simulation_completed: false`; it is retained unchanged for integrity, not used as a current evaluation-status claim. Those historical paths are not required for deployment.
 
-The full-suite campaign comprises 400 Precision, 300 Memory, 600 Generalization, 400 Long Horizon and 400 Open episodes. Frozen protocol snapshots are included under `evaluation/`. They record intended evaluation coverage and checkpoint hashes; this repository does not claim a completed full-suite score.
+## Download the exact checkpoint
 
-### Training and validation context
-
-Recovery after checkpoint 55 used ground-truth action supervision (1.0) and reliability-weighted teacher-video supervision (0.05), with K/V and teacher-action imitation losses both zero. Video weights trained; Action2B remained frozen. Curated 30,000-window subset, global batch 256, training seed 100008. The complete source config is preserved.
-
-At update 1,000, the fixed 140-window offline validation panel reported task-macro normalized native-action MSE of **0.0005726480308242441** against ground truth. The offline action metric scores up to the first 24 valid actions; it differs from the simulation execution interval of 16. `validation.json` contains the original metrics. Lower offline error alone does not establish improved robot success.
+```bash
+hf download bageldotcom/adaprune \
+  --revision 2651631c5d08819088475e8778676f5cf6d9afcf \
+  --include video.pt action.pt compact_architecture.json dataset_stats.json COMPLETE.json validation.json \
+  --local-dir /absolute/path/to/checkpoints/cp1000
+```
 
-The original `COMPLETE.json` has `simulation_completed: false`, reflecting its export-time status. It is preserved unedited; subsequent evaluation provenance is separate.
+There are approximately 29.0 GB of checkpoint tensors. Keep all six files together. The inference downloader checks their sizes and SHA-256 hashes and downloads approximately 16.7 GB of separately pinned VAE/conditioning/tokenizer assets. Allow extra disk space for caches and installation.
 
-## Download
+## Inference adapter and reevaluation
 
-Authenticate with an account that has access to this private repository, then run:
+The standalone XPolicyLab adapter is provided under [`inference/`](inference/), at release tag `adaprune-cp1000-inference-v1`. No private research repository is required.
 
 ```bash
 hf download bageldotcom/adaprune \
-  --revision 2651631c5d08819088475e8778676f5cf6d9afcf \
-  --local-dir /absolute/path/to/checkpoints/cp1000
-```
+  --revision adaprune-cp1000-inference-v1 \
+  --include 'inference/**' \
+  --local-dir /absolute/path/to/adaprune-release
 
-Use the code and setup link below to load these files. Keep the integrity receipts with the downloaded weights.
+git clone https://github.com/XPolicyLab/XPolicyLab.git /absolute/path/to/workspace/XPolicyLab
+git -C /absolute/path/to/workspace/XPolicyLab checkout 6baccd57ff958570d0db1588370289ee36c8bb70
+python /absolute/path/to/adaprune-release/inference/install_adapter.py /absolute/path/to/workspace/XPolicyLab
+```
 
-For the prepared XPolicyLab submission, the policy directory is `policy/bagel_robotics_1` and the checkpoint name is `cp1000`. Its pinned downloader fetches the six required bundle files and the separate VAE/conditioning assets, verifies their hashes, and supplies the standalone inference runtime. The adapter PR is being prepared; no public XPolicyLab PR link is available yet. The technical report and training/pruning/export code are planned for November 20, 2026. Evaluator access must be arranged while this repository remains private.
+Follow the included [policy README](inference/policy/bagel_robotics_1/README.md) to install the Python 3.10 environment, download/verify assets and run both debug checks. Follow [REEVALUATION.md](inference/policy/bagel_robotics_1/REEVALUATION.md) for task, suite and three-seed commands. Install RoboDojo and simulator assets separately using its [official guide](https://robodojo-benchmark.com/doc/usage/install-and-download/).
 
-## Lineage and code
+The adapter uses `policy_name=bagel_robotics_1`, `ckpt_name=cp1000`, ARX X5 end-effector actions, and one environment per policy process. Fixed inference settings: BF16; predict 32 actions and execute 16; ten Euler steps; shift 1; 90% new / 10% previous temporal blending. Qualification used a 96 GB RTX PRO 6000 Blackwell, Python 3.10.21 and PyTorch 2.11.0+cu130; a lower minimum GPU capacity has not been qualified.
 
-Mask learning update 1,500 selected the channels. The physically pruned video expert has **12,468,409,664 parameters**, compared with 16,394,878,784 before pruning (23.95% fewer video parameters). The frozen Action2B expert is separate from this count. Recovery checkpoint 55 already had 55 recovery updates with K/V supervision; the later no-K/V branch continued from it to update 1,000.
+These are compact-aware VPP2 artifacts, not Transformers `AutoModel` checkpoints. The XPolicyLab PR has not been created; publishing this inference release does not imply committee acceptance.
 
-[Historical research code (access may be restricted)](https://github.com/bageldotcom/bagel-robotics/tree/2fc74c246136e543d72ff9af1732936814d3baf8/workflows/distillation) · [Historical research integration PR](https://github.com/bageldotcom/bagel-robotics/pull/169)
+## Validation and release scope
 
-These artifacts require a compact-aware VPP2 loader; they are not Transformers `AutoModel` checkpoints. The prepared XPolicyLab adapter includes that inference runtime and does not require access to the historical research repository. Until its PR is available, reviewers need the standalone adapter package from Bagel Labs. External VAE/conditioning encoders and RoboDojo data remain separate dependencies. Original upstream model/data terms continue to apply; this backup does not assert a new redistribution license.
+The adapter's existing qualification includes 15 CPU tests, both real-checkpoint debug transports, four valid native simulator episodes, and eight exact action/history/reset/backend checks on the tested recordings. Full multi-seed official leaderboard validation remains separate. Historical research scores and scene exceptions are documented in the adapter's EVALUATION_RESULTS.md; they are not official leaderboard results.
 
-## Integrity and training continuation
+The technical report and training/pruning/export code are planned for **November 20, 2026**. This release contains inference code and weights; optimizer, scheduler, distributed RNG/sampler state, training datasets and simulator assets are not included. It does not support exact training continuation from weights alone.
 
-`UPLOAD_INVENTORY.json` records every uploaded artifact's original path, size, and SHA-256. Upload verifies the source checksum against its recorded receipt and the remote file checksum. Original files are never modified or deleted.
+## Attribution and terms
 
-**Optimizer, scheduler, distributed RNG/sampler state and ZeRO shards are not included in this model-weight upload.** Model weights alone cannot exactly resume training. `RESUME_STATE_LOCAL.json` identifies the preserved local full checkpoint and its checksums; it does not mean those shards are on Hugging Face.
+Derived from [VPP2](https://github.com/roboterax/video-prediction-policy-2) and Wan2.1 components. See [THIRD_PARTY_NOTICES.md](THIRD_PARTY_NOTICES.md) and the adapter's retained MIT/Apache-2.0 notices. Model and auxiliary-asset upstream terms continue to apply; an inference-code license is not a new blanket license for all weights or datasets.
diff --git a/RELEASE.json b/RELEASE.json
new file mode 100644
index 0000000000000000000000000000000000000000..bdfccdec5fc47aab9ac3ddd15f0cf03dc3395d36
--- /dev/null
+++ b/RELEASE.json
@@ -0,0 +1,36 @@
+{
+  "model_name": "Bagel-Robotics-1",
+  "release_name": "AdaPrune checkpoint 1000",
+  "repository": "bageldotcom/adaprune",
+  "checkpoint_revision": "2651631c5d08819088475e8778676f5cf6d9afcf",
+  "checkpoint_identity": "790bffc9464fa7a026aa43b217aec2033d186895cad643166daa0bcc01914d5c",
+  "files": {
+    "COMPLETE.json": {
+      "sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
+      "size": 2491
+    },
+    "action.pt": {
+      "sha256": "9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3",
+      "size": 4057916577
+    },
+    "compact_architecture.json": {
+      "sha256": "a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32",
+      "size": 6563270
+    },
+    "dataset_stats.json": {
+      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
+      "size": 100256
+    },
+    "validation.json": {
+      "sha256": "9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e",
+      "size": 88781
+    },
+    "video.pt": {
+      "sha256": "fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb",
+      "size": 24937214042
+    }
+  },
+  "scope": "Public inference checkpoint and supporting adapter; training/pruning/export source planned November 20, 2026",
+  "historical_provenance": "Original checkpoint revision and checksums preserved. Older commits retain original provenance and export records; the current release view contains inference materials.",
+  "official_leaderboard_result": false
+}
diff --git a/RESUME_STATE_LOCAL.json b/RESUME_STATE_LOCAL.json
deleted file mode 100644
index 38784bfe197976af073e251e3e77b43c445a8e49..0000000000000000000000000000000000000000
--- a/RESUME_STATE_LOCAL.json
+++ /dev/null
@@ -1,80 +0,0 @@
-{
-  "original_path": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103/recovery/checkpoints/update_001000",
-  "reason": "Full optimizer/resume shards exceed the personal free-tier private-storage allowance and are deferred pending destination selection.",
-  "receipt": {
-    "checkpoint_identity": "2899ec0a0f6af889a1babd95c7475e7bc6449b39b12b354384964f5abd30d388",
-    "files": {
-      "bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt": {
-        "sha256": "0d33035235009137414b82d1d8e08f8042e3b6b2a5751f8078ee7eb66d7c4dee",
-        "size": 17208428484
-      },
-      "bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt": {
-        "sha256": "85ea00ab82114fe05b62f0441b8615f5e2c569d4cdee1cd0d7eaee96243fed63",
-        "size": 17213174428
-      },
-      "bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt": {
-        "sha256": "37c505863a1129efe703fd34d992b45adc440e9dd93bd1306da9d39aa856a0a8",
-        "size": 17211998172
-      },
-      "bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt": {
-        "sha256": "923d3b5fde2d350eb134cc3bfc93a6a98d589b0ef062d8ee945c184cf4422539",
-        "size": 17214269733
-      },
-      "bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt": {
-        "sha256": "c362c831a50935204078a434f05862560d4657d61a2d822f4bade80ab8ec4751",
-        "size": 17211683350
-      },
-      "bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt": {
-        "sha256": "9ff16b0c5a15154c6e486d206379ac70f58fb8a156be12ed7ba777d537ccba76",
-        "size": 17205282759
-      },
-      "bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt": {
-        "sha256": "d3d625ded0abb49ae560a5a9e7356f2f345246f6d2deb1fbf492a8c3dde3a68d",
-        "size": 17193645344
-      },
-      "bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt": {
-        "sha256": "b64e40e2b433e0a95f28ccebd0f3fa899f2bedfb77ef62a9b0371e3f093a21a0",
-        "size": 17196229893
-      },
-      "zero_pp_rank_0_mp_rank_00_model_states.pt": {
-        "sha256": "4eec7353b2f34dbfff54df2013f5261bb20643e1509760766f02686d37d4a94e",
-        "size": 78779
-      },
-      "zero_pp_rank_1_mp_rank_00_model_states.pt": {
-        "sha256": "ab1b0fd39ea72506f2d8f74d570968e646c01ca82f3cd04705d888b137f74090",
-        "size": 78783
-      },
-      "zero_pp_rank_2_mp_rank_00_model_states.pt": {
-        "sha256": "30082af65f03a955dc837eb72eebf1dc2dc1aec23af1f5d20eef1a9f6edae2fe",
-        "size": 78780
-      },
-      "zero_pp_rank_3_mp_rank_00_model_states.pt": {
-        "sha256": "126b9d5641ba0e0c59967ef41285771ad5c76e70a8f8896c745fb9a4f1b54a81",
-        "size": 78781
-      },
-      "zero_pp_rank_4_mp_rank_00_model_states.pt": {
-        "sha256": "93278bc910a03046b2463d964d4187d2d6ae7ddd33b3f1f039aaf425b0dc8924",
-        "size": 78782
-      },
-      "zero_pp_rank_5_mp_rank_00_model_states.pt": {
-        "sha256": "a63df2fdeb27465b234d0f9f32283a86043d38a6e85a96fc89d16046fe5f4c22",
-        "size": 78770
-      },
-      "zero_pp_rank_6_mp_rank_00_model_states.pt": {
-        "sha256": "c7cad4d47a16c0a1bc33e39255001ac3c0b61e7fcdb36ccb337483e4d83a46a8",
-        "size": 78779
-      },
-      "zero_pp_rank_7_mp_rank_00_model_states.pt": {
-        "sha256": "43029cb15da85321fff5e2751376e7178dd239c852d92a6dbc3d73d1ab8db0f2",
-        "size": 78779
-      }
-    },
-    "identity": "790bffc9464fa7a026aa43b217aec2033d186895cad643166daa0bcc01914d5c",
-    "mask_identity": null,
-    "recovery_export_sha256": "2917adfeadbc2c6d4da60b7739dbee861d7f318f8fc66877f1ca6bd691938172",
-    "stage": "recovery",
-    "update": 1000
-  },
-  "total_shard_bytes": 137655342396,
-  "uploaded": false
-}
diff --git a/SHA256SUMS b/SHA256SUMS
new file mode 100644
index 0000000000000000000000000000000000000000..884ce4d321dbe2038c6a24f0b389b767df77b706
--- /dev/null
+++ b/SHA256SUMS
@@ -0,0 +1,6 @@
+eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1  COMPLETE.json
+9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3  action.pt
+a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32  compact_architecture.json
+2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16  dataset_stats.json
+9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e  validation.json
+fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb  video.pt
diff --git a/THIRD_PARTY_NOTICES.md b/THIRD_PARTY_NOTICES.md
new file mode 100644
index 0000000000000000000000000000000000000000..7c7140daa83b46bf89a8ef0a227cc2585c1d8932
--- /dev/null
+++ b/THIRD_PARTY_NOTICES.md
@@ -0,0 +1,13 @@
+# Source and artifact notices
+
+- The retained VPP2/FastWAM-derived runtime carries the original MIT notice in
+  `LICENSE`, copyright 2026 The FastWAM Authors. Bagel-specific packaging does not
+  remove that attribution. See `source-provenance.json` for frozen-source hashes.
+- Wan2.1/DiffSynth-derived model components retain Apache-2.0 terms; see
+  `licenses/Apache-2.0.txt` and upstream Wan2.1/DiffSynth source notices.
+- The XPolicyLab interface follows its upstream license, included in
+  `licenses/XPolicyLab.txt`. RoboDojo and simulator assets are not bundled.
+- Encoder weights, tokenizer files and model weights have their own upstream
+  terms. The code license does not assign a new license to those artifacts.
+- The existing preview PR by Soroush Mehraban was consulted as a reference; its
+  alternate checkpoint, code and release promises are not this submission.
diff --git a/UPLOAD_INVENTORY.json b/UPLOAD_INVENTORY.json
deleted file mode 100644
index 935a999bc30df7bc85fd7b8008c9789ffb2e9b60..0000000000000000000000000000000000000000
--- a/UPLOAD_INVENTORY.json
+++ /dev/null
@@ -1,123 +0,0 @@
-{
-  "files": [
-    {
-      "bytes": 4057916577,
-      "mtime_ns": 1793716074808787383,
-      "path": "action.pt",
-      "sha256": "9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3",
-      "source": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103/artifacts/best_models/update_001000/action.pt"
-    },
-    {
-      "bytes": 6563270,
-      "mtime_ns": 1793716074811787315,
-      "path": "compact_architecture.json",
-      "sha256": "a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32",
-      "source": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103/artifacts/best_models/update_001000/compact_architecture.json"
-    },
-    {
-      "bytes": 100256,
-      "mtime_ns": 1793716074808787383,
-      "path": "dataset_stats.json",
-      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
-      "source": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103/artifacts/best_models/update_001000/dataset_stats.json"
-    },
-    {
-      "bytes": 88781,
-      "mtime_ns": 1793716074812787292,
-      "path": "validation.json",
-      "sha256": "9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e",
-      "source": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103/artifacts/best_models/update_001000/validation.json"
-    },
-    {
-      "bytes": 24937214042,
-      "mtime_ns": 1793716059522133613,
-      "path": "video.pt",
-      "sha256": "fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb",
-      "source": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103/artifacts/best_models/update_001000/video.pt"
-    },
-    {
-      "bytes": 2491,
-      "mtime_ns": 1793822228420499831,
-      "path": "COMPLETE.json",
-      "sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/COMPLETE.json"
-    },
-    {
-      "bytes": 5403,
-      "mtime_ns": 1793822228422499788,
-      "path": "evaluation/full-suite-manifest.json",
-      "sha256": "07680ea0528ddc40a6851429d402af835e96963be981d1f99fc46d569a6ecf64",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/evaluation/full-suite-manifest.json"
-    },
-    {
-      "bytes": 80852,
-      "mtime_ns": 1793822228422499788,
-      "path": "evaluation/generalization-protocol.json",
-      "sha256": "0e16900a6fc62fe4a1bca7cddb04fe6015db82d889d9e7d969f72544dee18d4b",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/evaluation/generalization-protocol.json"
-    },
-    {
-      "bytes": 80848,
-      "mtime_ns": 1793822228423499766,
-      "path": "evaluation/long-horizon-protocol.json",
-      "sha256": "aa821db302916520d175f76cffe4e5cc18145242d406103005a07e654bf9e3ea",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/evaluation/long-horizon-protocol.json"
-    },
-    {
-      "bytes": 80938,
-      "mtime_ns": 1793822228423499766,
-      "path": "evaluation/memory-protocol.json",
-      "sha256": "731a4efed771eb6cf8061383272d605a86d2db3e9839080b8a77130889087506",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/evaluation/memory-protocol.json"
-    },
-    {
-      "bytes": 80832,
-      "mtime_ns": 1793822228424499744,
-      "path": "evaluation/open-protocol.json",
-      "sha256": "c6333ded75dff92991d01f6ab5fc4bf4eb883021d08cd0bb0b915da223005259",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/evaluation/open-protocol.json"
-    },
-    {
-      "bytes": 598,
-      "mtime_ns": 1793822228425499722,
-      "path": "evaluation/precision-completion.json",
-      "sha256": "751464aae8c9f83e80f92ab2fededc2aed7a22870aac49cd7aa946b2b4274df4",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/evaluation/precision-completion.json"
-    },
-    {
-      "bytes": 80563,
-      "mtime_ns": 1793822228425499722,
-      "path": "evaluation/precision-protocol.json",
-      "sha256": "09c98cae0e01a93c787a17e3e22f0ab482886fb761d4bcc9916b856dbdb7f25d",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/evaluation/precision-protocol.json"
-    },
-    {
-      "bytes": 11380,
-      "mtime_ns": 1793822228426499701,
-      "path": "provenance/config.json",
-      "sha256": "75d5278ad92fa2ba3d44113b74c346302bcfbb1ae4681ad33e74cd25bca714f8",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/provenance/config.json"
-    },
-    {
-      "bytes": 35190,
-      "mtime_ns": 1793822228426499701,
-      "path": "provenance/source_manifest.json",
-      "sha256": "e825a83c1f50d1ff8bbee371a860c901f48da3355050e1958bed81acaa5dc270",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/provenance/source_manifest.json"
-    },
-    {
-      "bytes": 3528,
-      "mtime_ns": 1793822228427499679,
-      "path": "RESUME_STATE_LOCAL.json",
-      "sha256": "970fb75ee3c422edab95e7267f68f42122ece34233a1d86f53db2fa356d9c226",
-      "source": "/home/praveen/hf-vpp2-distillation-upload-20261104/recovery/metadata/RESUME_STATE_LOCAL.json"
-    }
-  ],
-  "original_files_unchanged": true,
-  "prepared_at": "2026-11-04T19:57:08.429540+00:00",
-  "repo_id": "GnanaPraveen/vpp2-recovery-step-1000",
-  "schema": "vpp2-hf-artifact-inventory-v1",
-  "source_code_commit": "2fc74c246136e543d72ff9af1732936814d3baf8",
-  "total_bytes": 29002345549,
-  "training_state_included": false
-}
diff --git a/UPLOAD_VERIFICATION.json b/UPLOAD_VERIFICATION.json
deleted file mode 100644
index e2b6ee48bf6bfdbeabc1657f078e89969f64e5ce..0000000000000000000000000000000000000000
--- a/UPLOAD_VERIFICATION.json
+++ /dev/null
@@ -1,12 +0,0 @@
-{
-  "artifact_revision": "bc72b758356042311edf1592171d48c33e5eeb90",
-  "bytes": 29002345549,
-  "files": 16,
-  "inventory_sha256": "6b27ac39a5c9e96e3ea45be5da6d32006e393a75fd8b3eb921a714bc4c44e1b3",
-  "optimizer_resume_shards_uploaded": false,
-  "original_files_modified": false,
-  "private": true,
-  "repo_id": "GnanaPraveen/vpp2-recovery-step-1000",
-  "verification": "Source SHA-256 verified against recorded receipts; remote LFS SHA-256 or Git blob matched for every artifact; source size/mtime unchanged.",
-  "verified_at": "2026-11-04T20:00:51.244557+00:00"
-}
diff --git a/artifact-manifest.json b/artifact-manifest.json
new file mode 100644
index 0000000000000000000000000000000000000000..fcc1c37a1eaf549c4b77576d0dcbcfcc197c44aa
--- /dev/null
+++ b/artifact-manifest.json
@@ -0,0 +1,112 @@
+{
+  "schema": "bagel-cp1000-submission-v1",
+  "display_name": "Bagel-Robotics-1",
+  "checkpoint_identity": "790bffc9464fa7a026aa43b217aec2033d186895cad643166daa0bcc01914d5c",
+  "update": 1000,
+  "video_parameters": 12468409664,
+  "repo_id": "bageldotcom/adaprune",
+  "revision": "2651631c5d08819088475e8778676f5cf6d9afcf",
+  "private": false,
+  "files": {
+    "COMPLETE.json": {
+      "sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
+      "size": 2491
+    },
+    "action.pt": {
+      "sha256": "9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3",
+      "size": 4057916577
+    },
+    "compact_architecture.json": {
+      "sha256": "a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32",
+      "size": 6563270
+    },
+    "dataset_stats.json": {
+      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
+      "size": 100256
+    },
+    "validation.json": {
+      "sha256": "9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e",
+      "size": 88781
+    },
+    "video.pt": {
+      "sha256": "fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb",
+      "size": 24937214042
+    }
+  },
+  "wan": {
+    "repo": "Haodong082399/VPP2",
+    "revision": "04824ddb3aae11f13db50984ed7d540c3b77dd9f",
+    "files": [
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/LICENSE.txt",
+        "blobId": "261eeb9e9f8b2b4b0d119366dda99c6fd7d35c64",
+        "size": 11357
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth",
+        "blobId": "5897fba405232a6b07a947d6188d19a8e050ccfb",
+        "size": 507609880,
+        "lfs": {
+          "sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+          "size": 507609880,
+          "pointerSize": 134
+        }
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/config.json",
+        "blobId": "f5a12b285ec9c2497cafc4a4443fc6ba61b1a30b",
+        "size": 250
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/google/umt5-xxl/special_tokens_map.json",
+        "blobId": "14855e7052ffbb595057dfd791d293c1c940db2c",
+        "size": 6623
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/google/umt5-xxl/spiece.model",
+        "blobId": "2fe5f347e9f9367585589ae89e997dfbd5cf802c",
+        "size": 4548313,
+        "lfs": {
+          "sha256": "e3909a67b780650b35cf529ac782ad2b6b26e6d1f849d3fbb6a872905f452458",
+          "size": 4548313,
+          "pointerSize": 132
+        }
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/google/umt5-xxl/tokenizer.json",
+        "blobId": "7bd4ee3cf5633e49c0e06a3aac2a621559e6222a",
+        "size": 16837417,
+        "lfs": {
+          "sha256": "6e197b4d3dbd71da14b4eb255f4fa91c9c1f2068b20a2de2472967ca3d22602b",
+          "size": 16837417,
+          "pointerSize": 133
+        }
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/google/umt5-xxl/tokenizer_config.json",
+        "blobId": "4e1cc1cd85599ce0b47fd0a746af188fe4043ff2",
+        "size": 61728
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth",
+        "blobId": "cbdd1a5bd944b2b4c47e21435106316971ac76d8",
+        "size": 4772359047,
+        "lfs": {
+          "sha256": "628c9998b613391f193eb67ff68da9667d75f492911e4eb3decf23460a158c38",
+          "size": 4772359047,
+          "pointerSize": 135
+        }
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth",
+        "blobId": "d5dad910304ab4b909a2c8a225a71840606e6de4",
+        "size": 11361920418,
+        "lfs": {
+          "sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+          "size": 11361920418,
+          "pointerSize": 136
+        }
+      }
+    ]
+  }
+}
diff --git a/evaluation/full-suite-manifest.json b/evaluation/full-suite-manifest.json
deleted file mode 100644
index 446ab7bd56e9ad9f562e87943c1349a2d83d9894..0000000000000000000000000000000000000000
--- a/evaluation/full-suite-manifest.json
+++ /dev/null
@@ -1,146 +0,0 @@
-{
-  "baseline": {
-    "checkpoint": 55,
-    "episodes": 400,
-    "mean_partial_score": 0.39425,
-    "source_url": "https://app.notion.com/p/40ded7607a4081f48d77c47f8549784c",
-    "successes": 122
-  },
-  "checkpoint": 1000,
-  "decision_rule": "Finish all400 checkpoint1000 Precision episodes, then Memory, Generalization, Long Horizon and Open regardless of success rate or partial score; checkpoint950 cancelled by user.",
-  "execution_sources": {
-    "generalization.sbatch": "d6fc7abb1b6170704fa60e4ca75bb574952253316f59ca4ffc85d83bfdeff6ca",
-    "long_horizon.sbatch": "bdf112f24eb2fa5c85567731046544277cd209a561f50d65f54a07bb03a2a7e0",
-    "memory.sbatch": "efc073d50f80bb115c41cc9b8988c1afab1010e6c047f330cd71f12d8c3b1175",
-    "open.sbatch": "5dfd7afdcf8f4b5b97c5fd688b980f58bd61a1a4101a37c825005b55f2ed051c",
-    "precision_handoff.py": "9a783d94112f12542c6715aac55f802e66d392367c1c72cd8f9aa93f65374378",
-    "prepare_followup.py": "c2e457fdea0f9ef3a33f5c233ea019aea31ff89fb2982a5032cc4cadac2d01ac",
-    "suite_common.py": "157988d7033ea528b2dc9200359a4edfbb41e5d2f692da6487108669abf11b1f",
-    "suite_driver.py": "d4fc61af349bd499eddd5d061675405dedede1cda918765b1954d2ec85a55d8e",
-    "test_followup.py": "106e26b1be0059be2444054628000e1dee482ba315adec2ce0b9a46ba371941e"
-  },
-  "id": "c62f853a713ce9200730a1b937a619358a9c146f2d914ddddfe8988b27ae47d6",
-  "input_hashes": {
-    "inputs/authorization.json": "23aba269d4c3f71508be5fdc44e45fbfda7c358fef1ac4b761690f25cbbace45",
-    "inputs/generalization.json": "de5a880e127039c689082262619791f26eb66ae5ed1fe929b8308f93a818f6b5",
-    "inputs/long_horizon.json": "26a9c66ed3082a59ccbb11c1c43d5bc2ad0123ab66a3ffd27750e834c9de2d20",
-    "inputs/memory.json": "b961cfc82459c2a394922903cc6d275aaf647d8cb4561ebafc7dc259bc73928b",
-    "inputs/open.json": "3daaf3e86b2f7dacad8fbe063efc23aa06d805a5c50a60cac347395fec87f4a7",
-    "inputs/unconditional-authorization.json": "a33af04a144b2d59a15c8fe4a74fd93a6b5b94ab7539316abc09b553226bf43a"
-  },
-  "performance_condition": "none",
-  "precision_protocol": "266df4b5d9b8eac16145600a8c4f9223d15e03285d61473c3bbd8c43b694f211",
-  "precision_root": "/home/praveen/experiments/vpp2-recovery1000-950-precision-20261103",
-  "precision_tasks": [
-    "build_tower",
-    "insert_tubes",
-    "deposit_coin",
-    "fasten_screws",
-    "insert_key",
-    "play_Xylophone",
-    "plug_in_charger",
-    "pour_balls_into_vase"
-  ],
-  "replacements": [],
-  "revision_of": "f5491d41dd847cc86f52a9f9b8d7ffc9b8aa9642a595372aa20d51f5026b534f",
-  "schema": "checkpoint1000-unconditional-followup-v2",
-  "selection": "original",
-  "suite_order": [
-    "memory",
-    "generalization",
-    "long_horizon",
-    "open"
-  ],
-  "suites": {
-    "generalization": {
-      "episodes_file": "inputs/generalization.json",
-      "expected": 600,
-      "name": "generalization",
-      "overrides": {},
-      "reporting_holds": {},
-      "runtime": {
-        "protocol": "4d705e28dc3ee7b401ea3f851cfa2b205981a5009d65cc01f3bd467489d50ab6",
-        "root": "/data/praveen/vpp2-recovery1000-followup-20261104/generalization/runtime",
-        "update": 1000
-      },
-      "task_order": [
-        "stack_bowls",
-        "push_T",
-        "pack_objects_into_box",
-        "fold_clothes",
-        "hang_mugs",
-        "sweep_blocks",
-        "pour_liquid_into_cup",
-        "make_toast",
-        "arrange_largest_number",
-        "sort_nesting_dolls_by_size",
-        "store_laptop_and_headphones",
-        "stack_blocks"
-      ]
-    },
-    "long_horizon": {
-      "episodes_file": "inputs/long_horizon.json",
-      "expected": 400,
-      "name": "long_horizon",
-      "overrides": {},
-      "reporting_holds": {},
-      "runtime": {
-        "protocol": "c2867b1bb38b88d54cbe92103a5cf6d2020c3d44f67c89412182265c4bac8256",
-        "root": "/data/praveen/vpp2-recovery1000-followup-20261104/long_horizon/runtime",
-        "update": 1000
-      },
-      "task_order": [
-        "put_bottles_into_dustbin",
-        "fill_pen_holder",
-        "classify_objects",
-        "play_tic_tac_toe",
-        "fill_egg_holder",
-        "organize_table",
-        "make_kong",
-        "play_stacking_toy"
-      ]
-    },
-    "memory": {
-      "episodes_file": "inputs/memory.json",
-      "expected": 300,
-      "name": "memory",
-      "overrides": {},
-      "reporting_holds": {},
-      "runtime": {
-        "protocol": "35ef508554aee3ce8b89ddc3826fb6eda60bd022db79d6d91a60f6f37e7014f8",
-        "root": "/data/praveen/vpp2-recovery1000-followup-20261104/memory/runtime",
-        "update": 1000
-      },
-      "task_order": [
-        "cover_blocks",
-        "match_and_pick_from_conveyor",
-        "swap_blocks",
-        "swap_T",
-        "press_by_number",
-        "imitate_sorting_sequence"
-      ]
-    },
-    "open": {
-      "episodes_file": "inputs/open.json",
-      "expected": 400,
-      "name": "open",
-      "overrides": {},
-      "reporting_holds": {},
-      "runtime": {
-        "protocol": "92558f946ce17044692364bcb3515da3925bcf3c646266c8dc9ce1f38771384e",
-        "root": "/data/praveen/vpp2-recovery1000-followup-20261104/open/runtime",
-        "update": 1000
-      },
-      "task_order": [
-        "align_blocks",
-        "general_pickup",
-        "stack_blocks_by_language",
-        "solve_equation",
-        "classify_objects_by_language",
-        "pick_from_conveyor_by_image",
-        "store_tools_in_toolbox",
-        "pour_by_language"
-      ]
-    }
-  }
-}
diff --git a/evaluation/generalization-protocol.json b/evaluation/generalization-protocol.json
deleted file mode 100644
index 7dccca57339275a0350127e73f706fa49b4ad4e7..0000000000000000000000000000000000000000
--- a/evaluation/generalization-protocol.json
+++ /dev/null
@@ -1,706 +0,0 @@
-{
-  "active_policies": [
-    "vpp2"
-  ],
-  "caveat": "Same checkpoint55 policy/simulator code, benchmark layouts, seeds, r16 and new90/old10. Simulator physics are not claimed bitwise deterministic. Offline ranking used replan24 without ensemble.",
-  "checkpoint_files": {
-    "COMPLETE.json": {
-      "mtime_ns": 1793716093276368990,
-      "sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-      "size": 2491
-    },
-    "action.pt": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3",
-      "size": 4057916577
-    },
-    "compact_architecture.json": {
-      "mtime_ns": 1793716074811787315,
-      "sha256": "a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32",
-      "size": 6563270
-    },
-    "dataset_stats.json": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
-      "size": 100256
-    },
-    "validation.json": {
-      "mtime_ns": 1793716074812787292,
-      "sha256": "9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e",
-      "size": 88781
-    },
-    "video.pt": {
-      "mtime_ns": 1793716059522133613,
-      "sha256": "fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb",
-      "size": 24937214042
-    }
-  },
-  "checkpoint_receipt_sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-  "execution_order": "one report task at a time; eight shared GPU workers",
-  "fresh_policy_per_episode": true,
-  "fresh_simulator_per_episode": true,
-  "gpu_sharing": {
-    "minimum_free_gpu_mib": 61440,
-    "mode": "shared-shard",
-    "one_worker_per_assigned_gpu": true
-  },
-  "id": "4d705e28dc3ee7b401ea3f851cfa2b205981a5009d65cc01f3bd467489d50ab6",
-  "inherited_results": false,
-  "input_hashes": {
-    "inputs/episodes.json": "de5a880e127039c689082262619791f26eb66ae5ed1fe929b8308f93a818f6b5",
-    "inputs/master-episodes.json": "de5a880e127039c689082262619791f26eb66ae5ed1fe929b8308f93a818f6b5",
-    "inputs/vpp2-policy.json": "70bdf6403f8492f02317d39d061b3d11a8f5c5cedbfabcebe889608eb9041b02"
-  },
-  "native_inference": {
-    "horizon": 32,
-    "replan": 16,
-    "seed": 1,
-    "shift": 1,
-    "steps": 10,
-    "temporal_ensemble": {
-      "enabled": true,
-      "new_weight": 0.9,
-      "old_weight": 0.1
-    }
-  },
-  "native_scoring": true,
-  "order": [
-    "vpp2"
-  ],
-  "parent_evaluation": "/home/praveen/experiments/vpp2-recovery1000-950-precision-20261103/update_001000",
-  "parent_protocol": "266df4b5d9b8eac16145600a8c4f9223d15e03285d61473c3bbd8c43b694f211",
-  "primary_episodes": 600,
-  "repetitions": 1,
-  "schema": "checkpoint1000-followup-v1",
-  "scope": true,
-  "selection_manifest": "/data/praveen/vpp2-recovery1000-followup-20261104/inputs/generalization.json",
-  "selection_metric": "task_macro_ground_truth_native_action_mse",
-  "source_changes": [
-    "native_eval/scope.py"
-  ],
-  "source_hashes": {
-    "control/simulator_recovery.py": "fcee066ec9e75ec081b0a955b136e2ab898cd1ee529a2c1e74a453c350c05367",
-    "native_eval/__init__.py": "2deca42c85edcf3d80dba72439ad95641937396e27e1149d77182cfc082267cd",
-    "native_eval/body_coverage.py": "50ce610f86584965281893ff2f98a78643c7e3207014212aad4c13fce4c8a782",
-    "native_eval/camera_group_cleanup.py": "8df69fdba913ec41ff7b1256ab1b43921a0b6fbef7ffbb6dacbf2adaba0f7843",
-    "native_eval/camera_lifecycle.py": "b0952d95d1da8a1bd3e9f544a2eb1eb0f516159b3442666c7acd9bbaee71687a",
-    "native_eval/check.py": "9587d7b200969ccb78ddf750dd3d8616d5e2245e29baa3a0d798e27a55bdaa5e",
-    "native_eval/compact_worker.py": "129a2e26a85e021f7a0282201dec976104a8e1ea1f0a67e6b75ce760257ee46e",
-    "native_eval/completion_integrity.py": "3b737f6dcf9ae9908b1a3602f66fda46c3b8399ee47b1c61f5c0fb9741031029",
-    "native_eval/criteria.py": "97ebc94d91c62d793e614f178d2d120df989faafb54881dd1d3f5d703df8bb76",
-    "native_eval/diagnostics.py": "53b0deb571984fd02ff65c400d5f3b9bc2754b8207ea8edf4e5017c3588ed3cd",
-    "native_eval/ensemble_check.py": "0bef52411b2ec2dd7e1ad21b8690b1f88ed229107b6e46b8dc43c34ca92a3572",
-    "native_eval/evidence.py": "cf7f769cfd838227b26101d7656ebcaede1b1d45a80785a32efb26c9befef0e9",
-    "native_eval/frame_epoch.py": "7b1d3b10f46d5acf82f6e983ecd5f8995310b2fef9fee04bfae168fda02e3efc",
-    "native_eval/hardware.py": "e30ed3359d91595d3053e24124445498a3f1be9d5c0d52f97e1e7d7d88268a5f",
-    "native_eval/hydra_frame.py": "c1aefee86396025b405e050597af7b67bb07c0dd1e65ea4a9be1ed78f8d99072",
-    "native_eval/particle_evidence.py": "378d667b2ab80176045e0bdcdbe3cb39adb612a356b8f6b81d13ae531bde711f",
-    "native_eval/policy.py": "e1cebbfb03c137bd4bf9f054c69be74f9a7f1df5f7ca13c250e606c2346b55b0",
-    "native_eval/prepare.py": "486df6d062098ea07ac63c4d3be51b5f27baf524d0d004411275243783d02ac3",
-    "native_eval/process_lifetime.py": "7dd20bee91c51546b6da067b96012dc5d5a013e8389dafd567b72c1f742cf97d",
-    "native_eval/provenance.py": "8fc2da1678832be9e9b8b778ff0dd1604db5d464cf1a19ff6137acf6b83d3e11",
-    "native_eval/receipts.py": "9ba5800e6e7e796937992117569efe1b3d625900a8e2ebf7955940a4e8cc81e0",
-    "native_eval/replay_receipts.py": "1c60f9bddccf7245e32df97b9e90749bc3ddbe5d65f37ed2d40ba339b458017e",
-    "native_eval/replay_server.py": "1c2c19bae8d1cd3ff564f1775540718c33293663d30161ea7bd293b8f322cab3",
-    "native_eval/report.py": "9d9f0ddc794d2af2ad6e349f991450592a6fe98db292e1ac3ae4fcadea961fc4",
-    "native_eval/repro.py": "f0e4ecd0ba009774105cebeba31f49a70f34c4f88f11855c939368027789a928",
-    "native_eval/reset_coverage.py": "3c6a414878c4cc2ef4114170356db0a6b2193b03ce5c3b4f6293c10c7d3138b0",
-    "native_eval/run.py": "23d2d58fe1160a03bdcf9591b5d5fae671d21832008cc4958ac58e060b88caa1",
-    "native_eval/run_one.py": "dddeea1abb269293bd08c3f5a69ea0a668620b18d07154ca1c0740a0bb213b99",
-    "native_eval/scope.py": "8a25f7509d394956c17e9c24d1dc62d4a2b77b3ba204f9e8a3e1eaf9ac7a2dc2",
-    "native_eval/simulator.py": "d40972be52b7544f62653114217e213d19d4b386bf2991abfbbbc618ab09cf83",
-    "native_eval/status.py": "b49d03681dbc28b80f39f4d6163c3a6e7c5a1ca62f1f0ef502c77dfbf149ca02",
-    "native_eval/topology.py": "062e9246890ad1f2e8bb8b64b3bbe6598f9dce6fcc33ec3eaa106f3cbc0f0ae5",
-    "native_eval/transport.py": "eb6cc295d064a142372dd51859e91233f61f1ad2c8ac8bf3880739f3fe4abec4",
-    "native_eval/vpp2_pilot.py": "571ca3eeba11fbf2afd4a4058629a280bc0ad4bd3eecca178ddde4aa543bd8de",
-    "source/simulator/LICENSE": "7794bb06af8fe5485ca912454ad1665ccd7846c45f9c2b1258658e480948cdbe",
-    "source/simulator/XPolicyLab/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
-    "source/simulator/XPolicyLab/XPolicyLab.py": "a3909558697b73479f8baee03e15b4d74060a7c9ac2f31fcbc6cd659e7364862",
-    "source/simulator/XPolicyLab/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/client_server/__init__.py": "093e47e707f4a5598b0159b5ee371cd9a918b802c70a6e66154a18e9ef8fc5db",
-    "source/simulator/XPolicyLab/client_server/ws/__init__.py": "79905987a12ebf318dc0b2b9ec6f375d9b4561bac29d6c27d04e318d8f6e00b3",
-    "source/simulator/XPolicyLab/client_server/ws/__main__.py": "ee1c5807dcdf1c8f6721c5447463eb51babdb0f9ffbd91d1e7343834ca4aa19a",
-    "source/simulator/XPolicyLab/client_server/ws/model_client.py": "4d6e225ae5f577d862366350dca8742e91dd9527d6e76b8b2a6d48c991212d67",
-    "source/simulator/XPolicyLab/client_server/ws/model_server.py": "4988823b436cd05e6a4d8b13012cc30c7ffd57711b72497b1223a6b10dc0c043",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/__init__.py": "19aeae375e94e58d437d789d0aff07b232c13f375235384d784984631d25d5d9",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/client.py": "b309f58b1bddd99e56578026b61a15e6cfdc07e843e8a082bfd291d0e853ce5a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/codec.py": "8a97ef7e102fd954c266352d81eaceef0f1b0f5504571c03037683efe387865a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/exceptions.py": "73161f4cdcc7ffc8a87cec94f5459bfcb9113076be272d6cb2a3ff8d69e0c6ce",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/keepalive.py": "7f6b9e636eefcd6c80a4b62e404a6f89e2f7e1a3b6008500312a972d6990a21d",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/messages.py": "d05c8cbc20e0cbe1502551864907f25897035993f920748c9290c944c959eab7",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/schemas.py": "f1ca82eafa8dcad6d968508adb51c25342b291f7b1c553fff305c381a6098bd0",
-    "source/simulator/XPolicyLab/model_template.py": "4f8d1eeb8073aa7ae6093abf0f395cd5fef70ff4a4bd1fa6b2d83b356a90754a",
-    "source/simulator/XPolicyLab/policy/VPP2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.py": "ab94a8569e6168ab7cde6660890035e11b34961ff2648854a7b61d188c555b6e",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.yml": "d2c3f06eb2ffa2d1866767dc20db8d60e542fc2418dbc4cf86d4a0f86f944c5e",
-    "source/simulator/XPolicyLab/policy/VPP2/model.py": "699acc0c459c5997db4e3155e2b1b7eb64bf750552436074a369c4eb6e9ffef9",
-    "source/simulator/XPolicyLab/policy/VPP2/temporal_ensemble.py": "42be4b09c56377f4686b2e722de760fa3bafd1c6b92605ada269d94e19efc95e",
-    "source/simulator/XPolicyLab/policy/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.py": "42dc75140935046dfe83f19b3834bd7cf8372e1bdbe51f598b1f91b8d557b7fb",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.yml": "e59de6edeee891d2aa24902ae9034e9c10889fb20116bc9263f30d773a4b0d10",
-    "source/simulator/XPolicyLab/utils/get_free_port.sh": "f45bcb327b45c91367e3e5cdff72d3a42bfb7a09f50e4774701637c4c23fc1a4",
-    "source/simulator/XPolicyLab/utils/load_file.py": "488590d76084f419a36c2fb3e658eeb240ae554678899e9f2ce0cc10da5c4b57",
-    "source/simulator/XPolicyLab/utils/process_data.py": "e044e627eb2677907e9227eda8458c857208913bd484c5e922b53025010eef0b",
-    "source/simulator/XPolicyLab/utils/robot/_robot_info.json": "8a5df36e750971003b484ad770a7a243e523e986d2977a4ecdbeecd4a07946bf",
-    "source/simulator/XPolicyLab/utils/wait_for_policy_server.sh": "f75a8360d83f2960f365dae77804c94a28aca332c2fbc62cdb4a39ae220374ad",
-    "source/simulator/env/camera_manager/camera_manager.py": "1a4250cb3bdc7305fe362a7c6b113f57e6f3e9e4983b69bd61e2c84e8e995471",
-    "source/simulator/env/camera_manager/capture/camera_view.py": "be9ec0d4df8a532c35adc7090fbd7d6a76c163f90e8dac30496ee4f875929869",
-    "source/simulator/env/camera_manager/capture/render_sync.py": "48d70e9a95f3445ae464ab7ea00a00c0d4829752eb55c4ce52ef73ab1f453cae",
-    "source/simulator/env/camera_manager/capture/tiled_capture_manager.py": "959b4d87d96e676bc279d01e7755557bb8c6e73920a75944259f72b7ac385f8b",
-    "source/simulator/env/description_manager/desc_manager.py": "bb566028b523a4cc27363cee2e526c2d3e008e671c9773d4f0b8401f58f0388e",
-    "source/simulator/env/environment/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/env/environment/base_env.py": "ade58ebc19c79bdbf720def30786a0c397b1c7d2141fb6a1bfd63c80f8a3db8e",
-    "source/simulator/env/environment/isaac/__init__.py": "123023ddc75a0f9b1139c7322bee64a27f0e756b6e84242f43a1481428596126",
-    "source/simulator/env/environment/isaac/direct_rl_env.py": "77958f14c465932697efed302c8ef13054f2ec61ee3a4bc5ad66d0390b71df77",
-    "source/simulator/env/environment/isaac/isaac_rl_env.py": "144af40adc2113751bc1bd72caadab145e918b316de53ef3e4e42358c01a1fbb",
-    "source/simulator/env/environment/task_env.py": "ac816cf26efabbb8ac4276d2a2a09fbd557e3db805514db0c3dd7ac22b58e931",
-    "source/simulator/env/global_configs.py": "e96f6141b42e635d69bba5b4727fb1fec54f8965431b3a70d95fe5dcd3d0ee45",
-    "source/simulator/env/observation_manager/obs_manager.py": "771d6cda443b591f3e7850b901a06b4e653a2feefcbabd23608042377fe9be4b",
-    "source/simulator/env/planner_manager/.gitignore": "bdcb46018aafd15f5c2d40adbbccbb06f2dec5b9ec4311cbf776075bdffd4156",
-    "source/simulator/env/planner_manager/curobo_planner.py": "3cd830c002901959409ef6f64d54106d8106b48471b266409abc2199f3cb33e1",
-    "source/simulator/env/reward_manager/func_parser.py": "eb928a05c3eb6bae429a922c59e64272c2c3851c22d779143d45627e47b3a504",
-    "source/simulator/env/reward_manager/reward_manager.py": "bc56efd3aaa861b4b38960dc4fb7778a9b4da9cf131f815ac3170300dd233a41",
-    "source/simulator/env/robot_manager/.gitignore": "fd94daa58076711ae84376170199a63f66177471e587a76252399fe79eb231e2",
-    "source/simulator/env/robot_manager/control_manager.py": "81b7ecc3bed8248c16aa1517e686682d8e28ccfa9e550d45011b579a9219caab",
-    "source/simulator/env/robot_manager/frozen_robot_configs/franka.yml": "a3c944f8cdd07fc1a4eb54cb915f565763292acb24f9c4f725534271449f138c",
-    "source/simulator/env/robot_manager/frozen_robot_configs/x5.yml": "4df90d234656c34132ca573e69a55de5984270e103aac2ea8ab9956a913685b5",
-    "source/simulator/env/robot_manager/robot_class/franka.py": "dc49eb7738a8b58fb731595340c60a1ce32f52ef0dc63a3da84f9c25879af140",
-    "source/simulator/env/robot_manager/robot_class/x5.py": "4748cbe4d273449a48d09817767dec361f53f8e5cdc6e43368fbb8724141b654",
-    "source/simulator/env/robot_manager/robot_config/franka.py": "8bf57a07913573e74dad97aec6c50e58b67746fa90875915d74ce213856ce35a",
-    "source/simulator/env/robot_manager/robot_config/x5.py": "13d35630f833d1b0a1e3df0bea05f843e88aae1248b59af8aadfefa9cad29bbb",
-    "source/simulator/env/robot_manager/robot_manager.py": "19636ff932a0e2192341b57d8695d4102b870bd10c69c9343d712682fcc15082",
-    "source/simulator/env/scene_manager/layout_manager.py": "1b4c940200dd94a1d0300e103d168b9a823e0dc00ce7c921afe16e81d4e01ddc",
-    "source/simulator/env/scene_manager/objects/articulation.py": "0c542a1fc8ba7a61bf71b343745c71949c863291c1b167d4d7048fe2ef9ca4c4",
-    "source/simulator/env/scene_manager/objects/background.py": "04afea9b5c4e21a105ba3b08ce0e5b6692c082cf17ec1053fdffd5dba1c9511b",
-    "source/simulator/env/scene_manager/objects/dynamic.py": "7a4652c649f503681914e37c50c959c7ca8457918bb00628413f4288993f8ea7",
-    "source/simulator/env/scene_manager/objects/fluid.py": "12b183d4d798d6b7188fd42747ee1927747318c386a91135551d1fa0561c41b5",
-    "source/simulator/env/scene_manager/objects/garment.py": "85f38bd85b52da8dd2ce34ce3692d45848016a9f0f14ecf13114705d4a1ff46a",
-    "source/simulator/env/scene_manager/objects/geometry.py": "3100d61b152b894b3b59be0b3be1d2708f2d8f363ddb8311dabfb5411c731f3d",
-    "source/simulator/env/scene_manager/objects/ground.py": "12dd51addae6f5b7c6ad410b4ae7248231d216cf63513c564057d577a09a2972",
-    "source/simulator/env/scene_manager/objects/light.py": "ba8c8529c6608592d4d0eb9638820b70ae6b3c0cd616631bc6dabdb1c20e421a",
-    "source/simulator/env/scene_manager/objects/physics_material.py": "a1ac0dc9928e9253e644ceb9a604aed262514aaad0473aff303bfae448e76c29",
-    "source/simulator/env/scene_manager/objects/primitives.py": "848f4ab08d86307812ebaea18cc86a51638395f0e6ea194b412d45a699ced926",
-    "source/simulator/env/scene_manager/objects/rigid.py": "8260fd2d8350d6a354c8281685bc7e25531923327078752efe0538d9cd78450b",
-    "source/simulator/env/scene_manager/objects/room.py": "946aa1afc23ab9b8fd17394517a5f54605bbbc609582d0854fed2078b3b2d8ee",
-    "source/simulator/env/scene_manager/objects/table.py": "c0e71396b0d53a250f74c75490e91086d67962237400a49f886a13ec07875308",
-    "source/simulator/env/scene_manager/scene_manager.py": "37911b12601038c9ee9bd180cae7a4751ea649bcdaf1ba2628f22c65cfd7dc72",
-    "source/simulator/env/seed_manager/seed_manager.py": "99bcf5a90392c8927aa0dbd799c07e06b9b03befe8b8eabaa8744a3dc9605ba4",
-    "source/simulator/env/seeding.py": "3f88d2a635c528253439cb31a29e5b9d0f47f6aaef0e975d86d9a8cf174be26b",
-    "source/simulator/env_cfg/arx_x5.yml": "7691350159fdeca92fd1b0b2da36913464ccdea3f3251eefb0176aa51db0807f",
-    "source/simulator/env_cfg/camera/camera_config.yml": "c1752170da855e735beba27fddbf78dad16f2395e4ef8b24e9827a768e616a62",
-    "source/simulator/env_cfg/camera/template.py": "68287ed1daec3b8ed7f053bf97a4d48bbddeb8b5757b53e438ac3c4240a9e61b",
-    "source/simulator/env_cfg/robot/_robot_info.json": "4c9de628bc1572502fc48d876c023fc63e02760e1a11a2a5258dea30ffa330e8",
-    "source/simulator/env_cfg/robot/dual_x5.yml": "4eefaccc75c5f83443b7bb85ef1de68ee7a3772e27af2b04a5e8d8769ab9baa1",
-    "source/simulator/env_cfg/robot/dual_x5_and_franka_competition.yml": "ec893f9fde577d0d1209e7dcb85ec2dd74288b636d0c68316cb07bc817b743ab",
-    "source/simulator/env_cfg/scene/conveyor.yml": "c1dfd27faf978f90dfc4825cd83bd203823547c46793f4f9201ce9f4b43beee7",
-    "source/simulator/env_cfg/scene/default.yml": "729df024b9707a55449da94f980c1ca116035e09a1284e2fe071f1b38efd061a",
-    "source/simulator/env_cfg/sim/sim_config.yml": "e442ede1c3d90fcfb50b9461c854d31b22dad9bfbb23397db228b27a6a486963",
-    "source/simulator/scripts/eval_policy.sh": "786966648496b7b68aad81e000a3f92910302eb391dbd07fec0a2dc91107956e",
-    "source/simulator/scripts/internal/summarize_result.py": "9ec361af587e1788d83d3cf8670e05b817ff252d33c33829a6337dc16ae370d2",
-    "source/simulator/scripts/internal/task_inventory.py": "ec82252b0f87fa5a3dbec2f682c170b25cbe47fe181e431a81df2c4b0675084d",
-    "source/simulator/src/__init__.py": "75bfbedb80f2e1421ad787992d78c23349bc8738b9edef7a76a3af83c9586fd6",
-    "source/simulator/src/eval_client/eval_env.py": "c577b27dab3c87588940670da81bcff007cd64b687eee8e0cc6684e5e24966b1",
-    "source/simulator/src/eval_client/main.py": "de2216bb5df707e505f7e2f3d31702907a4c1ab65205c5ea35494021f168c80e",
-    "source/simulator/src/eval_client/physx_warning_monitor.py": "6a9f8e8d75b1031d2cf2feb6dd72d37370d85d5bd80c4bea798622e23ce8dae6",
-    "source/simulator/task/RoboDojo/config/_task.yml": "655a87cd59fa6272bf352d32deb8ade162737b49a9824e16d428fecf0803380b",
-    "source/simulator/task/RoboDojo/config/align_blocks.yml": "082c9d7a32968052b8ce954bee212c7ec0a96b756150701d4b531abdea59be2b",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number.yml": "3bf6b7e286ea328fb1e4a994a8d30af9afb4b0b4dcd7ceafed43cc6160c9978f",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number_random.yml": "145b26eb5c8ed7d084a4ddc6e21694c542f4a86150756188ea38dcf66e565603",
-    "source/simulator/task/RoboDojo/config/build_tower.yml": "e1bab8756bbc5664e3171e2389c060a79a3fbbf92a6af4ef555dcaf59fce275c",
-    "source/simulator/task/RoboDojo/config/classify_objects.yml": "80a754d2aedd6a89f58185eae36f5b2eb1a28250f40a67325f576fde34650384",
-    "source/simulator/task/RoboDojo/config/classify_objects_by_language.yml": "ac6c739ca82a64b53c799af79b44363db1a88246f76d867f8830729e5613507b",
-    "source/simulator/task/RoboDojo/config/cover_blocks.yml": "68f90c85fad2b152cf7853cf48888d09ce74bb27a692ff2feacd11160771c5a7",
-    "source/simulator/task/RoboDojo/config/deposit_coin.yml": "f547b12216fb9a5dfe2e7d7e882d89e1e53ee1a805c943ba09a3c2cdea8c17a3",
-    "source/simulator/task/RoboDojo/config/fasten_screws.yml": "5fbe4cf5b3ee2eafc2771137b13176f01617f8b2131f02503e676a7b62e2eabf",
-    "source/simulator/task/RoboDojo/config/fill_egg_holder.yml": "d79129278c9a5488d63bb1d3865f2c6b39f21f85957ec94ace091af9ae29e31f",
-    "source/simulator/task/RoboDojo/config/fill_pen_holder.yml": "ac72e5582495e8757c547997b317e2fe990370009527b1faa6bfbf4e1275537c",
-    "source/simulator/task/RoboDojo/config/fold_clothes.yml": "56447e3f592e0db388ebb7482ec758ecd5485ec81b9227c2dffe5390bd7d8c8e",
-    "source/simulator/task/RoboDojo/config/fold_clothes_random.yml": "eafe52c0abcc1889069383b008cbe95c6af2d18ceafb1aecc78f0fae3b491b04",
-    "source/simulator/task/RoboDojo/config/general_pickup.yml": "35de68117cb0ed596b30db1c62f1f9090f2cd1237242a03d5aaeb7186e791d93",
-    "source/simulator/task/RoboDojo/config/hang_mugs.yml": "0086d374d3f7e82abb85e43a17cdfa437e1ac645a0171fb47496c2b81427dce9",
-    "source/simulator/task/RoboDojo/config/hang_mugs_random.yml": "847dfd39059275a671b5beff0243bf7bbaeebd9e5c0612a1877cdf12b9007dc6",
-    "source/simulator/task/RoboDojo/config/imitate_sorting_sequence.yml": "d9e194d8c742288601c641ccb75052a67da46e536c0b06c7fb7311d9e2c61a42",
-    "source/simulator/task/RoboDojo/config/insert_key.yml": "c0b409d5c020e0ad706c378b6f63b39dfe7bed4682d978907c40e1cfa7623335",
-    "source/simulator/task/RoboDojo/config/insert_tubes.yml": "5c070e5c764a2ea93f8de0b52e6a664a23c717963fe09fe3af17918a4a1ee9ef",
-    "source/simulator/task/RoboDojo/config/make_kong.yml": "8a33c9e57b6e6c719c4800911c37272ba055ad0f4cda2f1de9d531ee2910351a",
-    "source/simulator/task/RoboDojo/config/make_toast.yml": "d190a26233db174538ff0052b4c2958a3b107f0ccc63d0fdfe28da4c7f5dc57b",
-    "source/simulator/task/RoboDojo/config/make_toast_random.yml": "ca2c1e55a0975c8b019523710e90131504d8d2afeb829be362e4d7927fe97b48",
-    "source/simulator/task/RoboDojo/config/match_and_pick_from_conveyor.yml": "0ae1c2537d9b1880748316928cfd7ba4c6421ecccfedb85c4c5395a7a86c5361",
-    "source/simulator/task/RoboDojo/config/organize_table.yml": "9ffa15dd279060d4f042285a2c53a341c7208e2f89a2f3dd36ee08b22aa453fa",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box.yml": "e232962b8394d08c04c5da14deb88d7b233afa00ab99af8001c084218951d7ca",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box_random.yml": "7ac7f034fcb169e33a7f0a03b2b77e54d0206ed715409359b579055a701196ef",
-    "source/simulator/task/RoboDojo/config/pick_from_conveyor_by_image.yml": "6b310b70a0ead9d09e7c221c6bbc952140a44b1d43034613bf02e954eebcf9dd",
-    "source/simulator/task/RoboDojo/config/play_Xylophone.yml": "28adedb0f03c49693af32cea37855f3f95ec3f76f031bfd571fb8bfffb2ec59c",
-    "source/simulator/task/RoboDojo/config/play_stacking_toy.yml": "a09afc106161064a2ea6a40882d717f7872bd4875309a97431d6863a9b7ffa1c",
-    "source/simulator/task/RoboDojo/config/play_tic_tac_toe.yml": "63f863a7fb81a040f65803c64f97c4e9577d4fe32cd0f92563bdc72d1aa336d2",
-    "source/simulator/task/RoboDojo/config/plug_in_charger.yml": "2046b47112049e038ecfa24a8c8044fb1db84f629a4a32eeb24822157d2d4dcb",
-    "source/simulator/task/RoboDojo/config/pour_balls_into_vase.yml": "084a707af61d24123912777b417fe0c0b230eb9653e686af42570791e924ff8b",
-    "source/simulator/task/RoboDojo/config/pour_by_language.yml": "efb2e3c3cb08dd6d2d2ae011b3e9a218b5f1f0a92f025b8e3b9e8abd7011c250",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup.yml": "32092b11667b3306579b04225303610c061585ace28531fa9996afc339c5a2ba",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup_random.yml": "1603feeab6f0f7dc9c4897e7f40f322b86112b6ab987dad0598f42ac1b35db69",
-    "source/simulator/task/RoboDojo/config/press_by_number.yml": "b784643495bcbca933f5a5fbc91b7dad3465fe888239b358d516977c78c38f63",
-    "source/simulator/task/RoboDojo/config/push_T.yml": "b55ea11002ff3e5dc63982ca8c1247fd7bee4f4a607f4d674e220b86cf120a25",
-    "source/simulator/task/RoboDojo/config/push_T_random.yml": "07609ec20cea40afa1d5720e895fb1b445411f261b2e5b6850e75d0f1930d32d",
-    "source/simulator/task/RoboDojo/config/put_bottles_into_dustbin.yml": "7eddf43e287b6e8aea686c6d686fb87e13ac1be1cb5aac16a8a2b0d2e9eb976e",
-    "source/simulator/task/RoboDojo/config/solve_equation.yml": "bdaad2ff1cc8b7ec6413a3715d49ea9cefd664e7e204544977832e18785f1e75",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size.yml": "8927cb2ebb18449554d851e629bd7d568e12e46c5c3c636c8e7122f4fab1ed1f",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size_random.yml": "f895d3ce7eaae99454706afccd09011f56d077998bcc1e95840116658b78f139",
-    "source/simulator/task/RoboDojo/config/stack_blocks.yml": "2802e88ebc3e2d3d05ce345d75e9ad6151a6690e4ef1e0021a5223d291a402d2",
-    "source/simulator/task/RoboDojo/config/stack_blocks_by_language.yml": "4fbfed4f7fc49120831087e709a8dbdfdb4504eb143196000d5790b54eb3bc40",
-    "source/simulator/task/RoboDojo/config/stack_blocks_random.yml": "e25d2a096601da30abdcfe1256a464eaf48c0c8f844eb2c737232626d49639d5",
-    "source/simulator/task/RoboDojo/config/stack_bowls.yml": "4e2aaa12b8b0f58ca801f1a93638c46d3b49ebad809c5a8011382ccda846dcb1",
-    "source/simulator/task/RoboDojo/config/stack_bowls_random.yml": "5c08af1bcc925d958d53de51e8ebd336006203f1bc81e4ef68f3719eadfabb49",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones.yml": "0fdaab803a9370a9e2d1fc8d86a5788daa3940c08fd6f767334f625642fca3e2",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones_random.yml": "93f0ca39bc8bb32df94e84a02f22ed4d7860f37cdc0b9dd17e44721445849576",
-    "source/simulator/task/RoboDojo/config/store_tools_in_toolbox.yml": "641c8ef92393bd986e408d26a84b1cfcdac27466a5ac5f0b3ebe4bb84430c527",
-    "source/simulator/task/RoboDojo/config/swap_T.yml": "1daf769911e13714b4aeabd208292931e32f4dd39968893152bb7bb477c46c9b",
-    "source/simulator/task/RoboDojo/config/swap_blocks.yml": "b6bb2a9df494719f26c6b1aab2e7576107ae76e2fbac8b63a5770f21e0e9f05f",
-    "source/simulator/task/RoboDojo/config/sweep_blocks.yml": "98aa25ac39be9d335c71b2b586bbba372789e9d61e652e1147152c87febbf1f5",
-    "source/simulator/task/RoboDojo/config/sweep_blocks_random.yml": "4ebff5e2cad07f958dc4b2793b86b1aa803c0809394818ee3752d300e58b6187",
-    "source/simulator/task/RoboDojo/task_registry.py": "11218028f772d9253ee709a7d85aa72d1e1fd87e4f8d33ee3787d6064438f9ab",
-    "source/simulator/task/RoboDojo/tasks/align_blocks.py": "c8597e96132cd7c9d30f3ae101f7bbe8ac28597b629599d75d442bbea4c6e74c",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number.py": "277acd20ae4345dc71267762e588b7b0343387c71685183890f044a4808b7a20",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number_random.py": "01757509978663bf7c0c95059e8ae18d45b8fe0953f6b54540e2bd6bf141182e",
-    "source/simulator/task/RoboDojo/tasks/build_tower.py": "814cf6ffb459871e5cf49f4e127903dbfbc30926f33db71de0b84a71ab235d6e",
-    "source/simulator/task/RoboDojo/tasks/classify_objects.py": "aef4333c088766375e21a928d53a9a93aaee285fecae41e22385acdc74dd11ac",
-    "source/simulator/task/RoboDojo/tasks/classify_objects_by_language.py": "242a8f048f487eb174ef246be78613bbb5f153549ee11a2650412718cf3b36c9",
-    "source/simulator/task/RoboDojo/tasks/cover_blocks.py": "9d14931b16842b93bea8abf9a07991a93a7e663b93c5fd1695108c8b922ab862",
-    "source/simulator/task/RoboDojo/tasks/deposit_coin.py": "f9868ef827ec17b8db6c4650cc5d9a9e593d20d00ee822e2212ce229ec01fb61",
-    "source/simulator/task/RoboDojo/tasks/fasten_screws.py": "2e187a28e33e44efb87baae6463b3ec35264ccd41e35c21a1be44f6361675185",
-    "source/simulator/task/RoboDojo/tasks/fill_egg_holder.py": "506dfb772f8882ae53966e3580affde19c3c0f11591c78cc5f5b1d997ba04f7e",
-    "source/simulator/task/RoboDojo/tasks/fill_pen_holder.py": "8706dce239487e078b21af0b95bcc5cd4eabdcdadcdcbc6400d6b138d982f1e6",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes.py": "ecea9c287d0579605330168e6f36734704fb14a6b1e5dec82b6f2ee3a56ec20a",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes_random.py": "9e40abe3f7201a8c13961edc1ca8075260c315584d3d6b93fa1df966d13d1e88",
-    "source/simulator/task/RoboDojo/tasks/general_pickup.py": "1a9eef7fe518f9d2c6805ad1a8f97c1045fa83726f84b9c63f67ffdfabbc40a7",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs.py": "4bc6586bd87de030b9c580fea2a45e7977c3b6a4d75cf3c43b546d349d42e906",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs_random.py": "ebfa4a95deb954aa7a9063a07fe366cb5ea1835d3f52342b10d531225dfeee92",
-    "source/simulator/task/RoboDojo/tasks/imitate_sorting_sequence.py": "18400cb3c5f5103492819df2dcdf05a77916106e8a503cb62b4ee6cdbaa9305b",
-    "source/simulator/task/RoboDojo/tasks/insert_key.py": "180978a2915744f0596fbca4ea0ad6dcb47dce005d43f06807249bc8b4d1a828",
-    "source/simulator/task/RoboDojo/tasks/insert_tubes.py": "9a68795ddbd0b816ffd7627e673390ecd83594c0e9d374d4e3abf3435ad439e1",
-    "source/simulator/task/RoboDojo/tasks/make_kong.py": "b9266689d65f9213b05219ad952344f9fd63be8892097ec0103af99791b245d5",
-    "source/simulator/task/RoboDojo/tasks/make_toast.py": "1047f484a73d1f92304393879b51f87e33d661b6ad7937198352fe37a0d3233d",
-    "source/simulator/task/RoboDojo/tasks/make_toast_random.py": "69fb879b89e98f7369971834eb6d1b95c5a65f5ffe978b2cffae0b12c6b756ca",
-    "source/simulator/task/RoboDojo/tasks/match_and_pick_from_conveyor.py": "53110b8a0c971bd3ae8878bfb7d0dcd9626bd9b610d4e7fc91772313b435a5c0",
-    "source/simulator/task/RoboDojo/tasks/organize_table.py": "aceb298e78f3f39e4cf9a9f4c5a4f747b2c55f1a6006de2065e110a05f4295d6",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box.py": "84a1cf1dd5e166f36291130f9d8cb5dfe4b3f4732a46674bef6374a7ea2d66be",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box_random.py": "4db259a1b97cdcc39ebc9794244fadc79d31c4ae64069c9b11a233356fcb655c",
-    "source/simulator/task/RoboDojo/tasks/pick_from_conveyor_by_image.py": "a0d29576d0aab8232d2fd3ebcc8c510e81c70b4cc04602a5be3e99ee2275e01d",
-    "source/simulator/task/RoboDojo/tasks/play_Xylophone.py": "b7a4b0bc0457bcf55d714ca1867476fc4aabc53084adfc78d3ff5d1437bf7847",
-    "source/simulator/task/RoboDojo/tasks/play_stacking_toy.py": "23361329ee7efe330da1f1c241415205bafed2b6c082a913659d868a80d338c3",
-    "source/simulator/task/RoboDojo/tasks/play_tic_tac_toe.py": "8f75ebfe55c997ad9bd3b342db46e843e87397db94356f31cadcbe0cb483bca7",
-    "source/simulator/task/RoboDojo/tasks/plug_in_charger.py": "aaa59ade46dc54f430e1148e3f6fae487970cd29eb2c07b5ff7b4045edb831b6",
-    "source/simulator/task/RoboDojo/tasks/pour_balls_into_vase.py": "eae388b93ec196c401f76bbab3b0119cb2164f90bece0225d4fb925d3f2b73ab",
-    "source/simulator/task/RoboDojo/tasks/pour_by_language.py": "f7a15a7fca2c649a7918d68960ede27e667f409a339b668dd9876db890724cfb",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup.py": "73f8c91b57c352fdc1e7ba3fe5354b3c10e4156d8e5ba777762ad0fa1aff5663",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup_random.py": "a1c4097b6e34b2c62885b3a9c7d2584a93d9e296dc41d311a80c2e6d541d53af",
-    "source/simulator/task/RoboDojo/tasks/press_by_number.py": "006f969087c00c417a63875a9de153f698c52d1f1ca0caf0aa16076aa15059c5",
-    "source/simulator/task/RoboDojo/tasks/push_T.py": "a01d753bc320ffbf492ccc29e5f7ab3b39e6c7d255cd25804db1934a30fa5b41",
-    "source/simulator/task/RoboDojo/tasks/push_T_random.py": "dc8d9c8e0d2e8dfffa2b7d433a8278c0eaf707cc4861b15035364c2e70061dab",
-    "source/simulator/task/RoboDojo/tasks/put_bottles_into_dustbin.py": "3ba6e19df0c5ff19ba961092f031ce14e91e72a9a4a5934eece6000a1ffa1c48",
-    "source/simulator/task/RoboDojo/tasks/solve_equation.py": "e287a2f36b82b28e8ca7824f63b1b723b02972ea6dfc115ff32b2882729751f7",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size.py": "1838fa6b515bad3796895cb5c1883c6a42789b7dada5434765764600e70beb49",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size_random.py": "b793ab5bd32bd2c10bd46874440904837f9c4e290e7508b06426e2e96425709a",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks.py": "50bd388f9e47002966fbb0146742467a820295b8686e0f86379f0efbcfffcc1b",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_by_language.py": "36a83925f35c7a4caac5add328dd35cb1d346bad833d3064f455d88221e7a41e",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_random.py": "19d4e5cd693e068ad64de75cbcd24977ab8d39e22fa17362a391dcd733bc9bb5",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls.py": "022c129ec3493183b1a26688a0402f5a00a1546ae011999e9c19f4c0029e2b7d",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls_random.py": "a0e20469eb44a9ee98650d09386814c1df5264c68ec139f06aabf0c86f092514",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones.py": "2bee4e31f0594900629fe7f43266e16fecb310b152930ce51524507136809a92",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones_random.py": "58c3c5113054a379214d64ae406ed8dde99603cf684ce9cd8c7bcb72ed334382",
-    "source/simulator/task/RoboDojo/tasks/store_tools_in_toolbox.py": "2969fc3eb171560b3c0bcea62b9f38b8b35f3aecc9bcc7085c3b1cec365b48fc",
-    "source/simulator/task/RoboDojo/tasks/swap_T.py": "57fc289b3957caa454e1d8e005e3a57c973d6ab40121845c94799e61e4bfafdf",
-    "source/simulator/task/RoboDojo/tasks/swap_blocks.py": "08ed24ae9dff27f4e5c850161a02f62ce237859c096d9b06a853a7890e0ad4ed",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks.py": "65786217e199c1e2b26ec8e2c232bf98ecb304a20939d0908fe12bc41e990614",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks_random.py": "68b6127038322cf001d5f614060b8a1d96e64a49efddfcdcae91127d5c206bc5",
-    "source/simulator/utils/cluttered_generator.py": "f59cfcfed2b4cce5ae9aeaf3339c86d17e237874e3526bded3cb05136dbcbd50",
-    "source/simulator/utils/ensure_usd_path.py": "7c8e71cf91636e630b1b174098fe0b38740857daeaf207e2fae772c4727b11bf",
-    "source/simulator/utils/load_file.py": "ef3a6dc96aa1a7442cfd24f7be9671052a98d2f20783c3e71032332704132943",
-    "source/simulator/utils/path.py": "1d427c7cf33e748e01d0d083c9432d8be8b5c3a1f35c5fb0f6ebb0714098f1d1",
-    "source/simulator/utils/pipeline_utils.py": "d5d8da29e6ca84ab04323a2f79130f930572785b4399634e06a42834ed4087ae",
-    "source/simulator/utils/rotations.py": "8e814f79938bdb9e85f1c024825617cc08d281b61728c6410b23ef1bd852c340",
-    "source/simulator/utils/save_file.py": "6cff5950f73bf7a1ed404f804d6f08af712974f22c3955de4cdf142f211aaf00",
-    "source/simulator/utils/transformer.py": "73de4f5e1e0ff60d4f2e1aaf6104368e8f919269facb68b4fa0344d35bc08cbe",
-    "source/simulator/utils/usd_schema.py": "260d30b0ada2976136cad3e49e810b4409304eb4a2bc60bd9ed710864d3953ac",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/METADATA": "614985b278f601b8ef863f82783b991cc3ada4b89f592eb71c7c48c5f2ef2298",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/RECORD": "c7660eb223c8080f75c49690eaf0a9c9c00b6195511791b33b15e6b24e42edd7",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/WHEEL": "942926c567d0b1273d09f20295879abd37507845f6b683eab9f87d808df91100",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/licenses/LICENSE": "fe1049884b1a0d9342901e88e07f32925d24b3121d9972b6a6805fb9824b095d",
-    "source/transport_deps/annotated_types/__init__.py": "a7104a4d439b27a9f74fc0be236b9ba1b7831e6044026802a205abc1298a9bc8",
-    "source/transport_deps/annotated_types/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types/test_cases.py": "d867472acb57b81ce97f78972828e6bc5669ca4db33fd1e6f645c6183846068e",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/AUTHORS.md": "5fe0446ddd6d1c70fa298747368582dc1d91c23d62c8fb7228595bc87fd53232",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/LICENSE.md": "509aa7af19ae2b5b1681b22b969627ee9e1fd15ff7ea5bf9f6e301fc44d5a6c2",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/METADATA": "b7ee8fa1bf018747082cd6d2fcaa14e5d0f5956b5477ab5810eb61ac4cc7bcf9",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/RECORD": "51331727bf8d9ff60510de5e7756c5dd72aba7663f3673e2827bf1ca5c09c7e3",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/WHEEL": "1d7faff7e9e8524c94a31c99d4f312b92edab94c43011e1505da182ea0f4c70b",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/top_level.txt": "19da698ad1490bdde1c7f18fabaceb2625aaf48b364d4dd7ad404c7e5eed75b9",
-    "source/transport_deps/msgpack_numpy.py": "0a45a7430e2cefa2e74235c4c6a34a22ead2fb57a7de8b2631f45df977701d63",
-    "source/transport_deps/pydantic-2.12.3.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic-2.12.3.dist-info/METADATA": "348336c208743aafeabfa5537c29f429035a3107b7a78bbb6a3bf7b3a42fe4aa",
-    "source/transport_deps/pydantic-2.12.3.dist-info/RECORD": "089c7ceb033486c4e6106749bb1a38c80b63905da74d842d067735e8b9c86e4d",
-    "source/transport_deps/pydantic-2.12.3.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic-2.12.3.dist-info/WHEEL": "aad0b0a12256807936d52d4a6f88a1773236ae527564a688bab4e3fe780e8724",
-    "source/transport_deps/pydantic-2.12.3.dist-info/licenses/LICENSE": "a9e186f3ca16b5eef84318e7a701721351a00cb7b8ae3a4394b67b49e3529ef3",
-    "source/transport_deps/pydantic/__init__.py": "e62127278c07bf5384cdd2903f368a69929f3b8a524000bae4e0eb608ebf4bc6",
-    "source/transport_deps/pydantic/_internal/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/_internal/_config.py": "4d66708377346d91e24f76e8479f9862a928b877308d175e9e6c861e87d5ec4d",
-    "source/transport_deps/pydantic/_internal/_core_metadata.py": "63f836b778bbba5b8afb05c266fcc97d114c3d4336dda0582df69ee05cc13f2d",
-    "source/transport_deps/pydantic/_internal/_core_utils.py": "d63aeee156c9d31eb747a76d55cb8e23e74a4130bf77f9529c9584050cc63444",
-    "source/transport_deps/pydantic/_internal/_dataclasses.py": "4e4d6611a7e169dd6457b2b9b4f5f9070c5649763b0be30ac1fd0e2c58089440",
-    "source/transport_deps/pydantic/_internal/_decorators.py": "3e7c80a0a4a0dc135b0954999f0ab0f67684835503b58bc34fd2e5ba280f88ef",
-    "source/transport_deps/pydantic/_internal/_decorators_v1.py": "b5f75f76940a6384765c23b0a876d97a840cbabe95362811adf86e757047c77f",
-    "source/transport_deps/pydantic/_internal/_discriminated_union.py": "68c9744914b24321df5b8fa49673131cdbf0491a2a1371f73d157fd39bd1b138",
-    "source/transport_deps/pydantic/_internal/_docs_extraction.py": "7f2ce74801e1e40ce88675d94ad5741ef1fe9d16dabcd1cfca0fa49f1f92fc41",
-    "source/transport_deps/pydantic/_internal/_fields.py": "6127c42aadb51608cb27a62a6172a1d1e104b399f130fbd0ea1a7da40f0dcdfc",
-    "source/transport_deps/pydantic/_internal/_forward_ref.py": "e67dd8efedc028b9fcfc54b761cecabad2e23d48725e602d90464e6859e89f03",
-    "source/transport_deps/pydantic/_internal/_generate_schema.py": "4d3e3dbf3633a87f74ad6aefe69b4da19823cceb11d0a3e54a4a8f545ae7ac1c",
-    "source/transport_deps/pydantic/_internal/_generics.py": "10baa34fa2cccd0cd600ad04079ffdaa47bf8806b3cf4390e20ba7a7fb8ab986",
-    "source/transport_deps/pydantic/_internal/_git.py": "2303e1dc33df6b65eadeb06e07d371f25b91d80d62ebd41d7937d659722e0958",
-    "source/transport_deps/pydantic/_internal/_import_utils.py": "4d18710f93ae63a094a2c8a805d05c254b34a26ec820e36265d60057bcd0f233",
-    "source/transport_deps/pydantic/_internal/_internal_dataclass.py": "fdb79d7355dbbaeca044688b66a914930c05a506a847584a2e5479d359f2c926",
-    "source/transport_deps/pydantic/_internal/_known_annotated_metadata.py": "25ceca4cd159a01ddffb489b3ff36024834e795bd8137ab73930500e354c9375",
-    "source/transport_deps/pydantic/_internal/_mock_val_ser.py": "c2645115206fa9f70b6c8e353ec165881e2ed80677989a597a21116c3df1293a",
-    "source/transport_deps/pydantic/_internal/_model_construction.py": "c24f9b3460c026f76e686be7d14d3ff33125d06111bb4b2126f37cfd97e461ac",
-    "source/transport_deps/pydantic/_internal/_namespace_utils.py": "865dfe4d102bf365368d3c8f3f7b7e42ab2f28b8abc6d90b7cd7eb37e7e9d31f",
-    "source/transport_deps/pydantic/_internal/_repr.py": "8d07e726ec83c50a5244d846dbd208f4f5fc7b836fdaa599ac4c3696a8a1dd41",
-    "source/transport_deps/pydantic/_internal/_schema_gather.py": "54b12fe754d811e78d0f6733b32ae626ad4c56726a4ce9a72da9fb546e3873c0",
-    "source/transport_deps/pydantic/_internal/_schema_generation_shared.py": "17fadb41bae4a289a0c6cb24747a4fd23509ed309f7b40400c0124aba089e273",
-    "source/transport_deps/pydantic/_internal/_serializers.py": "6085af4a6011e5f9db19258239076e5add72078650638d9e02bb9cf9e9eb6fa7",
-    "source/transport_deps/pydantic/_internal/_signature.py": "f049633c97b8a529daa6e8ab1b90e4040803d618201f1100cb314f1fef47d331",
-    "source/transport_deps/pydantic/_internal/_typing_extra.py": "755c481a1b0be37cda79b3423108752b94235fa2386bd3959cd01c2f6a486bfc",
-    "source/transport_deps/pydantic/_internal/_utils.py": "0878c7fb4ce75235fd479b144a33d3f23f3efbb3f37bcca36a27ac7130152e24",
-    "source/transport_deps/pydantic/_internal/_validate_call.py": "3df7559d2ce15ceac436d683a03c373c558f5580f95bf80d60f7bca77520e8b8",
-    "source/transport_deps/pydantic/_internal/_validators.py": "76fd1ad8d91ce33718aafdf51a1fd021dda571ff96d24429574a1237380475f8",
-    "source/transport_deps/pydantic/_migration.py": "545ef72d1094cf722b6f9c55b75de36f7340717567105e93d7e27438b7de6790",
-    "source/transport_deps/pydantic/alias_generators.py": "28cd67deee097cb48197552e620de1a181f35c90fecaf82b9eaf2ed5c73087f0",
-    "source/transport_deps/pydantic/aliases.py": "be1087ca84969d7f8427ec166f9aa3e31c91b2c8065a74d07f3429e06499f6e8",
-    "source/transport_deps/pydantic/annotated_handlers.py": "59fc854aac28108157061ed3ef73f270a9682350e25f8e465a2d3e24eb024786",
-    "source/transport_deps/pydantic/class_validators.py": "8bf5778fe3d87462d22e68ff20965e913463ba73bc488573f0b325aae3f23fb1",
-    "source/transport_deps/pydantic/color.py": "033a867d540717ddbf65cb4372e7b40cce324e9d8fead7a49304483535ab2c8a",
-    "source/transport_deps/pydantic/config.py": "b83676c4df89ec255035e9c269529c62c244cc83a92aa9607d0bc3774d35c229",
-    "source/transport_deps/pydantic/dataclasses.py": "dbf8cdfb9d595abbb222c7a78cb235f9585de2a206345ddeb9ef6319c769a872",
-    "source/transport_deps/pydantic/datetime_parse.py": "402f9680cc4caffc10fe6357512ec055ffb684b121befb0f6353d0ca148639d9",
-    "source/transport_deps/pydantic/decorator.py": "617fa3500a6ee4029a55628fa1a57e9fee25ed46d2ebd184b7d45ade1b3398a2",
-    "source/transport_deps/pydantic/deprecated/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/deprecated/class_validators.py": "10071a550339ce9db0066974c9b37ad827d07f226f0cbc7941df4f938fed6f85",
-    "source/transport_deps/pydantic/deprecated/config.py": "93f96c564e7ba5ac4b24e701b9e1f4edcbb53a012059d541c66ea57da0b70825",
-    "source/transport_deps/pydantic/deprecated/copy_internals.py": "1a177ebe431de446020a0c821ad3ca3b9f27a7d704281402eaa9017881052368",
-    "source/transport_deps/pydantic/deprecated/decorator.py": "4c19ba6c9ef026c36287ff16ab92330dcc0fdf69bdfef7f1b3de9d7ac2ee934d",
-    "source/transport_deps/pydantic/deprecated/json.py": "1e55821b7e5146bc72cee4d2e8b4d089907046198365699ea901fcacb5bac00f",
-    "source/transport_deps/pydantic/deprecated/parse.py": "1b377a6ff83ccc95dcb84ed046ae5a761c7f10c25a8577dca57085d1182baaa2",
-    "source/transport_deps/pydantic/deprecated/tools.py": "36b9bda054595a9f3e8e57ef3e02482c4b32c29e18cd90f9d972063c32f11dc2",
-    "source/transport_deps/pydantic/env_settings.py": "e881de79612a9543d152fdd5f805e217f5bdd5f836270fccdced25df827f7b20",
-    "source/transport_deps/pydantic/error_wrappers.py": "44aea6a804dcf72303f8a043f48252f47a4a0a9ad61ddf30a3ce019e6fb77d1f",
-    "source/transport_deps/pydantic/errors.py": "edcb41342b6de7b919171ef52ecd87f3a22e7d0011bf8c0f29ff0385db159f9c",
-    "source/transport_deps/pydantic/experimental/__init__.py": "413eeb298743b0289327d1848e6b107561d2730a4aaeb3641aaeafa8e34fe914",
-    "source/transport_deps/pydantic/experimental/arguments_schema.py": "1059e35ffba5a7eb4fc948d05f9a6642dba0d4e14bfc071cf1b70c6cb77e7d56",
-    "source/transport_deps/pydantic/experimental/missing_sentinel.py": "8507a382d174d3052e40c9e2f78dbd7af83e797c88c292af8ec0fca1fa9f8fec",
-    "source/transport_deps/pydantic/experimental/pipeline.py": "2affddbdc7b12ae99acdf44bd32ebd01ac9e03a1f7ed2ae6b19dd2525fe7d2a6",
-    "source/transport_deps/pydantic/fields.py": "5ae0c63af076d8a5aeb96ddf5e74b85af838a97fed769f17ec1ac09736b8b30f",
-    "source/transport_deps/pydantic/functional_serializers.py": "ac4cc7dfdd73ab2de8fdb5a4d9012ebf24a67103599b05e62502c2dd9185ed00",
-    "source/transport_deps/pydantic/functional_validators.py": "73ffbbc1e5a918d70e61f45f55416ed758ebc4c54c75f63f73ee22b2dc24f796",
-    "source/transport_deps/pydantic/generics.py": "d19a99f4ef5a9e7223ff7986051aa9b3886d7b2ddbe65575f9ddad5313cc3270",
-    "source/transport_deps/pydantic/json.py": "647f11908ee1e01cfece9f0e7530316c952856f72853e8e131d459d01fa4d317",
-    "source/transport_deps/pydantic/json_schema.py": "677786d509be226f8c417ebac28c963fd5b2c055272d116d68e0cbced559f937",
-    "source/transport_deps/pydantic/main.py": "efb414c13b5041d51a7e8af78abe70d820b5b34be2805b460747d2953bc5dbb2",
-    "source/transport_deps/pydantic/mypy.py": "a7a294d46c0f1dacc5ec4e6f26ad6e2dde2d1dde8313a6eb7b8fa6e4b9f6de6b",
-    "source/transport_deps/pydantic/networks.py": "2c995c70c79df7b1007a9293e8e98f23d73a8e1e62c1b063935fe35f5449c37d",
-    "source/transport_deps/pydantic/parse.py": "c2477cd9d82dbd6b43f3de54fc8e84d61b6a3251a1052604577f5cb8149ea370",
-    "source/transport_deps/pydantic/plugin/__init__.py": "6bb4f0dfae94e8ade496d08235963bea773db2cfbbb86190e344d769df4eca94",
-    "source/transport_deps/pydantic/plugin/_loader.py": "f502d79de2c49afca15e46bff63e0bae46e67b8a8fbfaa98a6196c8c5d8c1ac0",
-    "source/transport_deps/pydantic/plugin/_schema_validator.py": "41b9aab06df730199fb4d4369cd8ae376d8b85badec5403b8a90d5bf7274d815",
-    "source/transport_deps/pydantic/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/root_model.py": "06f98bb56e22d7574993e74b38cdeb97e8e725d40679e41314172610eabaa4c8",
-    "source/transport_deps/pydantic/schema.py": "56aaa3beafcb9daa5592779b51ddc1a7f275a768176599d94602f8f1b5441bba",
-    "source/transport_deps/pydantic/tools.py": "88742977c489e43093b4f5796ad015d3a4fcf5b8d268c15e659d8b5fd95ab196",
-    "source/transport_deps/pydantic/type_adapter.py": "553fbeca0e1adbbb21481cd61c13cacf8f777f7890f686dd9049218d92a513b4",
-    "source/transport_deps/pydantic/types.py": "9ea752f89d995ea4e1daa7b224ecc14c1b475b267961115ef2068c579f5df471",
-    "source/transport_deps/pydantic/typing.py": "3fb7de037e4cc1370bb11d6e2fb75bd12fa8c9d071a1b9976b9e580e80606a34",
-    "source/transport_deps/pydantic/utils.py": "d799d1d90a6a4c1165415e1336d4c8b4cc93271fdf6f257e80f988118d46a287",
-    "source/transport_deps/pydantic/v1/__init__.py": "14b43c212a7a3156517e39d2edf4389b51714311421764158a41380cafb83e11",
-    "source/transport_deps/pydantic/v1/_hypothesis_plugin.py": "e444b9c56bb0d45400b3298b7b3cbc4209d5cf466955f5378e4993ef81f6ed54",
-    "source/transport_deps/pydantic/v1/annotated_types.py": "ba4d8d000c6a88d10b2a388787287129a20e87c1759585bf2f3ad6dd7ee86417",
-    "source/transport_deps/pydantic/v1/class_validators.py": "50b39a21481850306c1cbec411511aadc33e52e6ca52082837c8526c3a27b051",
-    "source/transport_deps/pydantic/v1/color.py": "8990012d8a7a395a28d801643fd229ae2ff049cebe2a496ef18ba149aaed7798",
-    "source/transport_deps/pydantic/v1/config.py": "6ba3f459eafdc7871bc0a5bb5eff29a1252a33858ff912d6c17e9832962af400",
-    "source/transport_deps/pydantic/v1/dataclasses.py": "efce1caaf2276f020f5abf6eb1fa57ddc87bcf8b77276b532ba374ebbff0935a",
-    "source/transport_deps/pydantic/v1/datetime_parse.py": "e10cb5910a6adeb3556497881de48f0e9b92d81be1a7528fb73246d71bbe1f4d",
-    "source/transport_deps/pydantic/v1/decorator.py": "cda6b1c71a163c29bcd7c0f56ecd3285aa518d79b7d95f06d191d609d335b970",
-    "source/transport_deps/pydantic/v1/env_settings.py": "03d557c2d465d36018fa31f40b4a2ecb954dc377e2e85fe9933b870e380000e3",
-    "source/transport_deps/pydantic/v1/error_wrappers.py": "ebadb931fc3daa40b63708ad07f2450167bc07e5efeb3054eeb2fd936f2d7f2a",
-    "source/transport_deps/pydantic/v1/errors.py": "988c0f103e6f18ce50e6fe02e19d493e5753447fa89af132947ea4b0c84e98fc",
-    "source/transport_deps/pydantic/v1/fields.py": "56a5890ab894362132a6d5eba03549e74d49a550347a7d9500d724b2a5cbd9bf",
-    "source/transport_deps/pydantic/v1/generics.py": "5730bd61457e11b3e94376807e4d5c3457a3efdfc8cf39e443b5ab9934d94bd1",
-    "source/transport_deps/pydantic/v1/json.py": "590e47cbf848a5f751dd84bc93a37613a28c273b1d6c18bf95d58e3c9695f353",
-    "source/transport_deps/pydantic/v1/main.py": "cee36974de50d15d301b65144cab741d4cb75c9e0e02f3d26437625d8f85233b",
-    "source/transport_deps/pydantic/v1/mypy.py": "0a5f1745f0a621c544de3e40114e760bc88387c95c5fffc3de1cf68c85b1300b",
-    "source/transport_deps/pydantic/v1/networks.py": "1d836d2807ce98e9ca269b0383583a488923f563e153ffa2f25e5ed89281a46e",
-    "source/transport_deps/pydantic/v1/parse.py": "049b5daa2651b5abfd55114298ec6863e2889909a33f2f80fcda288e21546716",
-    "source/transport_deps/pydantic/v1/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/v1/schema.py": "6aa06e03ef9cabc8005648a6e4124f140487cce67c745b66157fdf346afa8a9e",
-    "source/transport_deps/pydantic/v1/tools.py": "d650dd5c79348cbe6e3f7bb9442600bd402518296000efb8e6592af63edfc810",
-    "source/transport_deps/pydantic/v1/types.py": "07397e45c9e2b4f0479eac308fd8afec98c7b8dd46a67587db874a905de5f5ef",
-    "source/transport_deps/pydantic/v1/typing.py": "ec67418356131d42d4f35b6107ff5c8d13437d97e7e248685fb9c6b70fe47821",
-    "source/transport_deps/pydantic/v1/utils.py": "339151c9f3546f50369a4f6568680255d7c71dbdc0b5082b8cee6a7f205fe310",
-    "source/transport_deps/pydantic/v1/validators.py": "9725249f5316847c65097e597c48058ff0801e88e888f71a41e31d10cf57be25",
-    "source/transport_deps/pydantic/v1/version.py": "1d79d75bed5b316e6a2a196be5180e10fa21ad90c20d2bb2cbcfa08bc182819a",
-    "source/transport_deps/pydantic/validate_call_decorator.py": "f23a8b9605d38d6123e1d5c3834c08dc41902a46f42670ac2ff2525236d4e528",
-    "source/transport_deps/pydantic/validators.py": "a706c825755bd42576980138c3f1067cd8fb0f0cec29a5b0fed4cbe9cbe24eeb",
-    "source/transport_deps/pydantic/version.py": "4e6124eaa03e540d4f56c737f5b66584f010032bda71de812b3c57007c96d395",
-    "source/transport_deps/pydantic/warnings.py": "5aed551b3aef159c384fac8220a1e31fb2d263ae878dbb720856e7e6e5a8309e",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/METADATA": "3434fb886fe254ed414541f305a86e43ff575093f8f6d15c89bd11144ed8e131",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/RECORD": "7f3a82de0054511dbd2f7309c5e5b6ddb3d96c6230bfa83fd73bbdffc51cd0f1",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/WHEEL": "7f32b0ec3e21425972f904759f2efc26a495ce223c174ce187bb87ac82f5240a",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/licenses/LICENSE": "2afdd30d54b4d62b6f488a6bcc1546e84ec5061f13f4209c03d012348783795a",
-    "source/transport_deps/pydantic_core/__init__.py": "9cad6292b75254af606a970aae9bff6e5ae9f0b08089cd63acafa391b60794d2",
-    "source/transport_deps/pydantic_core/_pydantic_core.cpython-310-x86_64-linux-gnu.so": "cd9bb47485cdffc363010507a214155966ca860b0b93bda39583492c6764e631",
-    "source/transport_deps/pydantic_core/_pydantic_core.pyi": "3ea1dbd4182f08cf9341f24b3c5cf7db77a05b3535ffe9e23525127a3617a11f",
-    "source/transport_deps/pydantic_core/core_schema.py": "bbdc850b72d685133a0e250fed263b33691dcdf3813492f3c0e31001e3d46005",
-    "source/transport_deps/pydantic_core/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/METADATA": "bad3d9e6342ef6750cf5128ec942327b8bebdce0b6e4d6f67cb15b27ae5118da",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/RECORD": "cf5e3948866214f2e9ab0c1d43c157e9f9c14f59cf8442936c6cb97d5f2b6a5d",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/WHEEL": "ccec20e2307acd7da453dd7437e70c6b08e2bc3ead3bc3445af135da37b56d59",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/licenses/LICENSE": "804b59b25f2c31bd278f9202a19ae49a3945aa2664387e2d0a128c7cacc61ec3",
-    "source/transport_deps/typing_inspection/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/introspection.py": "5888c45bd627da692d9527699e3f549adf35651596b1656772310e12181b4396",
-    "source/transport_deps/typing_inspection/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/typing_objects.py": "a778d6290d51581c9c9a23e0263af2a4604c76def5c4d40a7cd1a375b0046ff6",
-    "source/transport_deps/typing_inspection/typing_objects.pyi": "3bd24dd92d0270d31d64fd40abc1ef8693f5322aab9a78aeb3c5b62f992f84be",
-    "source/vpp2/src/vpp2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/adaptive_mask/__init__.py": "a73f5515ffb77d323f9c324776e9c166d13817c470506cfd5db4ef0546cbbdda",
-    "source/vpp2/src/vpp2/adaptive_mask/audit.py": "c949909a2e60c3c4389c0221c79459961f2c4925baa07f7985774a52257a1d17",
-    "source/vpp2/src/vpp2/adaptive_mask/budget.py": "74e32285903dbf673ecadb6a60d917a990b9c6619beac9f926c1b3cbdae3887c",
-    "source/vpp2/src/vpp2/adaptive_mask/controller.py": "328f636201a845ebdc211340a9014a5076fa895e10cf9c2e1f91b96761a2514b",
-    "source/vpp2/src/vpp2/adaptive_mask/data.py": "e3b633fedaa503b101c80610cbed0933e9e245acf56a6c0692c4d99620faf84c",
-    "source/vpp2/src/vpp2/adaptive_mask/equivalence.py": "8658cb69d549fa0575570364fe3aa1e0b121fcf8b9965b17cd174d459dddf5ab",
-    "source/vpp2/src/vpp2/adaptive_mask/execution.py": "efec52c9b2eabf0300c72edea18f9cd4cc16a0889496793abb6a50161f2328ef",
-    "source/vpp2/src/vpp2/adaptive_mask/export.py": "0ce51e74787d28bbe9f4e73a3774297b424aa131fa78968d24154ff56db98b0f",
-    "source/vpp2/src/vpp2/adaptive_mask/feedback.py": "519fa4e6416723fc000754815a2f05d3415299ae35724f37bb806ebc22393d38",
-    "source/vpp2/src/vpp2/adaptive_mask/geometry.py": "f1ce4110c3a3a893f4f240c49e2ccc4e93215ac4ba9e447de8042f9b19afab98",
-    "source/vpp2/src/vpp2/adaptive_mask/hidden.py": "968560bd6dae0fce59bd28084091b10d559cf165c81764f6ad24b7492101810d",
-    "source/vpp2/src/vpp2/adaptive_mask/losses.py": "4a7dea6f87ef0fa38a7202d85d98d37460f8f4dc13ca5336f32b1da8b84a4322",
-    "source/vpp2/src/vpp2/adaptive_mask/masking.py": "31b595916410c60f772fff1ab63e92c854c8fd0d2261e6b865fd05ce2c61b9c6",
-    "source/vpp2/src/vpp2/adaptive_mask/monitor.py": "48b71f499dd5a0a79557eeb918f5fb1a641dc8654e5d412aa628ca9c57b5dcf8",
-    "source/vpp2/src/vpp2/adaptive_mask/physical.py": "f54e8b4fae8026cc89f296d74bf4cbea06fa4e126905b572a22c7d8226f0238c",
-    "source/vpp2/src/vpp2/adaptive_mask/preflight.py": "b646e7ca785afc87356e3e1184e48f535792ebafc84936450fc83bfd3f9f5ea8",
-    "source/vpp2/src/vpp2/adaptive_mask/priority.py": "68ccff1a384081f0c9a82e8f54ad2c720293f8500e7440a469b5e8fac1ee608a",
-    "source/vpp2/src/vpp2/adaptive_mask/quality.py": "c3b3a2264ca0b4773805f4485ebfe91bb4ab861e226ff4bd27bd667ab372b59b",
-    "source/vpp2/src/vpp2/adaptive_mask/quality_policy.py": "f100b7d4a412592777d936a6cf3c73a9f787cc4a62d5f1b8a369fed74acf897f",
-    "source/vpp2/src/vpp2/adaptive_mask/reliability.py": "306b41720154810bd972fe4d9c4c388e6967539d076ddbedd3a432ac62d13104",
-    "source/vpp2/src/vpp2/adaptive_mask/report.py": "f71401e09297daed8bbb72cb3af3b887e5fdf5ec282f47637fbb01392d280171",
-    "source/vpp2/src/vpp2/adaptive_mask/runtime.py": "ed74bfedd147389ea01f3be8c2979fa06c7152326ccc29ee347175690cb9c6cc",
-    "source/vpp2/src/vpp2/adaptive_mask/state.py": "5e8f95021eeb35511f5481e4e91ae687fe605f9e88d17b3b98d377cc967c17e1",
-    "source/vpp2/src/vpp2/adaptive_mask/train.py": "ad95349bfb242bcecd7ddf82fc58b31d63f2938a2101f9560bbc9e305f5f5b20",
-    "source/vpp2/src/vpp2/benchmark/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/benchmark/audit.py": "8587be371653c38b9926aa0ace006a244f270e0ef0020147a7bf5b0fc85d4538",
-    "source/vpp2/src/vpp2/benchmark/packed.py": "5047fd02f3fcc68e41f543fe83bc9202769a05794d898524d6d25e308f2b5125",
-    "source/vpp2/src/vpp2/benchmark/scheduler_io.py": "d6df06259549faed2da9f2888b540eac8e1ff2895c339eea58b4ea6a7e3b3cb0",
-    "source/vpp2/src/vpp2/checkpoint_compat.py": "462c5f93812483b5871633b200bc2ffecb9c201a2caf6b14f8f36dbe95fba50a",
-    "source/vpp2/src/vpp2/cli.py": "54c19a0f7e25b53bbf56ec13b0f537c9d7b81282e1e80575bddfb470172164e3",
-    "source/vpp2/src/vpp2/compact_kv/__init__.py": "f34f2f77f68399f45f6656b562a5c331f2504672dc145df368e086532191f766",
-    "source/vpp2/src/vpp2/compact_kv/blocks.py": "662c1f35055c2be2e2a14f2d92615d046f5e454879e34af39e3c7aec3af34098",
-    "source/vpp2/src/vpp2/compact_kv/checkpointing.py": "3c33c156e96f26ceb22eee61cfb8b18c3cb9560db0ab405cfd4c46efe5aba353",
-    "source/vpp2/src/vpp2/compact_kv/eval_prepare.py": "bdb33444056245e3cc773b0849b4db608b454cd24a7ac5b752f1ba26ff978865",
-    "source/vpp2/src/vpp2/compact_kv/export_check.py": "e6cba70b47027882a34d587aba0da5f9371128ec2db2282b91ad2d49e40065b9",
-    "source/vpp2/src/vpp2/compact_kv/gpu_check.py": "fedb6a9ab69b51080161ff2a4feae3f85262cc50f932f48e8decf04c1d147a20",
-    "source/vpp2/src/vpp2/compact_kv/losses.py": "914437ebf3ad91fd2cffa893a65ecb04ab01f2b337658a781a472feeb50c0a06",
-    "source/vpp2/src/vpp2/compact_kv/model.py": "332ebaebf543d40a34d188ebbe81055ec3e4bcff164bfe90d15aa6cef0748b8d",
-    "source/vpp2/src/vpp2/compact_kv/prepare.py": "7b01764cad0fafc0cc9441952b2e928048b0a43a6105dbd0eb64c349f4b9bf1f",
-    "source/vpp2/src/vpp2/compact_kv/provenance.py": "103e0f7100be56233aa86041ba9bc427ccb0db4fe045c4e6b8d3481ab6bb42da",
-    "source/vpp2/src/vpp2/compact_kv/reporting.py": "0b9ff602fc3425d1b9c122a77da3a426814e11c9cbdcb6b9a75a2a48130087a7",
-    "source/vpp2/src/vpp2/compact_kv/sampling.py": "2e8d82704f07853a7b487d5aaedef1b2f1fcfc5247e9100a56ed21d69262eae2",
-    "source/vpp2/src/vpp2/compact_kv/spec.py": "9bc920fc7d4398e424a0bde38ce97f87473aa417cf68b48e2c292a59491bc7d7",
-    "source/vpp2/src/vpp2/compact_kv/train.py": "a9e4410c5cde5df002d9808554538c29b77ed9e2d2bedc9ac5f8465eed70cc3b",
-    "source/vpp2/src/vpp2/compact_kv/validation.py": "73b62d02ce1d518cba7b9462e85c3c8d272ae9974e1161b1b6adea6f0040e2b5",
-    "source/vpp2/src/vpp2/convert_robodojo.py": "7ec2d6bfffbb4f5891ee33a4bf630cf27b17628506aa3cee1c46b3050027b290",
-    "source/vpp2/src/vpp2/critical_windows/__init__.py": "22721b77e25ec4874c4853d2f2991a14dff6aaa665da500f2830826b06545478",
-    "source/vpp2/src/vpp2/critical_windows/contracts.py": "ded4a239f7599785d74ff57ff892c5aa74f0280d6759460c93942e91f505d47a",
-    "source/vpp2/src/vpp2/critical_windows/coverage.py": "6c18bf995aacf413bbe4536eaeed92ab4b2f15f9eaaaa03a2730256ea7922ce4",
-    "source/vpp2/src/vpp2/critical_windows/inference.py": "df301bd187b27d20b6623febcb5e85815e1b7d4a1e9524a8a65bdf5904bf7232",
-    "source/vpp2/src/vpp2/critical_windows/metrics.py": "ae2bab145555d55696ef479758a5bae6780527fd3085a4f46d035b667ff32736",
-    "source/vpp2/src/vpp2/critical_windows/packets.py": "8f7d2c411407d95b1ff4d0e5b9055589f6114b1bb1020eeb7e1b106f330b853a",
-    "source/vpp2/src/vpp2/critical_windows/prepare_staged.py": "694cb5699de2d9f5ce1bc4667627b47d6bf0698715b4aad86907139859caffe5",
-    "source/vpp2/src/vpp2/critical_windows/reselect.py": "8d0e31ca1c17906d74de6439276e49ff7ea78abfc66fa556d8a306070b3dbf4c",
-    "source/vpp2/src/vpp2/critical_windows/score.py": "32daceb9a3fd957d9be75823b4d519629bc60a64d1da62f465e172cb1ec98d55",
-    "source/vpp2/src/vpp2/critical_windows/selection.py": "a7c35f6719d3308bfb738965d17544a15985c6d684bb47eac7272d6155d96786",
-    "source/vpp2/src/vpp2/critical_windows/staged_review.py": "d5cfb3f49ec5acc6914da98f5413d1549b2695e7c85c9aed9c0db83dee6be754",
-    "source/vpp2/src/vpp2/data.py": "8574e1ae3c6ca6e4b84633ee1815f2b3f2ea17c1e55f12fda0b00c5c5a6a7b53",
-    "source/vpp2/src/vpp2/datasets/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/dataset_utils.py": "e1a1471b0aa70917eaf699f85a9acf3ac70d9fb6eb9ebaf02d2bf2f218d6ab2f",
-    "source/vpp2/src/vpp2/datasets/lerobot/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/base_lerobot_dataset.py": "729fb3fd4ca531212eadb4a4e9c7e9f97ad34cca091d66380acd6b179f0c29c6",
-    "source/vpp2/src/vpp2/datasets/lerobot/constants.py": "11ac9260341e1560c5eb0dd7acb79ed63c74180b8ab4113c27622ff213db417f",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_dataset.py": "bfe75bb9593ec646236777d90d9fa309fa0e1188db07e07a3be0810edbe3da56",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_sampling.py": "266b0b7b023f014f4ccd935b6015b0a05b4659f372264fdefd997ea9692b244a",
-    "source/vpp2/src/vpp2/datasets/lerobot/episode_instruction.py": "74767839a813c7902861f4561101541c282b72e261df1fd28fe8d06c91231b69",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/compute_stats.py": "7d7daab219332b5abdfd5ec67ccaf3502e8eb583a17a4e482c3f3d4f7a691e89",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/utils.py": "4fdb4ac0653201e6743f2161c485213429948c1e1d5b068d03b4d97803254006",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/video_utils.py": "4f66c410ae6aa5a98acfc3320f3dd7b017d0285117054f14226f9e5d82395087",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/lerobot_dataset.py": "5659e992010bb6674ec50ab0d33b28a8e9dab0ecc895c199662af4c40fe56629",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/base_processor.py": "d3327cbe398d77761df960014d41e70d349bb0fa42bca874333b9ee970222447",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/vpp2_processor.py": "43c5604cbfc315b7377316cb97c9a6f3322124404397abfd01c6c3d0090cd414",
-    "source/vpp2/src/vpp2/datasets/lerobot/robot_video_dataset.py": "8eabfbd649116b0b4d6c9704c0df34febbf8875f6ab863dd75f4107dbbcdef73",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/action_state_merger.py": "39fd44f254bd7fb71b9269d75d8f9b5a5bbbe167ea438dda92d40eda1d256734",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/image.py": "f494d262ce29770e959bdff60b997f7a07e5f00d10ef515eee7be813d1db038f",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/rotation.py": "feb59d26539be6cfdb86488f0c10c11488ed73901658bc657e67a755ddb9e67e",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/normalizer.py": "8207d813b698a91dfdbe6910a4aa1fe0e775a2a8edf5c05b2d7a42591e9f8adb",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/rotation.py": "8ea2c72565707f5979967a1abeb5edf1eaa2c3fec55a7b50e640de869fc9fbcf",
-    "source/vpp2/src/vpp2/datasets/lerobot/video_event_dataset.py": "c20d3f5e2bc05e1b7ed94ae38872883a976a179e09b7fa051e94f94e2836c02d",
-    "source/vpp2/src/vpp2/enhance_prompt.py": "f31b3190c24bad073c5e3121ded4f564232729566e8c1fcc3e76fe75b28faf45",
-    "source/vpp2/src/vpp2/evaluation.py": "9b76882149e3c3a8c417a0e23514af1bb305c01bf0b6d21fff1f70bc9843f55a",
-    "source/vpp2/src/vpp2/flexible_mask/__init__.py": "64bed1808976d1f3c708154c36a97288a75abc6463d49d7f0065cb8360a2fb16",
-    "source/vpp2/src/vpp2/flexible_mask/calibration_batch.py": "dc5d83fc9e2bb595b9e3aeb1923115ebb9345283290a7c3f95c79fdeef7d3e76",
-    "source/vpp2/src/vpp2/flexible_mask/checkpoint.py": "c8531727e2388ec3d7a9770a504ceda119afcf03d64b2462befcd45a48949f9d",
-    "source/vpp2/src/vpp2/flexible_mask/coefficient.py": "07a3cc8d8e6c8b60119b9f94ae85074da017b08158fa0a0912cc9051b6df8ef7",
-    "source/vpp2/src/vpp2/flexible_mask/config.py": "4e1e9b289e293145b4e895c8ef7347bc837d01c929e05fd5cbf65bdb3c20be5b",
-    "source/vpp2/src/vpp2/flexible_mask/continuation.py": "2016463b4efe01b5d0a9ee84f03c3d95ee16f8e24f230faa465598cc978da29b",
-    "source/vpp2/src/vpp2/flexible_mask/controller.py": "263f1c9091e0c4318aeefb61e91549ba905ff046a625e35e1b18c94da4b4668d",
-    "source/vpp2/src/vpp2/flexible_mask/data.py": "d245bafd6f04f33ae7cef7c01d5db876a1a2a1a138c4de8a139d3ea78095521c",
-    "source/vpp2/src/vpp2/flexible_mask/diagnose.py": "faa29a0fa57c66af3b2e817640fdedfddc998bf878357f484415b244555ae4a5",
-    "source/vpp2/src/vpp2/flexible_mask/export.py": "3064dfd82bb55a145c16c77bbe17244970d07d18e944b793f4b7368dbcede764",
-    "source/vpp2/src/vpp2/flexible_mask/gradient_probe.py": "6fe3f13540cb9d8dd20c0b6429268ca74625dfad1e41806412c0f980fac76509",
-    "source/vpp2/src/vpp2/flexible_mask/losses.py": "df264b71365885ea90329c0b0facc92db3d2eb6e04324ea90a4701a1cd457abf",
-    "source/vpp2/src/vpp2/flexible_mask/metrics.py": "e976dd066e4b5e7e7ae79db77b9d1be64d17c212c61d0bc113b87a048e8f2b6f",
-    "source/vpp2/src/vpp2/flexible_mask/performance.py": "95b3977fd739cc98f7de94b14fb490a7c132021e742e7179dc5b5dddd6557c68",
-    "source/vpp2/src/vpp2/flexible_mask/preflight.py": "f794b997a492911c4d77656310c89d3b85e54983b3b889ba9904be64fb961e43",
-    "source/vpp2/src/vpp2/flexible_mask/prepare.py": "d6cf82e27555ecc13973727f4b71ce7e40558fef7d95805665fc7e458a42fad4",
-    "source/vpp2/src/vpp2/flexible_mask/pressure.py": "6cf74edcf3625fce9eec689d70841c6c1481f7c1abd87e5bf70bb38eb34632a2",
-    "source/vpp2/src/vpp2/flexible_mask/recover.py": "418b2d5ecdd08b08686b2a07bbeb0f969471eb73c04ac6952dcfbcda3b123f30",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_best.py": "917d924c5708584a0192050a6d8210809e905938aaf877dd81cff3be287ea289",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_checkpointing.py": "6f6a61c2ee7076c6b0977d1acb489fee7df04322d7f6b13bee9bb8229be319df",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_export.py": "c50433aa1b7e528fb23c514dddb5b9197f44c662829838ecda106980ab0dd54d",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_handoff.py": "aa590f707738e4b7c7250f901e29795c7441430f961bfa764437cf34bdc514b1",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_kv_decay.py": "bf83761798deea4f89c32d14f43bbfb0f16f751b3230962456c073b843dcd08a",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_subset.py": "7f7154cab6d910b747a0189742ee625983d08f13de50ca60bd4154f61df2b0d3",
-    "source/vpp2/src/vpp2/flexible_mask/report.py": "5b49283d472bace3a6ad56ce568af04f2c3d04add1a3920ad5006fe98374bce6",
-    "source/vpp2/src/vpp2/flexible_mask/rng.py": "01680b5b7e8a29ce92e96d756e12ceb80bb0e0ce1ebb642a14fa2a3a34776daf",
-    "source/vpp2/src/vpp2/flexible_mask/runtime.py": "373a3ecbc2ab6592a359daf6adfd8c958cb657f393e698ca2cb834f06750be4e",
-    "source/vpp2/src/vpp2/flexible_mask/select.py": "b547857bad61ae7ebd25b888fba73c2ca8f50a8eb31551ca54a909c75c8984be",
-    "source/vpp2/src/vpp2/flexible_mask/selection.py": "b9453c8735051b54ae7dc8610d2ca332eae8ca1a3e099aaadaf5ab75bfb0f8b0",
-    "source/vpp2/src/vpp2/flexible_mask/speed_review.py": "9d7099d1e173026274eb439169df2ad4812ec1d56a5f0b21242f972cd835ffd2",
-    "source/vpp2/src/vpp2/flexible_mask/throughput.py": "ce21bdfda51ae10d1470ac537d910d9f46f60a38910888808ffb3f5a80174d58",
-    "source/vpp2/src/vpp2/flexible_mask/train.py": "86625613bbc692722c6d2d24278a586f88ce0d76263a9daf92562a9002a8e037",
-    "source/vpp2/src/vpp2/flexible_mask/verify_export.py": "fb68e62b0da98e6f96cfa694bd061162bbeec1763e5787e057a1cbac5b2ce8f6",
-    "source/vpp2/src/vpp2/full_mask/__init__.py": "27c9ede277380e8d64586371f2a03334e68ab2d3bfe516d95e8cb96099827929",
-    "source/vpp2/src/vpp2/full_mask/calibration.py": "5d97e0385eba82252e12cc1562d1a77da2f2bce29332d758c1b5b6b84c5bb47a",
-    "source/vpp2/src/vpp2/full_mask/conditioning_cache.py": "07adf95165db4c3c1c469695d6b7b6fe4cf4fa9aff0a3d360f4229bffe9be17e",
-    "source/vpp2/src/vpp2/full_mask/controller.py": "0e6a969b89a4275352cc5a1e00bdfe8e3f8754a6ebd7f9266b92179cbe6bdd10",
-    "source/vpp2/src/vpp2/full_mask/cpu_contracts.py": "1a46a53a68d2d8a202c5f98d9a78f6d5f5041dada0c2239ed81e7916a28f0758",
-    "source/vpp2/src/vpp2/full_mask/curation.py": "4ffd7b0212bdbca1daf73ba7c24423523437012150fc044a43bbb33cce747545",
-    "source/vpp2/src/vpp2/full_mask/curation_v3.py": "9dba386d66869de0fba321ade79b2692d47f111ccfebc7f367194cef7a4b4e8c",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_checks.py": "688ea4e7e44c4ecc1a3cd81cf682e2cbf611d9ad2d52744ba1308324da70febc",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_features.py": "11011a089f1586045e2972cf267773aaf4602cf08c8b0e755bbf017ede1e6e43",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_finalize.py": "a2a7812eb6c670686a7a4f4adff0a2c6363dbc078c71d948537f1c5fd546254f",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_probe.py": "2755e9cabd0fe02e70cb8330c083af881a5e4696697d58cf7aa01bd00ef57f6d",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_quality.py": "b9170398762112a23bb133dde53c6a49193c671c767647239c8b0ff15159530c",
-    "source/vpp2/src/vpp2/full_mask/data.py": "668a7c5f88a3ab17854eb0df240c974082f4aa64604cf11d3143b18337f73181",
-    "source/vpp2/src/vpp2/full_mask/distributed_smoke.py": "f699d4f330309f29a537fbabb61c60f61eb3322d7304605351dba3d2acb48e8e",
-    "source/vpp2/src/vpp2/full_mask/gpu_prepare.py": "92ce96dc4c1b6d6ebc22d99935472c15d86ba2a7c050e31640d2c32e08748ec5",
-    "source/vpp2/src/vpp2/full_mask/monitor.py": "2724a934d1f96444227c854f75ef84df6025c02485aa68eca34f3b28d509fe19",
-    "source/vpp2/src/vpp2/full_mask/native_monitor.py": "5572712a340f17692a669d2a4c41d1d691164925eb7bd06a56fff1dc2afd69fa",
-    "source/vpp2/src/vpp2/full_mask/panel.py": "2844b663340cbfb54f7622f3447a7831e2809050739a52dc076fd6ae32ed9e46",
-    "source/vpp2/src/vpp2/full_mask/review_decisions.py": "ed1184bcda7b5bb3fb11bb4262c4461c408e3d3ac0d23e80d2f8cafcf39c7e18",
-    "source/vpp2/src/vpp2/full_mask/review_report.py": "752ba9e8a3caf5b8cefa2d18e6fe14860e6f998f57c03af0940af0943b09a9b1",
-    "source/vpp2/src/vpp2/full_mask/runtime.py": "ae30b4d3fd3c888bef17672376dc3b34491cb296a1883959674ad465579efe7b",
-    "source/vpp2/src/vpp2/full_mask/search_quality.py": "ee3238d3366799e97d3f9a90c11f579ff8f51d5bd0642f52c37d9dcdbcf02e6e",
-    "source/vpp2/src/vpp2/full_mask/train.py": "f4127e6580522d4d9e5a50db10d201526e948d4d3479ddfbbc12e60ec61c7501",
-    "source/vpp2/src/vpp2/full_mask/verify_data.py": "72ee60ba2016d5f73eb2ccb01b666a5dac37138620efc4a79a20607df3658fb3",
-    "source/vpp2/src/vpp2/inference.py": "42b3057a85aaacbb87440bc740f13b09b9c82cd51d8c5af4e080716aabdcf38e",
-    "source/vpp2/src/vpp2/learned_width_kv/__init__.py": "0c59cb8424b6cd64508a005625e9edf63c84d24618a5d2caf6e6fa5d63cc859e",
-    "source/vpp2/src/vpp2/learned_width_kv/calibrate.py": "b89bbf103b49fd85f85f8c300a7868bb42ae7a13f29bd9878dbd5a11061ae51d",
-    "source/vpp2/src/vpp2/learned_width_kv/checkpointing.py": "7d30e46b95d91df8aff088346ae75809958622b3a8e8d953db4f30adfbaebadd",
-    "source/vpp2/src/vpp2/learned_width_kv/common.py": "52dff98c6317b34df2a13313a641998acf4f9e18ac31ea4e8e25227afa14620d",
-    "source/vpp2/src/vpp2/learned_width_kv/controller.py": "7b5a8793a540e1e7464185229189f4f1f1acadcd7cf3c2bf477bda5de7d18787",
-    "source/vpp2/src/vpp2/learned_width_kv/data_scope.py": "0ddde6805f646afbf5f5966c53530f829becd083802f0e0a8697dc1004e2003e",
-    "source/vpp2/src/vpp2/learned_width_kv/distributed_check.py": "f89f153a20730a6bc3c70133479f34918d30c5a48775bb3f6c14bb25701801ce",
-    "source/vpp2/src/vpp2/learned_width_kv/gradient_probe.py": "3bed964f6488e95b386cc58058e7f28aff50c9a98dd526f3365891a9049a6fab",
-    "source/vpp2/src/vpp2/learned_width_kv/lifecycle.py": "798afa47ec5dcc8749982da2d7853eff184e450a32d51bcf70eb4578f2bb0b09",
-    "source/vpp2/src/vpp2/learned_width_kv/masking.py": "3272ea09271130dcb3cab6935f6bb7d61bda66b9982c3baabcd61ee0efd914a0",
-    "source/vpp2/src/vpp2/learned_width_kv/performance.py": "a2fd9597c51bf0cf5c435b48e8f57792df24d2723a6d7b6b841d6fa24f17ab2f",
-    "source/vpp2/src/vpp2/learned_width_kv/reliability.py": "f74399fb67427093b3e4c7df9749379f9291b7ac58bd6632db226b6e78fdbfd0",
-    "source/vpp2/src/vpp2/learned_width_kv/reporting.py": "f22c1d8806a5b7757bb6dd99b15153d0305b2eeb0790bb37763337690d642036",
-    "source/vpp2/src/vpp2/learned_width_kv/runtime.py": "513da1e06ba9509164f017e0b8dbf6b219d7e1cff82edb2d2fd8d2d7839e666f",
-    "source/vpp2/src/vpp2/learned_width_kv/train.py": "5813ebd4fefec6d3a8cead9e0c645d10734f5ec88c4d74a5742587f2eb4de6c3",
-    "source/vpp2/src/vpp2/learned_width_kv/validation.py": "d5594ce875c8011bff12d3c77083dca68878178ed022b8079761267d04a82386",
-    "source/vpp2/src/vpp2/learned_width_kv/verification.py": "1bc43d48ef0ec3056af614adf43ca0744d94b1e77fe970168fef87c7cf2e6fe8",
-    "source/vpp2/src/vpp2/libero/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/libero/action_model.py": "178fbe4cf02b9a2e133ef31389b6bb2f9baf0504bd0818a265b6bff32ebc3914",
-    "source/vpp2/src/vpp2/libero/cli.py": "fcf68a88e2db0a9b786c6199d9b719802fe4b84514741658b49e2b06e2473e45",
-    "source/vpp2/src/vpp2/libero/data.py": "19b699381e7d6fa769b5090c7125ae01b7584576e969468660ee65b1585ef5f1",
-    "source/vpp2/src/vpp2/libero/diagnostics.py": "9540e31dfd2def06bec942ffe73262278700a17a340e14cb5bd41a295706e147",
-    "source/vpp2/src/vpp2/libero/eval_helpers.py": "ac338d4ade195ceb3a87620d162befa9679e2d7ed3d91b09927e5d59d86d97f4",
-    "source/vpp2/src/vpp2/libero/eval_libero_single.py": "87e6b457dc412ef4fe5d01aa48b1a6d20a28dbce3152848042e2103769e4efef",
-    "source/vpp2/src/vpp2/libero/eval_libero_task_manifest.py": "f0c721a7ddc768edddc8b0e065a4d8aeec35e1a546c88deebd9097ad93f6227b",
-    "source/vpp2/src/vpp2/libero/evaluate.py": "24216f9814f8304e077d72ff8d8c778fd588239313b34cd7d23b960c8f924c67",
-    "source/vpp2/src/vpp2/libero/libero_utils.py": "c7cc23976c86eabcada4846d46345d3605fc2ce03929761646486be3901697d4",
-    "source/vpp2/src/vpp2/libero/prepare_pro.py": "0999050352d924c11d29b22f63e902162d8cfb23f94815c93c01295c11ed81ae",
-    "source/vpp2/src/vpp2/libero/pro_alias.py": "90b5c8c8061bc727a6e413a8746fcc98cbd06e1c7c4175fd223b45af4f10b700",
-    "source/vpp2/src/vpp2/libero/pro_runtime.py": "adf94ed64b032aaeb0d88391e7b653f88f67e76b9411419ece25faa31ccc1340",
-    "source/vpp2/src/vpp2/libero/protocols.py": "68225c95eeeee25032966a7f90c1ba18067c53225397f59a463d764030fd0d93",
-    "source/vpp2/src/vpp2/libero/runtime.py": "1e2e2900a604d33cbdf8774cfa33dcee1065f21dc0b59fcf30799b5098c1662a",
-    "source/vpp2/src/vpp2/libero/trainer.py": "ea29b4e000047a6b0c8aee3cb6f32df9187987a657924085dbafd9628315107b",
-    "source/vpp2/src/vpp2/libero/worker.py": "9df6350c4ab92dea70863c08f1bdcebd0d77574a3a2d8d7ac9899a5a825be6b0",
-    "source/vpp2/src/vpp2/models/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/action_dit.py": "9a654edd860ce77c1308923bd338c7f14d09f68686642531d7143c16b9f6f331",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/compact_channels.py": "6f7ad787624844970985117002ab809bd777f7474969daca121e7101381c6d0c",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/gradient.py": "594249b6eb931048e352fd3ffef9b1ac1db9e4fc74659e04216d6b823dd174f3",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/io.py": "1b68cf2f2fd674c7a59140f8632d4464044bffc2e9277ef5f9a975e7c699966a",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/loader.py": "b9180b9c9177af192e5cb92387b4a09bd0dbc08fb75a94e843b67d315c94e339",
-    "source/vpp2/src/vpp2/models/wan21_14b/mot.py": "cc14a909bfb181a441bc9d8d6de12f3c31d89f3dc3415031e34a5f8c21410019",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/scheduler_continuous.py": "15de04574c1d8ccb794ca6560b1e501e5c83cc86104bb80f1eb3cc3b0324ffe0",
-    "source/vpp2/src/vpp2/models/wan21_14b/vpp2.py": "8c419af905a8ffa93d77c827a283b7aeff744f3a461c390783eef8a333c31756",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan21.py": "01b45cc71f69a6914f7b286559317c8e8485d0b0fec93bb28ebc37c183c295f4",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_clip.py": "f65e44c9d1b7811d2996b3fa22e6fe4c2ac5e6edf1fc2c14251cf582bf8d1f05",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_dit.py": "0d1b5837ea923508c05394d56aaa465813e7b4f5aea3b31ceceae39b2de9b0f9",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_text_encoder.py": "902a1e46b21b5c48e166fe03defabb94b3773c6bbe836fc09b9a0d0894eb1f6e",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_vae.py": "33f7d804677b44a6f3ab334684017dcdb1e9dfb748279de76ad94a5bb71a7e51",
-    "source/vpp2/src/vpp2/policy.py": "2a7696cd73ccb99575bf901185cf578c3775945f61f6e67aeb2b4ec6bfc4e30b",
-    "source/vpp2/src/vpp2/preflight.py": "7a91e486675a4cdf38565e0f4b58e60a68bba7de5654cfff736b5df58f5c53a6",
-    "source/vpp2/src/vpp2/robodojo.py": "183db869a622ef17ee92c487949db2e22a7d1365558fe703bfb8dd8fef83e9c6",
-    "source/vpp2/src/vpp2/runtime.py": "e26885a3732f5fdac3d8381da870a1673d0c03354f02f9deee7fe779d645f8a7",
-    "source/vpp2/src/vpp2/trainer.py": "feaa0ac4ed94a376836eb60cfe6f52460392b11f7cafafec9c6d9f151f5b61f1",
-    "source/vpp2/src/vpp2/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/utils/conditioning.py": "43d43f0f3ca0add64498c3d64a3791513a53e3fce0ea53f77ef09580a86b7b2d",
-    "source/vpp2/src/vpp2/utils/fs.py": "69e6df236b43360428a613686c8f37c531c0d16a3a2c5b06a725c46b49d657c5",
-    "source/vpp2/src/vpp2/utils/logging_config.py": "1de4abf66d9d201b49e7bff48291643b9dc5ec9b6e7918ca73e1bd18887b1c43",
-    "source/vpp2/src/vpp2/utils/lr_continuation.py": "72ea9f2af52041db19aab681d7a99fdf6d87d5ed5486f2499383a211c6c1ab76",
-    "source/vpp2/src/vpp2/utils/misc.py": "7e42c330c24604e118f1878473dc804aa8a272944df502ef1414c05be4d5bba2",
-    "source/vpp2/src/vpp2/utils/pytorch_utils.py": "9b0ba19de9a64458b12a0fb24f0811f49d9e9fca5d8b815f5f5ecd3b7886fd63",
-    "source/vpp2/src/vpp2/utils/samplers.py": "4d3bd9b2d498f4ad8f2cf9765d60c4bf5368fa6fa2a7d42d0db6210f66899652",
-    "source/vpp2/src/vpp2/utils/training_profiler.py": "91f6d98577e63d85e474c738a9bc1901edc194bdf318b5387d527c31b6708832",
-    "source/vpp2/src/vpp2/utils/video_caption.py": "d9572deb3690aecb0f16c052016cedc5f94da8ce8f016b4dff1a4e358e18d5e5",
-    "source/vpp2/src/vpp2/utils/video_io.py": "9e9898b021a44e87aa1b4e6d13e3755a9869a2e27a7eb53c72a6802808d5bd80",
-    "source/vpp2/src/vpp2/utils/video_metrics.py": "1840a37685d686336c7326a2a3d4fbe048d1e8e7e431fb14519d360a8b7aaae8",
-    "source/vpp2/src/vpp2/video_training.py": "20f55bf8b0a7dba6422a1115be1c8adf31a0db6a9f53a8f6f53f5d970188bfb2",
-    "source/vpp2/src/vpp2/weight_resize.py": "269edd4b9d32a1c7f363ee082dde7ce57f2176190daab6d5bf4a1c98ab43f0b4",
-    "source/vpp2/src/vpp2/weights.py": "5dbdff347de1b57e30433efca6da28334d0263d27aaa9b2278d95614db1b8641",
-    "source/vpp2/src/vpp2/width_kv/__init__.py": "0183a7106da7e45223550a552db2905d6b59c2a4259f8c86cb98629b03e66b43",
-    "source/vpp2/src/vpp2/width_kv/calibrate.py": "93a06910cb23a9e1dbd961ee9ef9794d70aaaa6186b08483260c8215109598f5",
-    "source/vpp2/src/vpp2/width_kv/common.py": "45f5f7087f5f3dc067358838e3fc738698b796defecb2ea721545d275f767766",
-    "source/vpp2/src/vpp2/width_kv/diagnostics.py": "cf184ae58f5e1c5f4453774eee1e10c85931e618bbbf768ab42f649cf301f5f1",
-    "source/vpp2/src/vpp2/width_kv/reproducibility.py": "89c5377e1314dd6e8de7f67c4385d30f5bd91fa4a5742449f1c1ff8dd578f451",
-    "source/vpp2/src/vpp2/width_kv/scoring.py": "7644358d0b2d29d3b91515fcea56090dbd2a7a0dad2d5e5c88e8d585bff15533",
-    "source/vpp2/src/vpp2/width_kv/slicing.py": "856203c14e0b06d19585a96392d00a00703779ebdb9dbdc9ab1323d50daf8ba4"
-  },
-  "suite": "generalization",
-  "update": 1000,
-  "validation_gt_action_mse": 0.0005726480308242441,
-  "video_parameters": 12468409664
-}
diff --git a/evaluation/long-horizon-protocol.json b/evaluation/long-horizon-protocol.json
deleted file mode 100644
index 0a95893421b8b417c5fd882be2b21ba4ee62a887..0000000000000000000000000000000000000000
--- a/evaluation/long-horizon-protocol.json
+++ /dev/null
@@ -1,706 +0,0 @@
-{
-  "active_policies": [
-    "vpp2"
-  ],
-  "caveat": "Same checkpoint55 policy/simulator code, benchmark layouts, seeds, r16 and new90/old10. Simulator physics are not claimed bitwise deterministic. Offline ranking used replan24 without ensemble.",
-  "checkpoint_files": {
-    "COMPLETE.json": {
-      "mtime_ns": 1793716093276368990,
-      "sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-      "size": 2491
-    },
-    "action.pt": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3",
-      "size": 4057916577
-    },
-    "compact_architecture.json": {
-      "mtime_ns": 1793716074811787315,
-      "sha256": "a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32",
-      "size": 6563270
-    },
-    "dataset_stats.json": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
-      "size": 100256
-    },
-    "validation.json": {
-      "mtime_ns": 1793716074812787292,
-      "sha256": "9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e",
-      "size": 88781
-    },
-    "video.pt": {
-      "mtime_ns": 1793716059522133613,
-      "sha256": "fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb",
-      "size": 24937214042
-    }
-  },
-  "checkpoint_receipt_sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-  "execution_order": "one report task at a time; eight shared GPU workers",
-  "fresh_policy_per_episode": true,
-  "fresh_simulator_per_episode": true,
-  "gpu_sharing": {
-    "minimum_free_gpu_mib": 61440,
-    "mode": "shared-shard",
-    "one_worker_per_assigned_gpu": true
-  },
-  "id": "c2867b1bb38b88d54cbe92103a5cf6d2020c3d44f67c89412182265c4bac8256",
-  "inherited_results": false,
-  "input_hashes": {
-    "inputs/episodes.json": "26a9c66ed3082a59ccbb11c1c43d5bc2ad0123ab66a3ffd27750e834c9de2d20",
-    "inputs/master-episodes.json": "26a9c66ed3082a59ccbb11c1c43d5bc2ad0123ab66a3ffd27750e834c9de2d20",
-    "inputs/vpp2-policy.json": "3acd3dd24eb03f6d0e51670aee1f6d7bd1c58d845432cb8435020582a778123a"
-  },
-  "native_inference": {
-    "horizon": 32,
-    "replan": 16,
-    "seed": 1,
-    "shift": 1,
-    "steps": 10,
-    "temporal_ensemble": {
-      "enabled": true,
-      "new_weight": 0.9,
-      "old_weight": 0.1
-    }
-  },
-  "native_scoring": true,
-  "order": [
-    "vpp2"
-  ],
-  "parent_evaluation": "/home/praveen/experiments/vpp2-recovery1000-950-precision-20261103/update_001000",
-  "parent_protocol": "266df4b5d9b8eac16145600a8c4f9223d15e03285d61473c3bbd8c43b694f211",
-  "primary_episodes": 400,
-  "repetitions": 1,
-  "schema": "checkpoint1000-followup-v1",
-  "scope": true,
-  "selection_manifest": "/data/praveen/vpp2-recovery1000-followup-20261104/inputs/long_horizon.json",
-  "selection_metric": "task_macro_ground_truth_native_action_mse",
-  "source_changes": [
-    "native_eval/scope.py"
-  ],
-  "source_hashes": {
-    "control/simulator_recovery.py": "fcee066ec9e75ec081b0a955b136e2ab898cd1ee529a2c1e74a453c350c05367",
-    "native_eval/__init__.py": "2deca42c85edcf3d80dba72439ad95641937396e27e1149d77182cfc082267cd",
-    "native_eval/body_coverage.py": "50ce610f86584965281893ff2f98a78643c7e3207014212aad4c13fce4c8a782",
-    "native_eval/camera_group_cleanup.py": "8df69fdba913ec41ff7b1256ab1b43921a0b6fbef7ffbb6dacbf2adaba0f7843",
-    "native_eval/camera_lifecycle.py": "b0952d95d1da8a1bd3e9f544a2eb1eb0f516159b3442666c7acd9bbaee71687a",
-    "native_eval/check.py": "9587d7b200969ccb78ddf750dd3d8616d5e2245e29baa3a0d798e27a55bdaa5e",
-    "native_eval/compact_worker.py": "129a2e26a85e021f7a0282201dec976104a8e1ea1f0a67e6b75ce760257ee46e",
-    "native_eval/completion_integrity.py": "3b737f6dcf9ae9908b1a3602f66fda46c3b8399ee47b1c61f5c0fb9741031029",
-    "native_eval/criteria.py": "97ebc94d91c62d793e614f178d2d120df989faafb54881dd1d3f5d703df8bb76",
-    "native_eval/diagnostics.py": "53b0deb571984fd02ff65c400d5f3b9bc2754b8207ea8edf4e5017c3588ed3cd",
-    "native_eval/ensemble_check.py": "0bef52411b2ec2dd7e1ad21b8690b1f88ed229107b6e46b8dc43c34ca92a3572",
-    "native_eval/evidence.py": "cf7f769cfd838227b26101d7656ebcaede1b1d45a80785a32efb26c9befef0e9",
-    "native_eval/frame_epoch.py": "7b1d3b10f46d5acf82f6e983ecd5f8995310b2fef9fee04bfae168fda02e3efc",
-    "native_eval/hardware.py": "e30ed3359d91595d3053e24124445498a3f1be9d5c0d52f97e1e7d7d88268a5f",
-    "native_eval/hydra_frame.py": "c1aefee86396025b405e050597af7b67bb07c0dd1e65ea4a9be1ed78f8d99072",
-    "native_eval/particle_evidence.py": "378d667b2ab80176045e0bdcdbe3cb39adb612a356b8f6b81d13ae531bde711f",
-    "native_eval/policy.py": "e1cebbfb03c137bd4bf9f054c69be74f9a7f1df5f7ca13c250e606c2346b55b0",
-    "native_eval/prepare.py": "486df6d062098ea07ac63c4d3be51b5f27baf524d0d004411275243783d02ac3",
-    "native_eval/process_lifetime.py": "7dd20bee91c51546b6da067b96012dc5d5a013e8389dafd567b72c1f742cf97d",
-    "native_eval/provenance.py": "8fc2da1678832be9e9b8b778ff0dd1604db5d464cf1a19ff6137acf6b83d3e11",
-    "native_eval/receipts.py": "9ba5800e6e7e796937992117569efe1b3d625900a8e2ebf7955940a4e8cc81e0",
-    "native_eval/replay_receipts.py": "1c60f9bddccf7245e32df97b9e90749bc3ddbe5d65f37ed2d40ba339b458017e",
-    "native_eval/replay_server.py": "1c2c19bae8d1cd3ff564f1775540718c33293663d30161ea7bd293b8f322cab3",
-    "native_eval/report.py": "9d9f0ddc794d2af2ad6e349f991450592a6fe98db292e1ac3ae4fcadea961fc4",
-    "native_eval/repro.py": "f0e4ecd0ba009774105cebeba31f49a70f34c4f88f11855c939368027789a928",
-    "native_eval/reset_coverage.py": "3c6a414878c4cc2ef4114170356db0a6b2193b03ce5c3b4f6293c10c7d3138b0",
-    "native_eval/run.py": "23d2d58fe1160a03bdcf9591b5d5fae671d21832008cc4958ac58e060b88caa1",
-    "native_eval/run_one.py": "dddeea1abb269293bd08c3f5a69ea0a668620b18d07154ca1c0740a0bb213b99",
-    "native_eval/scope.py": "d0ddf1b0896f798c6fd2c9c482874b92cfb57585579e648b619ecb0e5a00dcc9",
-    "native_eval/simulator.py": "d40972be52b7544f62653114217e213d19d4b386bf2991abfbbbc618ab09cf83",
-    "native_eval/status.py": "b49d03681dbc28b80f39f4d6163c3a6e7c5a1ca62f1f0ef502c77dfbf149ca02",
-    "native_eval/topology.py": "062e9246890ad1f2e8bb8b64b3bbe6598f9dce6fcc33ec3eaa106f3cbc0f0ae5",
-    "native_eval/transport.py": "eb6cc295d064a142372dd51859e91233f61f1ad2c8ac8bf3880739f3fe4abec4",
-    "native_eval/vpp2_pilot.py": "571ca3eeba11fbf2afd4a4058629a280bc0ad4bd3eecca178ddde4aa543bd8de",
-    "source/simulator/LICENSE": "7794bb06af8fe5485ca912454ad1665ccd7846c45f9c2b1258658e480948cdbe",
-    "source/simulator/XPolicyLab/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
-    "source/simulator/XPolicyLab/XPolicyLab.py": "a3909558697b73479f8baee03e15b4d74060a7c9ac2f31fcbc6cd659e7364862",
-    "source/simulator/XPolicyLab/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/client_server/__init__.py": "093e47e707f4a5598b0159b5ee371cd9a918b802c70a6e66154a18e9ef8fc5db",
-    "source/simulator/XPolicyLab/client_server/ws/__init__.py": "79905987a12ebf318dc0b2b9ec6f375d9b4561bac29d6c27d04e318d8f6e00b3",
-    "source/simulator/XPolicyLab/client_server/ws/__main__.py": "ee1c5807dcdf1c8f6721c5447463eb51babdb0f9ffbd91d1e7343834ca4aa19a",
-    "source/simulator/XPolicyLab/client_server/ws/model_client.py": "4d6e225ae5f577d862366350dca8742e91dd9527d6e76b8b2a6d48c991212d67",
-    "source/simulator/XPolicyLab/client_server/ws/model_server.py": "4988823b436cd05e6a4d8b13012cc30c7ffd57711b72497b1223a6b10dc0c043",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/__init__.py": "19aeae375e94e58d437d789d0aff07b232c13f375235384d784984631d25d5d9",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/client.py": "b309f58b1bddd99e56578026b61a15e6cfdc07e843e8a082bfd291d0e853ce5a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/codec.py": "8a97ef7e102fd954c266352d81eaceef0f1b0f5504571c03037683efe387865a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/exceptions.py": "73161f4cdcc7ffc8a87cec94f5459bfcb9113076be272d6cb2a3ff8d69e0c6ce",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/keepalive.py": "7f6b9e636eefcd6c80a4b62e404a6f89e2f7e1a3b6008500312a972d6990a21d",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/messages.py": "d05c8cbc20e0cbe1502551864907f25897035993f920748c9290c944c959eab7",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/schemas.py": "f1ca82eafa8dcad6d968508adb51c25342b291f7b1c553fff305c381a6098bd0",
-    "source/simulator/XPolicyLab/model_template.py": "4f8d1eeb8073aa7ae6093abf0f395cd5fef70ff4a4bd1fa6b2d83b356a90754a",
-    "source/simulator/XPolicyLab/policy/VPP2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.py": "ab94a8569e6168ab7cde6660890035e11b34961ff2648854a7b61d188c555b6e",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.yml": "d2c3f06eb2ffa2d1866767dc20db8d60e542fc2418dbc4cf86d4a0f86f944c5e",
-    "source/simulator/XPolicyLab/policy/VPP2/model.py": "699acc0c459c5997db4e3155e2b1b7eb64bf750552436074a369c4eb6e9ffef9",
-    "source/simulator/XPolicyLab/policy/VPP2/temporal_ensemble.py": "42be4b09c56377f4686b2e722de760fa3bafd1c6b92605ada269d94e19efc95e",
-    "source/simulator/XPolicyLab/policy/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.py": "42dc75140935046dfe83f19b3834bd7cf8372e1bdbe51f598b1f91b8d557b7fb",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.yml": "e59de6edeee891d2aa24902ae9034e9c10889fb20116bc9263f30d773a4b0d10",
-    "source/simulator/XPolicyLab/utils/get_free_port.sh": "f45bcb327b45c91367e3e5cdff72d3a42bfb7a09f50e4774701637c4c23fc1a4",
-    "source/simulator/XPolicyLab/utils/load_file.py": "488590d76084f419a36c2fb3e658eeb240ae554678899e9f2ce0cc10da5c4b57",
-    "source/simulator/XPolicyLab/utils/process_data.py": "e044e627eb2677907e9227eda8458c857208913bd484c5e922b53025010eef0b",
-    "source/simulator/XPolicyLab/utils/robot/_robot_info.json": "8a5df36e750971003b484ad770a7a243e523e986d2977a4ecdbeecd4a07946bf",
-    "source/simulator/XPolicyLab/utils/wait_for_policy_server.sh": "f75a8360d83f2960f365dae77804c94a28aca332c2fbc62cdb4a39ae220374ad",
-    "source/simulator/env/camera_manager/camera_manager.py": "1a4250cb3bdc7305fe362a7c6b113f57e6f3e9e4983b69bd61e2c84e8e995471",
-    "source/simulator/env/camera_manager/capture/camera_view.py": "be9ec0d4df8a532c35adc7090fbd7d6a76c163f90e8dac30496ee4f875929869",
-    "source/simulator/env/camera_manager/capture/render_sync.py": "48d70e9a95f3445ae464ab7ea00a00c0d4829752eb55c4ce52ef73ab1f453cae",
-    "source/simulator/env/camera_manager/capture/tiled_capture_manager.py": "959b4d87d96e676bc279d01e7755557bb8c6e73920a75944259f72b7ac385f8b",
-    "source/simulator/env/description_manager/desc_manager.py": "bb566028b523a4cc27363cee2e526c2d3e008e671c9773d4f0b8401f58f0388e",
-    "source/simulator/env/environment/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/env/environment/base_env.py": "ade58ebc19c79bdbf720def30786a0c397b1c7d2141fb6a1bfd63c80f8a3db8e",
-    "source/simulator/env/environment/isaac/__init__.py": "123023ddc75a0f9b1139c7322bee64a27f0e756b6e84242f43a1481428596126",
-    "source/simulator/env/environment/isaac/direct_rl_env.py": "77958f14c465932697efed302c8ef13054f2ec61ee3a4bc5ad66d0390b71df77",
-    "source/simulator/env/environment/isaac/isaac_rl_env.py": "144af40adc2113751bc1bd72caadab145e918b316de53ef3e4e42358c01a1fbb",
-    "source/simulator/env/environment/task_env.py": "ac816cf26efabbb8ac4276d2a2a09fbd557e3db805514db0c3dd7ac22b58e931",
-    "source/simulator/env/global_configs.py": "e96f6141b42e635d69bba5b4727fb1fec54f8965431b3a70d95fe5dcd3d0ee45",
-    "source/simulator/env/observation_manager/obs_manager.py": "771d6cda443b591f3e7850b901a06b4e653a2feefcbabd23608042377fe9be4b",
-    "source/simulator/env/planner_manager/.gitignore": "bdcb46018aafd15f5c2d40adbbccbb06f2dec5b9ec4311cbf776075bdffd4156",
-    "source/simulator/env/planner_manager/curobo_planner.py": "3cd830c002901959409ef6f64d54106d8106b48471b266409abc2199f3cb33e1",
-    "source/simulator/env/reward_manager/func_parser.py": "eb928a05c3eb6bae429a922c59e64272c2c3851c22d779143d45627e47b3a504",
-    "source/simulator/env/reward_manager/reward_manager.py": "bc56efd3aaa861b4b38960dc4fb7778a9b4da9cf131f815ac3170300dd233a41",
-    "source/simulator/env/robot_manager/.gitignore": "fd94daa58076711ae84376170199a63f66177471e587a76252399fe79eb231e2",
-    "source/simulator/env/robot_manager/control_manager.py": "81b7ecc3bed8248c16aa1517e686682d8e28ccfa9e550d45011b579a9219caab",
-    "source/simulator/env/robot_manager/frozen_robot_configs/franka.yml": "a3c944f8cdd07fc1a4eb54cb915f565763292acb24f9c4f725534271449f138c",
-    "source/simulator/env/robot_manager/frozen_robot_configs/x5.yml": "4df90d234656c34132ca573e69a55de5984270e103aac2ea8ab9956a913685b5",
-    "source/simulator/env/robot_manager/robot_class/franka.py": "dc49eb7738a8b58fb731595340c60a1ce32f52ef0dc63a3da84f9c25879af140",
-    "source/simulator/env/robot_manager/robot_class/x5.py": "4748cbe4d273449a48d09817767dec361f53f8e5cdc6e43368fbb8724141b654",
-    "source/simulator/env/robot_manager/robot_config/franka.py": "8bf57a07913573e74dad97aec6c50e58b67746fa90875915d74ce213856ce35a",
-    "source/simulator/env/robot_manager/robot_config/x5.py": "13d35630f833d1b0a1e3df0bea05f843e88aae1248b59af8aadfefa9cad29bbb",
-    "source/simulator/env/robot_manager/robot_manager.py": "19636ff932a0e2192341b57d8695d4102b870bd10c69c9343d712682fcc15082",
-    "source/simulator/env/scene_manager/layout_manager.py": "1b4c940200dd94a1d0300e103d168b9a823e0dc00ce7c921afe16e81d4e01ddc",
-    "source/simulator/env/scene_manager/objects/articulation.py": "0c542a1fc8ba7a61bf71b343745c71949c863291c1b167d4d7048fe2ef9ca4c4",
-    "source/simulator/env/scene_manager/objects/background.py": "04afea9b5c4e21a105ba3b08ce0e5b6692c082cf17ec1053fdffd5dba1c9511b",
-    "source/simulator/env/scene_manager/objects/dynamic.py": "7a4652c649f503681914e37c50c959c7ca8457918bb00628413f4288993f8ea7",
-    "source/simulator/env/scene_manager/objects/fluid.py": "12b183d4d798d6b7188fd42747ee1927747318c386a91135551d1fa0561c41b5",
-    "source/simulator/env/scene_manager/objects/garment.py": "85f38bd85b52da8dd2ce34ce3692d45848016a9f0f14ecf13114705d4a1ff46a",
-    "source/simulator/env/scene_manager/objects/geometry.py": "3100d61b152b894b3b59be0b3be1d2708f2d8f363ddb8311dabfb5411c731f3d",
-    "source/simulator/env/scene_manager/objects/ground.py": "12dd51addae6f5b7c6ad410b4ae7248231d216cf63513c564057d577a09a2972",
-    "source/simulator/env/scene_manager/objects/light.py": "ba8c8529c6608592d4d0eb9638820b70ae6b3c0cd616631bc6dabdb1c20e421a",
-    "source/simulator/env/scene_manager/objects/physics_material.py": "a1ac0dc9928e9253e644ceb9a604aed262514aaad0473aff303bfae448e76c29",
-    "source/simulator/env/scene_manager/objects/primitives.py": "848f4ab08d86307812ebaea18cc86a51638395f0e6ea194b412d45a699ced926",
-    "source/simulator/env/scene_manager/objects/rigid.py": "8260fd2d8350d6a354c8281685bc7e25531923327078752efe0538d9cd78450b",
-    "source/simulator/env/scene_manager/objects/room.py": "946aa1afc23ab9b8fd17394517a5f54605bbbc609582d0854fed2078b3b2d8ee",
-    "source/simulator/env/scene_manager/objects/table.py": "c0e71396b0d53a250f74c75490e91086d67962237400a49f886a13ec07875308",
-    "source/simulator/env/scene_manager/scene_manager.py": "37911b12601038c9ee9bd180cae7a4751ea649bcdaf1ba2628f22c65cfd7dc72",
-    "source/simulator/env/seed_manager/seed_manager.py": "99bcf5a90392c8927aa0dbd799c07e06b9b03befe8b8eabaa8744a3dc9605ba4",
-    "source/simulator/env/seeding.py": "3f88d2a635c528253439cb31a29e5b9d0f47f6aaef0e975d86d9a8cf174be26b",
-    "source/simulator/env_cfg/arx_x5.yml": "7691350159fdeca92fd1b0b2da36913464ccdea3f3251eefb0176aa51db0807f",
-    "source/simulator/env_cfg/camera/camera_config.yml": "c1752170da855e735beba27fddbf78dad16f2395e4ef8b24e9827a768e616a62",
-    "source/simulator/env_cfg/camera/template.py": "68287ed1daec3b8ed7f053bf97a4d48bbddeb8b5757b53e438ac3c4240a9e61b",
-    "source/simulator/env_cfg/robot/_robot_info.json": "4c9de628bc1572502fc48d876c023fc63e02760e1a11a2a5258dea30ffa330e8",
-    "source/simulator/env_cfg/robot/dual_x5.yml": "4eefaccc75c5f83443b7bb85ef1de68ee7a3772e27af2b04a5e8d8769ab9baa1",
-    "source/simulator/env_cfg/robot/dual_x5_and_franka_competition.yml": "ec893f9fde577d0d1209e7dcb85ec2dd74288b636d0c68316cb07bc817b743ab",
-    "source/simulator/env_cfg/scene/conveyor.yml": "c1dfd27faf978f90dfc4825cd83bd203823547c46793f4f9201ce9f4b43beee7",
-    "source/simulator/env_cfg/scene/default.yml": "729df024b9707a55449da94f980c1ca116035e09a1284e2fe071f1b38efd061a",
-    "source/simulator/env_cfg/sim/sim_config.yml": "e442ede1c3d90fcfb50b9461c854d31b22dad9bfbb23397db228b27a6a486963",
-    "source/simulator/scripts/eval_policy.sh": "786966648496b7b68aad81e000a3f92910302eb391dbd07fec0a2dc91107956e",
-    "source/simulator/scripts/internal/summarize_result.py": "9ec361af587e1788d83d3cf8670e05b817ff252d33c33829a6337dc16ae370d2",
-    "source/simulator/scripts/internal/task_inventory.py": "ec82252b0f87fa5a3dbec2f682c170b25cbe47fe181e431a81df2c4b0675084d",
-    "source/simulator/src/__init__.py": "75bfbedb80f2e1421ad787992d78c23349bc8738b9edef7a76a3af83c9586fd6",
-    "source/simulator/src/eval_client/eval_env.py": "c577b27dab3c87588940670da81bcff007cd64b687eee8e0cc6684e5e24966b1",
-    "source/simulator/src/eval_client/main.py": "de2216bb5df707e505f7e2f3d31702907a4c1ab65205c5ea35494021f168c80e",
-    "source/simulator/src/eval_client/physx_warning_monitor.py": "6a9f8e8d75b1031d2cf2feb6dd72d37370d85d5bd80c4bea798622e23ce8dae6",
-    "source/simulator/task/RoboDojo/config/_task.yml": "655a87cd59fa6272bf352d32deb8ade162737b49a9824e16d428fecf0803380b",
-    "source/simulator/task/RoboDojo/config/align_blocks.yml": "082c9d7a32968052b8ce954bee212c7ec0a96b756150701d4b531abdea59be2b",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number.yml": "3bf6b7e286ea328fb1e4a994a8d30af9afb4b0b4dcd7ceafed43cc6160c9978f",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number_random.yml": "145b26eb5c8ed7d084a4ddc6e21694c542f4a86150756188ea38dcf66e565603",
-    "source/simulator/task/RoboDojo/config/build_tower.yml": "e1bab8756bbc5664e3171e2389c060a79a3fbbf92a6af4ef555dcaf59fce275c",
-    "source/simulator/task/RoboDojo/config/classify_objects.yml": "80a754d2aedd6a89f58185eae36f5b2eb1a28250f40a67325f576fde34650384",
-    "source/simulator/task/RoboDojo/config/classify_objects_by_language.yml": "ac6c739ca82a64b53c799af79b44363db1a88246f76d867f8830729e5613507b",
-    "source/simulator/task/RoboDojo/config/cover_blocks.yml": "68f90c85fad2b152cf7853cf48888d09ce74bb27a692ff2feacd11160771c5a7",
-    "source/simulator/task/RoboDojo/config/deposit_coin.yml": "f547b12216fb9a5dfe2e7d7e882d89e1e53ee1a805c943ba09a3c2cdea8c17a3",
-    "source/simulator/task/RoboDojo/config/fasten_screws.yml": "5fbe4cf5b3ee2eafc2771137b13176f01617f8b2131f02503e676a7b62e2eabf",
-    "source/simulator/task/RoboDojo/config/fill_egg_holder.yml": "d79129278c9a5488d63bb1d3865f2c6b39f21f85957ec94ace091af9ae29e31f",
-    "source/simulator/task/RoboDojo/config/fill_pen_holder.yml": "ac72e5582495e8757c547997b317e2fe990370009527b1faa6bfbf4e1275537c",
-    "source/simulator/task/RoboDojo/config/fold_clothes.yml": "56447e3f592e0db388ebb7482ec758ecd5485ec81b9227c2dffe5390bd7d8c8e",
-    "source/simulator/task/RoboDojo/config/fold_clothes_random.yml": "eafe52c0abcc1889069383b008cbe95c6af2d18ceafb1aecc78f0fae3b491b04",
-    "source/simulator/task/RoboDojo/config/general_pickup.yml": "35de68117cb0ed596b30db1c62f1f9090f2cd1237242a03d5aaeb7186e791d93",
-    "source/simulator/task/RoboDojo/config/hang_mugs.yml": "0086d374d3f7e82abb85e43a17cdfa437e1ac645a0171fb47496c2b81427dce9",
-    "source/simulator/task/RoboDojo/config/hang_mugs_random.yml": "847dfd39059275a671b5beff0243bf7bbaeebd9e5c0612a1877cdf12b9007dc6",
-    "source/simulator/task/RoboDojo/config/imitate_sorting_sequence.yml": "d9e194d8c742288601c641ccb75052a67da46e536c0b06c7fb7311d9e2c61a42",
-    "source/simulator/task/RoboDojo/config/insert_key.yml": "c0b409d5c020e0ad706c378b6f63b39dfe7bed4682d978907c40e1cfa7623335",
-    "source/simulator/task/RoboDojo/config/insert_tubes.yml": "5c070e5c764a2ea93f8de0b52e6a664a23c717963fe09fe3af17918a4a1ee9ef",
-    "source/simulator/task/RoboDojo/config/make_kong.yml": "8a33c9e57b6e6c719c4800911c37272ba055ad0f4cda2f1de9d531ee2910351a",
-    "source/simulator/task/RoboDojo/config/make_toast.yml": "d190a26233db174538ff0052b4c2958a3b107f0ccc63d0fdfe28da4c7f5dc57b",
-    "source/simulator/task/RoboDojo/config/make_toast_random.yml": "ca2c1e55a0975c8b019523710e90131504d8d2afeb829be362e4d7927fe97b48",
-    "source/simulator/task/RoboDojo/config/match_and_pick_from_conveyor.yml": "0ae1c2537d9b1880748316928cfd7ba4c6421ecccfedb85c4c5395a7a86c5361",
-    "source/simulator/task/RoboDojo/config/organize_table.yml": "9ffa15dd279060d4f042285a2c53a341c7208e2f89a2f3dd36ee08b22aa453fa",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box.yml": "e232962b8394d08c04c5da14deb88d7b233afa00ab99af8001c084218951d7ca",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box_random.yml": "7ac7f034fcb169e33a7f0a03b2b77e54d0206ed715409359b579055a701196ef",
-    "source/simulator/task/RoboDojo/config/pick_from_conveyor_by_image.yml": "6b310b70a0ead9d09e7c221c6bbc952140a44b1d43034613bf02e954eebcf9dd",
-    "source/simulator/task/RoboDojo/config/play_Xylophone.yml": "28adedb0f03c49693af32cea37855f3f95ec3f76f031bfd571fb8bfffb2ec59c",
-    "source/simulator/task/RoboDojo/config/play_stacking_toy.yml": "a09afc106161064a2ea6a40882d717f7872bd4875309a97431d6863a9b7ffa1c",
-    "source/simulator/task/RoboDojo/config/play_tic_tac_toe.yml": "63f863a7fb81a040f65803c64f97c4e9577d4fe32cd0f92563bdc72d1aa336d2",
-    "source/simulator/task/RoboDojo/config/plug_in_charger.yml": "2046b47112049e038ecfa24a8c8044fb1db84f629a4a32eeb24822157d2d4dcb",
-    "source/simulator/task/RoboDojo/config/pour_balls_into_vase.yml": "084a707af61d24123912777b417fe0c0b230eb9653e686af42570791e924ff8b",
-    "source/simulator/task/RoboDojo/config/pour_by_language.yml": "efb2e3c3cb08dd6d2d2ae011b3e9a218b5f1f0a92f025b8e3b9e8abd7011c250",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup.yml": "32092b11667b3306579b04225303610c061585ace28531fa9996afc339c5a2ba",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup_random.yml": "1603feeab6f0f7dc9c4897e7f40f322b86112b6ab987dad0598f42ac1b35db69",
-    "source/simulator/task/RoboDojo/config/press_by_number.yml": "b784643495bcbca933f5a5fbc91b7dad3465fe888239b358d516977c78c38f63",
-    "source/simulator/task/RoboDojo/config/push_T.yml": "b55ea11002ff3e5dc63982ca8c1247fd7bee4f4a607f4d674e220b86cf120a25",
-    "source/simulator/task/RoboDojo/config/push_T_random.yml": "07609ec20cea40afa1d5720e895fb1b445411f261b2e5b6850e75d0f1930d32d",
-    "source/simulator/task/RoboDojo/config/put_bottles_into_dustbin.yml": "7eddf43e287b6e8aea686c6d686fb87e13ac1be1cb5aac16a8a2b0d2e9eb976e",
-    "source/simulator/task/RoboDojo/config/solve_equation.yml": "bdaad2ff1cc8b7ec6413a3715d49ea9cefd664e7e204544977832e18785f1e75",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size.yml": "8927cb2ebb18449554d851e629bd7d568e12e46c5c3c636c8e7122f4fab1ed1f",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size_random.yml": "f895d3ce7eaae99454706afccd09011f56d077998bcc1e95840116658b78f139",
-    "source/simulator/task/RoboDojo/config/stack_blocks.yml": "2802e88ebc3e2d3d05ce345d75e9ad6151a6690e4ef1e0021a5223d291a402d2",
-    "source/simulator/task/RoboDojo/config/stack_blocks_by_language.yml": "4fbfed4f7fc49120831087e709a8dbdfdb4504eb143196000d5790b54eb3bc40",
-    "source/simulator/task/RoboDojo/config/stack_blocks_random.yml": "e25d2a096601da30abdcfe1256a464eaf48c0c8f844eb2c737232626d49639d5",
-    "source/simulator/task/RoboDojo/config/stack_bowls.yml": "4e2aaa12b8b0f58ca801f1a93638c46d3b49ebad809c5a8011382ccda846dcb1",
-    "source/simulator/task/RoboDojo/config/stack_bowls_random.yml": "5c08af1bcc925d958d53de51e8ebd336006203f1bc81e4ef68f3719eadfabb49",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones.yml": "0fdaab803a9370a9e2d1fc8d86a5788daa3940c08fd6f767334f625642fca3e2",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones_random.yml": "93f0ca39bc8bb32df94e84a02f22ed4d7860f37cdc0b9dd17e44721445849576",
-    "source/simulator/task/RoboDojo/config/store_tools_in_toolbox.yml": "641c8ef92393bd986e408d26a84b1cfcdac27466a5ac5f0b3ebe4bb84430c527",
-    "source/simulator/task/RoboDojo/config/swap_T.yml": "1daf769911e13714b4aeabd208292931e32f4dd39968893152bb7bb477c46c9b",
-    "source/simulator/task/RoboDojo/config/swap_blocks.yml": "b6bb2a9df494719f26c6b1aab2e7576107ae76e2fbac8b63a5770f21e0e9f05f",
-    "source/simulator/task/RoboDojo/config/sweep_blocks.yml": "98aa25ac39be9d335c71b2b586bbba372789e9d61e652e1147152c87febbf1f5",
-    "source/simulator/task/RoboDojo/config/sweep_blocks_random.yml": "4ebff5e2cad07f958dc4b2793b86b1aa803c0809394818ee3752d300e58b6187",
-    "source/simulator/task/RoboDojo/task_registry.py": "11218028f772d9253ee709a7d85aa72d1e1fd87e4f8d33ee3787d6064438f9ab",
-    "source/simulator/task/RoboDojo/tasks/align_blocks.py": "c8597e96132cd7c9d30f3ae101f7bbe8ac28597b629599d75d442bbea4c6e74c",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number.py": "277acd20ae4345dc71267762e588b7b0343387c71685183890f044a4808b7a20",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number_random.py": "01757509978663bf7c0c95059e8ae18d45b8fe0953f6b54540e2bd6bf141182e",
-    "source/simulator/task/RoboDojo/tasks/build_tower.py": "814cf6ffb459871e5cf49f4e127903dbfbc30926f33db71de0b84a71ab235d6e",
-    "source/simulator/task/RoboDojo/tasks/classify_objects.py": "aef4333c088766375e21a928d53a9a93aaee285fecae41e22385acdc74dd11ac",
-    "source/simulator/task/RoboDojo/tasks/classify_objects_by_language.py": "242a8f048f487eb174ef246be78613bbb5f153549ee11a2650412718cf3b36c9",
-    "source/simulator/task/RoboDojo/tasks/cover_blocks.py": "9d14931b16842b93bea8abf9a07991a93a7e663b93c5fd1695108c8b922ab862",
-    "source/simulator/task/RoboDojo/tasks/deposit_coin.py": "f9868ef827ec17b8db6c4650cc5d9a9e593d20d00ee822e2212ce229ec01fb61",
-    "source/simulator/task/RoboDojo/tasks/fasten_screws.py": "2e187a28e33e44efb87baae6463b3ec35264ccd41e35c21a1be44f6361675185",
-    "source/simulator/task/RoboDojo/tasks/fill_egg_holder.py": "506dfb772f8882ae53966e3580affde19c3c0f11591c78cc5f5b1d997ba04f7e",
-    "source/simulator/task/RoboDojo/tasks/fill_pen_holder.py": "8706dce239487e078b21af0b95bcc5cd4eabdcdadcdcbc6400d6b138d982f1e6",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes.py": "ecea9c287d0579605330168e6f36734704fb14a6b1e5dec82b6f2ee3a56ec20a",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes_random.py": "9e40abe3f7201a8c13961edc1ca8075260c315584d3d6b93fa1df966d13d1e88",
-    "source/simulator/task/RoboDojo/tasks/general_pickup.py": "1a9eef7fe518f9d2c6805ad1a8f97c1045fa83726f84b9c63f67ffdfabbc40a7",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs.py": "4bc6586bd87de030b9c580fea2a45e7977c3b6a4d75cf3c43b546d349d42e906",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs_random.py": "ebfa4a95deb954aa7a9063a07fe366cb5ea1835d3f52342b10d531225dfeee92",
-    "source/simulator/task/RoboDojo/tasks/imitate_sorting_sequence.py": "18400cb3c5f5103492819df2dcdf05a77916106e8a503cb62b4ee6cdbaa9305b",
-    "source/simulator/task/RoboDojo/tasks/insert_key.py": "180978a2915744f0596fbca4ea0ad6dcb47dce005d43f06807249bc8b4d1a828",
-    "source/simulator/task/RoboDojo/tasks/insert_tubes.py": "9a68795ddbd0b816ffd7627e673390ecd83594c0e9d374d4e3abf3435ad439e1",
-    "source/simulator/task/RoboDojo/tasks/make_kong.py": "b9266689d65f9213b05219ad952344f9fd63be8892097ec0103af99791b245d5",
-    "source/simulator/task/RoboDojo/tasks/make_toast.py": "1047f484a73d1f92304393879b51f87e33d661b6ad7937198352fe37a0d3233d",
-    "source/simulator/task/RoboDojo/tasks/make_toast_random.py": "69fb879b89e98f7369971834eb6d1b95c5a65f5ffe978b2cffae0b12c6b756ca",
-    "source/simulator/task/RoboDojo/tasks/match_and_pick_from_conveyor.py": "53110b8a0c971bd3ae8878bfb7d0dcd9626bd9b610d4e7fc91772313b435a5c0",
-    "source/simulator/task/RoboDojo/tasks/organize_table.py": "aceb298e78f3f39e4cf9a9f4c5a4f747b2c55f1a6006de2065e110a05f4295d6",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box.py": "84a1cf1dd5e166f36291130f9d8cb5dfe4b3f4732a46674bef6374a7ea2d66be",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box_random.py": "4db259a1b97cdcc39ebc9794244fadc79d31c4ae64069c9b11a233356fcb655c",
-    "source/simulator/task/RoboDojo/tasks/pick_from_conveyor_by_image.py": "a0d29576d0aab8232d2fd3ebcc8c510e81c70b4cc04602a5be3e99ee2275e01d",
-    "source/simulator/task/RoboDojo/tasks/play_Xylophone.py": "b7a4b0bc0457bcf55d714ca1867476fc4aabc53084adfc78d3ff5d1437bf7847",
-    "source/simulator/task/RoboDojo/tasks/play_stacking_toy.py": "23361329ee7efe330da1f1c241415205bafed2b6c082a913659d868a80d338c3",
-    "source/simulator/task/RoboDojo/tasks/play_tic_tac_toe.py": "8f75ebfe55c997ad9bd3b342db46e843e87397db94356f31cadcbe0cb483bca7",
-    "source/simulator/task/RoboDojo/tasks/plug_in_charger.py": "aaa59ade46dc54f430e1148e3f6fae487970cd29eb2c07b5ff7b4045edb831b6",
-    "source/simulator/task/RoboDojo/tasks/pour_balls_into_vase.py": "eae388b93ec196c401f76bbab3b0119cb2164f90bece0225d4fb925d3f2b73ab",
-    "source/simulator/task/RoboDojo/tasks/pour_by_language.py": "f7a15a7fca2c649a7918d68960ede27e667f409a339b668dd9876db890724cfb",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup.py": "73f8c91b57c352fdc1e7ba3fe5354b3c10e4156d8e5ba777762ad0fa1aff5663",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup_random.py": "a1c4097b6e34b2c62885b3a9c7d2584a93d9e296dc41d311a80c2e6d541d53af",
-    "source/simulator/task/RoboDojo/tasks/press_by_number.py": "006f969087c00c417a63875a9de153f698c52d1f1ca0caf0aa16076aa15059c5",
-    "source/simulator/task/RoboDojo/tasks/push_T.py": "a01d753bc320ffbf492ccc29e5f7ab3b39e6c7d255cd25804db1934a30fa5b41",
-    "source/simulator/task/RoboDojo/tasks/push_T_random.py": "dc8d9c8e0d2e8dfffa2b7d433a8278c0eaf707cc4861b15035364c2e70061dab",
-    "source/simulator/task/RoboDojo/tasks/put_bottles_into_dustbin.py": "3ba6e19df0c5ff19ba961092f031ce14e91e72a9a4a5934eece6000a1ffa1c48",
-    "source/simulator/task/RoboDojo/tasks/solve_equation.py": "e287a2f36b82b28e8ca7824f63b1b723b02972ea6dfc115ff32b2882729751f7",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size.py": "1838fa6b515bad3796895cb5c1883c6a42789b7dada5434765764600e70beb49",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size_random.py": "b793ab5bd32bd2c10bd46874440904837f9c4e290e7508b06426e2e96425709a",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks.py": "50bd388f9e47002966fbb0146742467a820295b8686e0f86379f0efbcfffcc1b",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_by_language.py": "36a83925f35c7a4caac5add328dd35cb1d346bad833d3064f455d88221e7a41e",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_random.py": "19d4e5cd693e068ad64de75cbcd24977ab8d39e22fa17362a391dcd733bc9bb5",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls.py": "022c129ec3493183b1a26688a0402f5a00a1546ae011999e9c19f4c0029e2b7d",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls_random.py": "a0e20469eb44a9ee98650d09386814c1df5264c68ec139f06aabf0c86f092514",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones.py": "2bee4e31f0594900629fe7f43266e16fecb310b152930ce51524507136809a92",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones_random.py": "58c3c5113054a379214d64ae406ed8dde99603cf684ce9cd8c7bcb72ed334382",
-    "source/simulator/task/RoboDojo/tasks/store_tools_in_toolbox.py": "2969fc3eb171560b3c0bcea62b9f38b8b35f3aecc9bcc7085c3b1cec365b48fc",
-    "source/simulator/task/RoboDojo/tasks/swap_T.py": "57fc289b3957caa454e1d8e005e3a57c973d6ab40121845c94799e61e4bfafdf",
-    "source/simulator/task/RoboDojo/tasks/swap_blocks.py": "08ed24ae9dff27f4e5c850161a02f62ce237859c096d9b06a853a7890e0ad4ed",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks.py": "65786217e199c1e2b26ec8e2c232bf98ecb304a20939d0908fe12bc41e990614",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks_random.py": "68b6127038322cf001d5f614060b8a1d96e64a49efddfcdcae91127d5c206bc5",
-    "source/simulator/utils/cluttered_generator.py": "f59cfcfed2b4cce5ae9aeaf3339c86d17e237874e3526bded3cb05136dbcbd50",
-    "source/simulator/utils/ensure_usd_path.py": "7c8e71cf91636e630b1b174098fe0b38740857daeaf207e2fae772c4727b11bf",
-    "source/simulator/utils/load_file.py": "ef3a6dc96aa1a7442cfd24f7be9671052a98d2f20783c3e71032332704132943",
-    "source/simulator/utils/path.py": "1d427c7cf33e748e01d0d083c9432d8be8b5c3a1f35c5fb0f6ebb0714098f1d1",
-    "source/simulator/utils/pipeline_utils.py": "d5d8da29e6ca84ab04323a2f79130f930572785b4399634e06a42834ed4087ae",
-    "source/simulator/utils/rotations.py": "8e814f79938bdb9e85f1c024825617cc08d281b61728c6410b23ef1bd852c340",
-    "source/simulator/utils/save_file.py": "6cff5950f73bf7a1ed404f804d6f08af712974f22c3955de4cdf142f211aaf00",
-    "source/simulator/utils/transformer.py": "73de4f5e1e0ff60d4f2e1aaf6104368e8f919269facb68b4fa0344d35bc08cbe",
-    "source/simulator/utils/usd_schema.py": "260d30b0ada2976136cad3e49e810b4409304eb4a2bc60bd9ed710864d3953ac",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/METADATA": "614985b278f601b8ef863f82783b991cc3ada4b89f592eb71c7c48c5f2ef2298",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/RECORD": "c7660eb223c8080f75c49690eaf0a9c9c00b6195511791b33b15e6b24e42edd7",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/WHEEL": "942926c567d0b1273d09f20295879abd37507845f6b683eab9f87d808df91100",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/licenses/LICENSE": "fe1049884b1a0d9342901e88e07f32925d24b3121d9972b6a6805fb9824b095d",
-    "source/transport_deps/annotated_types/__init__.py": "a7104a4d439b27a9f74fc0be236b9ba1b7831e6044026802a205abc1298a9bc8",
-    "source/transport_deps/annotated_types/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types/test_cases.py": "d867472acb57b81ce97f78972828e6bc5669ca4db33fd1e6f645c6183846068e",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/AUTHORS.md": "5fe0446ddd6d1c70fa298747368582dc1d91c23d62c8fb7228595bc87fd53232",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/LICENSE.md": "509aa7af19ae2b5b1681b22b969627ee9e1fd15ff7ea5bf9f6e301fc44d5a6c2",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/METADATA": "b7ee8fa1bf018747082cd6d2fcaa14e5d0f5956b5477ab5810eb61ac4cc7bcf9",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/RECORD": "51331727bf8d9ff60510de5e7756c5dd72aba7663f3673e2827bf1ca5c09c7e3",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/WHEEL": "1d7faff7e9e8524c94a31c99d4f312b92edab94c43011e1505da182ea0f4c70b",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/top_level.txt": "19da698ad1490bdde1c7f18fabaceb2625aaf48b364d4dd7ad404c7e5eed75b9",
-    "source/transport_deps/msgpack_numpy.py": "0a45a7430e2cefa2e74235c4c6a34a22ead2fb57a7de8b2631f45df977701d63",
-    "source/transport_deps/pydantic-2.12.3.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic-2.12.3.dist-info/METADATA": "348336c208743aafeabfa5537c29f429035a3107b7a78bbb6a3bf7b3a42fe4aa",
-    "source/transport_deps/pydantic-2.12.3.dist-info/RECORD": "089c7ceb033486c4e6106749bb1a38c80b63905da74d842d067735e8b9c86e4d",
-    "source/transport_deps/pydantic-2.12.3.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic-2.12.3.dist-info/WHEEL": "aad0b0a12256807936d52d4a6f88a1773236ae527564a688bab4e3fe780e8724",
-    "source/transport_deps/pydantic-2.12.3.dist-info/licenses/LICENSE": "a9e186f3ca16b5eef84318e7a701721351a00cb7b8ae3a4394b67b49e3529ef3",
-    "source/transport_deps/pydantic/__init__.py": "e62127278c07bf5384cdd2903f368a69929f3b8a524000bae4e0eb608ebf4bc6",
-    "source/transport_deps/pydantic/_internal/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/_internal/_config.py": "4d66708377346d91e24f76e8479f9862a928b877308d175e9e6c861e87d5ec4d",
-    "source/transport_deps/pydantic/_internal/_core_metadata.py": "63f836b778bbba5b8afb05c266fcc97d114c3d4336dda0582df69ee05cc13f2d",
-    "source/transport_deps/pydantic/_internal/_core_utils.py": "d63aeee156c9d31eb747a76d55cb8e23e74a4130bf77f9529c9584050cc63444",
-    "source/transport_deps/pydantic/_internal/_dataclasses.py": "4e4d6611a7e169dd6457b2b9b4f5f9070c5649763b0be30ac1fd0e2c58089440",
-    "source/transport_deps/pydantic/_internal/_decorators.py": "3e7c80a0a4a0dc135b0954999f0ab0f67684835503b58bc34fd2e5ba280f88ef",
-    "source/transport_deps/pydantic/_internal/_decorators_v1.py": "b5f75f76940a6384765c23b0a876d97a840cbabe95362811adf86e757047c77f",
-    "source/transport_deps/pydantic/_internal/_discriminated_union.py": "68c9744914b24321df5b8fa49673131cdbf0491a2a1371f73d157fd39bd1b138",
-    "source/transport_deps/pydantic/_internal/_docs_extraction.py": "7f2ce74801e1e40ce88675d94ad5741ef1fe9d16dabcd1cfca0fa49f1f92fc41",
-    "source/transport_deps/pydantic/_internal/_fields.py": "6127c42aadb51608cb27a62a6172a1d1e104b399f130fbd0ea1a7da40f0dcdfc",
-    "source/transport_deps/pydantic/_internal/_forward_ref.py": "e67dd8efedc028b9fcfc54b761cecabad2e23d48725e602d90464e6859e89f03",
-    "source/transport_deps/pydantic/_internal/_generate_schema.py": "4d3e3dbf3633a87f74ad6aefe69b4da19823cceb11d0a3e54a4a8f545ae7ac1c",
-    "source/transport_deps/pydantic/_internal/_generics.py": "10baa34fa2cccd0cd600ad04079ffdaa47bf8806b3cf4390e20ba7a7fb8ab986",
-    "source/transport_deps/pydantic/_internal/_git.py": "2303e1dc33df6b65eadeb06e07d371f25b91d80d62ebd41d7937d659722e0958",
-    "source/transport_deps/pydantic/_internal/_import_utils.py": "4d18710f93ae63a094a2c8a805d05c254b34a26ec820e36265d60057bcd0f233",
-    "source/transport_deps/pydantic/_internal/_internal_dataclass.py": "fdb79d7355dbbaeca044688b66a914930c05a506a847584a2e5479d359f2c926",
-    "source/transport_deps/pydantic/_internal/_known_annotated_metadata.py": "25ceca4cd159a01ddffb489b3ff36024834e795bd8137ab73930500e354c9375",
-    "source/transport_deps/pydantic/_internal/_mock_val_ser.py": "c2645115206fa9f70b6c8e353ec165881e2ed80677989a597a21116c3df1293a",
-    "source/transport_deps/pydantic/_internal/_model_construction.py": "c24f9b3460c026f76e686be7d14d3ff33125d06111bb4b2126f37cfd97e461ac",
-    "source/transport_deps/pydantic/_internal/_namespace_utils.py": "865dfe4d102bf365368d3c8f3f7b7e42ab2f28b8abc6d90b7cd7eb37e7e9d31f",
-    "source/transport_deps/pydantic/_internal/_repr.py": "8d07e726ec83c50a5244d846dbd208f4f5fc7b836fdaa599ac4c3696a8a1dd41",
-    "source/transport_deps/pydantic/_internal/_schema_gather.py": "54b12fe754d811e78d0f6733b32ae626ad4c56726a4ce9a72da9fb546e3873c0",
-    "source/transport_deps/pydantic/_internal/_schema_generation_shared.py": "17fadb41bae4a289a0c6cb24747a4fd23509ed309f7b40400c0124aba089e273",
-    "source/transport_deps/pydantic/_internal/_serializers.py": "6085af4a6011e5f9db19258239076e5add72078650638d9e02bb9cf9e9eb6fa7",
-    "source/transport_deps/pydantic/_internal/_signature.py": "f049633c97b8a529daa6e8ab1b90e4040803d618201f1100cb314f1fef47d331",
-    "source/transport_deps/pydantic/_internal/_typing_extra.py": "755c481a1b0be37cda79b3423108752b94235fa2386bd3959cd01c2f6a486bfc",
-    "source/transport_deps/pydantic/_internal/_utils.py": "0878c7fb4ce75235fd479b144a33d3f23f3efbb3f37bcca36a27ac7130152e24",
-    "source/transport_deps/pydantic/_internal/_validate_call.py": "3df7559d2ce15ceac436d683a03c373c558f5580f95bf80d60f7bca77520e8b8",
-    "source/transport_deps/pydantic/_internal/_validators.py": "76fd1ad8d91ce33718aafdf51a1fd021dda571ff96d24429574a1237380475f8",
-    "source/transport_deps/pydantic/_migration.py": "545ef72d1094cf722b6f9c55b75de36f7340717567105e93d7e27438b7de6790",
-    "source/transport_deps/pydantic/alias_generators.py": "28cd67deee097cb48197552e620de1a181f35c90fecaf82b9eaf2ed5c73087f0",
-    "source/transport_deps/pydantic/aliases.py": "be1087ca84969d7f8427ec166f9aa3e31c91b2c8065a74d07f3429e06499f6e8",
-    "source/transport_deps/pydantic/annotated_handlers.py": "59fc854aac28108157061ed3ef73f270a9682350e25f8e465a2d3e24eb024786",
-    "source/transport_deps/pydantic/class_validators.py": "8bf5778fe3d87462d22e68ff20965e913463ba73bc488573f0b325aae3f23fb1",
-    "source/transport_deps/pydantic/color.py": "033a867d540717ddbf65cb4372e7b40cce324e9d8fead7a49304483535ab2c8a",
-    "source/transport_deps/pydantic/config.py": "b83676c4df89ec255035e9c269529c62c244cc83a92aa9607d0bc3774d35c229",
-    "source/transport_deps/pydantic/dataclasses.py": "dbf8cdfb9d595abbb222c7a78cb235f9585de2a206345ddeb9ef6319c769a872",
-    "source/transport_deps/pydantic/datetime_parse.py": "402f9680cc4caffc10fe6357512ec055ffb684b121befb0f6353d0ca148639d9",
-    "source/transport_deps/pydantic/decorator.py": "617fa3500a6ee4029a55628fa1a57e9fee25ed46d2ebd184b7d45ade1b3398a2",
-    "source/transport_deps/pydantic/deprecated/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/deprecated/class_validators.py": "10071a550339ce9db0066974c9b37ad827d07f226f0cbc7941df4f938fed6f85",
-    "source/transport_deps/pydantic/deprecated/config.py": "93f96c564e7ba5ac4b24e701b9e1f4edcbb53a012059d541c66ea57da0b70825",
-    "source/transport_deps/pydantic/deprecated/copy_internals.py": "1a177ebe431de446020a0c821ad3ca3b9f27a7d704281402eaa9017881052368",
-    "source/transport_deps/pydantic/deprecated/decorator.py": "4c19ba6c9ef026c36287ff16ab92330dcc0fdf69bdfef7f1b3de9d7ac2ee934d",
-    "source/transport_deps/pydantic/deprecated/json.py": "1e55821b7e5146bc72cee4d2e8b4d089907046198365699ea901fcacb5bac00f",
-    "source/transport_deps/pydantic/deprecated/parse.py": "1b377a6ff83ccc95dcb84ed046ae5a761c7f10c25a8577dca57085d1182baaa2",
-    "source/transport_deps/pydantic/deprecated/tools.py": "36b9bda054595a9f3e8e57ef3e02482c4b32c29e18cd90f9d972063c32f11dc2",
-    "source/transport_deps/pydantic/env_settings.py": "e881de79612a9543d152fdd5f805e217f5bdd5f836270fccdced25df827f7b20",
-    "source/transport_deps/pydantic/error_wrappers.py": "44aea6a804dcf72303f8a043f48252f47a4a0a9ad61ddf30a3ce019e6fb77d1f",
-    "source/transport_deps/pydantic/errors.py": "edcb41342b6de7b919171ef52ecd87f3a22e7d0011bf8c0f29ff0385db159f9c",
-    "source/transport_deps/pydantic/experimental/__init__.py": "413eeb298743b0289327d1848e6b107561d2730a4aaeb3641aaeafa8e34fe914",
-    "source/transport_deps/pydantic/experimental/arguments_schema.py": "1059e35ffba5a7eb4fc948d05f9a6642dba0d4e14bfc071cf1b70c6cb77e7d56",
-    "source/transport_deps/pydantic/experimental/missing_sentinel.py": "8507a382d174d3052e40c9e2f78dbd7af83e797c88c292af8ec0fca1fa9f8fec",
-    "source/transport_deps/pydantic/experimental/pipeline.py": "2affddbdc7b12ae99acdf44bd32ebd01ac9e03a1f7ed2ae6b19dd2525fe7d2a6",
-    "source/transport_deps/pydantic/fields.py": "5ae0c63af076d8a5aeb96ddf5e74b85af838a97fed769f17ec1ac09736b8b30f",
-    "source/transport_deps/pydantic/functional_serializers.py": "ac4cc7dfdd73ab2de8fdb5a4d9012ebf24a67103599b05e62502c2dd9185ed00",
-    "source/transport_deps/pydantic/functional_validators.py": "73ffbbc1e5a918d70e61f45f55416ed758ebc4c54c75f63f73ee22b2dc24f796",
-    "source/transport_deps/pydantic/generics.py": "d19a99f4ef5a9e7223ff7986051aa9b3886d7b2ddbe65575f9ddad5313cc3270",
-    "source/transport_deps/pydantic/json.py": "647f11908ee1e01cfece9f0e7530316c952856f72853e8e131d459d01fa4d317",
-    "source/transport_deps/pydantic/json_schema.py": "677786d509be226f8c417ebac28c963fd5b2c055272d116d68e0cbced559f937",
-    "source/transport_deps/pydantic/main.py": "efb414c13b5041d51a7e8af78abe70d820b5b34be2805b460747d2953bc5dbb2",
-    "source/transport_deps/pydantic/mypy.py": "a7a294d46c0f1dacc5ec4e6f26ad6e2dde2d1dde8313a6eb7b8fa6e4b9f6de6b",
-    "source/transport_deps/pydantic/networks.py": "2c995c70c79df7b1007a9293e8e98f23d73a8e1e62c1b063935fe35f5449c37d",
-    "source/transport_deps/pydantic/parse.py": "c2477cd9d82dbd6b43f3de54fc8e84d61b6a3251a1052604577f5cb8149ea370",
-    "source/transport_deps/pydantic/plugin/__init__.py": "6bb4f0dfae94e8ade496d08235963bea773db2cfbbb86190e344d769df4eca94",
-    "source/transport_deps/pydantic/plugin/_loader.py": "f502d79de2c49afca15e46bff63e0bae46e67b8a8fbfaa98a6196c8c5d8c1ac0",
-    "source/transport_deps/pydantic/plugin/_schema_validator.py": "41b9aab06df730199fb4d4369cd8ae376d8b85badec5403b8a90d5bf7274d815",
-    "source/transport_deps/pydantic/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/root_model.py": "06f98bb56e22d7574993e74b38cdeb97e8e725d40679e41314172610eabaa4c8",
-    "source/transport_deps/pydantic/schema.py": "56aaa3beafcb9daa5592779b51ddc1a7f275a768176599d94602f8f1b5441bba",
-    "source/transport_deps/pydantic/tools.py": "88742977c489e43093b4f5796ad015d3a4fcf5b8d268c15e659d8b5fd95ab196",
-    "source/transport_deps/pydantic/type_adapter.py": "553fbeca0e1adbbb21481cd61c13cacf8f777f7890f686dd9049218d92a513b4",
-    "source/transport_deps/pydantic/types.py": "9ea752f89d995ea4e1daa7b224ecc14c1b475b267961115ef2068c579f5df471",
-    "source/transport_deps/pydantic/typing.py": "3fb7de037e4cc1370bb11d6e2fb75bd12fa8c9d071a1b9976b9e580e80606a34",
-    "source/transport_deps/pydantic/utils.py": "d799d1d90a6a4c1165415e1336d4c8b4cc93271fdf6f257e80f988118d46a287",
-    "source/transport_deps/pydantic/v1/__init__.py": "14b43c212a7a3156517e39d2edf4389b51714311421764158a41380cafb83e11",
-    "source/transport_deps/pydantic/v1/_hypothesis_plugin.py": "e444b9c56bb0d45400b3298b7b3cbc4209d5cf466955f5378e4993ef81f6ed54",
-    "source/transport_deps/pydantic/v1/annotated_types.py": "ba4d8d000c6a88d10b2a388787287129a20e87c1759585bf2f3ad6dd7ee86417",
-    "source/transport_deps/pydantic/v1/class_validators.py": "50b39a21481850306c1cbec411511aadc33e52e6ca52082837c8526c3a27b051",
-    "source/transport_deps/pydantic/v1/color.py": "8990012d8a7a395a28d801643fd229ae2ff049cebe2a496ef18ba149aaed7798",
-    "source/transport_deps/pydantic/v1/config.py": "6ba3f459eafdc7871bc0a5bb5eff29a1252a33858ff912d6c17e9832962af400",
-    "source/transport_deps/pydantic/v1/dataclasses.py": "efce1caaf2276f020f5abf6eb1fa57ddc87bcf8b77276b532ba374ebbff0935a",
-    "source/transport_deps/pydantic/v1/datetime_parse.py": "e10cb5910a6adeb3556497881de48f0e9b92d81be1a7528fb73246d71bbe1f4d",
-    "source/transport_deps/pydantic/v1/decorator.py": "cda6b1c71a163c29bcd7c0f56ecd3285aa518d79b7d95f06d191d609d335b970",
-    "source/transport_deps/pydantic/v1/env_settings.py": "03d557c2d465d36018fa31f40b4a2ecb954dc377e2e85fe9933b870e380000e3",
-    "source/transport_deps/pydantic/v1/error_wrappers.py": "ebadb931fc3daa40b63708ad07f2450167bc07e5efeb3054eeb2fd936f2d7f2a",
-    "source/transport_deps/pydantic/v1/errors.py": "988c0f103e6f18ce50e6fe02e19d493e5753447fa89af132947ea4b0c84e98fc",
-    "source/transport_deps/pydantic/v1/fields.py": "56a5890ab894362132a6d5eba03549e74d49a550347a7d9500d724b2a5cbd9bf",
-    "source/transport_deps/pydantic/v1/generics.py": "5730bd61457e11b3e94376807e4d5c3457a3efdfc8cf39e443b5ab9934d94bd1",
-    "source/transport_deps/pydantic/v1/json.py": "590e47cbf848a5f751dd84bc93a37613a28c273b1d6c18bf95d58e3c9695f353",
-    "source/transport_deps/pydantic/v1/main.py": "cee36974de50d15d301b65144cab741d4cb75c9e0e02f3d26437625d8f85233b",
-    "source/transport_deps/pydantic/v1/mypy.py": "0a5f1745f0a621c544de3e40114e760bc88387c95c5fffc3de1cf68c85b1300b",
-    "source/transport_deps/pydantic/v1/networks.py": "1d836d2807ce98e9ca269b0383583a488923f563e153ffa2f25e5ed89281a46e",
-    "source/transport_deps/pydantic/v1/parse.py": "049b5daa2651b5abfd55114298ec6863e2889909a33f2f80fcda288e21546716",
-    "source/transport_deps/pydantic/v1/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/v1/schema.py": "6aa06e03ef9cabc8005648a6e4124f140487cce67c745b66157fdf346afa8a9e",
-    "source/transport_deps/pydantic/v1/tools.py": "d650dd5c79348cbe6e3f7bb9442600bd402518296000efb8e6592af63edfc810",
-    "source/transport_deps/pydantic/v1/types.py": "07397e45c9e2b4f0479eac308fd8afec98c7b8dd46a67587db874a905de5f5ef",
-    "source/transport_deps/pydantic/v1/typing.py": "ec67418356131d42d4f35b6107ff5c8d13437d97e7e248685fb9c6b70fe47821",
-    "source/transport_deps/pydantic/v1/utils.py": "339151c9f3546f50369a4f6568680255d7c71dbdc0b5082b8cee6a7f205fe310",
-    "source/transport_deps/pydantic/v1/validators.py": "9725249f5316847c65097e597c48058ff0801e88e888f71a41e31d10cf57be25",
-    "source/transport_deps/pydantic/v1/version.py": "1d79d75bed5b316e6a2a196be5180e10fa21ad90c20d2bb2cbcfa08bc182819a",
-    "source/transport_deps/pydantic/validate_call_decorator.py": "f23a8b9605d38d6123e1d5c3834c08dc41902a46f42670ac2ff2525236d4e528",
-    "source/transport_deps/pydantic/validators.py": "a706c825755bd42576980138c3f1067cd8fb0f0cec29a5b0fed4cbe9cbe24eeb",
-    "source/transport_deps/pydantic/version.py": "4e6124eaa03e540d4f56c737f5b66584f010032bda71de812b3c57007c96d395",
-    "source/transport_deps/pydantic/warnings.py": "5aed551b3aef159c384fac8220a1e31fb2d263ae878dbb720856e7e6e5a8309e",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/METADATA": "3434fb886fe254ed414541f305a86e43ff575093f8f6d15c89bd11144ed8e131",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/RECORD": "7f3a82de0054511dbd2f7309c5e5b6ddb3d96c6230bfa83fd73bbdffc51cd0f1",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/WHEEL": "7f32b0ec3e21425972f904759f2efc26a495ce223c174ce187bb87ac82f5240a",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/licenses/LICENSE": "2afdd30d54b4d62b6f488a6bcc1546e84ec5061f13f4209c03d012348783795a",
-    "source/transport_deps/pydantic_core/__init__.py": "9cad6292b75254af606a970aae9bff6e5ae9f0b08089cd63acafa391b60794d2",
-    "source/transport_deps/pydantic_core/_pydantic_core.cpython-310-x86_64-linux-gnu.so": "cd9bb47485cdffc363010507a214155966ca860b0b93bda39583492c6764e631",
-    "source/transport_deps/pydantic_core/_pydantic_core.pyi": "3ea1dbd4182f08cf9341f24b3c5cf7db77a05b3535ffe9e23525127a3617a11f",
-    "source/transport_deps/pydantic_core/core_schema.py": "bbdc850b72d685133a0e250fed263b33691dcdf3813492f3c0e31001e3d46005",
-    "source/transport_deps/pydantic_core/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/METADATA": "bad3d9e6342ef6750cf5128ec942327b8bebdce0b6e4d6f67cb15b27ae5118da",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/RECORD": "cf5e3948866214f2e9ab0c1d43c157e9f9c14f59cf8442936c6cb97d5f2b6a5d",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/WHEEL": "ccec20e2307acd7da453dd7437e70c6b08e2bc3ead3bc3445af135da37b56d59",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/licenses/LICENSE": "804b59b25f2c31bd278f9202a19ae49a3945aa2664387e2d0a128c7cacc61ec3",
-    "source/transport_deps/typing_inspection/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/introspection.py": "5888c45bd627da692d9527699e3f549adf35651596b1656772310e12181b4396",
-    "source/transport_deps/typing_inspection/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/typing_objects.py": "a778d6290d51581c9c9a23e0263af2a4604c76def5c4d40a7cd1a375b0046ff6",
-    "source/transport_deps/typing_inspection/typing_objects.pyi": "3bd24dd92d0270d31d64fd40abc1ef8693f5322aab9a78aeb3c5b62f992f84be",
-    "source/vpp2/src/vpp2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/adaptive_mask/__init__.py": "a73f5515ffb77d323f9c324776e9c166d13817c470506cfd5db4ef0546cbbdda",
-    "source/vpp2/src/vpp2/adaptive_mask/audit.py": "c949909a2e60c3c4389c0221c79459961f2c4925baa07f7985774a52257a1d17",
-    "source/vpp2/src/vpp2/adaptive_mask/budget.py": "74e32285903dbf673ecadb6a60d917a990b9c6619beac9f926c1b3cbdae3887c",
-    "source/vpp2/src/vpp2/adaptive_mask/controller.py": "328f636201a845ebdc211340a9014a5076fa895e10cf9c2e1f91b96761a2514b",
-    "source/vpp2/src/vpp2/adaptive_mask/data.py": "e3b633fedaa503b101c80610cbed0933e9e245acf56a6c0692c4d99620faf84c",
-    "source/vpp2/src/vpp2/adaptive_mask/equivalence.py": "8658cb69d549fa0575570364fe3aa1e0b121fcf8b9965b17cd174d459dddf5ab",
-    "source/vpp2/src/vpp2/adaptive_mask/execution.py": "efec52c9b2eabf0300c72edea18f9cd4cc16a0889496793abb6a50161f2328ef",
-    "source/vpp2/src/vpp2/adaptive_mask/export.py": "0ce51e74787d28bbe9f4e73a3774297b424aa131fa78968d24154ff56db98b0f",
-    "source/vpp2/src/vpp2/adaptive_mask/feedback.py": "519fa4e6416723fc000754815a2f05d3415299ae35724f37bb806ebc22393d38",
-    "source/vpp2/src/vpp2/adaptive_mask/geometry.py": "f1ce4110c3a3a893f4f240c49e2ccc4e93215ac4ba9e447de8042f9b19afab98",
-    "source/vpp2/src/vpp2/adaptive_mask/hidden.py": "968560bd6dae0fce59bd28084091b10d559cf165c81764f6ad24b7492101810d",
-    "source/vpp2/src/vpp2/adaptive_mask/losses.py": "4a7dea6f87ef0fa38a7202d85d98d37460f8f4dc13ca5336f32b1da8b84a4322",
-    "source/vpp2/src/vpp2/adaptive_mask/masking.py": "31b595916410c60f772fff1ab63e92c854c8fd0d2261e6b865fd05ce2c61b9c6",
-    "source/vpp2/src/vpp2/adaptive_mask/monitor.py": "48b71f499dd5a0a79557eeb918f5fb1a641dc8654e5d412aa628ca9c57b5dcf8",
-    "source/vpp2/src/vpp2/adaptive_mask/physical.py": "f54e8b4fae8026cc89f296d74bf4cbea06fa4e126905b572a22c7d8226f0238c",
-    "source/vpp2/src/vpp2/adaptive_mask/preflight.py": "b646e7ca785afc87356e3e1184e48f535792ebafc84936450fc83bfd3f9f5ea8",
-    "source/vpp2/src/vpp2/adaptive_mask/priority.py": "68ccff1a384081f0c9a82e8f54ad2c720293f8500e7440a469b5e8fac1ee608a",
-    "source/vpp2/src/vpp2/adaptive_mask/quality.py": "c3b3a2264ca0b4773805f4485ebfe91bb4ab861e226ff4bd27bd667ab372b59b",
-    "source/vpp2/src/vpp2/adaptive_mask/quality_policy.py": "f100b7d4a412592777d936a6cf3c73a9f787cc4a62d5f1b8a369fed74acf897f",
-    "source/vpp2/src/vpp2/adaptive_mask/reliability.py": "306b41720154810bd972fe4d9c4c388e6967539d076ddbedd3a432ac62d13104",
-    "source/vpp2/src/vpp2/adaptive_mask/report.py": "f71401e09297daed8bbb72cb3af3b887e5fdf5ec282f47637fbb01392d280171",
-    "source/vpp2/src/vpp2/adaptive_mask/runtime.py": "ed74bfedd147389ea01f3be8c2979fa06c7152326ccc29ee347175690cb9c6cc",
-    "source/vpp2/src/vpp2/adaptive_mask/state.py": "5e8f95021eeb35511f5481e4e91ae687fe605f9e88d17b3b98d377cc967c17e1",
-    "source/vpp2/src/vpp2/adaptive_mask/train.py": "ad95349bfb242bcecd7ddf82fc58b31d63f2938a2101f9560bbc9e305f5f5b20",
-    "source/vpp2/src/vpp2/benchmark/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/benchmark/audit.py": "8587be371653c38b9926aa0ace006a244f270e0ef0020147a7bf5b0fc85d4538",
-    "source/vpp2/src/vpp2/benchmark/packed.py": "5047fd02f3fcc68e41f543fe83bc9202769a05794d898524d6d25e308f2b5125",
-    "source/vpp2/src/vpp2/benchmark/scheduler_io.py": "d6df06259549faed2da9f2888b540eac8e1ff2895c339eea58b4ea6a7e3b3cb0",
-    "source/vpp2/src/vpp2/checkpoint_compat.py": "462c5f93812483b5871633b200bc2ffecb9c201a2caf6b14f8f36dbe95fba50a",
-    "source/vpp2/src/vpp2/cli.py": "54c19a0f7e25b53bbf56ec13b0f537c9d7b81282e1e80575bddfb470172164e3",
-    "source/vpp2/src/vpp2/compact_kv/__init__.py": "f34f2f77f68399f45f6656b562a5c331f2504672dc145df368e086532191f766",
-    "source/vpp2/src/vpp2/compact_kv/blocks.py": "662c1f35055c2be2e2a14f2d92615d046f5e454879e34af39e3c7aec3af34098",
-    "source/vpp2/src/vpp2/compact_kv/checkpointing.py": "3c33c156e96f26ceb22eee61cfb8b18c3cb9560db0ab405cfd4c46efe5aba353",
-    "source/vpp2/src/vpp2/compact_kv/eval_prepare.py": "bdb33444056245e3cc773b0849b4db608b454cd24a7ac5b752f1ba26ff978865",
-    "source/vpp2/src/vpp2/compact_kv/export_check.py": "e6cba70b47027882a34d587aba0da5f9371128ec2db2282b91ad2d49e40065b9",
-    "source/vpp2/src/vpp2/compact_kv/gpu_check.py": "fedb6a9ab69b51080161ff2a4feae3f85262cc50f932f48e8decf04c1d147a20",
-    "source/vpp2/src/vpp2/compact_kv/losses.py": "914437ebf3ad91fd2cffa893a65ecb04ab01f2b337658a781a472feeb50c0a06",
-    "source/vpp2/src/vpp2/compact_kv/model.py": "332ebaebf543d40a34d188ebbe81055ec3e4bcff164bfe90d15aa6cef0748b8d",
-    "source/vpp2/src/vpp2/compact_kv/prepare.py": "7b01764cad0fafc0cc9441952b2e928048b0a43a6105dbd0eb64c349f4b9bf1f",
-    "source/vpp2/src/vpp2/compact_kv/provenance.py": "103e0f7100be56233aa86041ba9bc427ccb0db4fe045c4e6b8d3481ab6bb42da",
-    "source/vpp2/src/vpp2/compact_kv/reporting.py": "0b9ff602fc3425d1b9c122a77da3a426814e11c9cbdcb6b9a75a2a48130087a7",
-    "source/vpp2/src/vpp2/compact_kv/sampling.py": "2e8d82704f07853a7b487d5aaedef1b2f1fcfc5247e9100a56ed21d69262eae2",
-    "source/vpp2/src/vpp2/compact_kv/spec.py": "9bc920fc7d4398e424a0bde38ce97f87473aa417cf68b48e2c292a59491bc7d7",
-    "source/vpp2/src/vpp2/compact_kv/train.py": "a9e4410c5cde5df002d9808554538c29b77ed9e2d2bedc9ac5f8465eed70cc3b",
-    "source/vpp2/src/vpp2/compact_kv/validation.py": "73b62d02ce1d518cba7b9462e85c3c8d272ae9974e1161b1b6adea6f0040e2b5",
-    "source/vpp2/src/vpp2/convert_robodojo.py": "7ec2d6bfffbb4f5891ee33a4bf630cf27b17628506aa3cee1c46b3050027b290",
-    "source/vpp2/src/vpp2/critical_windows/__init__.py": "22721b77e25ec4874c4853d2f2991a14dff6aaa665da500f2830826b06545478",
-    "source/vpp2/src/vpp2/critical_windows/contracts.py": "ded4a239f7599785d74ff57ff892c5aa74f0280d6759460c93942e91f505d47a",
-    "source/vpp2/src/vpp2/critical_windows/coverage.py": "6c18bf995aacf413bbe4536eaeed92ab4b2f15f9eaaaa03a2730256ea7922ce4",
-    "source/vpp2/src/vpp2/critical_windows/inference.py": "df301bd187b27d20b6623febcb5e85815e1b7d4a1e9524a8a65bdf5904bf7232",
-    "source/vpp2/src/vpp2/critical_windows/metrics.py": "ae2bab145555d55696ef479758a5bae6780527fd3085a4f46d035b667ff32736",
-    "source/vpp2/src/vpp2/critical_windows/packets.py": "8f7d2c411407d95b1ff4d0e5b9055589f6114b1bb1020eeb7e1b106f330b853a",
-    "source/vpp2/src/vpp2/critical_windows/prepare_staged.py": "694cb5699de2d9f5ce1bc4667627b47d6bf0698715b4aad86907139859caffe5",
-    "source/vpp2/src/vpp2/critical_windows/reselect.py": "8d0e31ca1c17906d74de6439276e49ff7ea78abfc66fa556d8a306070b3dbf4c",
-    "source/vpp2/src/vpp2/critical_windows/score.py": "32daceb9a3fd957d9be75823b4d519629bc60a64d1da62f465e172cb1ec98d55",
-    "source/vpp2/src/vpp2/critical_windows/selection.py": "a7c35f6719d3308bfb738965d17544a15985c6d684bb47eac7272d6155d96786",
-    "source/vpp2/src/vpp2/critical_windows/staged_review.py": "d5cfb3f49ec5acc6914da98f5413d1549b2695e7c85c9aed9c0db83dee6be754",
-    "source/vpp2/src/vpp2/data.py": "8574e1ae3c6ca6e4b84633ee1815f2b3f2ea17c1e55f12fda0b00c5c5a6a7b53",
-    "source/vpp2/src/vpp2/datasets/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/dataset_utils.py": "e1a1471b0aa70917eaf699f85a9acf3ac70d9fb6eb9ebaf02d2bf2f218d6ab2f",
-    "source/vpp2/src/vpp2/datasets/lerobot/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/base_lerobot_dataset.py": "729fb3fd4ca531212eadb4a4e9c7e9f97ad34cca091d66380acd6b179f0c29c6",
-    "source/vpp2/src/vpp2/datasets/lerobot/constants.py": "11ac9260341e1560c5eb0dd7acb79ed63c74180b8ab4113c27622ff213db417f",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_dataset.py": "bfe75bb9593ec646236777d90d9fa309fa0e1188db07e07a3be0810edbe3da56",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_sampling.py": "266b0b7b023f014f4ccd935b6015b0a05b4659f372264fdefd997ea9692b244a",
-    "source/vpp2/src/vpp2/datasets/lerobot/episode_instruction.py": "74767839a813c7902861f4561101541c282b72e261df1fd28fe8d06c91231b69",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/compute_stats.py": "7d7daab219332b5abdfd5ec67ccaf3502e8eb583a17a4e482c3f3d4f7a691e89",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/utils.py": "4fdb4ac0653201e6743f2161c485213429948c1e1d5b068d03b4d97803254006",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/video_utils.py": "4f66c410ae6aa5a98acfc3320f3dd7b017d0285117054f14226f9e5d82395087",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/lerobot_dataset.py": "5659e992010bb6674ec50ab0d33b28a8e9dab0ecc895c199662af4c40fe56629",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/base_processor.py": "d3327cbe398d77761df960014d41e70d349bb0fa42bca874333b9ee970222447",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/vpp2_processor.py": "43c5604cbfc315b7377316cb97c9a6f3322124404397abfd01c6c3d0090cd414",
-    "source/vpp2/src/vpp2/datasets/lerobot/robot_video_dataset.py": "8eabfbd649116b0b4d6c9704c0df34febbf8875f6ab863dd75f4107dbbcdef73",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/action_state_merger.py": "39fd44f254bd7fb71b9269d75d8f9b5a5bbbe167ea438dda92d40eda1d256734",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/image.py": "f494d262ce29770e959bdff60b997f7a07e5f00d10ef515eee7be813d1db038f",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/rotation.py": "feb59d26539be6cfdb86488f0c10c11488ed73901658bc657e67a755ddb9e67e",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/normalizer.py": "8207d813b698a91dfdbe6910a4aa1fe0e775a2a8edf5c05b2d7a42591e9f8adb",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/rotation.py": "8ea2c72565707f5979967a1abeb5edf1eaa2c3fec55a7b50e640de869fc9fbcf",
-    "source/vpp2/src/vpp2/datasets/lerobot/video_event_dataset.py": "c20d3f5e2bc05e1b7ed94ae38872883a976a179e09b7fa051e94f94e2836c02d",
-    "source/vpp2/src/vpp2/enhance_prompt.py": "f31b3190c24bad073c5e3121ded4f564232729566e8c1fcc3e76fe75b28faf45",
-    "source/vpp2/src/vpp2/evaluation.py": "9b76882149e3c3a8c417a0e23514af1bb305c01bf0b6d21fff1f70bc9843f55a",
-    "source/vpp2/src/vpp2/flexible_mask/__init__.py": "64bed1808976d1f3c708154c36a97288a75abc6463d49d7f0065cb8360a2fb16",
-    "source/vpp2/src/vpp2/flexible_mask/calibration_batch.py": "dc5d83fc9e2bb595b9e3aeb1923115ebb9345283290a7c3f95c79fdeef7d3e76",
-    "source/vpp2/src/vpp2/flexible_mask/checkpoint.py": "c8531727e2388ec3d7a9770a504ceda119afcf03d64b2462befcd45a48949f9d",
-    "source/vpp2/src/vpp2/flexible_mask/coefficient.py": "07a3cc8d8e6c8b60119b9f94ae85074da017b08158fa0a0912cc9051b6df8ef7",
-    "source/vpp2/src/vpp2/flexible_mask/config.py": "4e1e9b289e293145b4e895c8ef7347bc837d01c929e05fd5cbf65bdb3c20be5b",
-    "source/vpp2/src/vpp2/flexible_mask/continuation.py": "2016463b4efe01b5d0a9ee84f03c3d95ee16f8e24f230faa465598cc978da29b",
-    "source/vpp2/src/vpp2/flexible_mask/controller.py": "263f1c9091e0c4318aeefb61e91549ba905ff046a625e35e1b18c94da4b4668d",
-    "source/vpp2/src/vpp2/flexible_mask/data.py": "d245bafd6f04f33ae7cef7c01d5db876a1a2a1a138c4de8a139d3ea78095521c",
-    "source/vpp2/src/vpp2/flexible_mask/diagnose.py": "faa29a0fa57c66af3b2e817640fdedfddc998bf878357f484415b244555ae4a5",
-    "source/vpp2/src/vpp2/flexible_mask/export.py": "3064dfd82bb55a145c16c77bbe17244970d07d18e944b793f4b7368dbcede764",
-    "source/vpp2/src/vpp2/flexible_mask/gradient_probe.py": "6fe3f13540cb9d8dd20c0b6429268ca74625dfad1e41806412c0f980fac76509",
-    "source/vpp2/src/vpp2/flexible_mask/losses.py": "df264b71365885ea90329c0b0facc92db3d2eb6e04324ea90a4701a1cd457abf",
-    "source/vpp2/src/vpp2/flexible_mask/metrics.py": "e976dd066e4b5e7e7ae79db77b9d1be64d17c212c61d0bc113b87a048e8f2b6f",
-    "source/vpp2/src/vpp2/flexible_mask/performance.py": "95b3977fd739cc98f7de94b14fb490a7c132021e742e7179dc5b5dddd6557c68",
-    "source/vpp2/src/vpp2/flexible_mask/preflight.py": "f794b997a492911c4d77656310c89d3b85e54983b3b889ba9904be64fb961e43",
-    "source/vpp2/src/vpp2/flexible_mask/prepare.py": "d6cf82e27555ecc13973727f4b71ce7e40558fef7d95805665fc7e458a42fad4",
-    "source/vpp2/src/vpp2/flexible_mask/pressure.py": "6cf74edcf3625fce9eec689d70841c6c1481f7c1abd87e5bf70bb38eb34632a2",
-    "source/vpp2/src/vpp2/flexible_mask/recover.py": "418b2d5ecdd08b08686b2a07bbeb0f969471eb73c04ac6952dcfbcda3b123f30",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_best.py": "917d924c5708584a0192050a6d8210809e905938aaf877dd81cff3be287ea289",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_checkpointing.py": "6f6a61c2ee7076c6b0977d1acb489fee7df04322d7f6b13bee9bb8229be319df",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_export.py": "c50433aa1b7e528fb23c514dddb5b9197f44c662829838ecda106980ab0dd54d",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_handoff.py": "aa590f707738e4b7c7250f901e29795c7441430f961bfa764437cf34bdc514b1",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_kv_decay.py": "bf83761798deea4f89c32d14f43bbfb0f16f751b3230962456c073b843dcd08a",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_subset.py": "7f7154cab6d910b747a0189742ee625983d08f13de50ca60bd4154f61df2b0d3",
-    "source/vpp2/src/vpp2/flexible_mask/report.py": "5b49283d472bace3a6ad56ce568af04f2c3d04add1a3920ad5006fe98374bce6",
-    "source/vpp2/src/vpp2/flexible_mask/rng.py": "01680b5b7e8a29ce92e96d756e12ceb80bb0e0ce1ebb642a14fa2a3a34776daf",
-    "source/vpp2/src/vpp2/flexible_mask/runtime.py": "373a3ecbc2ab6592a359daf6adfd8c958cb657f393e698ca2cb834f06750be4e",
-    "source/vpp2/src/vpp2/flexible_mask/select.py": "b547857bad61ae7ebd25b888fba73c2ca8f50a8eb31551ca54a909c75c8984be",
-    "source/vpp2/src/vpp2/flexible_mask/selection.py": "b9453c8735051b54ae7dc8610d2ca332eae8ca1a3e099aaadaf5ab75bfb0f8b0",
-    "source/vpp2/src/vpp2/flexible_mask/speed_review.py": "9d7099d1e173026274eb439169df2ad4812ec1d56a5f0b21242f972cd835ffd2",
-    "source/vpp2/src/vpp2/flexible_mask/throughput.py": "ce21bdfda51ae10d1470ac537d910d9f46f60a38910888808ffb3f5a80174d58",
-    "source/vpp2/src/vpp2/flexible_mask/train.py": "86625613bbc692722c6d2d24278a586f88ce0d76263a9daf92562a9002a8e037",
-    "source/vpp2/src/vpp2/flexible_mask/verify_export.py": "fb68e62b0da98e6f96cfa694bd061162bbeec1763e5787e057a1cbac5b2ce8f6",
-    "source/vpp2/src/vpp2/full_mask/__init__.py": "27c9ede277380e8d64586371f2a03334e68ab2d3bfe516d95e8cb96099827929",
-    "source/vpp2/src/vpp2/full_mask/calibration.py": "5d97e0385eba82252e12cc1562d1a77da2f2bce29332d758c1b5b6b84c5bb47a",
-    "source/vpp2/src/vpp2/full_mask/conditioning_cache.py": "07adf95165db4c3c1c469695d6b7b6fe4cf4fa9aff0a3d360f4229bffe9be17e",
-    "source/vpp2/src/vpp2/full_mask/controller.py": "0e6a969b89a4275352cc5a1e00bdfe8e3f8754a6ebd7f9266b92179cbe6bdd10",
-    "source/vpp2/src/vpp2/full_mask/cpu_contracts.py": "1a46a53a68d2d8a202c5f98d9a78f6d5f5041dada0c2239ed81e7916a28f0758",
-    "source/vpp2/src/vpp2/full_mask/curation.py": "4ffd7b0212bdbca1daf73ba7c24423523437012150fc044a43bbb33cce747545",
-    "source/vpp2/src/vpp2/full_mask/curation_v3.py": "9dba386d66869de0fba321ade79b2692d47f111ccfebc7f367194cef7a4b4e8c",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_checks.py": "688ea4e7e44c4ecc1a3cd81cf682e2cbf611d9ad2d52744ba1308324da70febc",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_features.py": "11011a089f1586045e2972cf267773aaf4602cf08c8b0e755bbf017ede1e6e43",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_finalize.py": "a2a7812eb6c670686a7a4f4adff0a2c6363dbc078c71d948537f1c5fd546254f",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_probe.py": "2755e9cabd0fe02e70cb8330c083af881a5e4696697d58cf7aa01bd00ef57f6d",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_quality.py": "b9170398762112a23bb133dde53c6a49193c671c767647239c8b0ff15159530c",
-    "source/vpp2/src/vpp2/full_mask/data.py": "668a7c5f88a3ab17854eb0df240c974082f4aa64604cf11d3143b18337f73181",
-    "source/vpp2/src/vpp2/full_mask/distributed_smoke.py": "f699d4f330309f29a537fbabb61c60f61eb3322d7304605351dba3d2acb48e8e",
-    "source/vpp2/src/vpp2/full_mask/gpu_prepare.py": "92ce96dc4c1b6d6ebc22d99935472c15d86ba2a7c050e31640d2c32e08748ec5",
-    "source/vpp2/src/vpp2/full_mask/monitor.py": "2724a934d1f96444227c854f75ef84df6025c02485aa68eca34f3b28d509fe19",
-    "source/vpp2/src/vpp2/full_mask/native_monitor.py": "5572712a340f17692a669d2a4c41d1d691164925eb7bd06a56fff1dc2afd69fa",
-    "source/vpp2/src/vpp2/full_mask/panel.py": "2844b663340cbfb54f7622f3447a7831e2809050739a52dc076fd6ae32ed9e46",
-    "source/vpp2/src/vpp2/full_mask/review_decisions.py": "ed1184bcda7b5bb3fb11bb4262c4461c408e3d3ac0d23e80d2f8cafcf39c7e18",
-    "source/vpp2/src/vpp2/full_mask/review_report.py": "752ba9e8a3caf5b8cefa2d18e6fe14860e6f998f57c03af0940af0943b09a9b1",
-    "source/vpp2/src/vpp2/full_mask/runtime.py": "ae30b4d3fd3c888bef17672376dc3b34491cb296a1883959674ad465579efe7b",
-    "source/vpp2/src/vpp2/full_mask/search_quality.py": "ee3238d3366799e97d3f9a90c11f579ff8f51d5bd0642f52c37d9dcdbcf02e6e",
-    "source/vpp2/src/vpp2/full_mask/train.py": "f4127e6580522d4d9e5a50db10d201526e948d4d3479ddfbbc12e60ec61c7501",
-    "source/vpp2/src/vpp2/full_mask/verify_data.py": "72ee60ba2016d5f73eb2ccb01b666a5dac37138620efc4a79a20607df3658fb3",
-    "source/vpp2/src/vpp2/inference.py": "42b3057a85aaacbb87440bc740f13b09b9c82cd51d8c5af4e080716aabdcf38e",
-    "source/vpp2/src/vpp2/learned_width_kv/__init__.py": "0c59cb8424b6cd64508a005625e9edf63c84d24618a5d2caf6e6fa5d63cc859e",
-    "source/vpp2/src/vpp2/learned_width_kv/calibrate.py": "b89bbf103b49fd85f85f8c300a7868bb42ae7a13f29bd9878dbd5a11061ae51d",
-    "source/vpp2/src/vpp2/learned_width_kv/checkpointing.py": "7d30e46b95d91df8aff088346ae75809958622b3a8e8d953db4f30adfbaebadd",
-    "source/vpp2/src/vpp2/learned_width_kv/common.py": "52dff98c6317b34df2a13313a641998acf4f9e18ac31ea4e8e25227afa14620d",
-    "source/vpp2/src/vpp2/learned_width_kv/controller.py": "7b5a8793a540e1e7464185229189f4f1f1acadcd7cf3c2bf477bda5de7d18787",
-    "source/vpp2/src/vpp2/learned_width_kv/data_scope.py": "0ddde6805f646afbf5f5966c53530f829becd083802f0e0a8697dc1004e2003e",
-    "source/vpp2/src/vpp2/learned_width_kv/distributed_check.py": "f89f153a20730a6bc3c70133479f34918d30c5a48775bb3f6c14bb25701801ce",
-    "source/vpp2/src/vpp2/learned_width_kv/gradient_probe.py": "3bed964f6488e95b386cc58058e7f28aff50c9a98dd526f3365891a9049a6fab",
-    "source/vpp2/src/vpp2/learned_width_kv/lifecycle.py": "798afa47ec5dcc8749982da2d7853eff184e450a32d51bcf70eb4578f2bb0b09",
-    "source/vpp2/src/vpp2/learned_width_kv/masking.py": "3272ea09271130dcb3cab6935f6bb7d61bda66b9982c3baabcd61ee0efd914a0",
-    "source/vpp2/src/vpp2/learned_width_kv/performance.py": "a2fd9597c51bf0cf5c435b48e8f57792df24d2723a6d7b6b841d6fa24f17ab2f",
-    "source/vpp2/src/vpp2/learned_width_kv/reliability.py": "f74399fb67427093b3e4c7df9749379f9291b7ac58bd6632db226b6e78fdbfd0",
-    "source/vpp2/src/vpp2/learned_width_kv/reporting.py": "f22c1d8806a5b7757bb6dd99b15153d0305b2eeb0790bb37763337690d642036",
-    "source/vpp2/src/vpp2/learned_width_kv/runtime.py": "513da1e06ba9509164f017e0b8dbf6b219d7e1cff82edb2d2fd8d2d7839e666f",
-    "source/vpp2/src/vpp2/learned_width_kv/train.py": "5813ebd4fefec6d3a8cead9e0c645d10734f5ec88c4d74a5742587f2eb4de6c3",
-    "source/vpp2/src/vpp2/learned_width_kv/validation.py": "d5594ce875c8011bff12d3c77083dca68878178ed022b8079761267d04a82386",
-    "source/vpp2/src/vpp2/learned_width_kv/verification.py": "1bc43d48ef0ec3056af614adf43ca0744d94b1e77fe970168fef87c7cf2e6fe8",
-    "source/vpp2/src/vpp2/libero/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/libero/action_model.py": "178fbe4cf02b9a2e133ef31389b6bb2f9baf0504bd0818a265b6bff32ebc3914",
-    "source/vpp2/src/vpp2/libero/cli.py": "fcf68a88e2db0a9b786c6199d9b719802fe4b84514741658b49e2b06e2473e45",
-    "source/vpp2/src/vpp2/libero/data.py": "19b699381e7d6fa769b5090c7125ae01b7584576e969468660ee65b1585ef5f1",
-    "source/vpp2/src/vpp2/libero/diagnostics.py": "9540e31dfd2def06bec942ffe73262278700a17a340e14cb5bd41a295706e147",
-    "source/vpp2/src/vpp2/libero/eval_helpers.py": "ac338d4ade195ceb3a87620d162befa9679e2d7ed3d91b09927e5d59d86d97f4",
-    "source/vpp2/src/vpp2/libero/eval_libero_single.py": "87e6b457dc412ef4fe5d01aa48b1a6d20a28dbce3152848042e2103769e4efef",
-    "source/vpp2/src/vpp2/libero/eval_libero_task_manifest.py": "f0c721a7ddc768edddc8b0e065a4d8aeec35e1a546c88deebd9097ad93f6227b",
-    "source/vpp2/src/vpp2/libero/evaluate.py": "24216f9814f8304e077d72ff8d8c778fd588239313b34cd7d23b960c8f924c67",
-    "source/vpp2/src/vpp2/libero/libero_utils.py": "c7cc23976c86eabcada4846d46345d3605fc2ce03929761646486be3901697d4",
-    "source/vpp2/src/vpp2/libero/prepare_pro.py": "0999050352d924c11d29b22f63e902162d8cfb23f94815c93c01295c11ed81ae",
-    "source/vpp2/src/vpp2/libero/pro_alias.py": "90b5c8c8061bc727a6e413a8746fcc98cbd06e1c7c4175fd223b45af4f10b700",
-    "source/vpp2/src/vpp2/libero/pro_runtime.py": "adf94ed64b032aaeb0d88391e7b653f88f67e76b9411419ece25faa31ccc1340",
-    "source/vpp2/src/vpp2/libero/protocols.py": "68225c95eeeee25032966a7f90c1ba18067c53225397f59a463d764030fd0d93",
-    "source/vpp2/src/vpp2/libero/runtime.py": "1e2e2900a604d33cbdf8774cfa33dcee1065f21dc0b59fcf30799b5098c1662a",
-    "source/vpp2/src/vpp2/libero/trainer.py": "ea29b4e000047a6b0c8aee3cb6f32df9187987a657924085dbafd9628315107b",
-    "source/vpp2/src/vpp2/libero/worker.py": "9df6350c4ab92dea70863c08f1bdcebd0d77574a3a2d8d7ac9899a5a825be6b0",
-    "source/vpp2/src/vpp2/models/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/action_dit.py": "9a654edd860ce77c1308923bd338c7f14d09f68686642531d7143c16b9f6f331",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/compact_channels.py": "6f7ad787624844970985117002ab809bd777f7474969daca121e7101381c6d0c",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/gradient.py": "594249b6eb931048e352fd3ffef9b1ac1db9e4fc74659e04216d6b823dd174f3",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/io.py": "1b68cf2f2fd674c7a59140f8632d4464044bffc2e9277ef5f9a975e7c699966a",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/loader.py": "b9180b9c9177af192e5cb92387b4a09bd0dbc08fb75a94e843b67d315c94e339",
-    "source/vpp2/src/vpp2/models/wan21_14b/mot.py": "cc14a909bfb181a441bc9d8d6de12f3c31d89f3dc3415031e34a5f8c21410019",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/scheduler_continuous.py": "15de04574c1d8ccb794ca6560b1e501e5c83cc86104bb80f1eb3cc3b0324ffe0",
-    "source/vpp2/src/vpp2/models/wan21_14b/vpp2.py": "8c419af905a8ffa93d77c827a283b7aeff744f3a461c390783eef8a333c31756",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan21.py": "01b45cc71f69a6914f7b286559317c8e8485d0b0fec93bb28ebc37c183c295f4",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_clip.py": "f65e44c9d1b7811d2996b3fa22e6fe4c2ac5e6edf1fc2c14251cf582bf8d1f05",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_dit.py": "0d1b5837ea923508c05394d56aaa465813e7b4f5aea3b31ceceae39b2de9b0f9",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_text_encoder.py": "902a1e46b21b5c48e166fe03defabb94b3773c6bbe836fc09b9a0d0894eb1f6e",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_vae.py": "33f7d804677b44a6f3ab334684017dcdb1e9dfb748279de76ad94a5bb71a7e51",
-    "source/vpp2/src/vpp2/policy.py": "2a7696cd73ccb99575bf901185cf578c3775945f61f6e67aeb2b4ec6bfc4e30b",
-    "source/vpp2/src/vpp2/preflight.py": "7a91e486675a4cdf38565e0f4b58e60a68bba7de5654cfff736b5df58f5c53a6",
-    "source/vpp2/src/vpp2/robodojo.py": "183db869a622ef17ee92c487949db2e22a7d1365558fe703bfb8dd8fef83e9c6",
-    "source/vpp2/src/vpp2/runtime.py": "e26885a3732f5fdac3d8381da870a1673d0c03354f02f9deee7fe779d645f8a7",
-    "source/vpp2/src/vpp2/trainer.py": "feaa0ac4ed94a376836eb60cfe6f52460392b11f7cafafec9c6d9f151f5b61f1",
-    "source/vpp2/src/vpp2/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/utils/conditioning.py": "43d43f0f3ca0add64498c3d64a3791513a53e3fce0ea53f77ef09580a86b7b2d",
-    "source/vpp2/src/vpp2/utils/fs.py": "69e6df236b43360428a613686c8f37c531c0d16a3a2c5b06a725c46b49d657c5",
-    "source/vpp2/src/vpp2/utils/logging_config.py": "1de4abf66d9d201b49e7bff48291643b9dc5ec9b6e7918ca73e1bd18887b1c43",
-    "source/vpp2/src/vpp2/utils/lr_continuation.py": "72ea9f2af52041db19aab681d7a99fdf6d87d5ed5486f2499383a211c6c1ab76",
-    "source/vpp2/src/vpp2/utils/misc.py": "7e42c330c24604e118f1878473dc804aa8a272944df502ef1414c05be4d5bba2",
-    "source/vpp2/src/vpp2/utils/pytorch_utils.py": "9b0ba19de9a64458b12a0fb24f0811f49d9e9fca5d8b815f5f5ecd3b7886fd63",
-    "source/vpp2/src/vpp2/utils/samplers.py": "4d3bd9b2d498f4ad8f2cf9765d60c4bf5368fa6fa2a7d42d0db6210f66899652",
-    "source/vpp2/src/vpp2/utils/training_profiler.py": "91f6d98577e63d85e474c738a9bc1901edc194bdf318b5387d527c31b6708832",
-    "source/vpp2/src/vpp2/utils/video_caption.py": "d9572deb3690aecb0f16c052016cedc5f94da8ce8f016b4dff1a4e358e18d5e5",
-    "source/vpp2/src/vpp2/utils/video_io.py": "9e9898b021a44e87aa1b4e6d13e3755a9869a2e27a7eb53c72a6802808d5bd80",
-    "source/vpp2/src/vpp2/utils/video_metrics.py": "1840a37685d686336c7326a2a3d4fbe048d1e8e7e431fb14519d360a8b7aaae8",
-    "source/vpp2/src/vpp2/video_training.py": "20f55bf8b0a7dba6422a1115be1c8adf31a0db6a9f53a8f6f53f5d970188bfb2",
-    "source/vpp2/src/vpp2/weight_resize.py": "269edd4b9d32a1c7f363ee082dde7ce57f2176190daab6d5bf4a1c98ab43f0b4",
-    "source/vpp2/src/vpp2/weights.py": "5dbdff347de1b57e30433efca6da28334d0263d27aaa9b2278d95614db1b8641",
-    "source/vpp2/src/vpp2/width_kv/__init__.py": "0183a7106da7e45223550a552db2905d6b59c2a4259f8c86cb98629b03e66b43",
-    "source/vpp2/src/vpp2/width_kv/calibrate.py": "93a06910cb23a9e1dbd961ee9ef9794d70aaaa6186b08483260c8215109598f5",
-    "source/vpp2/src/vpp2/width_kv/common.py": "45f5f7087f5f3dc067358838e3fc738698b796defecb2ea721545d275f767766",
-    "source/vpp2/src/vpp2/width_kv/diagnostics.py": "cf184ae58f5e1c5f4453774eee1e10c85931e618bbbf768ab42f649cf301f5f1",
-    "source/vpp2/src/vpp2/width_kv/reproducibility.py": "89c5377e1314dd6e8de7f67c4385d30f5bd91fa4a5742449f1c1ff8dd578f451",
-    "source/vpp2/src/vpp2/width_kv/scoring.py": "7644358d0b2d29d3b91515fcea56090dbd2a7a0dad2d5e5c88e8d585bff15533",
-    "source/vpp2/src/vpp2/width_kv/slicing.py": "856203c14e0b06d19585a96392d00a00703779ebdb9dbdc9ab1323d50daf8ba4"
-  },
-  "suite": "long_horizon",
-  "update": 1000,
-  "validation_gt_action_mse": 0.0005726480308242441,
-  "video_parameters": 12468409664
-}
diff --git a/evaluation/memory-protocol.json b/evaluation/memory-protocol.json
deleted file mode 100644
index 0d250765619a8a895fdabcbb0c916ff33a1d9c7d..0000000000000000000000000000000000000000
--- a/evaluation/memory-protocol.json
+++ /dev/null
@@ -1,707 +0,0 @@
-{
-  "active_policies": [
-    "vpp2"
-  ],
-  "caveat": "Same checkpoint55 policy/simulator code, benchmark layouts, seeds, r16 and new90/old10. Simulator physics are not claimed bitwise deterministic. Offline ranking used replan24 without ensemble.",
-  "checkpoint_files": {
-    "COMPLETE.json": {
-      "mtime_ns": 1793716093276368990,
-      "sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-      "size": 2491
-    },
-    "action.pt": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3",
-      "size": 4057916577
-    },
-    "compact_architecture.json": {
-      "mtime_ns": 1793716074811787315,
-      "sha256": "a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32",
-      "size": 6563270
-    },
-    "dataset_stats.json": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
-      "size": 100256
-    },
-    "validation.json": {
-      "mtime_ns": 1793716074812787292,
-      "sha256": "9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e",
-      "size": 88781
-    },
-    "video.pt": {
-      "mtime_ns": 1793716059522133613,
-      "sha256": "fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb",
-      "size": 24937214042
-    }
-  },
-  "checkpoint_receipt_sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-  "execution_order": "one report task at a time; eight shared GPU workers",
-  "fresh_policy_per_episode": true,
-  "fresh_simulator_per_episode": true,
-  "gpu_sharing": {
-    "minimum_free_gpu_mib": 61440,
-    "mode": "shared-shard",
-    "one_worker_per_assigned_gpu": true
-  },
-  "id": "35ef508554aee3ce8b89ddc3826fb6eda60bd022db79d6d91a60f6f37e7014f8",
-  "inherited_results": false,
-  "input_hashes": {
-    "inputs/episodes.json": "b961cfc82459c2a394922903cc6d275aaf647d8cb4561ebafc7dc259bc73928b",
-    "inputs/master-episodes.json": "b961cfc82459c2a394922903cc6d275aaf647d8cb4561ebafc7dc259bc73928b",
-    "inputs/support-assets.json": "57c577018fddc4e95cfb79b45d526b77a87da916cb6a4117aec5aada2d4dae58",
-    "inputs/vpp2-policy.json": "9d7a376622b8eae0a9b1d45f8cdf3f606ab6ba59490fa4c63f56353a7f8e8ef8"
-  },
-  "native_inference": {
-    "horizon": 32,
-    "replan": 16,
-    "seed": 1,
-    "shift": 1,
-    "steps": 10,
-    "temporal_ensemble": {
-      "enabled": true,
-      "new_weight": 0.9,
-      "old_weight": 0.1
-    }
-  },
-  "native_scoring": true,
-  "order": [
-    "vpp2"
-  ],
-  "parent_evaluation": "/home/praveen/experiments/vpp2-recovery1000-950-precision-20261103/update_001000",
-  "parent_protocol": "266df4b5d9b8eac16145600a8c4f9223d15e03285d61473c3bbd8c43b694f211",
-  "primary_episodes": 300,
-  "repetitions": 1,
-  "schema": "checkpoint1000-followup-v1",
-  "scope": true,
-  "selection_manifest": "/data/praveen/vpp2-recovery1000-followup-20261104/inputs/memory.json",
-  "selection_metric": "task_macro_ground_truth_native_action_mse",
-  "source_changes": [
-    "native_eval/scope.py"
-  ],
-  "source_hashes": {
-    "control/simulator_recovery.py": "fcee066ec9e75ec081b0a955b136e2ab898cd1ee529a2c1e74a453c350c05367",
-    "native_eval/__init__.py": "2deca42c85edcf3d80dba72439ad95641937396e27e1149d77182cfc082267cd",
-    "native_eval/body_coverage.py": "50ce610f86584965281893ff2f98a78643c7e3207014212aad4c13fce4c8a782",
-    "native_eval/camera_group_cleanup.py": "8df69fdba913ec41ff7b1256ab1b43921a0b6fbef7ffbb6dacbf2adaba0f7843",
-    "native_eval/camera_lifecycle.py": "b0952d95d1da8a1bd3e9f544a2eb1eb0f516159b3442666c7acd9bbaee71687a",
-    "native_eval/check.py": "9587d7b200969ccb78ddf750dd3d8616d5e2245e29baa3a0d798e27a55bdaa5e",
-    "native_eval/compact_worker.py": "129a2e26a85e021f7a0282201dec976104a8e1ea1f0a67e6b75ce760257ee46e",
-    "native_eval/completion_integrity.py": "3b737f6dcf9ae9908b1a3602f66fda46c3b8399ee47b1c61f5c0fb9741031029",
-    "native_eval/criteria.py": "97ebc94d91c62d793e614f178d2d120df989faafb54881dd1d3f5d703df8bb76",
-    "native_eval/diagnostics.py": "53b0deb571984fd02ff65c400d5f3b9bc2754b8207ea8edf4e5017c3588ed3cd",
-    "native_eval/ensemble_check.py": "0bef52411b2ec2dd7e1ad21b8690b1f88ed229107b6e46b8dc43c34ca92a3572",
-    "native_eval/evidence.py": "cf7f769cfd838227b26101d7656ebcaede1b1d45a80785a32efb26c9befef0e9",
-    "native_eval/frame_epoch.py": "7b1d3b10f46d5acf82f6e983ecd5f8995310b2fef9fee04bfae168fda02e3efc",
-    "native_eval/hardware.py": "e30ed3359d91595d3053e24124445498a3f1be9d5c0d52f97e1e7d7d88268a5f",
-    "native_eval/hydra_frame.py": "c1aefee86396025b405e050597af7b67bb07c0dd1e65ea4a9be1ed78f8d99072",
-    "native_eval/particle_evidence.py": "378d667b2ab80176045e0bdcdbe3cb39adb612a356b8f6b81d13ae531bde711f",
-    "native_eval/policy.py": "e1cebbfb03c137bd4bf9f054c69be74f9a7f1df5f7ca13c250e606c2346b55b0",
-    "native_eval/prepare.py": "486df6d062098ea07ac63c4d3be51b5f27baf524d0d004411275243783d02ac3",
-    "native_eval/process_lifetime.py": "7dd20bee91c51546b6da067b96012dc5d5a013e8389dafd567b72c1f742cf97d",
-    "native_eval/provenance.py": "8fc2da1678832be9e9b8b778ff0dd1604db5d464cf1a19ff6137acf6b83d3e11",
-    "native_eval/receipts.py": "9ba5800e6e7e796937992117569efe1b3d625900a8e2ebf7955940a4e8cc81e0",
-    "native_eval/replay_receipts.py": "1c60f9bddccf7245e32df97b9e90749bc3ddbe5d65f37ed2d40ba339b458017e",
-    "native_eval/replay_server.py": "1c2c19bae8d1cd3ff564f1775540718c33293663d30161ea7bd293b8f322cab3",
-    "native_eval/report.py": "9d9f0ddc794d2af2ad6e349f991450592a6fe98db292e1ac3ae4fcadea961fc4",
-    "native_eval/repro.py": "f0e4ecd0ba009774105cebeba31f49a70f34c4f88f11855c939368027789a928",
-    "native_eval/reset_coverage.py": "3c6a414878c4cc2ef4114170356db0a6b2193b03ce5c3b4f6293c10c7d3138b0",
-    "native_eval/run.py": "23d2d58fe1160a03bdcf9591b5d5fae671d21832008cc4958ac58e060b88caa1",
-    "native_eval/run_one.py": "dddeea1abb269293bd08c3f5a69ea0a668620b18d07154ca1c0740a0bb213b99",
-    "native_eval/scope.py": "f27b394ac956aaa512f815e13c94a044e434eedc10a493812b4cb99591038b40",
-    "native_eval/simulator.py": "d40972be52b7544f62653114217e213d19d4b386bf2991abfbbbc618ab09cf83",
-    "native_eval/status.py": "b49d03681dbc28b80f39f4d6163c3a6e7c5a1ca62f1f0ef502c77dfbf149ca02",
-    "native_eval/topology.py": "062e9246890ad1f2e8bb8b64b3bbe6598f9dce6fcc33ec3eaa106f3cbc0f0ae5",
-    "native_eval/transport.py": "eb6cc295d064a142372dd51859e91233f61f1ad2c8ac8bf3880739f3fe4abec4",
-    "native_eval/vpp2_pilot.py": "571ca3eeba11fbf2afd4a4058629a280bc0ad4bd3eecca178ddde4aa543bd8de",
-    "source/simulator/LICENSE": "7794bb06af8fe5485ca912454ad1665ccd7846c45f9c2b1258658e480948cdbe",
-    "source/simulator/XPolicyLab/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
-    "source/simulator/XPolicyLab/XPolicyLab.py": "a3909558697b73479f8baee03e15b4d74060a7c9ac2f31fcbc6cd659e7364862",
-    "source/simulator/XPolicyLab/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/client_server/__init__.py": "093e47e707f4a5598b0159b5ee371cd9a918b802c70a6e66154a18e9ef8fc5db",
-    "source/simulator/XPolicyLab/client_server/ws/__init__.py": "79905987a12ebf318dc0b2b9ec6f375d9b4561bac29d6c27d04e318d8f6e00b3",
-    "source/simulator/XPolicyLab/client_server/ws/__main__.py": "ee1c5807dcdf1c8f6721c5447463eb51babdb0f9ffbd91d1e7343834ca4aa19a",
-    "source/simulator/XPolicyLab/client_server/ws/model_client.py": "4d6e225ae5f577d862366350dca8742e91dd9527d6e76b8b2a6d48c991212d67",
-    "source/simulator/XPolicyLab/client_server/ws/model_server.py": "4988823b436cd05e6a4d8b13012cc30c7ffd57711b72497b1223a6b10dc0c043",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/__init__.py": "19aeae375e94e58d437d789d0aff07b232c13f375235384d784984631d25d5d9",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/client.py": "b309f58b1bddd99e56578026b61a15e6cfdc07e843e8a082bfd291d0e853ce5a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/codec.py": "8a97ef7e102fd954c266352d81eaceef0f1b0f5504571c03037683efe387865a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/exceptions.py": "73161f4cdcc7ffc8a87cec94f5459bfcb9113076be272d6cb2a3ff8d69e0c6ce",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/keepalive.py": "7f6b9e636eefcd6c80a4b62e404a6f89e2f7e1a3b6008500312a972d6990a21d",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/messages.py": "d05c8cbc20e0cbe1502551864907f25897035993f920748c9290c944c959eab7",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/schemas.py": "f1ca82eafa8dcad6d968508adb51c25342b291f7b1c553fff305c381a6098bd0",
-    "source/simulator/XPolicyLab/model_template.py": "4f8d1eeb8073aa7ae6093abf0f395cd5fef70ff4a4bd1fa6b2d83b356a90754a",
-    "source/simulator/XPolicyLab/policy/VPP2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.py": "ab94a8569e6168ab7cde6660890035e11b34961ff2648854a7b61d188c555b6e",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.yml": "d2c3f06eb2ffa2d1866767dc20db8d60e542fc2418dbc4cf86d4a0f86f944c5e",
-    "source/simulator/XPolicyLab/policy/VPP2/model.py": "699acc0c459c5997db4e3155e2b1b7eb64bf750552436074a369c4eb6e9ffef9",
-    "source/simulator/XPolicyLab/policy/VPP2/temporal_ensemble.py": "42be4b09c56377f4686b2e722de760fa3bafd1c6b92605ada269d94e19efc95e",
-    "source/simulator/XPolicyLab/policy/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.py": "42dc75140935046dfe83f19b3834bd7cf8372e1bdbe51f598b1f91b8d557b7fb",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.yml": "e59de6edeee891d2aa24902ae9034e9c10889fb20116bc9263f30d773a4b0d10",
-    "source/simulator/XPolicyLab/utils/get_free_port.sh": "f45bcb327b45c91367e3e5cdff72d3a42bfb7a09f50e4774701637c4c23fc1a4",
-    "source/simulator/XPolicyLab/utils/load_file.py": "488590d76084f419a36c2fb3e658eeb240ae554678899e9f2ce0cc10da5c4b57",
-    "source/simulator/XPolicyLab/utils/process_data.py": "e044e627eb2677907e9227eda8458c857208913bd484c5e922b53025010eef0b",
-    "source/simulator/XPolicyLab/utils/robot/_robot_info.json": "8a5df36e750971003b484ad770a7a243e523e986d2977a4ecdbeecd4a07946bf",
-    "source/simulator/XPolicyLab/utils/wait_for_policy_server.sh": "f75a8360d83f2960f365dae77804c94a28aca332c2fbc62cdb4a39ae220374ad",
-    "source/simulator/env/camera_manager/camera_manager.py": "1a4250cb3bdc7305fe362a7c6b113f57e6f3e9e4983b69bd61e2c84e8e995471",
-    "source/simulator/env/camera_manager/capture/camera_view.py": "be9ec0d4df8a532c35adc7090fbd7d6a76c163f90e8dac30496ee4f875929869",
-    "source/simulator/env/camera_manager/capture/render_sync.py": "48d70e9a95f3445ae464ab7ea00a00c0d4829752eb55c4ce52ef73ab1f453cae",
-    "source/simulator/env/camera_manager/capture/tiled_capture_manager.py": "959b4d87d96e676bc279d01e7755557bb8c6e73920a75944259f72b7ac385f8b",
-    "source/simulator/env/description_manager/desc_manager.py": "bb566028b523a4cc27363cee2e526c2d3e008e671c9773d4f0b8401f58f0388e",
-    "source/simulator/env/environment/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/env/environment/base_env.py": "ade58ebc19c79bdbf720def30786a0c397b1c7d2141fb6a1bfd63c80f8a3db8e",
-    "source/simulator/env/environment/isaac/__init__.py": "123023ddc75a0f9b1139c7322bee64a27f0e756b6e84242f43a1481428596126",
-    "source/simulator/env/environment/isaac/direct_rl_env.py": "77958f14c465932697efed302c8ef13054f2ec61ee3a4bc5ad66d0390b71df77",
-    "source/simulator/env/environment/isaac/isaac_rl_env.py": "144af40adc2113751bc1bd72caadab145e918b316de53ef3e4e42358c01a1fbb",
-    "source/simulator/env/environment/task_env.py": "ac816cf26efabbb8ac4276d2a2a09fbd557e3db805514db0c3dd7ac22b58e931",
-    "source/simulator/env/global_configs.py": "e96f6141b42e635d69bba5b4727fb1fec54f8965431b3a70d95fe5dcd3d0ee45",
-    "source/simulator/env/observation_manager/obs_manager.py": "771d6cda443b591f3e7850b901a06b4e653a2feefcbabd23608042377fe9be4b",
-    "source/simulator/env/planner_manager/.gitignore": "bdcb46018aafd15f5c2d40adbbccbb06f2dec5b9ec4311cbf776075bdffd4156",
-    "source/simulator/env/planner_manager/curobo_planner.py": "3cd830c002901959409ef6f64d54106d8106b48471b266409abc2199f3cb33e1",
-    "source/simulator/env/reward_manager/func_parser.py": "eb928a05c3eb6bae429a922c59e64272c2c3851c22d779143d45627e47b3a504",
-    "source/simulator/env/reward_manager/reward_manager.py": "bc56efd3aaa861b4b38960dc4fb7778a9b4da9cf131f815ac3170300dd233a41",
-    "source/simulator/env/robot_manager/.gitignore": "fd94daa58076711ae84376170199a63f66177471e587a76252399fe79eb231e2",
-    "source/simulator/env/robot_manager/control_manager.py": "81b7ecc3bed8248c16aa1517e686682d8e28ccfa9e550d45011b579a9219caab",
-    "source/simulator/env/robot_manager/frozen_robot_configs/franka.yml": "a3c944f8cdd07fc1a4eb54cb915f565763292acb24f9c4f725534271449f138c",
-    "source/simulator/env/robot_manager/frozen_robot_configs/x5.yml": "4df90d234656c34132ca573e69a55de5984270e103aac2ea8ab9956a913685b5",
-    "source/simulator/env/robot_manager/robot_class/franka.py": "dc49eb7738a8b58fb731595340c60a1ce32f52ef0dc63a3da84f9c25879af140",
-    "source/simulator/env/robot_manager/robot_class/x5.py": "4748cbe4d273449a48d09817767dec361f53f8e5cdc6e43368fbb8724141b654",
-    "source/simulator/env/robot_manager/robot_config/franka.py": "8bf57a07913573e74dad97aec6c50e58b67746fa90875915d74ce213856ce35a",
-    "source/simulator/env/robot_manager/robot_config/x5.py": "13d35630f833d1b0a1e3df0bea05f843e88aae1248b59af8aadfefa9cad29bbb",
-    "source/simulator/env/robot_manager/robot_manager.py": "19636ff932a0e2192341b57d8695d4102b870bd10c69c9343d712682fcc15082",
-    "source/simulator/env/scene_manager/layout_manager.py": "1b4c940200dd94a1d0300e103d168b9a823e0dc00ce7c921afe16e81d4e01ddc",
-    "source/simulator/env/scene_manager/objects/articulation.py": "0c542a1fc8ba7a61bf71b343745c71949c863291c1b167d4d7048fe2ef9ca4c4",
-    "source/simulator/env/scene_manager/objects/background.py": "04afea9b5c4e21a105ba3b08ce0e5b6692c082cf17ec1053fdffd5dba1c9511b",
-    "source/simulator/env/scene_manager/objects/dynamic.py": "7a4652c649f503681914e37c50c959c7ca8457918bb00628413f4288993f8ea7",
-    "source/simulator/env/scene_manager/objects/fluid.py": "12b183d4d798d6b7188fd42747ee1927747318c386a91135551d1fa0561c41b5",
-    "source/simulator/env/scene_manager/objects/garment.py": "85f38bd85b52da8dd2ce34ce3692d45848016a9f0f14ecf13114705d4a1ff46a",
-    "source/simulator/env/scene_manager/objects/geometry.py": "3100d61b152b894b3b59be0b3be1d2708f2d8f363ddb8311dabfb5411c731f3d",
-    "source/simulator/env/scene_manager/objects/ground.py": "12dd51addae6f5b7c6ad410b4ae7248231d216cf63513c564057d577a09a2972",
-    "source/simulator/env/scene_manager/objects/light.py": "ba8c8529c6608592d4d0eb9638820b70ae6b3c0cd616631bc6dabdb1c20e421a",
-    "source/simulator/env/scene_manager/objects/physics_material.py": "a1ac0dc9928e9253e644ceb9a604aed262514aaad0473aff303bfae448e76c29",
-    "source/simulator/env/scene_manager/objects/primitives.py": "848f4ab08d86307812ebaea18cc86a51638395f0e6ea194b412d45a699ced926",
-    "source/simulator/env/scene_manager/objects/rigid.py": "8260fd2d8350d6a354c8281685bc7e25531923327078752efe0538d9cd78450b",
-    "source/simulator/env/scene_manager/objects/room.py": "946aa1afc23ab9b8fd17394517a5f54605bbbc609582d0854fed2078b3b2d8ee",
-    "source/simulator/env/scene_manager/objects/table.py": "c0e71396b0d53a250f74c75490e91086d67962237400a49f886a13ec07875308",
-    "source/simulator/env/scene_manager/scene_manager.py": "37911b12601038c9ee9bd180cae7a4751ea649bcdaf1ba2628f22c65cfd7dc72",
-    "source/simulator/env/seed_manager/seed_manager.py": "99bcf5a90392c8927aa0dbd799c07e06b9b03befe8b8eabaa8744a3dc9605ba4",
-    "source/simulator/env/seeding.py": "3f88d2a635c528253439cb31a29e5b9d0f47f6aaef0e975d86d9a8cf174be26b",
-    "source/simulator/env_cfg/arx_x5.yml": "7691350159fdeca92fd1b0b2da36913464ccdea3f3251eefb0176aa51db0807f",
-    "source/simulator/env_cfg/camera/camera_config.yml": "c1752170da855e735beba27fddbf78dad16f2395e4ef8b24e9827a768e616a62",
-    "source/simulator/env_cfg/camera/template.py": "68287ed1daec3b8ed7f053bf97a4d48bbddeb8b5757b53e438ac3c4240a9e61b",
-    "source/simulator/env_cfg/robot/_robot_info.json": "4c9de628bc1572502fc48d876c023fc63e02760e1a11a2a5258dea30ffa330e8",
-    "source/simulator/env_cfg/robot/dual_x5.yml": "4eefaccc75c5f83443b7bb85ef1de68ee7a3772e27af2b04a5e8d8769ab9baa1",
-    "source/simulator/env_cfg/robot/dual_x5_and_franka_competition.yml": "ec893f9fde577d0d1209e7dcb85ec2dd74288b636d0c68316cb07bc817b743ab",
-    "source/simulator/env_cfg/scene/conveyor.yml": "c1dfd27faf978f90dfc4825cd83bd203823547c46793f4f9201ce9f4b43beee7",
-    "source/simulator/env_cfg/scene/default.yml": "729df024b9707a55449da94f980c1ca116035e09a1284e2fe071f1b38efd061a",
-    "source/simulator/env_cfg/sim/sim_config.yml": "e442ede1c3d90fcfb50b9461c854d31b22dad9bfbb23397db228b27a6a486963",
-    "source/simulator/scripts/eval_policy.sh": "786966648496b7b68aad81e000a3f92910302eb391dbd07fec0a2dc91107956e",
-    "source/simulator/scripts/internal/summarize_result.py": "9ec361af587e1788d83d3cf8670e05b817ff252d33c33829a6337dc16ae370d2",
-    "source/simulator/scripts/internal/task_inventory.py": "ec82252b0f87fa5a3dbec2f682c170b25cbe47fe181e431a81df2c4b0675084d",
-    "source/simulator/src/__init__.py": "75bfbedb80f2e1421ad787992d78c23349bc8738b9edef7a76a3af83c9586fd6",
-    "source/simulator/src/eval_client/eval_env.py": "c577b27dab3c87588940670da81bcff007cd64b687eee8e0cc6684e5e24966b1",
-    "source/simulator/src/eval_client/main.py": "de2216bb5df707e505f7e2f3d31702907a4c1ab65205c5ea35494021f168c80e",
-    "source/simulator/src/eval_client/physx_warning_monitor.py": "6a9f8e8d75b1031d2cf2feb6dd72d37370d85d5bd80c4bea798622e23ce8dae6",
-    "source/simulator/task/RoboDojo/config/_task.yml": "655a87cd59fa6272bf352d32deb8ade162737b49a9824e16d428fecf0803380b",
-    "source/simulator/task/RoboDojo/config/align_blocks.yml": "082c9d7a32968052b8ce954bee212c7ec0a96b756150701d4b531abdea59be2b",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number.yml": "3bf6b7e286ea328fb1e4a994a8d30af9afb4b0b4dcd7ceafed43cc6160c9978f",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number_random.yml": "145b26eb5c8ed7d084a4ddc6e21694c542f4a86150756188ea38dcf66e565603",
-    "source/simulator/task/RoboDojo/config/build_tower.yml": "e1bab8756bbc5664e3171e2389c060a79a3fbbf92a6af4ef555dcaf59fce275c",
-    "source/simulator/task/RoboDojo/config/classify_objects.yml": "80a754d2aedd6a89f58185eae36f5b2eb1a28250f40a67325f576fde34650384",
-    "source/simulator/task/RoboDojo/config/classify_objects_by_language.yml": "ac6c739ca82a64b53c799af79b44363db1a88246f76d867f8830729e5613507b",
-    "source/simulator/task/RoboDojo/config/cover_blocks.yml": "68f90c85fad2b152cf7853cf48888d09ce74bb27a692ff2feacd11160771c5a7",
-    "source/simulator/task/RoboDojo/config/deposit_coin.yml": "f547b12216fb9a5dfe2e7d7e882d89e1e53ee1a805c943ba09a3c2cdea8c17a3",
-    "source/simulator/task/RoboDojo/config/fasten_screws.yml": "5fbe4cf5b3ee2eafc2771137b13176f01617f8b2131f02503e676a7b62e2eabf",
-    "source/simulator/task/RoboDojo/config/fill_egg_holder.yml": "d79129278c9a5488d63bb1d3865f2c6b39f21f85957ec94ace091af9ae29e31f",
-    "source/simulator/task/RoboDojo/config/fill_pen_holder.yml": "ac72e5582495e8757c547997b317e2fe990370009527b1faa6bfbf4e1275537c",
-    "source/simulator/task/RoboDojo/config/fold_clothes.yml": "56447e3f592e0db388ebb7482ec758ecd5485ec81b9227c2dffe5390bd7d8c8e",
-    "source/simulator/task/RoboDojo/config/fold_clothes_random.yml": "eafe52c0abcc1889069383b008cbe95c6af2d18ceafb1aecc78f0fae3b491b04",
-    "source/simulator/task/RoboDojo/config/general_pickup.yml": "35de68117cb0ed596b30db1c62f1f9090f2cd1237242a03d5aaeb7186e791d93",
-    "source/simulator/task/RoboDojo/config/hang_mugs.yml": "0086d374d3f7e82abb85e43a17cdfa437e1ac645a0171fb47496c2b81427dce9",
-    "source/simulator/task/RoboDojo/config/hang_mugs_random.yml": "847dfd39059275a671b5beff0243bf7bbaeebd9e5c0612a1877cdf12b9007dc6",
-    "source/simulator/task/RoboDojo/config/imitate_sorting_sequence.yml": "d9e194d8c742288601c641ccb75052a67da46e536c0b06c7fb7311d9e2c61a42",
-    "source/simulator/task/RoboDojo/config/insert_key.yml": "c0b409d5c020e0ad706c378b6f63b39dfe7bed4682d978907c40e1cfa7623335",
-    "source/simulator/task/RoboDojo/config/insert_tubes.yml": "5c070e5c764a2ea93f8de0b52e6a664a23c717963fe09fe3af17918a4a1ee9ef",
-    "source/simulator/task/RoboDojo/config/make_kong.yml": "8a33c9e57b6e6c719c4800911c37272ba055ad0f4cda2f1de9d531ee2910351a",
-    "source/simulator/task/RoboDojo/config/make_toast.yml": "d190a26233db174538ff0052b4c2958a3b107f0ccc63d0fdfe28da4c7f5dc57b",
-    "source/simulator/task/RoboDojo/config/make_toast_random.yml": "ca2c1e55a0975c8b019523710e90131504d8d2afeb829be362e4d7927fe97b48",
-    "source/simulator/task/RoboDojo/config/match_and_pick_from_conveyor.yml": "0ae1c2537d9b1880748316928cfd7ba4c6421ecccfedb85c4c5395a7a86c5361",
-    "source/simulator/task/RoboDojo/config/organize_table.yml": "9ffa15dd279060d4f042285a2c53a341c7208e2f89a2f3dd36ee08b22aa453fa",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box.yml": "e232962b8394d08c04c5da14deb88d7b233afa00ab99af8001c084218951d7ca",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box_random.yml": "7ac7f034fcb169e33a7f0a03b2b77e54d0206ed715409359b579055a701196ef",
-    "source/simulator/task/RoboDojo/config/pick_from_conveyor_by_image.yml": "6b310b70a0ead9d09e7c221c6bbc952140a44b1d43034613bf02e954eebcf9dd",
-    "source/simulator/task/RoboDojo/config/play_Xylophone.yml": "28adedb0f03c49693af32cea37855f3f95ec3f76f031bfd571fb8bfffb2ec59c",
-    "source/simulator/task/RoboDojo/config/play_stacking_toy.yml": "a09afc106161064a2ea6a40882d717f7872bd4875309a97431d6863a9b7ffa1c",
-    "source/simulator/task/RoboDojo/config/play_tic_tac_toe.yml": "63f863a7fb81a040f65803c64f97c4e9577d4fe32cd0f92563bdc72d1aa336d2",
-    "source/simulator/task/RoboDojo/config/plug_in_charger.yml": "2046b47112049e038ecfa24a8c8044fb1db84f629a4a32eeb24822157d2d4dcb",
-    "source/simulator/task/RoboDojo/config/pour_balls_into_vase.yml": "084a707af61d24123912777b417fe0c0b230eb9653e686af42570791e924ff8b",
-    "source/simulator/task/RoboDojo/config/pour_by_language.yml": "efb2e3c3cb08dd6d2d2ae011b3e9a218b5f1f0a92f025b8e3b9e8abd7011c250",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup.yml": "32092b11667b3306579b04225303610c061585ace28531fa9996afc339c5a2ba",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup_random.yml": "1603feeab6f0f7dc9c4897e7f40f322b86112b6ab987dad0598f42ac1b35db69",
-    "source/simulator/task/RoboDojo/config/press_by_number.yml": "b784643495bcbca933f5a5fbc91b7dad3465fe888239b358d516977c78c38f63",
-    "source/simulator/task/RoboDojo/config/push_T.yml": "b55ea11002ff3e5dc63982ca8c1247fd7bee4f4a607f4d674e220b86cf120a25",
-    "source/simulator/task/RoboDojo/config/push_T_random.yml": "07609ec20cea40afa1d5720e895fb1b445411f261b2e5b6850e75d0f1930d32d",
-    "source/simulator/task/RoboDojo/config/put_bottles_into_dustbin.yml": "7eddf43e287b6e8aea686c6d686fb87e13ac1be1cb5aac16a8a2b0d2e9eb976e",
-    "source/simulator/task/RoboDojo/config/solve_equation.yml": "bdaad2ff1cc8b7ec6413a3715d49ea9cefd664e7e204544977832e18785f1e75",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size.yml": "8927cb2ebb18449554d851e629bd7d568e12e46c5c3c636c8e7122f4fab1ed1f",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size_random.yml": "f895d3ce7eaae99454706afccd09011f56d077998bcc1e95840116658b78f139",
-    "source/simulator/task/RoboDojo/config/stack_blocks.yml": "2802e88ebc3e2d3d05ce345d75e9ad6151a6690e4ef1e0021a5223d291a402d2",
-    "source/simulator/task/RoboDojo/config/stack_blocks_by_language.yml": "4fbfed4f7fc49120831087e709a8dbdfdb4504eb143196000d5790b54eb3bc40",
-    "source/simulator/task/RoboDojo/config/stack_blocks_random.yml": "e25d2a096601da30abdcfe1256a464eaf48c0c8f844eb2c737232626d49639d5",
-    "source/simulator/task/RoboDojo/config/stack_bowls.yml": "4e2aaa12b8b0f58ca801f1a93638c46d3b49ebad809c5a8011382ccda846dcb1",
-    "source/simulator/task/RoboDojo/config/stack_bowls_random.yml": "5c08af1bcc925d958d53de51e8ebd336006203f1bc81e4ef68f3719eadfabb49",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones.yml": "0fdaab803a9370a9e2d1fc8d86a5788daa3940c08fd6f767334f625642fca3e2",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones_random.yml": "93f0ca39bc8bb32df94e84a02f22ed4d7860f37cdc0b9dd17e44721445849576",
-    "source/simulator/task/RoboDojo/config/store_tools_in_toolbox.yml": "641c8ef92393bd986e408d26a84b1cfcdac27466a5ac5f0b3ebe4bb84430c527",
-    "source/simulator/task/RoboDojo/config/swap_T.yml": "1daf769911e13714b4aeabd208292931e32f4dd39968893152bb7bb477c46c9b",
-    "source/simulator/task/RoboDojo/config/swap_blocks.yml": "b6bb2a9df494719f26c6b1aab2e7576107ae76e2fbac8b63a5770f21e0e9f05f",
-    "source/simulator/task/RoboDojo/config/sweep_blocks.yml": "98aa25ac39be9d335c71b2b586bbba372789e9d61e652e1147152c87febbf1f5",
-    "source/simulator/task/RoboDojo/config/sweep_blocks_random.yml": "4ebff5e2cad07f958dc4b2793b86b1aa803c0809394818ee3752d300e58b6187",
-    "source/simulator/task/RoboDojo/task_registry.py": "11218028f772d9253ee709a7d85aa72d1e1fd87e4f8d33ee3787d6064438f9ab",
-    "source/simulator/task/RoboDojo/tasks/align_blocks.py": "c8597e96132cd7c9d30f3ae101f7bbe8ac28597b629599d75d442bbea4c6e74c",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number.py": "277acd20ae4345dc71267762e588b7b0343387c71685183890f044a4808b7a20",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number_random.py": "01757509978663bf7c0c95059e8ae18d45b8fe0953f6b54540e2bd6bf141182e",
-    "source/simulator/task/RoboDojo/tasks/build_tower.py": "814cf6ffb459871e5cf49f4e127903dbfbc30926f33db71de0b84a71ab235d6e",
-    "source/simulator/task/RoboDojo/tasks/classify_objects.py": "aef4333c088766375e21a928d53a9a93aaee285fecae41e22385acdc74dd11ac",
-    "source/simulator/task/RoboDojo/tasks/classify_objects_by_language.py": "242a8f048f487eb174ef246be78613bbb5f153549ee11a2650412718cf3b36c9",
-    "source/simulator/task/RoboDojo/tasks/cover_blocks.py": "9d14931b16842b93bea8abf9a07991a93a7e663b93c5fd1695108c8b922ab862",
-    "source/simulator/task/RoboDojo/tasks/deposit_coin.py": "f9868ef827ec17b8db6c4650cc5d9a9e593d20d00ee822e2212ce229ec01fb61",
-    "source/simulator/task/RoboDojo/tasks/fasten_screws.py": "2e187a28e33e44efb87baae6463b3ec35264ccd41e35c21a1be44f6361675185",
-    "source/simulator/task/RoboDojo/tasks/fill_egg_holder.py": "506dfb772f8882ae53966e3580affde19c3c0f11591c78cc5f5b1d997ba04f7e",
-    "source/simulator/task/RoboDojo/tasks/fill_pen_holder.py": "8706dce239487e078b21af0b95bcc5cd4eabdcdadcdcbc6400d6b138d982f1e6",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes.py": "ecea9c287d0579605330168e6f36734704fb14a6b1e5dec82b6f2ee3a56ec20a",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes_random.py": "9e40abe3f7201a8c13961edc1ca8075260c315584d3d6b93fa1df966d13d1e88",
-    "source/simulator/task/RoboDojo/tasks/general_pickup.py": "1a9eef7fe518f9d2c6805ad1a8f97c1045fa83726f84b9c63f67ffdfabbc40a7",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs.py": "4bc6586bd87de030b9c580fea2a45e7977c3b6a4d75cf3c43b546d349d42e906",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs_random.py": "ebfa4a95deb954aa7a9063a07fe366cb5ea1835d3f52342b10d531225dfeee92",
-    "source/simulator/task/RoboDojo/tasks/imitate_sorting_sequence.py": "18400cb3c5f5103492819df2dcdf05a77916106e8a503cb62b4ee6cdbaa9305b",
-    "source/simulator/task/RoboDojo/tasks/insert_key.py": "180978a2915744f0596fbca4ea0ad6dcb47dce005d43f06807249bc8b4d1a828",
-    "source/simulator/task/RoboDojo/tasks/insert_tubes.py": "9a68795ddbd0b816ffd7627e673390ecd83594c0e9d374d4e3abf3435ad439e1",
-    "source/simulator/task/RoboDojo/tasks/make_kong.py": "b9266689d65f9213b05219ad952344f9fd63be8892097ec0103af99791b245d5",
-    "source/simulator/task/RoboDojo/tasks/make_toast.py": "1047f484a73d1f92304393879b51f87e33d661b6ad7937198352fe37a0d3233d",
-    "source/simulator/task/RoboDojo/tasks/make_toast_random.py": "69fb879b89e98f7369971834eb6d1b95c5a65f5ffe978b2cffae0b12c6b756ca",
-    "source/simulator/task/RoboDojo/tasks/match_and_pick_from_conveyor.py": "53110b8a0c971bd3ae8878bfb7d0dcd9626bd9b610d4e7fc91772313b435a5c0",
-    "source/simulator/task/RoboDojo/tasks/organize_table.py": "aceb298e78f3f39e4cf9a9f4c5a4f747b2c55f1a6006de2065e110a05f4295d6",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box.py": "84a1cf1dd5e166f36291130f9d8cb5dfe4b3f4732a46674bef6374a7ea2d66be",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box_random.py": "4db259a1b97cdcc39ebc9794244fadc79d31c4ae64069c9b11a233356fcb655c",
-    "source/simulator/task/RoboDojo/tasks/pick_from_conveyor_by_image.py": "a0d29576d0aab8232d2fd3ebcc8c510e81c70b4cc04602a5be3e99ee2275e01d",
-    "source/simulator/task/RoboDojo/tasks/play_Xylophone.py": "b7a4b0bc0457bcf55d714ca1867476fc4aabc53084adfc78d3ff5d1437bf7847",
-    "source/simulator/task/RoboDojo/tasks/play_stacking_toy.py": "23361329ee7efe330da1f1c241415205bafed2b6c082a913659d868a80d338c3",
-    "source/simulator/task/RoboDojo/tasks/play_tic_tac_toe.py": "8f75ebfe55c997ad9bd3b342db46e843e87397db94356f31cadcbe0cb483bca7",
-    "source/simulator/task/RoboDojo/tasks/plug_in_charger.py": "aaa59ade46dc54f430e1148e3f6fae487970cd29eb2c07b5ff7b4045edb831b6",
-    "source/simulator/task/RoboDojo/tasks/pour_balls_into_vase.py": "eae388b93ec196c401f76bbab3b0119cb2164f90bece0225d4fb925d3f2b73ab",
-    "source/simulator/task/RoboDojo/tasks/pour_by_language.py": "f7a15a7fca2c649a7918d68960ede27e667f409a339b668dd9876db890724cfb",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup.py": "73f8c91b57c352fdc1e7ba3fe5354b3c10e4156d8e5ba777762ad0fa1aff5663",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup_random.py": "a1c4097b6e34b2c62885b3a9c7d2584a93d9e296dc41d311a80c2e6d541d53af",
-    "source/simulator/task/RoboDojo/tasks/press_by_number.py": "006f969087c00c417a63875a9de153f698c52d1f1ca0caf0aa16076aa15059c5",
-    "source/simulator/task/RoboDojo/tasks/push_T.py": "a01d753bc320ffbf492ccc29e5f7ab3b39e6c7d255cd25804db1934a30fa5b41",
-    "source/simulator/task/RoboDojo/tasks/push_T_random.py": "dc8d9c8e0d2e8dfffa2b7d433a8278c0eaf707cc4861b15035364c2e70061dab",
-    "source/simulator/task/RoboDojo/tasks/put_bottles_into_dustbin.py": "3ba6e19df0c5ff19ba961092f031ce14e91e72a9a4a5934eece6000a1ffa1c48",
-    "source/simulator/task/RoboDojo/tasks/solve_equation.py": "e287a2f36b82b28e8ca7824f63b1b723b02972ea6dfc115ff32b2882729751f7",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size.py": "1838fa6b515bad3796895cb5c1883c6a42789b7dada5434765764600e70beb49",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size_random.py": "b793ab5bd32bd2c10bd46874440904837f9c4e290e7508b06426e2e96425709a",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks.py": "50bd388f9e47002966fbb0146742467a820295b8686e0f86379f0efbcfffcc1b",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_by_language.py": "36a83925f35c7a4caac5add328dd35cb1d346bad833d3064f455d88221e7a41e",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_random.py": "19d4e5cd693e068ad64de75cbcd24977ab8d39e22fa17362a391dcd733bc9bb5",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls.py": "022c129ec3493183b1a26688a0402f5a00a1546ae011999e9c19f4c0029e2b7d",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls_random.py": "a0e20469eb44a9ee98650d09386814c1df5264c68ec139f06aabf0c86f092514",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones.py": "2bee4e31f0594900629fe7f43266e16fecb310b152930ce51524507136809a92",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones_random.py": "58c3c5113054a379214d64ae406ed8dde99603cf684ce9cd8c7bcb72ed334382",
-    "source/simulator/task/RoboDojo/tasks/store_tools_in_toolbox.py": "2969fc3eb171560b3c0bcea62b9f38b8b35f3aecc9bcc7085c3b1cec365b48fc",
-    "source/simulator/task/RoboDojo/tasks/swap_T.py": "57fc289b3957caa454e1d8e005e3a57c973d6ab40121845c94799e61e4bfafdf",
-    "source/simulator/task/RoboDojo/tasks/swap_blocks.py": "08ed24ae9dff27f4e5c850161a02f62ce237859c096d9b06a853a7890e0ad4ed",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks.py": "65786217e199c1e2b26ec8e2c232bf98ecb304a20939d0908fe12bc41e990614",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks_random.py": "68b6127038322cf001d5f614060b8a1d96e64a49efddfcdcae91127d5c206bc5",
-    "source/simulator/utils/cluttered_generator.py": "f59cfcfed2b4cce5ae9aeaf3339c86d17e237874e3526bded3cb05136dbcbd50",
-    "source/simulator/utils/ensure_usd_path.py": "7c8e71cf91636e630b1b174098fe0b38740857daeaf207e2fae772c4727b11bf",
-    "source/simulator/utils/load_file.py": "ef3a6dc96aa1a7442cfd24f7be9671052a98d2f20783c3e71032332704132943",
-    "source/simulator/utils/path.py": "1d427c7cf33e748e01d0d083c9432d8be8b5c3a1f35c5fb0f6ebb0714098f1d1",
-    "source/simulator/utils/pipeline_utils.py": "d5d8da29e6ca84ab04323a2f79130f930572785b4399634e06a42834ed4087ae",
-    "source/simulator/utils/rotations.py": "8e814f79938bdb9e85f1c024825617cc08d281b61728c6410b23ef1bd852c340",
-    "source/simulator/utils/save_file.py": "6cff5950f73bf7a1ed404f804d6f08af712974f22c3955de4cdf142f211aaf00",
-    "source/simulator/utils/transformer.py": "73de4f5e1e0ff60d4f2e1aaf6104368e8f919269facb68b4fa0344d35bc08cbe",
-    "source/simulator/utils/usd_schema.py": "260d30b0ada2976136cad3e49e810b4409304eb4a2bc60bd9ed710864d3953ac",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/METADATA": "614985b278f601b8ef863f82783b991cc3ada4b89f592eb71c7c48c5f2ef2298",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/RECORD": "c7660eb223c8080f75c49690eaf0a9c9c00b6195511791b33b15e6b24e42edd7",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/WHEEL": "942926c567d0b1273d09f20295879abd37507845f6b683eab9f87d808df91100",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/licenses/LICENSE": "fe1049884b1a0d9342901e88e07f32925d24b3121d9972b6a6805fb9824b095d",
-    "source/transport_deps/annotated_types/__init__.py": "a7104a4d439b27a9f74fc0be236b9ba1b7831e6044026802a205abc1298a9bc8",
-    "source/transport_deps/annotated_types/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types/test_cases.py": "d867472acb57b81ce97f78972828e6bc5669ca4db33fd1e6f645c6183846068e",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/AUTHORS.md": "5fe0446ddd6d1c70fa298747368582dc1d91c23d62c8fb7228595bc87fd53232",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/LICENSE.md": "509aa7af19ae2b5b1681b22b969627ee9e1fd15ff7ea5bf9f6e301fc44d5a6c2",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/METADATA": "b7ee8fa1bf018747082cd6d2fcaa14e5d0f5956b5477ab5810eb61ac4cc7bcf9",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/RECORD": "51331727bf8d9ff60510de5e7756c5dd72aba7663f3673e2827bf1ca5c09c7e3",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/WHEEL": "1d7faff7e9e8524c94a31c99d4f312b92edab94c43011e1505da182ea0f4c70b",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/top_level.txt": "19da698ad1490bdde1c7f18fabaceb2625aaf48b364d4dd7ad404c7e5eed75b9",
-    "source/transport_deps/msgpack_numpy.py": "0a45a7430e2cefa2e74235c4c6a34a22ead2fb57a7de8b2631f45df977701d63",
-    "source/transport_deps/pydantic-2.12.3.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic-2.12.3.dist-info/METADATA": "348336c208743aafeabfa5537c29f429035a3107b7a78bbb6a3bf7b3a42fe4aa",
-    "source/transport_deps/pydantic-2.12.3.dist-info/RECORD": "089c7ceb033486c4e6106749bb1a38c80b63905da74d842d067735e8b9c86e4d",
-    "source/transport_deps/pydantic-2.12.3.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic-2.12.3.dist-info/WHEEL": "aad0b0a12256807936d52d4a6f88a1773236ae527564a688bab4e3fe780e8724",
-    "source/transport_deps/pydantic-2.12.3.dist-info/licenses/LICENSE": "a9e186f3ca16b5eef84318e7a701721351a00cb7b8ae3a4394b67b49e3529ef3",
-    "source/transport_deps/pydantic/__init__.py": "e62127278c07bf5384cdd2903f368a69929f3b8a524000bae4e0eb608ebf4bc6",
-    "source/transport_deps/pydantic/_internal/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/_internal/_config.py": "4d66708377346d91e24f76e8479f9862a928b877308d175e9e6c861e87d5ec4d",
-    "source/transport_deps/pydantic/_internal/_core_metadata.py": "63f836b778bbba5b8afb05c266fcc97d114c3d4336dda0582df69ee05cc13f2d",
-    "source/transport_deps/pydantic/_internal/_core_utils.py": "d63aeee156c9d31eb747a76d55cb8e23e74a4130bf77f9529c9584050cc63444",
-    "source/transport_deps/pydantic/_internal/_dataclasses.py": "4e4d6611a7e169dd6457b2b9b4f5f9070c5649763b0be30ac1fd0e2c58089440",
-    "source/transport_deps/pydantic/_internal/_decorators.py": "3e7c80a0a4a0dc135b0954999f0ab0f67684835503b58bc34fd2e5ba280f88ef",
-    "source/transport_deps/pydantic/_internal/_decorators_v1.py": "b5f75f76940a6384765c23b0a876d97a840cbabe95362811adf86e757047c77f",
-    "source/transport_deps/pydantic/_internal/_discriminated_union.py": "68c9744914b24321df5b8fa49673131cdbf0491a2a1371f73d157fd39bd1b138",
-    "source/transport_deps/pydantic/_internal/_docs_extraction.py": "7f2ce74801e1e40ce88675d94ad5741ef1fe9d16dabcd1cfca0fa49f1f92fc41",
-    "source/transport_deps/pydantic/_internal/_fields.py": "6127c42aadb51608cb27a62a6172a1d1e104b399f130fbd0ea1a7da40f0dcdfc",
-    "source/transport_deps/pydantic/_internal/_forward_ref.py": "e67dd8efedc028b9fcfc54b761cecabad2e23d48725e602d90464e6859e89f03",
-    "source/transport_deps/pydantic/_internal/_generate_schema.py": "4d3e3dbf3633a87f74ad6aefe69b4da19823cceb11d0a3e54a4a8f545ae7ac1c",
-    "source/transport_deps/pydantic/_internal/_generics.py": "10baa34fa2cccd0cd600ad04079ffdaa47bf8806b3cf4390e20ba7a7fb8ab986",
-    "source/transport_deps/pydantic/_internal/_git.py": "2303e1dc33df6b65eadeb06e07d371f25b91d80d62ebd41d7937d659722e0958",
-    "source/transport_deps/pydantic/_internal/_import_utils.py": "4d18710f93ae63a094a2c8a805d05c254b34a26ec820e36265d60057bcd0f233",
-    "source/transport_deps/pydantic/_internal/_internal_dataclass.py": "fdb79d7355dbbaeca044688b66a914930c05a506a847584a2e5479d359f2c926",
-    "source/transport_deps/pydantic/_internal/_known_annotated_metadata.py": "25ceca4cd159a01ddffb489b3ff36024834e795bd8137ab73930500e354c9375",
-    "source/transport_deps/pydantic/_internal/_mock_val_ser.py": "c2645115206fa9f70b6c8e353ec165881e2ed80677989a597a21116c3df1293a",
-    "source/transport_deps/pydantic/_internal/_model_construction.py": "c24f9b3460c026f76e686be7d14d3ff33125d06111bb4b2126f37cfd97e461ac",
-    "source/transport_deps/pydantic/_internal/_namespace_utils.py": "865dfe4d102bf365368d3c8f3f7b7e42ab2f28b8abc6d90b7cd7eb37e7e9d31f",
-    "source/transport_deps/pydantic/_internal/_repr.py": "8d07e726ec83c50a5244d846dbd208f4f5fc7b836fdaa599ac4c3696a8a1dd41",
-    "source/transport_deps/pydantic/_internal/_schema_gather.py": "54b12fe754d811e78d0f6733b32ae626ad4c56726a4ce9a72da9fb546e3873c0",
-    "source/transport_deps/pydantic/_internal/_schema_generation_shared.py": "17fadb41bae4a289a0c6cb24747a4fd23509ed309f7b40400c0124aba089e273",
-    "source/transport_deps/pydantic/_internal/_serializers.py": "6085af4a6011e5f9db19258239076e5add72078650638d9e02bb9cf9e9eb6fa7",
-    "source/transport_deps/pydantic/_internal/_signature.py": "f049633c97b8a529daa6e8ab1b90e4040803d618201f1100cb314f1fef47d331",
-    "source/transport_deps/pydantic/_internal/_typing_extra.py": "755c481a1b0be37cda79b3423108752b94235fa2386bd3959cd01c2f6a486bfc",
-    "source/transport_deps/pydantic/_internal/_utils.py": "0878c7fb4ce75235fd479b144a33d3f23f3efbb3f37bcca36a27ac7130152e24",
-    "source/transport_deps/pydantic/_internal/_validate_call.py": "3df7559d2ce15ceac436d683a03c373c558f5580f95bf80d60f7bca77520e8b8",
-    "source/transport_deps/pydantic/_internal/_validators.py": "76fd1ad8d91ce33718aafdf51a1fd021dda571ff96d24429574a1237380475f8",
-    "source/transport_deps/pydantic/_migration.py": "545ef72d1094cf722b6f9c55b75de36f7340717567105e93d7e27438b7de6790",
-    "source/transport_deps/pydantic/alias_generators.py": "28cd67deee097cb48197552e620de1a181f35c90fecaf82b9eaf2ed5c73087f0",
-    "source/transport_deps/pydantic/aliases.py": "be1087ca84969d7f8427ec166f9aa3e31c91b2c8065a74d07f3429e06499f6e8",
-    "source/transport_deps/pydantic/annotated_handlers.py": "59fc854aac28108157061ed3ef73f270a9682350e25f8e465a2d3e24eb024786",
-    "source/transport_deps/pydantic/class_validators.py": "8bf5778fe3d87462d22e68ff20965e913463ba73bc488573f0b325aae3f23fb1",
-    "source/transport_deps/pydantic/color.py": "033a867d540717ddbf65cb4372e7b40cce324e9d8fead7a49304483535ab2c8a",
-    "source/transport_deps/pydantic/config.py": "b83676c4df89ec255035e9c269529c62c244cc83a92aa9607d0bc3774d35c229",
-    "source/transport_deps/pydantic/dataclasses.py": "dbf8cdfb9d595abbb222c7a78cb235f9585de2a206345ddeb9ef6319c769a872",
-    "source/transport_deps/pydantic/datetime_parse.py": "402f9680cc4caffc10fe6357512ec055ffb684b121befb0f6353d0ca148639d9",
-    "source/transport_deps/pydantic/decorator.py": "617fa3500a6ee4029a55628fa1a57e9fee25ed46d2ebd184b7d45ade1b3398a2",
-    "source/transport_deps/pydantic/deprecated/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/deprecated/class_validators.py": "10071a550339ce9db0066974c9b37ad827d07f226f0cbc7941df4f938fed6f85",
-    "source/transport_deps/pydantic/deprecated/config.py": "93f96c564e7ba5ac4b24e701b9e1f4edcbb53a012059d541c66ea57da0b70825",
-    "source/transport_deps/pydantic/deprecated/copy_internals.py": "1a177ebe431de446020a0c821ad3ca3b9f27a7d704281402eaa9017881052368",
-    "source/transport_deps/pydantic/deprecated/decorator.py": "4c19ba6c9ef026c36287ff16ab92330dcc0fdf69bdfef7f1b3de9d7ac2ee934d",
-    "source/transport_deps/pydantic/deprecated/json.py": "1e55821b7e5146bc72cee4d2e8b4d089907046198365699ea901fcacb5bac00f",
-    "source/transport_deps/pydantic/deprecated/parse.py": "1b377a6ff83ccc95dcb84ed046ae5a761c7f10c25a8577dca57085d1182baaa2",
-    "source/transport_deps/pydantic/deprecated/tools.py": "36b9bda054595a9f3e8e57ef3e02482c4b32c29e18cd90f9d972063c32f11dc2",
-    "source/transport_deps/pydantic/env_settings.py": "e881de79612a9543d152fdd5f805e217f5bdd5f836270fccdced25df827f7b20",
-    "source/transport_deps/pydantic/error_wrappers.py": "44aea6a804dcf72303f8a043f48252f47a4a0a9ad61ddf30a3ce019e6fb77d1f",
-    "source/transport_deps/pydantic/errors.py": "edcb41342b6de7b919171ef52ecd87f3a22e7d0011bf8c0f29ff0385db159f9c",
-    "source/transport_deps/pydantic/experimental/__init__.py": "413eeb298743b0289327d1848e6b107561d2730a4aaeb3641aaeafa8e34fe914",
-    "source/transport_deps/pydantic/experimental/arguments_schema.py": "1059e35ffba5a7eb4fc948d05f9a6642dba0d4e14bfc071cf1b70c6cb77e7d56",
-    "source/transport_deps/pydantic/experimental/missing_sentinel.py": "8507a382d174d3052e40c9e2f78dbd7af83e797c88c292af8ec0fca1fa9f8fec",
-    "source/transport_deps/pydantic/experimental/pipeline.py": "2affddbdc7b12ae99acdf44bd32ebd01ac9e03a1f7ed2ae6b19dd2525fe7d2a6",
-    "source/transport_deps/pydantic/fields.py": "5ae0c63af076d8a5aeb96ddf5e74b85af838a97fed769f17ec1ac09736b8b30f",
-    "source/transport_deps/pydantic/functional_serializers.py": "ac4cc7dfdd73ab2de8fdb5a4d9012ebf24a67103599b05e62502c2dd9185ed00",
-    "source/transport_deps/pydantic/functional_validators.py": "73ffbbc1e5a918d70e61f45f55416ed758ebc4c54c75f63f73ee22b2dc24f796",
-    "source/transport_deps/pydantic/generics.py": "d19a99f4ef5a9e7223ff7986051aa9b3886d7b2ddbe65575f9ddad5313cc3270",
-    "source/transport_deps/pydantic/json.py": "647f11908ee1e01cfece9f0e7530316c952856f72853e8e131d459d01fa4d317",
-    "source/transport_deps/pydantic/json_schema.py": "677786d509be226f8c417ebac28c963fd5b2c055272d116d68e0cbced559f937",
-    "source/transport_deps/pydantic/main.py": "efb414c13b5041d51a7e8af78abe70d820b5b34be2805b460747d2953bc5dbb2",
-    "source/transport_deps/pydantic/mypy.py": "a7a294d46c0f1dacc5ec4e6f26ad6e2dde2d1dde8313a6eb7b8fa6e4b9f6de6b",
-    "source/transport_deps/pydantic/networks.py": "2c995c70c79df7b1007a9293e8e98f23d73a8e1e62c1b063935fe35f5449c37d",
-    "source/transport_deps/pydantic/parse.py": "c2477cd9d82dbd6b43f3de54fc8e84d61b6a3251a1052604577f5cb8149ea370",
-    "source/transport_deps/pydantic/plugin/__init__.py": "6bb4f0dfae94e8ade496d08235963bea773db2cfbbb86190e344d769df4eca94",
-    "source/transport_deps/pydantic/plugin/_loader.py": "f502d79de2c49afca15e46bff63e0bae46e67b8a8fbfaa98a6196c8c5d8c1ac0",
-    "source/transport_deps/pydantic/plugin/_schema_validator.py": "41b9aab06df730199fb4d4369cd8ae376d8b85badec5403b8a90d5bf7274d815",
-    "source/transport_deps/pydantic/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/root_model.py": "06f98bb56e22d7574993e74b38cdeb97e8e725d40679e41314172610eabaa4c8",
-    "source/transport_deps/pydantic/schema.py": "56aaa3beafcb9daa5592779b51ddc1a7f275a768176599d94602f8f1b5441bba",
-    "source/transport_deps/pydantic/tools.py": "88742977c489e43093b4f5796ad015d3a4fcf5b8d268c15e659d8b5fd95ab196",
-    "source/transport_deps/pydantic/type_adapter.py": "553fbeca0e1adbbb21481cd61c13cacf8f777f7890f686dd9049218d92a513b4",
-    "source/transport_deps/pydantic/types.py": "9ea752f89d995ea4e1daa7b224ecc14c1b475b267961115ef2068c579f5df471",
-    "source/transport_deps/pydantic/typing.py": "3fb7de037e4cc1370bb11d6e2fb75bd12fa8c9d071a1b9976b9e580e80606a34",
-    "source/transport_deps/pydantic/utils.py": "d799d1d90a6a4c1165415e1336d4c8b4cc93271fdf6f257e80f988118d46a287",
-    "source/transport_deps/pydantic/v1/__init__.py": "14b43c212a7a3156517e39d2edf4389b51714311421764158a41380cafb83e11",
-    "source/transport_deps/pydantic/v1/_hypothesis_plugin.py": "e444b9c56bb0d45400b3298b7b3cbc4209d5cf466955f5378e4993ef81f6ed54",
-    "source/transport_deps/pydantic/v1/annotated_types.py": "ba4d8d000c6a88d10b2a388787287129a20e87c1759585bf2f3ad6dd7ee86417",
-    "source/transport_deps/pydantic/v1/class_validators.py": "50b39a21481850306c1cbec411511aadc33e52e6ca52082837c8526c3a27b051",
-    "source/transport_deps/pydantic/v1/color.py": "8990012d8a7a395a28d801643fd229ae2ff049cebe2a496ef18ba149aaed7798",
-    "source/transport_deps/pydantic/v1/config.py": "6ba3f459eafdc7871bc0a5bb5eff29a1252a33858ff912d6c17e9832962af400",
-    "source/transport_deps/pydantic/v1/dataclasses.py": "efce1caaf2276f020f5abf6eb1fa57ddc87bcf8b77276b532ba374ebbff0935a",
-    "source/transport_deps/pydantic/v1/datetime_parse.py": "e10cb5910a6adeb3556497881de48f0e9b92d81be1a7528fb73246d71bbe1f4d",
-    "source/transport_deps/pydantic/v1/decorator.py": "cda6b1c71a163c29bcd7c0f56ecd3285aa518d79b7d95f06d191d609d335b970",
-    "source/transport_deps/pydantic/v1/env_settings.py": "03d557c2d465d36018fa31f40b4a2ecb954dc377e2e85fe9933b870e380000e3",
-    "source/transport_deps/pydantic/v1/error_wrappers.py": "ebadb931fc3daa40b63708ad07f2450167bc07e5efeb3054eeb2fd936f2d7f2a",
-    "source/transport_deps/pydantic/v1/errors.py": "988c0f103e6f18ce50e6fe02e19d493e5753447fa89af132947ea4b0c84e98fc",
-    "source/transport_deps/pydantic/v1/fields.py": "56a5890ab894362132a6d5eba03549e74d49a550347a7d9500d724b2a5cbd9bf",
-    "source/transport_deps/pydantic/v1/generics.py": "5730bd61457e11b3e94376807e4d5c3457a3efdfc8cf39e443b5ab9934d94bd1",
-    "source/transport_deps/pydantic/v1/json.py": "590e47cbf848a5f751dd84bc93a37613a28c273b1d6c18bf95d58e3c9695f353",
-    "source/transport_deps/pydantic/v1/main.py": "cee36974de50d15d301b65144cab741d4cb75c9e0e02f3d26437625d8f85233b",
-    "source/transport_deps/pydantic/v1/mypy.py": "0a5f1745f0a621c544de3e40114e760bc88387c95c5fffc3de1cf68c85b1300b",
-    "source/transport_deps/pydantic/v1/networks.py": "1d836d2807ce98e9ca269b0383583a488923f563e153ffa2f25e5ed89281a46e",
-    "source/transport_deps/pydantic/v1/parse.py": "049b5daa2651b5abfd55114298ec6863e2889909a33f2f80fcda288e21546716",
-    "source/transport_deps/pydantic/v1/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/v1/schema.py": "6aa06e03ef9cabc8005648a6e4124f140487cce67c745b66157fdf346afa8a9e",
-    "source/transport_deps/pydantic/v1/tools.py": "d650dd5c79348cbe6e3f7bb9442600bd402518296000efb8e6592af63edfc810",
-    "source/transport_deps/pydantic/v1/types.py": "07397e45c9e2b4f0479eac308fd8afec98c7b8dd46a67587db874a905de5f5ef",
-    "source/transport_deps/pydantic/v1/typing.py": "ec67418356131d42d4f35b6107ff5c8d13437d97e7e248685fb9c6b70fe47821",
-    "source/transport_deps/pydantic/v1/utils.py": "339151c9f3546f50369a4f6568680255d7c71dbdc0b5082b8cee6a7f205fe310",
-    "source/transport_deps/pydantic/v1/validators.py": "9725249f5316847c65097e597c48058ff0801e88e888f71a41e31d10cf57be25",
-    "source/transport_deps/pydantic/v1/version.py": "1d79d75bed5b316e6a2a196be5180e10fa21ad90c20d2bb2cbcfa08bc182819a",
-    "source/transport_deps/pydantic/validate_call_decorator.py": "f23a8b9605d38d6123e1d5c3834c08dc41902a46f42670ac2ff2525236d4e528",
-    "source/transport_deps/pydantic/validators.py": "a706c825755bd42576980138c3f1067cd8fb0f0cec29a5b0fed4cbe9cbe24eeb",
-    "source/transport_deps/pydantic/version.py": "4e6124eaa03e540d4f56c737f5b66584f010032bda71de812b3c57007c96d395",
-    "source/transport_deps/pydantic/warnings.py": "5aed551b3aef159c384fac8220a1e31fb2d263ae878dbb720856e7e6e5a8309e",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/METADATA": "3434fb886fe254ed414541f305a86e43ff575093f8f6d15c89bd11144ed8e131",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/RECORD": "7f3a82de0054511dbd2f7309c5e5b6ddb3d96c6230bfa83fd73bbdffc51cd0f1",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/WHEEL": "7f32b0ec3e21425972f904759f2efc26a495ce223c174ce187bb87ac82f5240a",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/licenses/LICENSE": "2afdd30d54b4d62b6f488a6bcc1546e84ec5061f13f4209c03d012348783795a",
-    "source/transport_deps/pydantic_core/__init__.py": "9cad6292b75254af606a970aae9bff6e5ae9f0b08089cd63acafa391b60794d2",
-    "source/transport_deps/pydantic_core/_pydantic_core.cpython-310-x86_64-linux-gnu.so": "cd9bb47485cdffc363010507a214155966ca860b0b93bda39583492c6764e631",
-    "source/transport_deps/pydantic_core/_pydantic_core.pyi": "3ea1dbd4182f08cf9341f24b3c5cf7db77a05b3535ffe9e23525127a3617a11f",
-    "source/transport_deps/pydantic_core/core_schema.py": "bbdc850b72d685133a0e250fed263b33691dcdf3813492f3c0e31001e3d46005",
-    "source/transport_deps/pydantic_core/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/METADATA": "bad3d9e6342ef6750cf5128ec942327b8bebdce0b6e4d6f67cb15b27ae5118da",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/RECORD": "cf5e3948866214f2e9ab0c1d43c157e9f9c14f59cf8442936c6cb97d5f2b6a5d",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/WHEEL": "ccec20e2307acd7da453dd7437e70c6b08e2bc3ead3bc3445af135da37b56d59",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/licenses/LICENSE": "804b59b25f2c31bd278f9202a19ae49a3945aa2664387e2d0a128c7cacc61ec3",
-    "source/transport_deps/typing_inspection/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/introspection.py": "5888c45bd627da692d9527699e3f549adf35651596b1656772310e12181b4396",
-    "source/transport_deps/typing_inspection/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/typing_objects.py": "a778d6290d51581c9c9a23e0263af2a4604c76def5c4d40a7cd1a375b0046ff6",
-    "source/transport_deps/typing_inspection/typing_objects.pyi": "3bd24dd92d0270d31d64fd40abc1ef8693f5322aab9a78aeb3c5b62f992f84be",
-    "source/vpp2/src/vpp2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/adaptive_mask/__init__.py": "a73f5515ffb77d323f9c324776e9c166d13817c470506cfd5db4ef0546cbbdda",
-    "source/vpp2/src/vpp2/adaptive_mask/audit.py": "c949909a2e60c3c4389c0221c79459961f2c4925baa07f7985774a52257a1d17",
-    "source/vpp2/src/vpp2/adaptive_mask/budget.py": "74e32285903dbf673ecadb6a60d917a990b9c6619beac9f926c1b3cbdae3887c",
-    "source/vpp2/src/vpp2/adaptive_mask/controller.py": "328f636201a845ebdc211340a9014a5076fa895e10cf9c2e1f91b96761a2514b",
-    "source/vpp2/src/vpp2/adaptive_mask/data.py": "e3b633fedaa503b101c80610cbed0933e9e245acf56a6c0692c4d99620faf84c",
-    "source/vpp2/src/vpp2/adaptive_mask/equivalence.py": "8658cb69d549fa0575570364fe3aa1e0b121fcf8b9965b17cd174d459dddf5ab",
-    "source/vpp2/src/vpp2/adaptive_mask/execution.py": "efec52c9b2eabf0300c72edea18f9cd4cc16a0889496793abb6a50161f2328ef",
-    "source/vpp2/src/vpp2/adaptive_mask/export.py": "0ce51e74787d28bbe9f4e73a3774297b424aa131fa78968d24154ff56db98b0f",
-    "source/vpp2/src/vpp2/adaptive_mask/feedback.py": "519fa4e6416723fc000754815a2f05d3415299ae35724f37bb806ebc22393d38",
-    "source/vpp2/src/vpp2/adaptive_mask/geometry.py": "f1ce4110c3a3a893f4f240c49e2ccc4e93215ac4ba9e447de8042f9b19afab98",
-    "source/vpp2/src/vpp2/adaptive_mask/hidden.py": "968560bd6dae0fce59bd28084091b10d559cf165c81764f6ad24b7492101810d",
-    "source/vpp2/src/vpp2/adaptive_mask/losses.py": "4a7dea6f87ef0fa38a7202d85d98d37460f8f4dc13ca5336f32b1da8b84a4322",
-    "source/vpp2/src/vpp2/adaptive_mask/masking.py": "31b595916410c60f772fff1ab63e92c854c8fd0d2261e6b865fd05ce2c61b9c6",
-    "source/vpp2/src/vpp2/adaptive_mask/monitor.py": "48b71f499dd5a0a79557eeb918f5fb1a641dc8654e5d412aa628ca9c57b5dcf8",
-    "source/vpp2/src/vpp2/adaptive_mask/physical.py": "f54e8b4fae8026cc89f296d74bf4cbea06fa4e126905b572a22c7d8226f0238c",
-    "source/vpp2/src/vpp2/adaptive_mask/preflight.py": "b646e7ca785afc87356e3e1184e48f535792ebafc84936450fc83bfd3f9f5ea8",
-    "source/vpp2/src/vpp2/adaptive_mask/priority.py": "68ccff1a384081f0c9a82e8f54ad2c720293f8500e7440a469b5e8fac1ee608a",
-    "source/vpp2/src/vpp2/adaptive_mask/quality.py": "c3b3a2264ca0b4773805f4485ebfe91bb4ab861e226ff4bd27bd667ab372b59b",
-    "source/vpp2/src/vpp2/adaptive_mask/quality_policy.py": "f100b7d4a412592777d936a6cf3c73a9f787cc4a62d5f1b8a369fed74acf897f",
-    "source/vpp2/src/vpp2/adaptive_mask/reliability.py": "306b41720154810bd972fe4d9c4c388e6967539d076ddbedd3a432ac62d13104",
-    "source/vpp2/src/vpp2/adaptive_mask/report.py": "f71401e09297daed8bbb72cb3af3b887e5fdf5ec282f47637fbb01392d280171",
-    "source/vpp2/src/vpp2/adaptive_mask/runtime.py": "ed74bfedd147389ea01f3be8c2979fa06c7152326ccc29ee347175690cb9c6cc",
-    "source/vpp2/src/vpp2/adaptive_mask/state.py": "5e8f95021eeb35511f5481e4e91ae687fe605f9e88d17b3b98d377cc967c17e1",
-    "source/vpp2/src/vpp2/adaptive_mask/train.py": "ad95349bfb242bcecd7ddf82fc58b31d63f2938a2101f9560bbc9e305f5f5b20",
-    "source/vpp2/src/vpp2/benchmark/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/benchmark/audit.py": "8587be371653c38b9926aa0ace006a244f270e0ef0020147a7bf5b0fc85d4538",
-    "source/vpp2/src/vpp2/benchmark/packed.py": "5047fd02f3fcc68e41f543fe83bc9202769a05794d898524d6d25e308f2b5125",
-    "source/vpp2/src/vpp2/benchmark/scheduler_io.py": "d6df06259549faed2da9f2888b540eac8e1ff2895c339eea58b4ea6a7e3b3cb0",
-    "source/vpp2/src/vpp2/checkpoint_compat.py": "462c5f93812483b5871633b200bc2ffecb9c201a2caf6b14f8f36dbe95fba50a",
-    "source/vpp2/src/vpp2/cli.py": "54c19a0f7e25b53bbf56ec13b0f537c9d7b81282e1e80575bddfb470172164e3",
-    "source/vpp2/src/vpp2/compact_kv/__init__.py": "f34f2f77f68399f45f6656b562a5c331f2504672dc145df368e086532191f766",
-    "source/vpp2/src/vpp2/compact_kv/blocks.py": "662c1f35055c2be2e2a14f2d92615d046f5e454879e34af39e3c7aec3af34098",
-    "source/vpp2/src/vpp2/compact_kv/checkpointing.py": "3c33c156e96f26ceb22eee61cfb8b18c3cb9560db0ab405cfd4c46efe5aba353",
-    "source/vpp2/src/vpp2/compact_kv/eval_prepare.py": "bdb33444056245e3cc773b0849b4db608b454cd24a7ac5b752f1ba26ff978865",
-    "source/vpp2/src/vpp2/compact_kv/export_check.py": "e6cba70b47027882a34d587aba0da5f9371128ec2db2282b91ad2d49e40065b9",
-    "source/vpp2/src/vpp2/compact_kv/gpu_check.py": "fedb6a9ab69b51080161ff2a4feae3f85262cc50f932f48e8decf04c1d147a20",
-    "source/vpp2/src/vpp2/compact_kv/losses.py": "914437ebf3ad91fd2cffa893a65ecb04ab01f2b337658a781a472feeb50c0a06",
-    "source/vpp2/src/vpp2/compact_kv/model.py": "332ebaebf543d40a34d188ebbe81055ec3e4bcff164bfe90d15aa6cef0748b8d",
-    "source/vpp2/src/vpp2/compact_kv/prepare.py": "7b01764cad0fafc0cc9441952b2e928048b0a43a6105dbd0eb64c349f4b9bf1f",
-    "source/vpp2/src/vpp2/compact_kv/provenance.py": "103e0f7100be56233aa86041ba9bc427ccb0db4fe045c4e6b8d3481ab6bb42da",
-    "source/vpp2/src/vpp2/compact_kv/reporting.py": "0b9ff602fc3425d1b9c122a77da3a426814e11c9cbdcb6b9a75a2a48130087a7",
-    "source/vpp2/src/vpp2/compact_kv/sampling.py": "2e8d82704f07853a7b487d5aaedef1b2f1fcfc5247e9100a56ed21d69262eae2",
-    "source/vpp2/src/vpp2/compact_kv/spec.py": "9bc920fc7d4398e424a0bde38ce97f87473aa417cf68b48e2c292a59491bc7d7",
-    "source/vpp2/src/vpp2/compact_kv/train.py": "a9e4410c5cde5df002d9808554538c29b77ed9e2d2bedc9ac5f8465eed70cc3b",
-    "source/vpp2/src/vpp2/compact_kv/validation.py": "73b62d02ce1d518cba7b9462e85c3c8d272ae9974e1161b1b6adea6f0040e2b5",
-    "source/vpp2/src/vpp2/convert_robodojo.py": "7ec2d6bfffbb4f5891ee33a4bf630cf27b17628506aa3cee1c46b3050027b290",
-    "source/vpp2/src/vpp2/critical_windows/__init__.py": "22721b77e25ec4874c4853d2f2991a14dff6aaa665da500f2830826b06545478",
-    "source/vpp2/src/vpp2/critical_windows/contracts.py": "ded4a239f7599785d74ff57ff892c5aa74f0280d6759460c93942e91f505d47a",
-    "source/vpp2/src/vpp2/critical_windows/coverage.py": "6c18bf995aacf413bbe4536eaeed92ab4b2f15f9eaaaa03a2730256ea7922ce4",
-    "source/vpp2/src/vpp2/critical_windows/inference.py": "df301bd187b27d20b6623febcb5e85815e1b7d4a1e9524a8a65bdf5904bf7232",
-    "source/vpp2/src/vpp2/critical_windows/metrics.py": "ae2bab145555d55696ef479758a5bae6780527fd3085a4f46d035b667ff32736",
-    "source/vpp2/src/vpp2/critical_windows/packets.py": "8f7d2c411407d95b1ff4d0e5b9055589f6114b1bb1020eeb7e1b106f330b853a",
-    "source/vpp2/src/vpp2/critical_windows/prepare_staged.py": "694cb5699de2d9f5ce1bc4667627b47d6bf0698715b4aad86907139859caffe5",
-    "source/vpp2/src/vpp2/critical_windows/reselect.py": "8d0e31ca1c17906d74de6439276e49ff7ea78abfc66fa556d8a306070b3dbf4c",
-    "source/vpp2/src/vpp2/critical_windows/score.py": "32daceb9a3fd957d9be75823b4d519629bc60a64d1da62f465e172cb1ec98d55",
-    "source/vpp2/src/vpp2/critical_windows/selection.py": "a7c35f6719d3308bfb738965d17544a15985c6d684bb47eac7272d6155d96786",
-    "source/vpp2/src/vpp2/critical_windows/staged_review.py": "d5cfb3f49ec5acc6914da98f5413d1549b2695e7c85c9aed9c0db83dee6be754",
-    "source/vpp2/src/vpp2/data.py": "8574e1ae3c6ca6e4b84633ee1815f2b3f2ea17c1e55f12fda0b00c5c5a6a7b53",
-    "source/vpp2/src/vpp2/datasets/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/dataset_utils.py": "e1a1471b0aa70917eaf699f85a9acf3ac70d9fb6eb9ebaf02d2bf2f218d6ab2f",
-    "source/vpp2/src/vpp2/datasets/lerobot/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/base_lerobot_dataset.py": "729fb3fd4ca531212eadb4a4e9c7e9f97ad34cca091d66380acd6b179f0c29c6",
-    "source/vpp2/src/vpp2/datasets/lerobot/constants.py": "11ac9260341e1560c5eb0dd7acb79ed63c74180b8ab4113c27622ff213db417f",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_dataset.py": "bfe75bb9593ec646236777d90d9fa309fa0e1188db07e07a3be0810edbe3da56",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_sampling.py": "266b0b7b023f014f4ccd935b6015b0a05b4659f372264fdefd997ea9692b244a",
-    "source/vpp2/src/vpp2/datasets/lerobot/episode_instruction.py": "74767839a813c7902861f4561101541c282b72e261df1fd28fe8d06c91231b69",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/compute_stats.py": "7d7daab219332b5abdfd5ec67ccaf3502e8eb583a17a4e482c3f3d4f7a691e89",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/utils.py": "4fdb4ac0653201e6743f2161c485213429948c1e1d5b068d03b4d97803254006",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/video_utils.py": "4f66c410ae6aa5a98acfc3320f3dd7b017d0285117054f14226f9e5d82395087",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/lerobot_dataset.py": "5659e992010bb6674ec50ab0d33b28a8e9dab0ecc895c199662af4c40fe56629",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/base_processor.py": "d3327cbe398d77761df960014d41e70d349bb0fa42bca874333b9ee970222447",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/vpp2_processor.py": "43c5604cbfc315b7377316cb97c9a6f3322124404397abfd01c6c3d0090cd414",
-    "source/vpp2/src/vpp2/datasets/lerobot/robot_video_dataset.py": "8eabfbd649116b0b4d6c9704c0df34febbf8875f6ab863dd75f4107dbbcdef73",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/action_state_merger.py": "39fd44f254bd7fb71b9269d75d8f9b5a5bbbe167ea438dda92d40eda1d256734",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/image.py": "f494d262ce29770e959bdff60b997f7a07e5f00d10ef515eee7be813d1db038f",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/rotation.py": "feb59d26539be6cfdb86488f0c10c11488ed73901658bc657e67a755ddb9e67e",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/normalizer.py": "8207d813b698a91dfdbe6910a4aa1fe0e775a2a8edf5c05b2d7a42591e9f8adb",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/rotation.py": "8ea2c72565707f5979967a1abeb5edf1eaa2c3fec55a7b50e640de869fc9fbcf",
-    "source/vpp2/src/vpp2/datasets/lerobot/video_event_dataset.py": "c20d3f5e2bc05e1b7ed94ae38872883a976a179e09b7fa051e94f94e2836c02d",
-    "source/vpp2/src/vpp2/enhance_prompt.py": "f31b3190c24bad073c5e3121ded4f564232729566e8c1fcc3e76fe75b28faf45",
-    "source/vpp2/src/vpp2/evaluation.py": "9b76882149e3c3a8c417a0e23514af1bb305c01bf0b6d21fff1f70bc9843f55a",
-    "source/vpp2/src/vpp2/flexible_mask/__init__.py": "64bed1808976d1f3c708154c36a97288a75abc6463d49d7f0065cb8360a2fb16",
-    "source/vpp2/src/vpp2/flexible_mask/calibration_batch.py": "dc5d83fc9e2bb595b9e3aeb1923115ebb9345283290a7c3f95c79fdeef7d3e76",
-    "source/vpp2/src/vpp2/flexible_mask/checkpoint.py": "c8531727e2388ec3d7a9770a504ceda119afcf03d64b2462befcd45a48949f9d",
-    "source/vpp2/src/vpp2/flexible_mask/coefficient.py": "07a3cc8d8e6c8b60119b9f94ae85074da017b08158fa0a0912cc9051b6df8ef7",
-    "source/vpp2/src/vpp2/flexible_mask/config.py": "4e1e9b289e293145b4e895c8ef7347bc837d01c929e05fd5cbf65bdb3c20be5b",
-    "source/vpp2/src/vpp2/flexible_mask/continuation.py": "2016463b4efe01b5d0a9ee84f03c3d95ee16f8e24f230faa465598cc978da29b",
-    "source/vpp2/src/vpp2/flexible_mask/controller.py": "263f1c9091e0c4318aeefb61e91549ba905ff046a625e35e1b18c94da4b4668d",
-    "source/vpp2/src/vpp2/flexible_mask/data.py": "d245bafd6f04f33ae7cef7c01d5db876a1a2a1a138c4de8a139d3ea78095521c",
-    "source/vpp2/src/vpp2/flexible_mask/diagnose.py": "faa29a0fa57c66af3b2e817640fdedfddc998bf878357f484415b244555ae4a5",
-    "source/vpp2/src/vpp2/flexible_mask/export.py": "3064dfd82bb55a145c16c77bbe17244970d07d18e944b793f4b7368dbcede764",
-    "source/vpp2/src/vpp2/flexible_mask/gradient_probe.py": "6fe3f13540cb9d8dd20c0b6429268ca74625dfad1e41806412c0f980fac76509",
-    "source/vpp2/src/vpp2/flexible_mask/losses.py": "df264b71365885ea90329c0b0facc92db3d2eb6e04324ea90a4701a1cd457abf",
-    "source/vpp2/src/vpp2/flexible_mask/metrics.py": "e976dd066e4b5e7e7ae79db77b9d1be64d17c212c61d0bc113b87a048e8f2b6f",
-    "source/vpp2/src/vpp2/flexible_mask/performance.py": "95b3977fd739cc98f7de94b14fb490a7c132021e742e7179dc5b5dddd6557c68",
-    "source/vpp2/src/vpp2/flexible_mask/preflight.py": "f794b997a492911c4d77656310c89d3b85e54983b3b889ba9904be64fb961e43",
-    "source/vpp2/src/vpp2/flexible_mask/prepare.py": "d6cf82e27555ecc13973727f4b71ce7e40558fef7d95805665fc7e458a42fad4",
-    "source/vpp2/src/vpp2/flexible_mask/pressure.py": "6cf74edcf3625fce9eec689d70841c6c1481f7c1abd87e5bf70bb38eb34632a2",
-    "source/vpp2/src/vpp2/flexible_mask/recover.py": "418b2d5ecdd08b08686b2a07bbeb0f969471eb73c04ac6952dcfbcda3b123f30",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_best.py": "917d924c5708584a0192050a6d8210809e905938aaf877dd81cff3be287ea289",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_checkpointing.py": "6f6a61c2ee7076c6b0977d1acb489fee7df04322d7f6b13bee9bb8229be319df",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_export.py": "c50433aa1b7e528fb23c514dddb5b9197f44c662829838ecda106980ab0dd54d",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_handoff.py": "aa590f707738e4b7c7250f901e29795c7441430f961bfa764437cf34bdc514b1",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_kv_decay.py": "bf83761798deea4f89c32d14f43bbfb0f16f751b3230962456c073b843dcd08a",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_subset.py": "7f7154cab6d910b747a0189742ee625983d08f13de50ca60bd4154f61df2b0d3",
-    "source/vpp2/src/vpp2/flexible_mask/report.py": "5b49283d472bace3a6ad56ce568af04f2c3d04add1a3920ad5006fe98374bce6",
-    "source/vpp2/src/vpp2/flexible_mask/rng.py": "01680b5b7e8a29ce92e96d756e12ceb80bb0e0ce1ebb642a14fa2a3a34776daf",
-    "source/vpp2/src/vpp2/flexible_mask/runtime.py": "373a3ecbc2ab6592a359daf6adfd8c958cb657f393e698ca2cb834f06750be4e",
-    "source/vpp2/src/vpp2/flexible_mask/select.py": "b547857bad61ae7ebd25b888fba73c2ca8f50a8eb31551ca54a909c75c8984be",
-    "source/vpp2/src/vpp2/flexible_mask/selection.py": "b9453c8735051b54ae7dc8610d2ca332eae8ca1a3e099aaadaf5ab75bfb0f8b0",
-    "source/vpp2/src/vpp2/flexible_mask/speed_review.py": "9d7099d1e173026274eb439169df2ad4812ec1d56a5f0b21242f972cd835ffd2",
-    "source/vpp2/src/vpp2/flexible_mask/throughput.py": "ce21bdfda51ae10d1470ac537d910d9f46f60a38910888808ffb3f5a80174d58",
-    "source/vpp2/src/vpp2/flexible_mask/train.py": "86625613bbc692722c6d2d24278a586f88ce0d76263a9daf92562a9002a8e037",
-    "source/vpp2/src/vpp2/flexible_mask/verify_export.py": "fb68e62b0da98e6f96cfa694bd061162bbeec1763e5787e057a1cbac5b2ce8f6",
-    "source/vpp2/src/vpp2/full_mask/__init__.py": "27c9ede277380e8d64586371f2a03334e68ab2d3bfe516d95e8cb96099827929",
-    "source/vpp2/src/vpp2/full_mask/calibration.py": "5d97e0385eba82252e12cc1562d1a77da2f2bce29332d758c1b5b6b84c5bb47a",
-    "source/vpp2/src/vpp2/full_mask/conditioning_cache.py": "07adf95165db4c3c1c469695d6b7b6fe4cf4fa9aff0a3d360f4229bffe9be17e",
-    "source/vpp2/src/vpp2/full_mask/controller.py": "0e6a969b89a4275352cc5a1e00bdfe8e3f8754a6ebd7f9266b92179cbe6bdd10",
-    "source/vpp2/src/vpp2/full_mask/cpu_contracts.py": "1a46a53a68d2d8a202c5f98d9a78f6d5f5041dada0c2239ed81e7916a28f0758",
-    "source/vpp2/src/vpp2/full_mask/curation.py": "4ffd7b0212bdbca1daf73ba7c24423523437012150fc044a43bbb33cce747545",
-    "source/vpp2/src/vpp2/full_mask/curation_v3.py": "9dba386d66869de0fba321ade79b2692d47f111ccfebc7f367194cef7a4b4e8c",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_checks.py": "688ea4e7e44c4ecc1a3cd81cf682e2cbf611d9ad2d52744ba1308324da70febc",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_features.py": "11011a089f1586045e2972cf267773aaf4602cf08c8b0e755bbf017ede1e6e43",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_finalize.py": "a2a7812eb6c670686a7a4f4adff0a2c6363dbc078c71d948537f1c5fd546254f",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_probe.py": "2755e9cabd0fe02e70cb8330c083af881a5e4696697d58cf7aa01bd00ef57f6d",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_quality.py": "b9170398762112a23bb133dde53c6a49193c671c767647239c8b0ff15159530c",
-    "source/vpp2/src/vpp2/full_mask/data.py": "668a7c5f88a3ab17854eb0df240c974082f4aa64604cf11d3143b18337f73181",
-    "source/vpp2/src/vpp2/full_mask/distributed_smoke.py": "f699d4f330309f29a537fbabb61c60f61eb3322d7304605351dba3d2acb48e8e",
-    "source/vpp2/src/vpp2/full_mask/gpu_prepare.py": "92ce96dc4c1b6d6ebc22d99935472c15d86ba2a7c050e31640d2c32e08748ec5",
-    "source/vpp2/src/vpp2/full_mask/monitor.py": "2724a934d1f96444227c854f75ef84df6025c02485aa68eca34f3b28d509fe19",
-    "source/vpp2/src/vpp2/full_mask/native_monitor.py": "5572712a340f17692a669d2a4c41d1d691164925eb7bd06a56fff1dc2afd69fa",
-    "source/vpp2/src/vpp2/full_mask/panel.py": "2844b663340cbfb54f7622f3447a7831e2809050739a52dc076fd6ae32ed9e46",
-    "source/vpp2/src/vpp2/full_mask/review_decisions.py": "ed1184bcda7b5bb3fb11bb4262c4461c408e3d3ac0d23e80d2f8cafcf39c7e18",
-    "source/vpp2/src/vpp2/full_mask/review_report.py": "752ba9e8a3caf5b8cefa2d18e6fe14860e6f998f57c03af0940af0943b09a9b1",
-    "source/vpp2/src/vpp2/full_mask/runtime.py": "ae30b4d3fd3c888bef17672376dc3b34491cb296a1883959674ad465579efe7b",
-    "source/vpp2/src/vpp2/full_mask/search_quality.py": "ee3238d3366799e97d3f9a90c11f579ff8f51d5bd0642f52c37d9dcdbcf02e6e",
-    "source/vpp2/src/vpp2/full_mask/train.py": "f4127e6580522d4d9e5a50db10d201526e948d4d3479ddfbbc12e60ec61c7501",
-    "source/vpp2/src/vpp2/full_mask/verify_data.py": "72ee60ba2016d5f73eb2ccb01b666a5dac37138620efc4a79a20607df3658fb3",
-    "source/vpp2/src/vpp2/inference.py": "42b3057a85aaacbb87440bc740f13b09b9c82cd51d8c5af4e080716aabdcf38e",
-    "source/vpp2/src/vpp2/learned_width_kv/__init__.py": "0c59cb8424b6cd64508a005625e9edf63c84d24618a5d2caf6e6fa5d63cc859e",
-    "source/vpp2/src/vpp2/learned_width_kv/calibrate.py": "b89bbf103b49fd85f85f8c300a7868bb42ae7a13f29bd9878dbd5a11061ae51d",
-    "source/vpp2/src/vpp2/learned_width_kv/checkpointing.py": "7d30e46b95d91df8aff088346ae75809958622b3a8e8d953db4f30adfbaebadd",
-    "source/vpp2/src/vpp2/learned_width_kv/common.py": "52dff98c6317b34df2a13313a641998acf4f9e18ac31ea4e8e25227afa14620d",
-    "source/vpp2/src/vpp2/learned_width_kv/controller.py": "7b5a8793a540e1e7464185229189f4f1f1acadcd7cf3c2bf477bda5de7d18787",
-    "source/vpp2/src/vpp2/learned_width_kv/data_scope.py": "0ddde6805f646afbf5f5966c53530f829becd083802f0e0a8697dc1004e2003e",
-    "source/vpp2/src/vpp2/learned_width_kv/distributed_check.py": "f89f153a20730a6bc3c70133479f34918d30c5a48775bb3f6c14bb25701801ce",
-    "source/vpp2/src/vpp2/learned_width_kv/gradient_probe.py": "3bed964f6488e95b386cc58058e7f28aff50c9a98dd526f3365891a9049a6fab",
-    "source/vpp2/src/vpp2/learned_width_kv/lifecycle.py": "798afa47ec5dcc8749982da2d7853eff184e450a32d51bcf70eb4578f2bb0b09",
-    "source/vpp2/src/vpp2/learned_width_kv/masking.py": "3272ea09271130dcb3cab6935f6bb7d61bda66b9982c3baabcd61ee0efd914a0",
-    "source/vpp2/src/vpp2/learned_width_kv/performance.py": "a2fd9597c51bf0cf5c435b48e8f57792df24d2723a6d7b6b841d6fa24f17ab2f",
-    "source/vpp2/src/vpp2/learned_width_kv/reliability.py": "f74399fb67427093b3e4c7df9749379f9291b7ac58bd6632db226b6e78fdbfd0",
-    "source/vpp2/src/vpp2/learned_width_kv/reporting.py": "f22c1d8806a5b7757bb6dd99b15153d0305b2eeb0790bb37763337690d642036",
-    "source/vpp2/src/vpp2/learned_width_kv/runtime.py": "513da1e06ba9509164f017e0b8dbf6b219d7e1cff82edb2d2fd8d2d7839e666f",
-    "source/vpp2/src/vpp2/learned_width_kv/train.py": "5813ebd4fefec6d3a8cead9e0c645d10734f5ec88c4d74a5742587f2eb4de6c3",
-    "source/vpp2/src/vpp2/learned_width_kv/validation.py": "d5594ce875c8011bff12d3c77083dca68878178ed022b8079761267d04a82386",
-    "source/vpp2/src/vpp2/learned_width_kv/verification.py": "1bc43d48ef0ec3056af614adf43ca0744d94b1e77fe970168fef87c7cf2e6fe8",
-    "source/vpp2/src/vpp2/libero/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/libero/action_model.py": "178fbe4cf02b9a2e133ef31389b6bb2f9baf0504bd0818a265b6bff32ebc3914",
-    "source/vpp2/src/vpp2/libero/cli.py": "fcf68a88e2db0a9b786c6199d9b719802fe4b84514741658b49e2b06e2473e45",
-    "source/vpp2/src/vpp2/libero/data.py": "19b699381e7d6fa769b5090c7125ae01b7584576e969468660ee65b1585ef5f1",
-    "source/vpp2/src/vpp2/libero/diagnostics.py": "9540e31dfd2def06bec942ffe73262278700a17a340e14cb5bd41a295706e147",
-    "source/vpp2/src/vpp2/libero/eval_helpers.py": "ac338d4ade195ceb3a87620d162befa9679e2d7ed3d91b09927e5d59d86d97f4",
-    "source/vpp2/src/vpp2/libero/eval_libero_single.py": "87e6b457dc412ef4fe5d01aa48b1a6d20a28dbce3152848042e2103769e4efef",
-    "source/vpp2/src/vpp2/libero/eval_libero_task_manifest.py": "f0c721a7ddc768edddc8b0e065a4d8aeec35e1a546c88deebd9097ad93f6227b",
-    "source/vpp2/src/vpp2/libero/evaluate.py": "24216f9814f8304e077d72ff8d8c778fd588239313b34cd7d23b960c8f924c67",
-    "source/vpp2/src/vpp2/libero/libero_utils.py": "c7cc23976c86eabcada4846d46345d3605fc2ce03929761646486be3901697d4",
-    "source/vpp2/src/vpp2/libero/prepare_pro.py": "0999050352d924c11d29b22f63e902162d8cfb23f94815c93c01295c11ed81ae",
-    "source/vpp2/src/vpp2/libero/pro_alias.py": "90b5c8c8061bc727a6e413a8746fcc98cbd06e1c7c4175fd223b45af4f10b700",
-    "source/vpp2/src/vpp2/libero/pro_runtime.py": "adf94ed64b032aaeb0d88391e7b653f88f67e76b9411419ece25faa31ccc1340",
-    "source/vpp2/src/vpp2/libero/protocols.py": "68225c95eeeee25032966a7f90c1ba18067c53225397f59a463d764030fd0d93",
-    "source/vpp2/src/vpp2/libero/runtime.py": "1e2e2900a604d33cbdf8774cfa33dcee1065f21dc0b59fcf30799b5098c1662a",
-    "source/vpp2/src/vpp2/libero/trainer.py": "ea29b4e000047a6b0c8aee3cb6f32df9187987a657924085dbafd9628315107b",
-    "source/vpp2/src/vpp2/libero/worker.py": "9df6350c4ab92dea70863c08f1bdcebd0d77574a3a2d8d7ac9899a5a825be6b0",
-    "source/vpp2/src/vpp2/models/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/action_dit.py": "9a654edd860ce77c1308923bd338c7f14d09f68686642531d7143c16b9f6f331",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/compact_channels.py": "6f7ad787624844970985117002ab809bd777f7474969daca121e7101381c6d0c",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/gradient.py": "594249b6eb931048e352fd3ffef9b1ac1db9e4fc74659e04216d6b823dd174f3",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/io.py": "1b68cf2f2fd674c7a59140f8632d4464044bffc2e9277ef5f9a975e7c699966a",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/loader.py": "b9180b9c9177af192e5cb92387b4a09bd0dbc08fb75a94e843b67d315c94e339",
-    "source/vpp2/src/vpp2/models/wan21_14b/mot.py": "cc14a909bfb181a441bc9d8d6de12f3c31d89f3dc3415031e34a5f8c21410019",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/scheduler_continuous.py": "15de04574c1d8ccb794ca6560b1e501e5c83cc86104bb80f1eb3cc3b0324ffe0",
-    "source/vpp2/src/vpp2/models/wan21_14b/vpp2.py": "8c419af905a8ffa93d77c827a283b7aeff744f3a461c390783eef8a333c31756",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan21.py": "01b45cc71f69a6914f7b286559317c8e8485d0b0fec93bb28ebc37c183c295f4",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_clip.py": "f65e44c9d1b7811d2996b3fa22e6fe4c2ac5e6edf1fc2c14251cf582bf8d1f05",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_dit.py": "0d1b5837ea923508c05394d56aaa465813e7b4f5aea3b31ceceae39b2de9b0f9",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_text_encoder.py": "902a1e46b21b5c48e166fe03defabb94b3773c6bbe836fc09b9a0d0894eb1f6e",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_vae.py": "33f7d804677b44a6f3ab334684017dcdb1e9dfb748279de76ad94a5bb71a7e51",
-    "source/vpp2/src/vpp2/policy.py": "2a7696cd73ccb99575bf901185cf578c3775945f61f6e67aeb2b4ec6bfc4e30b",
-    "source/vpp2/src/vpp2/preflight.py": "7a91e486675a4cdf38565e0f4b58e60a68bba7de5654cfff736b5df58f5c53a6",
-    "source/vpp2/src/vpp2/robodojo.py": "183db869a622ef17ee92c487949db2e22a7d1365558fe703bfb8dd8fef83e9c6",
-    "source/vpp2/src/vpp2/runtime.py": "e26885a3732f5fdac3d8381da870a1673d0c03354f02f9deee7fe779d645f8a7",
-    "source/vpp2/src/vpp2/trainer.py": "feaa0ac4ed94a376836eb60cfe6f52460392b11f7cafafec9c6d9f151f5b61f1",
-    "source/vpp2/src/vpp2/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/utils/conditioning.py": "43d43f0f3ca0add64498c3d64a3791513a53e3fce0ea53f77ef09580a86b7b2d",
-    "source/vpp2/src/vpp2/utils/fs.py": "69e6df236b43360428a613686c8f37c531c0d16a3a2c5b06a725c46b49d657c5",
-    "source/vpp2/src/vpp2/utils/logging_config.py": "1de4abf66d9d201b49e7bff48291643b9dc5ec9b6e7918ca73e1bd18887b1c43",
-    "source/vpp2/src/vpp2/utils/lr_continuation.py": "72ea9f2af52041db19aab681d7a99fdf6d87d5ed5486f2499383a211c6c1ab76",
-    "source/vpp2/src/vpp2/utils/misc.py": "7e42c330c24604e118f1878473dc804aa8a272944df502ef1414c05be4d5bba2",
-    "source/vpp2/src/vpp2/utils/pytorch_utils.py": "9b0ba19de9a64458b12a0fb24f0811f49d9e9fca5d8b815f5f5ecd3b7886fd63",
-    "source/vpp2/src/vpp2/utils/samplers.py": "4d3bd9b2d498f4ad8f2cf9765d60c4bf5368fa6fa2a7d42d0db6210f66899652",
-    "source/vpp2/src/vpp2/utils/training_profiler.py": "91f6d98577e63d85e474c738a9bc1901edc194bdf318b5387d527c31b6708832",
-    "source/vpp2/src/vpp2/utils/video_caption.py": "d9572deb3690aecb0f16c052016cedc5f94da8ce8f016b4dff1a4e358e18d5e5",
-    "source/vpp2/src/vpp2/utils/video_io.py": "9e9898b021a44e87aa1b4e6d13e3755a9869a2e27a7eb53c72a6802808d5bd80",
-    "source/vpp2/src/vpp2/utils/video_metrics.py": "1840a37685d686336c7326a2a3d4fbe048d1e8e7e431fb14519d360a8b7aaae8",
-    "source/vpp2/src/vpp2/video_training.py": "20f55bf8b0a7dba6422a1115be1c8adf31a0db6a9f53a8f6f53f5d970188bfb2",
-    "source/vpp2/src/vpp2/weight_resize.py": "269edd4b9d32a1c7f363ee082dde7ce57f2176190daab6d5bf4a1c98ab43f0b4",
-    "source/vpp2/src/vpp2/weights.py": "5dbdff347de1b57e30433efca6da28334d0263d27aaa9b2278d95614db1b8641",
-    "source/vpp2/src/vpp2/width_kv/__init__.py": "0183a7106da7e45223550a552db2905d6b59c2a4259f8c86cb98629b03e66b43",
-    "source/vpp2/src/vpp2/width_kv/calibrate.py": "93a06910cb23a9e1dbd961ee9ef9794d70aaaa6186b08483260c8215109598f5",
-    "source/vpp2/src/vpp2/width_kv/common.py": "45f5f7087f5f3dc067358838e3fc738698b796defecb2ea721545d275f767766",
-    "source/vpp2/src/vpp2/width_kv/diagnostics.py": "cf184ae58f5e1c5f4453774eee1e10c85931e618bbbf768ab42f649cf301f5f1",
-    "source/vpp2/src/vpp2/width_kv/reproducibility.py": "89c5377e1314dd6e8de7f67c4385d30f5bd91fa4a5742449f1c1ff8dd578f451",
-    "source/vpp2/src/vpp2/width_kv/scoring.py": "7644358d0b2d29d3b91515fcea56090dbd2a7a0dad2d5e5c88e8d585bff15533",
-    "source/vpp2/src/vpp2/width_kv/slicing.py": "856203c14e0b06d19585a96392d00a00703779ebdb9dbdc9ab1323d50daf8ba4"
-  },
-  "suite": "memory",
-  "update": 1000,
-  "validation_gt_action_mse": 0.0005726480308242441,
-  "video_parameters": 12468409664
-}
diff --git a/evaluation/open-protocol.json b/evaluation/open-protocol.json
deleted file mode 100644
index d82958f15a029dd72001be805a329660aae81198..0000000000000000000000000000000000000000
--- a/evaluation/open-protocol.json
+++ /dev/null
@@ -1,706 +0,0 @@
-{
-  "active_policies": [
-    "vpp2"
-  ],
-  "caveat": "Same checkpoint55 policy/simulator code, benchmark layouts, seeds, r16 and new90/old10. Simulator physics are not claimed bitwise deterministic. Offline ranking used replan24 without ensemble.",
-  "checkpoint_files": {
-    "COMPLETE.json": {
-      "mtime_ns": 1793716093276368990,
-      "sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-      "size": 2491
-    },
-    "action.pt": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3",
-      "size": 4057916577
-    },
-    "compact_architecture.json": {
-      "mtime_ns": 1793716074811787315,
-      "sha256": "a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32",
-      "size": 6563270
-    },
-    "dataset_stats.json": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
-      "size": 100256
-    },
-    "validation.json": {
-      "mtime_ns": 1793716074812787292,
-      "sha256": "9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e",
-      "size": 88781
-    },
-    "video.pt": {
-      "mtime_ns": 1793716059522133613,
-      "sha256": "fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb",
-      "size": 24937214042
-    }
-  },
-  "checkpoint_receipt_sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-  "execution_order": "one report task at a time; eight shared GPU workers",
-  "fresh_policy_per_episode": true,
-  "fresh_simulator_per_episode": true,
-  "gpu_sharing": {
-    "minimum_free_gpu_mib": 61440,
-    "mode": "shared-shard",
-    "one_worker_per_assigned_gpu": true
-  },
-  "id": "92558f946ce17044692364bcb3515da3925bcf3c646266c8dc9ce1f38771384e",
-  "inherited_results": false,
-  "input_hashes": {
-    "inputs/episodes.json": "3daaf3e86b2f7dacad8fbe063efc23aa06d805a5c50a60cac347395fec87f4a7",
-    "inputs/master-episodes.json": "3daaf3e86b2f7dacad8fbe063efc23aa06d805a5c50a60cac347395fec87f4a7",
-    "inputs/vpp2-policy.json": "6cd61ca72b51b0be68fbfee10bcb9c0556322daf8757a0ef03982b3f080d9708"
-  },
-  "native_inference": {
-    "horizon": 32,
-    "replan": 16,
-    "seed": 1,
-    "shift": 1,
-    "steps": 10,
-    "temporal_ensemble": {
-      "enabled": true,
-      "new_weight": 0.9,
-      "old_weight": 0.1
-    }
-  },
-  "native_scoring": true,
-  "order": [
-    "vpp2"
-  ],
-  "parent_evaluation": "/home/praveen/experiments/vpp2-recovery1000-950-precision-20261103/update_001000",
-  "parent_protocol": "266df4b5d9b8eac16145600a8c4f9223d15e03285d61473c3bbd8c43b694f211",
-  "primary_episodes": 400,
-  "repetitions": 1,
-  "schema": "checkpoint1000-followup-v1",
-  "scope": true,
-  "selection_manifest": "/data/praveen/vpp2-recovery1000-followup-20261104/inputs/open.json",
-  "selection_metric": "task_macro_ground_truth_native_action_mse",
-  "source_changes": [
-    "native_eval/scope.py"
-  ],
-  "source_hashes": {
-    "control/simulator_recovery.py": "fcee066ec9e75ec081b0a955b136e2ab898cd1ee529a2c1e74a453c350c05367",
-    "native_eval/__init__.py": "2deca42c85edcf3d80dba72439ad95641937396e27e1149d77182cfc082267cd",
-    "native_eval/body_coverage.py": "50ce610f86584965281893ff2f98a78643c7e3207014212aad4c13fce4c8a782",
-    "native_eval/camera_group_cleanup.py": "8df69fdba913ec41ff7b1256ab1b43921a0b6fbef7ffbb6dacbf2adaba0f7843",
-    "native_eval/camera_lifecycle.py": "b0952d95d1da8a1bd3e9f544a2eb1eb0f516159b3442666c7acd9bbaee71687a",
-    "native_eval/check.py": "9587d7b200969ccb78ddf750dd3d8616d5e2245e29baa3a0d798e27a55bdaa5e",
-    "native_eval/compact_worker.py": "129a2e26a85e021f7a0282201dec976104a8e1ea1f0a67e6b75ce760257ee46e",
-    "native_eval/completion_integrity.py": "3b737f6dcf9ae9908b1a3602f66fda46c3b8399ee47b1c61f5c0fb9741031029",
-    "native_eval/criteria.py": "97ebc94d91c62d793e614f178d2d120df989faafb54881dd1d3f5d703df8bb76",
-    "native_eval/diagnostics.py": "53b0deb571984fd02ff65c400d5f3b9bc2754b8207ea8edf4e5017c3588ed3cd",
-    "native_eval/ensemble_check.py": "0bef52411b2ec2dd7e1ad21b8690b1f88ed229107b6e46b8dc43c34ca92a3572",
-    "native_eval/evidence.py": "cf7f769cfd838227b26101d7656ebcaede1b1d45a80785a32efb26c9befef0e9",
-    "native_eval/frame_epoch.py": "7b1d3b10f46d5acf82f6e983ecd5f8995310b2fef9fee04bfae168fda02e3efc",
-    "native_eval/hardware.py": "e30ed3359d91595d3053e24124445498a3f1be9d5c0d52f97e1e7d7d88268a5f",
-    "native_eval/hydra_frame.py": "c1aefee86396025b405e050597af7b67bb07c0dd1e65ea4a9be1ed78f8d99072",
-    "native_eval/particle_evidence.py": "378d667b2ab80176045e0bdcdbe3cb39adb612a356b8f6b81d13ae531bde711f",
-    "native_eval/policy.py": "e1cebbfb03c137bd4bf9f054c69be74f9a7f1df5f7ca13c250e606c2346b55b0",
-    "native_eval/prepare.py": "486df6d062098ea07ac63c4d3be51b5f27baf524d0d004411275243783d02ac3",
-    "native_eval/process_lifetime.py": "7dd20bee91c51546b6da067b96012dc5d5a013e8389dafd567b72c1f742cf97d",
-    "native_eval/provenance.py": "8fc2da1678832be9e9b8b778ff0dd1604db5d464cf1a19ff6137acf6b83d3e11",
-    "native_eval/receipts.py": "9ba5800e6e7e796937992117569efe1b3d625900a8e2ebf7955940a4e8cc81e0",
-    "native_eval/replay_receipts.py": "1c60f9bddccf7245e32df97b9e90749bc3ddbe5d65f37ed2d40ba339b458017e",
-    "native_eval/replay_server.py": "1c2c19bae8d1cd3ff564f1775540718c33293663d30161ea7bd293b8f322cab3",
-    "native_eval/report.py": "9d9f0ddc794d2af2ad6e349f991450592a6fe98db292e1ac3ae4fcadea961fc4",
-    "native_eval/repro.py": "f0e4ecd0ba009774105cebeba31f49a70f34c4f88f11855c939368027789a928",
-    "native_eval/reset_coverage.py": "3c6a414878c4cc2ef4114170356db0a6b2193b03ce5c3b4f6293c10c7d3138b0",
-    "native_eval/run.py": "23d2d58fe1160a03bdcf9591b5d5fae671d21832008cc4958ac58e060b88caa1",
-    "native_eval/run_one.py": "dddeea1abb269293bd08c3f5a69ea0a668620b18d07154ca1c0740a0bb213b99",
-    "native_eval/scope.py": "0f2738c8fb3c682eafed843eebc5b9101eedfbbec64fe85b03575347ece0a530",
-    "native_eval/simulator.py": "d40972be52b7544f62653114217e213d19d4b386bf2991abfbbbc618ab09cf83",
-    "native_eval/status.py": "b49d03681dbc28b80f39f4d6163c3a6e7c5a1ca62f1f0ef502c77dfbf149ca02",
-    "native_eval/topology.py": "062e9246890ad1f2e8bb8b64b3bbe6598f9dce6fcc33ec3eaa106f3cbc0f0ae5",
-    "native_eval/transport.py": "eb6cc295d064a142372dd51859e91233f61f1ad2c8ac8bf3880739f3fe4abec4",
-    "native_eval/vpp2_pilot.py": "571ca3eeba11fbf2afd4a4058629a280bc0ad4bd3eecca178ddde4aa543bd8de",
-    "source/simulator/LICENSE": "7794bb06af8fe5485ca912454ad1665ccd7846c45f9c2b1258658e480948cdbe",
-    "source/simulator/XPolicyLab/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
-    "source/simulator/XPolicyLab/XPolicyLab.py": "a3909558697b73479f8baee03e15b4d74060a7c9ac2f31fcbc6cd659e7364862",
-    "source/simulator/XPolicyLab/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/client_server/__init__.py": "093e47e707f4a5598b0159b5ee371cd9a918b802c70a6e66154a18e9ef8fc5db",
-    "source/simulator/XPolicyLab/client_server/ws/__init__.py": "79905987a12ebf318dc0b2b9ec6f375d9b4561bac29d6c27d04e318d8f6e00b3",
-    "source/simulator/XPolicyLab/client_server/ws/__main__.py": "ee1c5807dcdf1c8f6721c5447463eb51babdb0f9ffbd91d1e7343834ca4aa19a",
-    "source/simulator/XPolicyLab/client_server/ws/model_client.py": "4d6e225ae5f577d862366350dca8742e91dd9527d6e76b8b2a6d48c991212d67",
-    "source/simulator/XPolicyLab/client_server/ws/model_server.py": "4988823b436cd05e6a4d8b13012cc30c7ffd57711b72497b1223a6b10dc0c043",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/__init__.py": "19aeae375e94e58d437d789d0aff07b232c13f375235384d784984631d25d5d9",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/client.py": "b309f58b1bddd99e56578026b61a15e6cfdc07e843e8a082bfd291d0e853ce5a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/codec.py": "8a97ef7e102fd954c266352d81eaceef0f1b0f5504571c03037683efe387865a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/exceptions.py": "73161f4cdcc7ffc8a87cec94f5459bfcb9113076be272d6cb2a3ff8d69e0c6ce",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/keepalive.py": "7f6b9e636eefcd6c80a4b62e404a6f89e2f7e1a3b6008500312a972d6990a21d",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/messages.py": "d05c8cbc20e0cbe1502551864907f25897035993f920748c9290c944c959eab7",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/schemas.py": "f1ca82eafa8dcad6d968508adb51c25342b291f7b1c553fff305c381a6098bd0",
-    "source/simulator/XPolicyLab/model_template.py": "4f8d1eeb8073aa7ae6093abf0f395cd5fef70ff4a4bd1fa6b2d83b356a90754a",
-    "source/simulator/XPolicyLab/policy/VPP2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.py": "ab94a8569e6168ab7cde6660890035e11b34961ff2648854a7b61d188c555b6e",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.yml": "d2c3f06eb2ffa2d1866767dc20db8d60e542fc2418dbc4cf86d4a0f86f944c5e",
-    "source/simulator/XPolicyLab/policy/VPP2/model.py": "699acc0c459c5997db4e3155e2b1b7eb64bf750552436074a369c4eb6e9ffef9",
-    "source/simulator/XPolicyLab/policy/VPP2/temporal_ensemble.py": "42be4b09c56377f4686b2e722de760fa3bafd1c6b92605ada269d94e19efc95e",
-    "source/simulator/XPolicyLab/policy/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.py": "42dc75140935046dfe83f19b3834bd7cf8372e1bdbe51f598b1f91b8d557b7fb",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.yml": "e59de6edeee891d2aa24902ae9034e9c10889fb20116bc9263f30d773a4b0d10",
-    "source/simulator/XPolicyLab/utils/get_free_port.sh": "f45bcb327b45c91367e3e5cdff72d3a42bfb7a09f50e4774701637c4c23fc1a4",
-    "source/simulator/XPolicyLab/utils/load_file.py": "488590d76084f419a36c2fb3e658eeb240ae554678899e9f2ce0cc10da5c4b57",
-    "source/simulator/XPolicyLab/utils/process_data.py": "e044e627eb2677907e9227eda8458c857208913bd484c5e922b53025010eef0b",
-    "source/simulator/XPolicyLab/utils/robot/_robot_info.json": "8a5df36e750971003b484ad770a7a243e523e986d2977a4ecdbeecd4a07946bf",
-    "source/simulator/XPolicyLab/utils/wait_for_policy_server.sh": "f75a8360d83f2960f365dae77804c94a28aca332c2fbc62cdb4a39ae220374ad",
-    "source/simulator/env/camera_manager/camera_manager.py": "1a4250cb3bdc7305fe362a7c6b113f57e6f3e9e4983b69bd61e2c84e8e995471",
-    "source/simulator/env/camera_manager/capture/camera_view.py": "be9ec0d4df8a532c35adc7090fbd7d6a76c163f90e8dac30496ee4f875929869",
-    "source/simulator/env/camera_manager/capture/render_sync.py": "48d70e9a95f3445ae464ab7ea00a00c0d4829752eb55c4ce52ef73ab1f453cae",
-    "source/simulator/env/camera_manager/capture/tiled_capture_manager.py": "959b4d87d96e676bc279d01e7755557bb8c6e73920a75944259f72b7ac385f8b",
-    "source/simulator/env/description_manager/desc_manager.py": "bb566028b523a4cc27363cee2e526c2d3e008e671c9773d4f0b8401f58f0388e",
-    "source/simulator/env/environment/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/env/environment/base_env.py": "ade58ebc19c79bdbf720def30786a0c397b1c7d2141fb6a1bfd63c80f8a3db8e",
-    "source/simulator/env/environment/isaac/__init__.py": "123023ddc75a0f9b1139c7322bee64a27f0e756b6e84242f43a1481428596126",
-    "source/simulator/env/environment/isaac/direct_rl_env.py": "77958f14c465932697efed302c8ef13054f2ec61ee3a4bc5ad66d0390b71df77",
-    "source/simulator/env/environment/isaac/isaac_rl_env.py": "144af40adc2113751bc1bd72caadab145e918b316de53ef3e4e42358c01a1fbb",
-    "source/simulator/env/environment/task_env.py": "ac816cf26efabbb8ac4276d2a2a09fbd557e3db805514db0c3dd7ac22b58e931",
-    "source/simulator/env/global_configs.py": "e96f6141b42e635d69bba5b4727fb1fec54f8965431b3a70d95fe5dcd3d0ee45",
-    "source/simulator/env/observation_manager/obs_manager.py": "771d6cda443b591f3e7850b901a06b4e653a2feefcbabd23608042377fe9be4b",
-    "source/simulator/env/planner_manager/.gitignore": "bdcb46018aafd15f5c2d40adbbccbb06f2dec5b9ec4311cbf776075bdffd4156",
-    "source/simulator/env/planner_manager/curobo_planner.py": "3cd830c002901959409ef6f64d54106d8106b48471b266409abc2199f3cb33e1",
-    "source/simulator/env/reward_manager/func_parser.py": "eb928a05c3eb6bae429a922c59e64272c2c3851c22d779143d45627e47b3a504",
-    "source/simulator/env/reward_manager/reward_manager.py": "bc56efd3aaa861b4b38960dc4fb7778a9b4da9cf131f815ac3170300dd233a41",
-    "source/simulator/env/robot_manager/.gitignore": "fd94daa58076711ae84376170199a63f66177471e587a76252399fe79eb231e2",
-    "source/simulator/env/robot_manager/control_manager.py": "81b7ecc3bed8248c16aa1517e686682d8e28ccfa9e550d45011b579a9219caab",
-    "source/simulator/env/robot_manager/frozen_robot_configs/franka.yml": "a3c944f8cdd07fc1a4eb54cb915f565763292acb24f9c4f725534271449f138c",
-    "source/simulator/env/robot_manager/frozen_robot_configs/x5.yml": "4df90d234656c34132ca573e69a55de5984270e103aac2ea8ab9956a913685b5",
-    "source/simulator/env/robot_manager/robot_class/franka.py": "dc49eb7738a8b58fb731595340c60a1ce32f52ef0dc63a3da84f9c25879af140",
-    "source/simulator/env/robot_manager/robot_class/x5.py": "4748cbe4d273449a48d09817767dec361f53f8e5cdc6e43368fbb8724141b654",
-    "source/simulator/env/robot_manager/robot_config/franka.py": "8bf57a07913573e74dad97aec6c50e58b67746fa90875915d74ce213856ce35a",
-    "source/simulator/env/robot_manager/robot_config/x5.py": "13d35630f833d1b0a1e3df0bea05f843e88aae1248b59af8aadfefa9cad29bbb",
-    "source/simulator/env/robot_manager/robot_manager.py": "19636ff932a0e2192341b57d8695d4102b870bd10c69c9343d712682fcc15082",
-    "source/simulator/env/scene_manager/layout_manager.py": "1b4c940200dd94a1d0300e103d168b9a823e0dc00ce7c921afe16e81d4e01ddc",
-    "source/simulator/env/scene_manager/objects/articulation.py": "0c542a1fc8ba7a61bf71b343745c71949c863291c1b167d4d7048fe2ef9ca4c4",
-    "source/simulator/env/scene_manager/objects/background.py": "04afea9b5c4e21a105ba3b08ce0e5b6692c082cf17ec1053fdffd5dba1c9511b",
-    "source/simulator/env/scene_manager/objects/dynamic.py": "7a4652c649f503681914e37c50c959c7ca8457918bb00628413f4288993f8ea7",
-    "source/simulator/env/scene_manager/objects/fluid.py": "12b183d4d798d6b7188fd42747ee1927747318c386a91135551d1fa0561c41b5",
-    "source/simulator/env/scene_manager/objects/garment.py": "85f38bd85b52da8dd2ce34ce3692d45848016a9f0f14ecf13114705d4a1ff46a",
-    "source/simulator/env/scene_manager/objects/geometry.py": "3100d61b152b894b3b59be0b3be1d2708f2d8f363ddb8311dabfb5411c731f3d",
-    "source/simulator/env/scene_manager/objects/ground.py": "12dd51addae6f5b7c6ad410b4ae7248231d216cf63513c564057d577a09a2972",
-    "source/simulator/env/scene_manager/objects/light.py": "ba8c8529c6608592d4d0eb9638820b70ae6b3c0cd616631bc6dabdb1c20e421a",
-    "source/simulator/env/scene_manager/objects/physics_material.py": "a1ac0dc9928e9253e644ceb9a604aed262514aaad0473aff303bfae448e76c29",
-    "source/simulator/env/scene_manager/objects/primitives.py": "848f4ab08d86307812ebaea18cc86a51638395f0e6ea194b412d45a699ced926",
-    "source/simulator/env/scene_manager/objects/rigid.py": "8260fd2d8350d6a354c8281685bc7e25531923327078752efe0538d9cd78450b",
-    "source/simulator/env/scene_manager/objects/room.py": "946aa1afc23ab9b8fd17394517a5f54605bbbc609582d0854fed2078b3b2d8ee",
-    "source/simulator/env/scene_manager/objects/table.py": "c0e71396b0d53a250f74c75490e91086d67962237400a49f886a13ec07875308",
-    "source/simulator/env/scene_manager/scene_manager.py": "37911b12601038c9ee9bd180cae7a4751ea649bcdaf1ba2628f22c65cfd7dc72",
-    "source/simulator/env/seed_manager/seed_manager.py": "99bcf5a90392c8927aa0dbd799c07e06b9b03befe8b8eabaa8744a3dc9605ba4",
-    "source/simulator/env/seeding.py": "3f88d2a635c528253439cb31a29e5b9d0f47f6aaef0e975d86d9a8cf174be26b",
-    "source/simulator/env_cfg/arx_x5.yml": "7691350159fdeca92fd1b0b2da36913464ccdea3f3251eefb0176aa51db0807f",
-    "source/simulator/env_cfg/camera/camera_config.yml": "c1752170da855e735beba27fddbf78dad16f2395e4ef8b24e9827a768e616a62",
-    "source/simulator/env_cfg/camera/template.py": "68287ed1daec3b8ed7f053bf97a4d48bbddeb8b5757b53e438ac3c4240a9e61b",
-    "source/simulator/env_cfg/robot/_robot_info.json": "4c9de628bc1572502fc48d876c023fc63e02760e1a11a2a5258dea30ffa330e8",
-    "source/simulator/env_cfg/robot/dual_x5.yml": "4eefaccc75c5f83443b7bb85ef1de68ee7a3772e27af2b04a5e8d8769ab9baa1",
-    "source/simulator/env_cfg/robot/dual_x5_and_franka_competition.yml": "ec893f9fde577d0d1209e7dcb85ec2dd74288b636d0c68316cb07bc817b743ab",
-    "source/simulator/env_cfg/scene/conveyor.yml": "c1dfd27faf978f90dfc4825cd83bd203823547c46793f4f9201ce9f4b43beee7",
-    "source/simulator/env_cfg/scene/default.yml": "729df024b9707a55449da94f980c1ca116035e09a1284e2fe071f1b38efd061a",
-    "source/simulator/env_cfg/sim/sim_config.yml": "e442ede1c3d90fcfb50b9461c854d31b22dad9bfbb23397db228b27a6a486963",
-    "source/simulator/scripts/eval_policy.sh": "786966648496b7b68aad81e000a3f92910302eb391dbd07fec0a2dc91107956e",
-    "source/simulator/scripts/internal/summarize_result.py": "9ec361af587e1788d83d3cf8670e05b817ff252d33c33829a6337dc16ae370d2",
-    "source/simulator/scripts/internal/task_inventory.py": "ec82252b0f87fa5a3dbec2f682c170b25cbe47fe181e431a81df2c4b0675084d",
-    "source/simulator/src/__init__.py": "75bfbedb80f2e1421ad787992d78c23349bc8738b9edef7a76a3af83c9586fd6",
-    "source/simulator/src/eval_client/eval_env.py": "c577b27dab3c87588940670da81bcff007cd64b687eee8e0cc6684e5e24966b1",
-    "source/simulator/src/eval_client/main.py": "de2216bb5df707e505f7e2f3d31702907a4c1ab65205c5ea35494021f168c80e",
-    "source/simulator/src/eval_client/physx_warning_monitor.py": "6a9f8e8d75b1031d2cf2feb6dd72d37370d85d5bd80c4bea798622e23ce8dae6",
-    "source/simulator/task/RoboDojo/config/_task.yml": "655a87cd59fa6272bf352d32deb8ade162737b49a9824e16d428fecf0803380b",
-    "source/simulator/task/RoboDojo/config/align_blocks.yml": "082c9d7a32968052b8ce954bee212c7ec0a96b756150701d4b531abdea59be2b",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number.yml": "3bf6b7e286ea328fb1e4a994a8d30af9afb4b0b4dcd7ceafed43cc6160c9978f",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number_random.yml": "145b26eb5c8ed7d084a4ddc6e21694c542f4a86150756188ea38dcf66e565603",
-    "source/simulator/task/RoboDojo/config/build_tower.yml": "e1bab8756bbc5664e3171e2389c060a79a3fbbf92a6af4ef555dcaf59fce275c",
-    "source/simulator/task/RoboDojo/config/classify_objects.yml": "80a754d2aedd6a89f58185eae36f5b2eb1a28250f40a67325f576fde34650384",
-    "source/simulator/task/RoboDojo/config/classify_objects_by_language.yml": "ac6c739ca82a64b53c799af79b44363db1a88246f76d867f8830729e5613507b",
-    "source/simulator/task/RoboDojo/config/cover_blocks.yml": "68f90c85fad2b152cf7853cf48888d09ce74bb27a692ff2feacd11160771c5a7",
-    "source/simulator/task/RoboDojo/config/deposit_coin.yml": "f547b12216fb9a5dfe2e7d7e882d89e1e53ee1a805c943ba09a3c2cdea8c17a3",
-    "source/simulator/task/RoboDojo/config/fasten_screws.yml": "5fbe4cf5b3ee2eafc2771137b13176f01617f8b2131f02503e676a7b62e2eabf",
-    "source/simulator/task/RoboDojo/config/fill_egg_holder.yml": "d79129278c9a5488d63bb1d3865f2c6b39f21f85957ec94ace091af9ae29e31f",
-    "source/simulator/task/RoboDojo/config/fill_pen_holder.yml": "ac72e5582495e8757c547997b317e2fe990370009527b1faa6bfbf4e1275537c",
-    "source/simulator/task/RoboDojo/config/fold_clothes.yml": "56447e3f592e0db388ebb7482ec758ecd5485ec81b9227c2dffe5390bd7d8c8e",
-    "source/simulator/task/RoboDojo/config/fold_clothes_random.yml": "eafe52c0abcc1889069383b008cbe95c6af2d18ceafb1aecc78f0fae3b491b04",
-    "source/simulator/task/RoboDojo/config/general_pickup.yml": "35de68117cb0ed596b30db1c62f1f9090f2cd1237242a03d5aaeb7186e791d93",
-    "source/simulator/task/RoboDojo/config/hang_mugs.yml": "0086d374d3f7e82abb85e43a17cdfa437e1ac645a0171fb47496c2b81427dce9",
-    "source/simulator/task/RoboDojo/config/hang_mugs_random.yml": "847dfd39059275a671b5beff0243bf7bbaeebd9e5c0612a1877cdf12b9007dc6",
-    "source/simulator/task/RoboDojo/config/imitate_sorting_sequence.yml": "d9e194d8c742288601c641ccb75052a67da46e536c0b06c7fb7311d9e2c61a42",
-    "source/simulator/task/RoboDojo/config/insert_key.yml": "c0b409d5c020e0ad706c378b6f63b39dfe7bed4682d978907c40e1cfa7623335",
-    "source/simulator/task/RoboDojo/config/insert_tubes.yml": "5c070e5c764a2ea93f8de0b52e6a664a23c717963fe09fe3af17918a4a1ee9ef",
-    "source/simulator/task/RoboDojo/config/make_kong.yml": "8a33c9e57b6e6c719c4800911c37272ba055ad0f4cda2f1de9d531ee2910351a",
-    "source/simulator/task/RoboDojo/config/make_toast.yml": "d190a26233db174538ff0052b4c2958a3b107f0ccc63d0fdfe28da4c7f5dc57b",
-    "source/simulator/task/RoboDojo/config/make_toast_random.yml": "ca2c1e55a0975c8b019523710e90131504d8d2afeb829be362e4d7927fe97b48",
-    "source/simulator/task/RoboDojo/config/match_and_pick_from_conveyor.yml": "0ae1c2537d9b1880748316928cfd7ba4c6421ecccfedb85c4c5395a7a86c5361",
-    "source/simulator/task/RoboDojo/config/organize_table.yml": "9ffa15dd279060d4f042285a2c53a341c7208e2f89a2f3dd36ee08b22aa453fa",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box.yml": "e232962b8394d08c04c5da14deb88d7b233afa00ab99af8001c084218951d7ca",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box_random.yml": "7ac7f034fcb169e33a7f0a03b2b77e54d0206ed715409359b579055a701196ef",
-    "source/simulator/task/RoboDojo/config/pick_from_conveyor_by_image.yml": "6b310b70a0ead9d09e7c221c6bbc952140a44b1d43034613bf02e954eebcf9dd",
-    "source/simulator/task/RoboDojo/config/play_Xylophone.yml": "28adedb0f03c49693af32cea37855f3f95ec3f76f031bfd571fb8bfffb2ec59c",
-    "source/simulator/task/RoboDojo/config/play_stacking_toy.yml": "a09afc106161064a2ea6a40882d717f7872bd4875309a97431d6863a9b7ffa1c",
-    "source/simulator/task/RoboDojo/config/play_tic_tac_toe.yml": "63f863a7fb81a040f65803c64f97c4e9577d4fe32cd0f92563bdc72d1aa336d2",
-    "source/simulator/task/RoboDojo/config/plug_in_charger.yml": "2046b47112049e038ecfa24a8c8044fb1db84f629a4a32eeb24822157d2d4dcb",
-    "source/simulator/task/RoboDojo/config/pour_balls_into_vase.yml": "084a707af61d24123912777b417fe0c0b230eb9653e686af42570791e924ff8b",
-    "source/simulator/task/RoboDojo/config/pour_by_language.yml": "efb2e3c3cb08dd6d2d2ae011b3e9a218b5f1f0a92f025b8e3b9e8abd7011c250",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup.yml": "32092b11667b3306579b04225303610c061585ace28531fa9996afc339c5a2ba",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup_random.yml": "1603feeab6f0f7dc9c4897e7f40f322b86112b6ab987dad0598f42ac1b35db69",
-    "source/simulator/task/RoboDojo/config/press_by_number.yml": "b784643495bcbca933f5a5fbc91b7dad3465fe888239b358d516977c78c38f63",
-    "source/simulator/task/RoboDojo/config/push_T.yml": "b55ea11002ff3e5dc63982ca8c1247fd7bee4f4a607f4d674e220b86cf120a25",
-    "source/simulator/task/RoboDojo/config/push_T_random.yml": "07609ec20cea40afa1d5720e895fb1b445411f261b2e5b6850e75d0f1930d32d",
-    "source/simulator/task/RoboDojo/config/put_bottles_into_dustbin.yml": "7eddf43e287b6e8aea686c6d686fb87e13ac1be1cb5aac16a8a2b0d2e9eb976e",
-    "source/simulator/task/RoboDojo/config/solve_equation.yml": "bdaad2ff1cc8b7ec6413a3715d49ea9cefd664e7e204544977832e18785f1e75",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size.yml": "8927cb2ebb18449554d851e629bd7d568e12e46c5c3c636c8e7122f4fab1ed1f",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size_random.yml": "f895d3ce7eaae99454706afccd09011f56d077998bcc1e95840116658b78f139",
-    "source/simulator/task/RoboDojo/config/stack_blocks.yml": "2802e88ebc3e2d3d05ce345d75e9ad6151a6690e4ef1e0021a5223d291a402d2",
-    "source/simulator/task/RoboDojo/config/stack_blocks_by_language.yml": "4fbfed4f7fc49120831087e709a8dbdfdb4504eb143196000d5790b54eb3bc40",
-    "source/simulator/task/RoboDojo/config/stack_blocks_random.yml": "e25d2a096601da30abdcfe1256a464eaf48c0c8f844eb2c737232626d49639d5",
-    "source/simulator/task/RoboDojo/config/stack_bowls.yml": "4e2aaa12b8b0f58ca801f1a93638c46d3b49ebad809c5a8011382ccda846dcb1",
-    "source/simulator/task/RoboDojo/config/stack_bowls_random.yml": "5c08af1bcc925d958d53de51e8ebd336006203f1bc81e4ef68f3719eadfabb49",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones.yml": "0fdaab803a9370a9e2d1fc8d86a5788daa3940c08fd6f767334f625642fca3e2",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones_random.yml": "93f0ca39bc8bb32df94e84a02f22ed4d7860f37cdc0b9dd17e44721445849576",
-    "source/simulator/task/RoboDojo/config/store_tools_in_toolbox.yml": "641c8ef92393bd986e408d26a84b1cfcdac27466a5ac5f0b3ebe4bb84430c527",
-    "source/simulator/task/RoboDojo/config/swap_T.yml": "1daf769911e13714b4aeabd208292931e32f4dd39968893152bb7bb477c46c9b",
-    "source/simulator/task/RoboDojo/config/swap_blocks.yml": "b6bb2a9df494719f26c6b1aab2e7576107ae76e2fbac8b63a5770f21e0e9f05f",
-    "source/simulator/task/RoboDojo/config/sweep_blocks.yml": "98aa25ac39be9d335c71b2b586bbba372789e9d61e652e1147152c87febbf1f5",
-    "source/simulator/task/RoboDojo/config/sweep_blocks_random.yml": "4ebff5e2cad07f958dc4b2793b86b1aa803c0809394818ee3752d300e58b6187",
-    "source/simulator/task/RoboDojo/task_registry.py": "11218028f772d9253ee709a7d85aa72d1e1fd87e4f8d33ee3787d6064438f9ab",
-    "source/simulator/task/RoboDojo/tasks/align_blocks.py": "c8597e96132cd7c9d30f3ae101f7bbe8ac28597b629599d75d442bbea4c6e74c",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number.py": "277acd20ae4345dc71267762e588b7b0343387c71685183890f044a4808b7a20",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number_random.py": "01757509978663bf7c0c95059e8ae18d45b8fe0953f6b54540e2bd6bf141182e",
-    "source/simulator/task/RoboDojo/tasks/build_tower.py": "814cf6ffb459871e5cf49f4e127903dbfbc30926f33db71de0b84a71ab235d6e",
-    "source/simulator/task/RoboDojo/tasks/classify_objects.py": "aef4333c088766375e21a928d53a9a93aaee285fecae41e22385acdc74dd11ac",
-    "source/simulator/task/RoboDojo/tasks/classify_objects_by_language.py": "242a8f048f487eb174ef246be78613bbb5f153549ee11a2650412718cf3b36c9",
-    "source/simulator/task/RoboDojo/tasks/cover_blocks.py": "9d14931b16842b93bea8abf9a07991a93a7e663b93c5fd1695108c8b922ab862",
-    "source/simulator/task/RoboDojo/tasks/deposit_coin.py": "f9868ef827ec17b8db6c4650cc5d9a9e593d20d00ee822e2212ce229ec01fb61",
-    "source/simulator/task/RoboDojo/tasks/fasten_screws.py": "2e187a28e33e44efb87baae6463b3ec35264ccd41e35c21a1be44f6361675185",
-    "source/simulator/task/RoboDojo/tasks/fill_egg_holder.py": "506dfb772f8882ae53966e3580affde19c3c0f11591c78cc5f5b1d997ba04f7e",
-    "source/simulator/task/RoboDojo/tasks/fill_pen_holder.py": "8706dce239487e078b21af0b95bcc5cd4eabdcdadcdcbc6400d6b138d982f1e6",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes.py": "ecea9c287d0579605330168e6f36734704fb14a6b1e5dec82b6f2ee3a56ec20a",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes_random.py": "9e40abe3f7201a8c13961edc1ca8075260c315584d3d6b93fa1df966d13d1e88",
-    "source/simulator/task/RoboDojo/tasks/general_pickup.py": "1a9eef7fe518f9d2c6805ad1a8f97c1045fa83726f84b9c63f67ffdfabbc40a7",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs.py": "4bc6586bd87de030b9c580fea2a45e7977c3b6a4d75cf3c43b546d349d42e906",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs_random.py": "ebfa4a95deb954aa7a9063a07fe366cb5ea1835d3f52342b10d531225dfeee92",
-    "source/simulator/task/RoboDojo/tasks/imitate_sorting_sequence.py": "18400cb3c5f5103492819df2dcdf05a77916106e8a503cb62b4ee6cdbaa9305b",
-    "source/simulator/task/RoboDojo/tasks/insert_key.py": "180978a2915744f0596fbca4ea0ad6dcb47dce005d43f06807249bc8b4d1a828",
-    "source/simulator/task/RoboDojo/tasks/insert_tubes.py": "9a68795ddbd0b816ffd7627e673390ecd83594c0e9d374d4e3abf3435ad439e1",
-    "source/simulator/task/RoboDojo/tasks/make_kong.py": "b9266689d65f9213b05219ad952344f9fd63be8892097ec0103af99791b245d5",
-    "source/simulator/task/RoboDojo/tasks/make_toast.py": "1047f484a73d1f92304393879b51f87e33d661b6ad7937198352fe37a0d3233d",
-    "source/simulator/task/RoboDojo/tasks/make_toast_random.py": "69fb879b89e98f7369971834eb6d1b95c5a65f5ffe978b2cffae0b12c6b756ca",
-    "source/simulator/task/RoboDojo/tasks/match_and_pick_from_conveyor.py": "53110b8a0c971bd3ae8878bfb7d0dcd9626bd9b610d4e7fc91772313b435a5c0",
-    "source/simulator/task/RoboDojo/tasks/organize_table.py": "aceb298e78f3f39e4cf9a9f4c5a4f747b2c55f1a6006de2065e110a05f4295d6",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box.py": "84a1cf1dd5e166f36291130f9d8cb5dfe4b3f4732a46674bef6374a7ea2d66be",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box_random.py": "4db259a1b97cdcc39ebc9794244fadc79d31c4ae64069c9b11a233356fcb655c",
-    "source/simulator/task/RoboDojo/tasks/pick_from_conveyor_by_image.py": "a0d29576d0aab8232d2fd3ebcc8c510e81c70b4cc04602a5be3e99ee2275e01d",
-    "source/simulator/task/RoboDojo/tasks/play_Xylophone.py": "b7a4b0bc0457bcf55d714ca1867476fc4aabc53084adfc78d3ff5d1437bf7847",
-    "source/simulator/task/RoboDojo/tasks/play_stacking_toy.py": "23361329ee7efe330da1f1c241415205bafed2b6c082a913659d868a80d338c3",
-    "source/simulator/task/RoboDojo/tasks/play_tic_tac_toe.py": "8f75ebfe55c997ad9bd3b342db46e843e87397db94356f31cadcbe0cb483bca7",
-    "source/simulator/task/RoboDojo/tasks/plug_in_charger.py": "aaa59ade46dc54f430e1148e3f6fae487970cd29eb2c07b5ff7b4045edb831b6",
-    "source/simulator/task/RoboDojo/tasks/pour_balls_into_vase.py": "eae388b93ec196c401f76bbab3b0119cb2164f90bece0225d4fb925d3f2b73ab",
-    "source/simulator/task/RoboDojo/tasks/pour_by_language.py": "f7a15a7fca2c649a7918d68960ede27e667f409a339b668dd9876db890724cfb",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup.py": "73f8c91b57c352fdc1e7ba3fe5354b3c10e4156d8e5ba777762ad0fa1aff5663",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup_random.py": "a1c4097b6e34b2c62885b3a9c7d2584a93d9e296dc41d311a80c2e6d541d53af",
-    "source/simulator/task/RoboDojo/tasks/press_by_number.py": "006f969087c00c417a63875a9de153f698c52d1f1ca0caf0aa16076aa15059c5",
-    "source/simulator/task/RoboDojo/tasks/push_T.py": "a01d753bc320ffbf492ccc29e5f7ab3b39e6c7d255cd25804db1934a30fa5b41",
-    "source/simulator/task/RoboDojo/tasks/push_T_random.py": "dc8d9c8e0d2e8dfffa2b7d433a8278c0eaf707cc4861b15035364c2e70061dab",
-    "source/simulator/task/RoboDojo/tasks/put_bottles_into_dustbin.py": "3ba6e19df0c5ff19ba961092f031ce14e91e72a9a4a5934eece6000a1ffa1c48",
-    "source/simulator/task/RoboDojo/tasks/solve_equation.py": "e287a2f36b82b28e8ca7824f63b1b723b02972ea6dfc115ff32b2882729751f7",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size.py": "1838fa6b515bad3796895cb5c1883c6a42789b7dada5434765764600e70beb49",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size_random.py": "b793ab5bd32bd2c10bd46874440904837f9c4e290e7508b06426e2e96425709a",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks.py": "50bd388f9e47002966fbb0146742467a820295b8686e0f86379f0efbcfffcc1b",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_by_language.py": "36a83925f35c7a4caac5add328dd35cb1d346bad833d3064f455d88221e7a41e",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_random.py": "19d4e5cd693e068ad64de75cbcd24977ab8d39e22fa17362a391dcd733bc9bb5",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls.py": "022c129ec3493183b1a26688a0402f5a00a1546ae011999e9c19f4c0029e2b7d",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls_random.py": "a0e20469eb44a9ee98650d09386814c1df5264c68ec139f06aabf0c86f092514",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones.py": "2bee4e31f0594900629fe7f43266e16fecb310b152930ce51524507136809a92",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones_random.py": "58c3c5113054a379214d64ae406ed8dde99603cf684ce9cd8c7bcb72ed334382",
-    "source/simulator/task/RoboDojo/tasks/store_tools_in_toolbox.py": "2969fc3eb171560b3c0bcea62b9f38b8b35f3aecc9bcc7085c3b1cec365b48fc",
-    "source/simulator/task/RoboDojo/tasks/swap_T.py": "57fc289b3957caa454e1d8e005e3a57c973d6ab40121845c94799e61e4bfafdf",
-    "source/simulator/task/RoboDojo/tasks/swap_blocks.py": "08ed24ae9dff27f4e5c850161a02f62ce237859c096d9b06a853a7890e0ad4ed",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks.py": "65786217e199c1e2b26ec8e2c232bf98ecb304a20939d0908fe12bc41e990614",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks_random.py": "68b6127038322cf001d5f614060b8a1d96e64a49efddfcdcae91127d5c206bc5",
-    "source/simulator/utils/cluttered_generator.py": "f59cfcfed2b4cce5ae9aeaf3339c86d17e237874e3526bded3cb05136dbcbd50",
-    "source/simulator/utils/ensure_usd_path.py": "7c8e71cf91636e630b1b174098fe0b38740857daeaf207e2fae772c4727b11bf",
-    "source/simulator/utils/load_file.py": "ef3a6dc96aa1a7442cfd24f7be9671052a98d2f20783c3e71032332704132943",
-    "source/simulator/utils/path.py": "1d427c7cf33e748e01d0d083c9432d8be8b5c3a1f35c5fb0f6ebb0714098f1d1",
-    "source/simulator/utils/pipeline_utils.py": "d5d8da29e6ca84ab04323a2f79130f930572785b4399634e06a42834ed4087ae",
-    "source/simulator/utils/rotations.py": "8e814f79938bdb9e85f1c024825617cc08d281b61728c6410b23ef1bd852c340",
-    "source/simulator/utils/save_file.py": "6cff5950f73bf7a1ed404f804d6f08af712974f22c3955de4cdf142f211aaf00",
-    "source/simulator/utils/transformer.py": "73de4f5e1e0ff60d4f2e1aaf6104368e8f919269facb68b4fa0344d35bc08cbe",
-    "source/simulator/utils/usd_schema.py": "260d30b0ada2976136cad3e49e810b4409304eb4a2bc60bd9ed710864d3953ac",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/METADATA": "614985b278f601b8ef863f82783b991cc3ada4b89f592eb71c7c48c5f2ef2298",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/RECORD": "c7660eb223c8080f75c49690eaf0a9c9c00b6195511791b33b15e6b24e42edd7",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/WHEEL": "942926c567d0b1273d09f20295879abd37507845f6b683eab9f87d808df91100",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/licenses/LICENSE": "fe1049884b1a0d9342901e88e07f32925d24b3121d9972b6a6805fb9824b095d",
-    "source/transport_deps/annotated_types/__init__.py": "a7104a4d439b27a9f74fc0be236b9ba1b7831e6044026802a205abc1298a9bc8",
-    "source/transport_deps/annotated_types/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types/test_cases.py": "d867472acb57b81ce97f78972828e6bc5669ca4db33fd1e6f645c6183846068e",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/AUTHORS.md": "5fe0446ddd6d1c70fa298747368582dc1d91c23d62c8fb7228595bc87fd53232",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/LICENSE.md": "509aa7af19ae2b5b1681b22b969627ee9e1fd15ff7ea5bf9f6e301fc44d5a6c2",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/METADATA": "b7ee8fa1bf018747082cd6d2fcaa14e5d0f5956b5477ab5810eb61ac4cc7bcf9",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/RECORD": "51331727bf8d9ff60510de5e7756c5dd72aba7663f3673e2827bf1ca5c09c7e3",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/WHEEL": "1d7faff7e9e8524c94a31c99d4f312b92edab94c43011e1505da182ea0f4c70b",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/top_level.txt": "19da698ad1490bdde1c7f18fabaceb2625aaf48b364d4dd7ad404c7e5eed75b9",
-    "source/transport_deps/msgpack_numpy.py": "0a45a7430e2cefa2e74235c4c6a34a22ead2fb57a7de8b2631f45df977701d63",
-    "source/transport_deps/pydantic-2.12.3.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic-2.12.3.dist-info/METADATA": "348336c208743aafeabfa5537c29f429035a3107b7a78bbb6a3bf7b3a42fe4aa",
-    "source/transport_deps/pydantic-2.12.3.dist-info/RECORD": "089c7ceb033486c4e6106749bb1a38c80b63905da74d842d067735e8b9c86e4d",
-    "source/transport_deps/pydantic-2.12.3.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic-2.12.3.dist-info/WHEEL": "aad0b0a12256807936d52d4a6f88a1773236ae527564a688bab4e3fe780e8724",
-    "source/transport_deps/pydantic-2.12.3.dist-info/licenses/LICENSE": "a9e186f3ca16b5eef84318e7a701721351a00cb7b8ae3a4394b67b49e3529ef3",
-    "source/transport_deps/pydantic/__init__.py": "e62127278c07bf5384cdd2903f368a69929f3b8a524000bae4e0eb608ebf4bc6",
-    "source/transport_deps/pydantic/_internal/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/_internal/_config.py": "4d66708377346d91e24f76e8479f9862a928b877308d175e9e6c861e87d5ec4d",
-    "source/transport_deps/pydantic/_internal/_core_metadata.py": "63f836b778bbba5b8afb05c266fcc97d114c3d4336dda0582df69ee05cc13f2d",
-    "source/transport_deps/pydantic/_internal/_core_utils.py": "d63aeee156c9d31eb747a76d55cb8e23e74a4130bf77f9529c9584050cc63444",
-    "source/transport_deps/pydantic/_internal/_dataclasses.py": "4e4d6611a7e169dd6457b2b9b4f5f9070c5649763b0be30ac1fd0e2c58089440",
-    "source/transport_deps/pydantic/_internal/_decorators.py": "3e7c80a0a4a0dc135b0954999f0ab0f67684835503b58bc34fd2e5ba280f88ef",
-    "source/transport_deps/pydantic/_internal/_decorators_v1.py": "b5f75f76940a6384765c23b0a876d97a840cbabe95362811adf86e757047c77f",
-    "source/transport_deps/pydantic/_internal/_discriminated_union.py": "68c9744914b24321df5b8fa49673131cdbf0491a2a1371f73d157fd39bd1b138",
-    "source/transport_deps/pydantic/_internal/_docs_extraction.py": "7f2ce74801e1e40ce88675d94ad5741ef1fe9d16dabcd1cfca0fa49f1f92fc41",
-    "source/transport_deps/pydantic/_internal/_fields.py": "6127c42aadb51608cb27a62a6172a1d1e104b399f130fbd0ea1a7da40f0dcdfc",
-    "source/transport_deps/pydantic/_internal/_forward_ref.py": "e67dd8efedc028b9fcfc54b761cecabad2e23d48725e602d90464e6859e89f03",
-    "source/transport_deps/pydantic/_internal/_generate_schema.py": "4d3e3dbf3633a87f74ad6aefe69b4da19823cceb11d0a3e54a4a8f545ae7ac1c",
-    "source/transport_deps/pydantic/_internal/_generics.py": "10baa34fa2cccd0cd600ad04079ffdaa47bf8806b3cf4390e20ba7a7fb8ab986",
-    "source/transport_deps/pydantic/_internal/_git.py": "2303e1dc33df6b65eadeb06e07d371f25b91d80d62ebd41d7937d659722e0958",
-    "source/transport_deps/pydantic/_internal/_import_utils.py": "4d18710f93ae63a094a2c8a805d05c254b34a26ec820e36265d60057bcd0f233",
-    "source/transport_deps/pydantic/_internal/_internal_dataclass.py": "fdb79d7355dbbaeca044688b66a914930c05a506a847584a2e5479d359f2c926",
-    "source/transport_deps/pydantic/_internal/_known_annotated_metadata.py": "25ceca4cd159a01ddffb489b3ff36024834e795bd8137ab73930500e354c9375",
-    "source/transport_deps/pydantic/_internal/_mock_val_ser.py": "c2645115206fa9f70b6c8e353ec165881e2ed80677989a597a21116c3df1293a",
-    "source/transport_deps/pydantic/_internal/_model_construction.py": "c24f9b3460c026f76e686be7d14d3ff33125d06111bb4b2126f37cfd97e461ac",
-    "source/transport_deps/pydantic/_internal/_namespace_utils.py": "865dfe4d102bf365368d3c8f3f7b7e42ab2f28b8abc6d90b7cd7eb37e7e9d31f",
-    "source/transport_deps/pydantic/_internal/_repr.py": "8d07e726ec83c50a5244d846dbd208f4f5fc7b836fdaa599ac4c3696a8a1dd41",
-    "source/transport_deps/pydantic/_internal/_schema_gather.py": "54b12fe754d811e78d0f6733b32ae626ad4c56726a4ce9a72da9fb546e3873c0",
-    "source/transport_deps/pydantic/_internal/_schema_generation_shared.py": "17fadb41bae4a289a0c6cb24747a4fd23509ed309f7b40400c0124aba089e273",
-    "source/transport_deps/pydantic/_internal/_serializers.py": "6085af4a6011e5f9db19258239076e5add72078650638d9e02bb9cf9e9eb6fa7",
-    "source/transport_deps/pydantic/_internal/_signature.py": "f049633c97b8a529daa6e8ab1b90e4040803d618201f1100cb314f1fef47d331",
-    "source/transport_deps/pydantic/_internal/_typing_extra.py": "755c481a1b0be37cda79b3423108752b94235fa2386bd3959cd01c2f6a486bfc",
-    "source/transport_deps/pydantic/_internal/_utils.py": "0878c7fb4ce75235fd479b144a33d3f23f3efbb3f37bcca36a27ac7130152e24",
-    "source/transport_deps/pydantic/_internal/_validate_call.py": "3df7559d2ce15ceac436d683a03c373c558f5580f95bf80d60f7bca77520e8b8",
-    "source/transport_deps/pydantic/_internal/_validators.py": "76fd1ad8d91ce33718aafdf51a1fd021dda571ff96d24429574a1237380475f8",
-    "source/transport_deps/pydantic/_migration.py": "545ef72d1094cf722b6f9c55b75de36f7340717567105e93d7e27438b7de6790",
-    "source/transport_deps/pydantic/alias_generators.py": "28cd67deee097cb48197552e620de1a181f35c90fecaf82b9eaf2ed5c73087f0",
-    "source/transport_deps/pydantic/aliases.py": "be1087ca84969d7f8427ec166f9aa3e31c91b2c8065a74d07f3429e06499f6e8",
-    "source/transport_deps/pydantic/annotated_handlers.py": "59fc854aac28108157061ed3ef73f270a9682350e25f8e465a2d3e24eb024786",
-    "source/transport_deps/pydantic/class_validators.py": "8bf5778fe3d87462d22e68ff20965e913463ba73bc488573f0b325aae3f23fb1",
-    "source/transport_deps/pydantic/color.py": "033a867d540717ddbf65cb4372e7b40cce324e9d8fead7a49304483535ab2c8a",
-    "source/transport_deps/pydantic/config.py": "b83676c4df89ec255035e9c269529c62c244cc83a92aa9607d0bc3774d35c229",
-    "source/transport_deps/pydantic/dataclasses.py": "dbf8cdfb9d595abbb222c7a78cb235f9585de2a206345ddeb9ef6319c769a872",
-    "source/transport_deps/pydantic/datetime_parse.py": "402f9680cc4caffc10fe6357512ec055ffb684b121befb0f6353d0ca148639d9",
-    "source/transport_deps/pydantic/decorator.py": "617fa3500a6ee4029a55628fa1a57e9fee25ed46d2ebd184b7d45ade1b3398a2",
-    "source/transport_deps/pydantic/deprecated/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/deprecated/class_validators.py": "10071a550339ce9db0066974c9b37ad827d07f226f0cbc7941df4f938fed6f85",
-    "source/transport_deps/pydantic/deprecated/config.py": "93f96c564e7ba5ac4b24e701b9e1f4edcbb53a012059d541c66ea57da0b70825",
-    "source/transport_deps/pydantic/deprecated/copy_internals.py": "1a177ebe431de446020a0c821ad3ca3b9f27a7d704281402eaa9017881052368",
-    "source/transport_deps/pydantic/deprecated/decorator.py": "4c19ba6c9ef026c36287ff16ab92330dcc0fdf69bdfef7f1b3de9d7ac2ee934d",
-    "source/transport_deps/pydantic/deprecated/json.py": "1e55821b7e5146bc72cee4d2e8b4d089907046198365699ea901fcacb5bac00f",
-    "source/transport_deps/pydantic/deprecated/parse.py": "1b377a6ff83ccc95dcb84ed046ae5a761c7f10c25a8577dca57085d1182baaa2",
-    "source/transport_deps/pydantic/deprecated/tools.py": "36b9bda054595a9f3e8e57ef3e02482c4b32c29e18cd90f9d972063c32f11dc2",
-    "source/transport_deps/pydantic/env_settings.py": "e881de79612a9543d152fdd5f805e217f5bdd5f836270fccdced25df827f7b20",
-    "source/transport_deps/pydantic/error_wrappers.py": "44aea6a804dcf72303f8a043f48252f47a4a0a9ad61ddf30a3ce019e6fb77d1f",
-    "source/transport_deps/pydantic/errors.py": "edcb41342b6de7b919171ef52ecd87f3a22e7d0011bf8c0f29ff0385db159f9c",
-    "source/transport_deps/pydantic/experimental/__init__.py": "413eeb298743b0289327d1848e6b107561d2730a4aaeb3641aaeafa8e34fe914",
-    "source/transport_deps/pydantic/experimental/arguments_schema.py": "1059e35ffba5a7eb4fc948d05f9a6642dba0d4e14bfc071cf1b70c6cb77e7d56",
-    "source/transport_deps/pydantic/experimental/missing_sentinel.py": "8507a382d174d3052e40c9e2f78dbd7af83e797c88c292af8ec0fca1fa9f8fec",
-    "source/transport_deps/pydantic/experimental/pipeline.py": "2affddbdc7b12ae99acdf44bd32ebd01ac9e03a1f7ed2ae6b19dd2525fe7d2a6",
-    "source/transport_deps/pydantic/fields.py": "5ae0c63af076d8a5aeb96ddf5e74b85af838a97fed769f17ec1ac09736b8b30f",
-    "source/transport_deps/pydantic/functional_serializers.py": "ac4cc7dfdd73ab2de8fdb5a4d9012ebf24a67103599b05e62502c2dd9185ed00",
-    "source/transport_deps/pydantic/functional_validators.py": "73ffbbc1e5a918d70e61f45f55416ed758ebc4c54c75f63f73ee22b2dc24f796",
-    "source/transport_deps/pydantic/generics.py": "d19a99f4ef5a9e7223ff7986051aa9b3886d7b2ddbe65575f9ddad5313cc3270",
-    "source/transport_deps/pydantic/json.py": "647f11908ee1e01cfece9f0e7530316c952856f72853e8e131d459d01fa4d317",
-    "source/transport_deps/pydantic/json_schema.py": "677786d509be226f8c417ebac28c963fd5b2c055272d116d68e0cbced559f937",
-    "source/transport_deps/pydantic/main.py": "efb414c13b5041d51a7e8af78abe70d820b5b34be2805b460747d2953bc5dbb2",
-    "source/transport_deps/pydantic/mypy.py": "a7a294d46c0f1dacc5ec4e6f26ad6e2dde2d1dde8313a6eb7b8fa6e4b9f6de6b",
-    "source/transport_deps/pydantic/networks.py": "2c995c70c79df7b1007a9293e8e98f23d73a8e1e62c1b063935fe35f5449c37d",
-    "source/transport_deps/pydantic/parse.py": "c2477cd9d82dbd6b43f3de54fc8e84d61b6a3251a1052604577f5cb8149ea370",
-    "source/transport_deps/pydantic/plugin/__init__.py": "6bb4f0dfae94e8ade496d08235963bea773db2cfbbb86190e344d769df4eca94",
-    "source/transport_deps/pydantic/plugin/_loader.py": "f502d79de2c49afca15e46bff63e0bae46e67b8a8fbfaa98a6196c8c5d8c1ac0",
-    "source/transport_deps/pydantic/plugin/_schema_validator.py": "41b9aab06df730199fb4d4369cd8ae376d8b85badec5403b8a90d5bf7274d815",
-    "source/transport_deps/pydantic/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/root_model.py": "06f98bb56e22d7574993e74b38cdeb97e8e725d40679e41314172610eabaa4c8",
-    "source/transport_deps/pydantic/schema.py": "56aaa3beafcb9daa5592779b51ddc1a7f275a768176599d94602f8f1b5441bba",
-    "source/transport_deps/pydantic/tools.py": "88742977c489e43093b4f5796ad015d3a4fcf5b8d268c15e659d8b5fd95ab196",
-    "source/transport_deps/pydantic/type_adapter.py": "553fbeca0e1adbbb21481cd61c13cacf8f777f7890f686dd9049218d92a513b4",
-    "source/transport_deps/pydantic/types.py": "9ea752f89d995ea4e1daa7b224ecc14c1b475b267961115ef2068c579f5df471",
-    "source/transport_deps/pydantic/typing.py": "3fb7de037e4cc1370bb11d6e2fb75bd12fa8c9d071a1b9976b9e580e80606a34",
-    "source/transport_deps/pydantic/utils.py": "d799d1d90a6a4c1165415e1336d4c8b4cc93271fdf6f257e80f988118d46a287",
-    "source/transport_deps/pydantic/v1/__init__.py": "14b43c212a7a3156517e39d2edf4389b51714311421764158a41380cafb83e11",
-    "source/transport_deps/pydantic/v1/_hypothesis_plugin.py": "e444b9c56bb0d45400b3298b7b3cbc4209d5cf466955f5378e4993ef81f6ed54",
-    "source/transport_deps/pydantic/v1/annotated_types.py": "ba4d8d000c6a88d10b2a388787287129a20e87c1759585bf2f3ad6dd7ee86417",
-    "source/transport_deps/pydantic/v1/class_validators.py": "50b39a21481850306c1cbec411511aadc33e52e6ca52082837c8526c3a27b051",
-    "source/transport_deps/pydantic/v1/color.py": "8990012d8a7a395a28d801643fd229ae2ff049cebe2a496ef18ba149aaed7798",
-    "source/transport_deps/pydantic/v1/config.py": "6ba3f459eafdc7871bc0a5bb5eff29a1252a33858ff912d6c17e9832962af400",
-    "source/transport_deps/pydantic/v1/dataclasses.py": "efce1caaf2276f020f5abf6eb1fa57ddc87bcf8b77276b532ba374ebbff0935a",
-    "source/transport_deps/pydantic/v1/datetime_parse.py": "e10cb5910a6adeb3556497881de48f0e9b92d81be1a7528fb73246d71bbe1f4d",
-    "source/transport_deps/pydantic/v1/decorator.py": "cda6b1c71a163c29bcd7c0f56ecd3285aa518d79b7d95f06d191d609d335b970",
-    "source/transport_deps/pydantic/v1/env_settings.py": "03d557c2d465d36018fa31f40b4a2ecb954dc377e2e85fe9933b870e380000e3",
-    "source/transport_deps/pydantic/v1/error_wrappers.py": "ebadb931fc3daa40b63708ad07f2450167bc07e5efeb3054eeb2fd936f2d7f2a",
-    "source/transport_deps/pydantic/v1/errors.py": "988c0f103e6f18ce50e6fe02e19d493e5753447fa89af132947ea4b0c84e98fc",
-    "source/transport_deps/pydantic/v1/fields.py": "56a5890ab894362132a6d5eba03549e74d49a550347a7d9500d724b2a5cbd9bf",
-    "source/transport_deps/pydantic/v1/generics.py": "5730bd61457e11b3e94376807e4d5c3457a3efdfc8cf39e443b5ab9934d94bd1",
-    "source/transport_deps/pydantic/v1/json.py": "590e47cbf848a5f751dd84bc93a37613a28c273b1d6c18bf95d58e3c9695f353",
-    "source/transport_deps/pydantic/v1/main.py": "cee36974de50d15d301b65144cab741d4cb75c9e0e02f3d26437625d8f85233b",
-    "source/transport_deps/pydantic/v1/mypy.py": "0a5f1745f0a621c544de3e40114e760bc88387c95c5fffc3de1cf68c85b1300b",
-    "source/transport_deps/pydantic/v1/networks.py": "1d836d2807ce98e9ca269b0383583a488923f563e153ffa2f25e5ed89281a46e",
-    "source/transport_deps/pydantic/v1/parse.py": "049b5daa2651b5abfd55114298ec6863e2889909a33f2f80fcda288e21546716",
-    "source/transport_deps/pydantic/v1/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/v1/schema.py": "6aa06e03ef9cabc8005648a6e4124f140487cce67c745b66157fdf346afa8a9e",
-    "source/transport_deps/pydantic/v1/tools.py": "d650dd5c79348cbe6e3f7bb9442600bd402518296000efb8e6592af63edfc810",
-    "source/transport_deps/pydantic/v1/types.py": "07397e45c9e2b4f0479eac308fd8afec98c7b8dd46a67587db874a905de5f5ef",
-    "source/transport_deps/pydantic/v1/typing.py": "ec67418356131d42d4f35b6107ff5c8d13437d97e7e248685fb9c6b70fe47821",
-    "source/transport_deps/pydantic/v1/utils.py": "339151c9f3546f50369a4f6568680255d7c71dbdc0b5082b8cee6a7f205fe310",
-    "source/transport_deps/pydantic/v1/validators.py": "9725249f5316847c65097e597c48058ff0801e88e888f71a41e31d10cf57be25",
-    "source/transport_deps/pydantic/v1/version.py": "1d79d75bed5b316e6a2a196be5180e10fa21ad90c20d2bb2cbcfa08bc182819a",
-    "source/transport_deps/pydantic/validate_call_decorator.py": "f23a8b9605d38d6123e1d5c3834c08dc41902a46f42670ac2ff2525236d4e528",
-    "source/transport_deps/pydantic/validators.py": "a706c825755bd42576980138c3f1067cd8fb0f0cec29a5b0fed4cbe9cbe24eeb",
-    "source/transport_deps/pydantic/version.py": "4e6124eaa03e540d4f56c737f5b66584f010032bda71de812b3c57007c96d395",
-    "source/transport_deps/pydantic/warnings.py": "5aed551b3aef159c384fac8220a1e31fb2d263ae878dbb720856e7e6e5a8309e",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/METADATA": "3434fb886fe254ed414541f305a86e43ff575093f8f6d15c89bd11144ed8e131",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/RECORD": "7f3a82de0054511dbd2f7309c5e5b6ddb3d96c6230bfa83fd73bbdffc51cd0f1",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/WHEEL": "7f32b0ec3e21425972f904759f2efc26a495ce223c174ce187bb87ac82f5240a",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/licenses/LICENSE": "2afdd30d54b4d62b6f488a6bcc1546e84ec5061f13f4209c03d012348783795a",
-    "source/transport_deps/pydantic_core/__init__.py": "9cad6292b75254af606a970aae9bff6e5ae9f0b08089cd63acafa391b60794d2",
-    "source/transport_deps/pydantic_core/_pydantic_core.cpython-310-x86_64-linux-gnu.so": "cd9bb47485cdffc363010507a214155966ca860b0b93bda39583492c6764e631",
-    "source/transport_deps/pydantic_core/_pydantic_core.pyi": "3ea1dbd4182f08cf9341f24b3c5cf7db77a05b3535ffe9e23525127a3617a11f",
-    "source/transport_deps/pydantic_core/core_schema.py": "bbdc850b72d685133a0e250fed263b33691dcdf3813492f3c0e31001e3d46005",
-    "source/transport_deps/pydantic_core/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/METADATA": "bad3d9e6342ef6750cf5128ec942327b8bebdce0b6e4d6f67cb15b27ae5118da",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/RECORD": "cf5e3948866214f2e9ab0c1d43c157e9f9c14f59cf8442936c6cb97d5f2b6a5d",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/WHEEL": "ccec20e2307acd7da453dd7437e70c6b08e2bc3ead3bc3445af135da37b56d59",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/licenses/LICENSE": "804b59b25f2c31bd278f9202a19ae49a3945aa2664387e2d0a128c7cacc61ec3",
-    "source/transport_deps/typing_inspection/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/introspection.py": "5888c45bd627da692d9527699e3f549adf35651596b1656772310e12181b4396",
-    "source/transport_deps/typing_inspection/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/typing_objects.py": "a778d6290d51581c9c9a23e0263af2a4604c76def5c4d40a7cd1a375b0046ff6",
-    "source/transport_deps/typing_inspection/typing_objects.pyi": "3bd24dd92d0270d31d64fd40abc1ef8693f5322aab9a78aeb3c5b62f992f84be",
-    "source/vpp2/src/vpp2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/adaptive_mask/__init__.py": "a73f5515ffb77d323f9c324776e9c166d13817c470506cfd5db4ef0546cbbdda",
-    "source/vpp2/src/vpp2/adaptive_mask/audit.py": "c949909a2e60c3c4389c0221c79459961f2c4925baa07f7985774a52257a1d17",
-    "source/vpp2/src/vpp2/adaptive_mask/budget.py": "74e32285903dbf673ecadb6a60d917a990b9c6619beac9f926c1b3cbdae3887c",
-    "source/vpp2/src/vpp2/adaptive_mask/controller.py": "328f636201a845ebdc211340a9014a5076fa895e10cf9c2e1f91b96761a2514b",
-    "source/vpp2/src/vpp2/adaptive_mask/data.py": "e3b633fedaa503b101c80610cbed0933e9e245acf56a6c0692c4d99620faf84c",
-    "source/vpp2/src/vpp2/adaptive_mask/equivalence.py": "8658cb69d549fa0575570364fe3aa1e0b121fcf8b9965b17cd174d459dddf5ab",
-    "source/vpp2/src/vpp2/adaptive_mask/execution.py": "efec52c9b2eabf0300c72edea18f9cd4cc16a0889496793abb6a50161f2328ef",
-    "source/vpp2/src/vpp2/adaptive_mask/export.py": "0ce51e74787d28bbe9f4e73a3774297b424aa131fa78968d24154ff56db98b0f",
-    "source/vpp2/src/vpp2/adaptive_mask/feedback.py": "519fa4e6416723fc000754815a2f05d3415299ae35724f37bb806ebc22393d38",
-    "source/vpp2/src/vpp2/adaptive_mask/geometry.py": "f1ce4110c3a3a893f4f240c49e2ccc4e93215ac4ba9e447de8042f9b19afab98",
-    "source/vpp2/src/vpp2/adaptive_mask/hidden.py": "968560bd6dae0fce59bd28084091b10d559cf165c81764f6ad24b7492101810d",
-    "source/vpp2/src/vpp2/adaptive_mask/losses.py": "4a7dea6f87ef0fa38a7202d85d98d37460f8f4dc13ca5336f32b1da8b84a4322",
-    "source/vpp2/src/vpp2/adaptive_mask/masking.py": "31b595916410c60f772fff1ab63e92c854c8fd0d2261e6b865fd05ce2c61b9c6",
-    "source/vpp2/src/vpp2/adaptive_mask/monitor.py": "48b71f499dd5a0a79557eeb918f5fb1a641dc8654e5d412aa628ca9c57b5dcf8",
-    "source/vpp2/src/vpp2/adaptive_mask/physical.py": "f54e8b4fae8026cc89f296d74bf4cbea06fa4e126905b572a22c7d8226f0238c",
-    "source/vpp2/src/vpp2/adaptive_mask/preflight.py": "b646e7ca785afc87356e3e1184e48f535792ebafc84936450fc83bfd3f9f5ea8",
-    "source/vpp2/src/vpp2/adaptive_mask/priority.py": "68ccff1a384081f0c9a82e8f54ad2c720293f8500e7440a469b5e8fac1ee608a",
-    "source/vpp2/src/vpp2/adaptive_mask/quality.py": "c3b3a2264ca0b4773805f4485ebfe91bb4ab861e226ff4bd27bd667ab372b59b",
-    "source/vpp2/src/vpp2/adaptive_mask/quality_policy.py": "f100b7d4a412592777d936a6cf3c73a9f787cc4a62d5f1b8a369fed74acf897f",
-    "source/vpp2/src/vpp2/adaptive_mask/reliability.py": "306b41720154810bd972fe4d9c4c388e6967539d076ddbedd3a432ac62d13104",
-    "source/vpp2/src/vpp2/adaptive_mask/report.py": "f71401e09297daed8bbb72cb3af3b887e5fdf5ec282f47637fbb01392d280171",
-    "source/vpp2/src/vpp2/adaptive_mask/runtime.py": "ed74bfedd147389ea01f3be8c2979fa06c7152326ccc29ee347175690cb9c6cc",
-    "source/vpp2/src/vpp2/adaptive_mask/state.py": "5e8f95021eeb35511f5481e4e91ae687fe605f9e88d17b3b98d377cc967c17e1",
-    "source/vpp2/src/vpp2/adaptive_mask/train.py": "ad95349bfb242bcecd7ddf82fc58b31d63f2938a2101f9560bbc9e305f5f5b20",
-    "source/vpp2/src/vpp2/benchmark/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/benchmark/audit.py": "8587be371653c38b9926aa0ace006a244f270e0ef0020147a7bf5b0fc85d4538",
-    "source/vpp2/src/vpp2/benchmark/packed.py": "5047fd02f3fcc68e41f543fe83bc9202769a05794d898524d6d25e308f2b5125",
-    "source/vpp2/src/vpp2/benchmark/scheduler_io.py": "d6df06259549faed2da9f2888b540eac8e1ff2895c339eea58b4ea6a7e3b3cb0",
-    "source/vpp2/src/vpp2/checkpoint_compat.py": "462c5f93812483b5871633b200bc2ffecb9c201a2caf6b14f8f36dbe95fba50a",
-    "source/vpp2/src/vpp2/cli.py": "54c19a0f7e25b53bbf56ec13b0f537c9d7b81282e1e80575bddfb470172164e3",
-    "source/vpp2/src/vpp2/compact_kv/__init__.py": "f34f2f77f68399f45f6656b562a5c331f2504672dc145df368e086532191f766",
-    "source/vpp2/src/vpp2/compact_kv/blocks.py": "662c1f35055c2be2e2a14f2d92615d046f5e454879e34af39e3c7aec3af34098",
-    "source/vpp2/src/vpp2/compact_kv/checkpointing.py": "3c33c156e96f26ceb22eee61cfb8b18c3cb9560db0ab405cfd4c46efe5aba353",
-    "source/vpp2/src/vpp2/compact_kv/eval_prepare.py": "bdb33444056245e3cc773b0849b4db608b454cd24a7ac5b752f1ba26ff978865",
-    "source/vpp2/src/vpp2/compact_kv/export_check.py": "e6cba70b47027882a34d587aba0da5f9371128ec2db2282b91ad2d49e40065b9",
-    "source/vpp2/src/vpp2/compact_kv/gpu_check.py": "fedb6a9ab69b51080161ff2a4feae3f85262cc50f932f48e8decf04c1d147a20",
-    "source/vpp2/src/vpp2/compact_kv/losses.py": "914437ebf3ad91fd2cffa893a65ecb04ab01f2b337658a781a472feeb50c0a06",
-    "source/vpp2/src/vpp2/compact_kv/model.py": "332ebaebf543d40a34d188ebbe81055ec3e4bcff164bfe90d15aa6cef0748b8d",
-    "source/vpp2/src/vpp2/compact_kv/prepare.py": "7b01764cad0fafc0cc9441952b2e928048b0a43a6105dbd0eb64c349f4b9bf1f",
-    "source/vpp2/src/vpp2/compact_kv/provenance.py": "103e0f7100be56233aa86041ba9bc427ccb0db4fe045c4e6b8d3481ab6bb42da",
-    "source/vpp2/src/vpp2/compact_kv/reporting.py": "0b9ff602fc3425d1b9c122a77da3a426814e11c9cbdcb6b9a75a2a48130087a7",
-    "source/vpp2/src/vpp2/compact_kv/sampling.py": "2e8d82704f07853a7b487d5aaedef1b2f1fcfc5247e9100a56ed21d69262eae2",
-    "source/vpp2/src/vpp2/compact_kv/spec.py": "9bc920fc7d4398e424a0bde38ce97f87473aa417cf68b48e2c292a59491bc7d7",
-    "source/vpp2/src/vpp2/compact_kv/train.py": "a9e4410c5cde5df002d9808554538c29b77ed9e2d2bedc9ac5f8465eed70cc3b",
-    "source/vpp2/src/vpp2/compact_kv/validation.py": "73b62d02ce1d518cba7b9462e85c3c8d272ae9974e1161b1b6adea6f0040e2b5",
-    "source/vpp2/src/vpp2/convert_robodojo.py": "7ec2d6bfffbb4f5891ee33a4bf630cf27b17628506aa3cee1c46b3050027b290",
-    "source/vpp2/src/vpp2/critical_windows/__init__.py": "22721b77e25ec4874c4853d2f2991a14dff6aaa665da500f2830826b06545478",
-    "source/vpp2/src/vpp2/critical_windows/contracts.py": "ded4a239f7599785d74ff57ff892c5aa74f0280d6759460c93942e91f505d47a",
-    "source/vpp2/src/vpp2/critical_windows/coverage.py": "6c18bf995aacf413bbe4536eaeed92ab4b2f15f9eaaaa03a2730256ea7922ce4",
-    "source/vpp2/src/vpp2/critical_windows/inference.py": "df301bd187b27d20b6623febcb5e85815e1b7d4a1e9524a8a65bdf5904bf7232",
-    "source/vpp2/src/vpp2/critical_windows/metrics.py": "ae2bab145555d55696ef479758a5bae6780527fd3085a4f46d035b667ff32736",
-    "source/vpp2/src/vpp2/critical_windows/packets.py": "8f7d2c411407d95b1ff4d0e5b9055589f6114b1bb1020eeb7e1b106f330b853a",
-    "source/vpp2/src/vpp2/critical_windows/prepare_staged.py": "694cb5699de2d9f5ce1bc4667627b47d6bf0698715b4aad86907139859caffe5",
-    "source/vpp2/src/vpp2/critical_windows/reselect.py": "8d0e31ca1c17906d74de6439276e49ff7ea78abfc66fa556d8a306070b3dbf4c",
-    "source/vpp2/src/vpp2/critical_windows/score.py": "32daceb9a3fd957d9be75823b4d519629bc60a64d1da62f465e172cb1ec98d55",
-    "source/vpp2/src/vpp2/critical_windows/selection.py": "a7c35f6719d3308bfb738965d17544a15985c6d684bb47eac7272d6155d96786",
-    "source/vpp2/src/vpp2/critical_windows/staged_review.py": "d5cfb3f49ec5acc6914da98f5413d1549b2695e7c85c9aed9c0db83dee6be754",
-    "source/vpp2/src/vpp2/data.py": "8574e1ae3c6ca6e4b84633ee1815f2b3f2ea17c1e55f12fda0b00c5c5a6a7b53",
-    "source/vpp2/src/vpp2/datasets/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/dataset_utils.py": "e1a1471b0aa70917eaf699f85a9acf3ac70d9fb6eb9ebaf02d2bf2f218d6ab2f",
-    "source/vpp2/src/vpp2/datasets/lerobot/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/base_lerobot_dataset.py": "729fb3fd4ca531212eadb4a4e9c7e9f97ad34cca091d66380acd6b179f0c29c6",
-    "source/vpp2/src/vpp2/datasets/lerobot/constants.py": "11ac9260341e1560c5eb0dd7acb79ed63c74180b8ab4113c27622ff213db417f",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_dataset.py": "bfe75bb9593ec646236777d90d9fa309fa0e1188db07e07a3be0810edbe3da56",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_sampling.py": "266b0b7b023f014f4ccd935b6015b0a05b4659f372264fdefd997ea9692b244a",
-    "source/vpp2/src/vpp2/datasets/lerobot/episode_instruction.py": "74767839a813c7902861f4561101541c282b72e261df1fd28fe8d06c91231b69",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/compute_stats.py": "7d7daab219332b5abdfd5ec67ccaf3502e8eb583a17a4e482c3f3d4f7a691e89",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/utils.py": "4fdb4ac0653201e6743f2161c485213429948c1e1d5b068d03b4d97803254006",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/video_utils.py": "4f66c410ae6aa5a98acfc3320f3dd7b017d0285117054f14226f9e5d82395087",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/lerobot_dataset.py": "5659e992010bb6674ec50ab0d33b28a8e9dab0ecc895c199662af4c40fe56629",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/base_processor.py": "d3327cbe398d77761df960014d41e70d349bb0fa42bca874333b9ee970222447",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/vpp2_processor.py": "43c5604cbfc315b7377316cb97c9a6f3322124404397abfd01c6c3d0090cd414",
-    "source/vpp2/src/vpp2/datasets/lerobot/robot_video_dataset.py": "8eabfbd649116b0b4d6c9704c0df34febbf8875f6ab863dd75f4107dbbcdef73",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/action_state_merger.py": "39fd44f254bd7fb71b9269d75d8f9b5a5bbbe167ea438dda92d40eda1d256734",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/image.py": "f494d262ce29770e959bdff60b997f7a07e5f00d10ef515eee7be813d1db038f",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/rotation.py": "feb59d26539be6cfdb86488f0c10c11488ed73901658bc657e67a755ddb9e67e",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/normalizer.py": "8207d813b698a91dfdbe6910a4aa1fe0e775a2a8edf5c05b2d7a42591e9f8adb",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/rotation.py": "8ea2c72565707f5979967a1abeb5edf1eaa2c3fec55a7b50e640de869fc9fbcf",
-    "source/vpp2/src/vpp2/datasets/lerobot/video_event_dataset.py": "c20d3f5e2bc05e1b7ed94ae38872883a976a179e09b7fa051e94f94e2836c02d",
-    "source/vpp2/src/vpp2/enhance_prompt.py": "f31b3190c24bad073c5e3121ded4f564232729566e8c1fcc3e76fe75b28faf45",
-    "source/vpp2/src/vpp2/evaluation.py": "9b76882149e3c3a8c417a0e23514af1bb305c01bf0b6d21fff1f70bc9843f55a",
-    "source/vpp2/src/vpp2/flexible_mask/__init__.py": "64bed1808976d1f3c708154c36a97288a75abc6463d49d7f0065cb8360a2fb16",
-    "source/vpp2/src/vpp2/flexible_mask/calibration_batch.py": "dc5d83fc9e2bb595b9e3aeb1923115ebb9345283290a7c3f95c79fdeef7d3e76",
-    "source/vpp2/src/vpp2/flexible_mask/checkpoint.py": "c8531727e2388ec3d7a9770a504ceda119afcf03d64b2462befcd45a48949f9d",
-    "source/vpp2/src/vpp2/flexible_mask/coefficient.py": "07a3cc8d8e6c8b60119b9f94ae85074da017b08158fa0a0912cc9051b6df8ef7",
-    "source/vpp2/src/vpp2/flexible_mask/config.py": "4e1e9b289e293145b4e895c8ef7347bc837d01c929e05fd5cbf65bdb3c20be5b",
-    "source/vpp2/src/vpp2/flexible_mask/continuation.py": "2016463b4efe01b5d0a9ee84f03c3d95ee16f8e24f230faa465598cc978da29b",
-    "source/vpp2/src/vpp2/flexible_mask/controller.py": "263f1c9091e0c4318aeefb61e91549ba905ff046a625e35e1b18c94da4b4668d",
-    "source/vpp2/src/vpp2/flexible_mask/data.py": "d245bafd6f04f33ae7cef7c01d5db876a1a2a1a138c4de8a139d3ea78095521c",
-    "source/vpp2/src/vpp2/flexible_mask/diagnose.py": "faa29a0fa57c66af3b2e817640fdedfddc998bf878357f484415b244555ae4a5",
-    "source/vpp2/src/vpp2/flexible_mask/export.py": "3064dfd82bb55a145c16c77bbe17244970d07d18e944b793f4b7368dbcede764",
-    "source/vpp2/src/vpp2/flexible_mask/gradient_probe.py": "6fe3f13540cb9d8dd20c0b6429268ca74625dfad1e41806412c0f980fac76509",
-    "source/vpp2/src/vpp2/flexible_mask/losses.py": "df264b71365885ea90329c0b0facc92db3d2eb6e04324ea90a4701a1cd457abf",
-    "source/vpp2/src/vpp2/flexible_mask/metrics.py": "e976dd066e4b5e7e7ae79db77b9d1be64d17c212c61d0bc113b87a048e8f2b6f",
-    "source/vpp2/src/vpp2/flexible_mask/performance.py": "95b3977fd739cc98f7de94b14fb490a7c132021e742e7179dc5b5dddd6557c68",
-    "source/vpp2/src/vpp2/flexible_mask/preflight.py": "f794b997a492911c4d77656310c89d3b85e54983b3b889ba9904be64fb961e43",
-    "source/vpp2/src/vpp2/flexible_mask/prepare.py": "d6cf82e27555ecc13973727f4b71ce7e40558fef7d95805665fc7e458a42fad4",
-    "source/vpp2/src/vpp2/flexible_mask/pressure.py": "6cf74edcf3625fce9eec689d70841c6c1481f7c1abd87e5bf70bb38eb34632a2",
-    "source/vpp2/src/vpp2/flexible_mask/recover.py": "418b2d5ecdd08b08686b2a07bbeb0f969471eb73c04ac6952dcfbcda3b123f30",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_best.py": "917d924c5708584a0192050a6d8210809e905938aaf877dd81cff3be287ea289",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_checkpointing.py": "6f6a61c2ee7076c6b0977d1acb489fee7df04322d7f6b13bee9bb8229be319df",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_export.py": "c50433aa1b7e528fb23c514dddb5b9197f44c662829838ecda106980ab0dd54d",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_handoff.py": "aa590f707738e4b7c7250f901e29795c7441430f961bfa764437cf34bdc514b1",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_kv_decay.py": "bf83761798deea4f89c32d14f43bbfb0f16f751b3230962456c073b843dcd08a",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_subset.py": "7f7154cab6d910b747a0189742ee625983d08f13de50ca60bd4154f61df2b0d3",
-    "source/vpp2/src/vpp2/flexible_mask/report.py": "5b49283d472bace3a6ad56ce568af04f2c3d04add1a3920ad5006fe98374bce6",
-    "source/vpp2/src/vpp2/flexible_mask/rng.py": "01680b5b7e8a29ce92e96d756e12ceb80bb0e0ce1ebb642a14fa2a3a34776daf",
-    "source/vpp2/src/vpp2/flexible_mask/runtime.py": "373a3ecbc2ab6592a359daf6adfd8c958cb657f393e698ca2cb834f06750be4e",
-    "source/vpp2/src/vpp2/flexible_mask/select.py": "b547857bad61ae7ebd25b888fba73c2ca8f50a8eb31551ca54a909c75c8984be",
-    "source/vpp2/src/vpp2/flexible_mask/selection.py": "b9453c8735051b54ae7dc8610d2ca332eae8ca1a3e099aaadaf5ab75bfb0f8b0",
-    "source/vpp2/src/vpp2/flexible_mask/speed_review.py": "9d7099d1e173026274eb439169df2ad4812ec1d56a5f0b21242f972cd835ffd2",
-    "source/vpp2/src/vpp2/flexible_mask/throughput.py": "ce21bdfda51ae10d1470ac537d910d9f46f60a38910888808ffb3f5a80174d58",
-    "source/vpp2/src/vpp2/flexible_mask/train.py": "86625613bbc692722c6d2d24278a586f88ce0d76263a9daf92562a9002a8e037",
-    "source/vpp2/src/vpp2/flexible_mask/verify_export.py": "fb68e62b0da98e6f96cfa694bd061162bbeec1763e5787e057a1cbac5b2ce8f6",
-    "source/vpp2/src/vpp2/full_mask/__init__.py": "27c9ede277380e8d64586371f2a03334e68ab2d3bfe516d95e8cb96099827929",
-    "source/vpp2/src/vpp2/full_mask/calibration.py": "5d97e0385eba82252e12cc1562d1a77da2f2bce29332d758c1b5b6b84c5bb47a",
-    "source/vpp2/src/vpp2/full_mask/conditioning_cache.py": "07adf95165db4c3c1c469695d6b7b6fe4cf4fa9aff0a3d360f4229bffe9be17e",
-    "source/vpp2/src/vpp2/full_mask/controller.py": "0e6a969b89a4275352cc5a1e00bdfe8e3f8754a6ebd7f9266b92179cbe6bdd10",
-    "source/vpp2/src/vpp2/full_mask/cpu_contracts.py": "1a46a53a68d2d8a202c5f98d9a78f6d5f5041dada0c2239ed81e7916a28f0758",
-    "source/vpp2/src/vpp2/full_mask/curation.py": "4ffd7b0212bdbca1daf73ba7c24423523437012150fc044a43bbb33cce747545",
-    "source/vpp2/src/vpp2/full_mask/curation_v3.py": "9dba386d66869de0fba321ade79b2692d47f111ccfebc7f367194cef7a4b4e8c",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_checks.py": "688ea4e7e44c4ecc1a3cd81cf682e2cbf611d9ad2d52744ba1308324da70febc",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_features.py": "11011a089f1586045e2972cf267773aaf4602cf08c8b0e755bbf017ede1e6e43",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_finalize.py": "a2a7812eb6c670686a7a4f4adff0a2c6363dbc078c71d948537f1c5fd546254f",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_probe.py": "2755e9cabd0fe02e70cb8330c083af881a5e4696697d58cf7aa01bd00ef57f6d",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_quality.py": "b9170398762112a23bb133dde53c6a49193c671c767647239c8b0ff15159530c",
-    "source/vpp2/src/vpp2/full_mask/data.py": "668a7c5f88a3ab17854eb0df240c974082f4aa64604cf11d3143b18337f73181",
-    "source/vpp2/src/vpp2/full_mask/distributed_smoke.py": "f699d4f330309f29a537fbabb61c60f61eb3322d7304605351dba3d2acb48e8e",
-    "source/vpp2/src/vpp2/full_mask/gpu_prepare.py": "92ce96dc4c1b6d6ebc22d99935472c15d86ba2a7c050e31640d2c32e08748ec5",
-    "source/vpp2/src/vpp2/full_mask/monitor.py": "2724a934d1f96444227c854f75ef84df6025c02485aa68eca34f3b28d509fe19",
-    "source/vpp2/src/vpp2/full_mask/native_monitor.py": "5572712a340f17692a669d2a4c41d1d691164925eb7bd06a56fff1dc2afd69fa",
-    "source/vpp2/src/vpp2/full_mask/panel.py": "2844b663340cbfb54f7622f3447a7831e2809050739a52dc076fd6ae32ed9e46",
-    "source/vpp2/src/vpp2/full_mask/review_decisions.py": "ed1184bcda7b5bb3fb11bb4262c4461c408e3d3ac0d23e80d2f8cafcf39c7e18",
-    "source/vpp2/src/vpp2/full_mask/review_report.py": "752ba9e8a3caf5b8cefa2d18e6fe14860e6f998f57c03af0940af0943b09a9b1",
-    "source/vpp2/src/vpp2/full_mask/runtime.py": "ae30b4d3fd3c888bef17672376dc3b34491cb296a1883959674ad465579efe7b",
-    "source/vpp2/src/vpp2/full_mask/search_quality.py": "ee3238d3366799e97d3f9a90c11f579ff8f51d5bd0642f52c37d9dcdbcf02e6e",
-    "source/vpp2/src/vpp2/full_mask/train.py": "f4127e6580522d4d9e5a50db10d201526e948d4d3479ddfbbc12e60ec61c7501",
-    "source/vpp2/src/vpp2/full_mask/verify_data.py": "72ee60ba2016d5f73eb2ccb01b666a5dac37138620efc4a79a20607df3658fb3",
-    "source/vpp2/src/vpp2/inference.py": "42b3057a85aaacbb87440bc740f13b09b9c82cd51d8c5af4e080716aabdcf38e",
-    "source/vpp2/src/vpp2/learned_width_kv/__init__.py": "0c59cb8424b6cd64508a005625e9edf63c84d24618a5d2caf6e6fa5d63cc859e",
-    "source/vpp2/src/vpp2/learned_width_kv/calibrate.py": "b89bbf103b49fd85f85f8c300a7868bb42ae7a13f29bd9878dbd5a11061ae51d",
-    "source/vpp2/src/vpp2/learned_width_kv/checkpointing.py": "7d30e46b95d91df8aff088346ae75809958622b3a8e8d953db4f30adfbaebadd",
-    "source/vpp2/src/vpp2/learned_width_kv/common.py": "52dff98c6317b34df2a13313a641998acf4f9e18ac31ea4e8e25227afa14620d",
-    "source/vpp2/src/vpp2/learned_width_kv/controller.py": "7b5a8793a540e1e7464185229189f4f1f1acadcd7cf3c2bf477bda5de7d18787",
-    "source/vpp2/src/vpp2/learned_width_kv/data_scope.py": "0ddde6805f646afbf5f5966c53530f829becd083802f0e0a8697dc1004e2003e",
-    "source/vpp2/src/vpp2/learned_width_kv/distributed_check.py": "f89f153a20730a6bc3c70133479f34918d30c5a48775bb3f6c14bb25701801ce",
-    "source/vpp2/src/vpp2/learned_width_kv/gradient_probe.py": "3bed964f6488e95b386cc58058e7f28aff50c9a98dd526f3365891a9049a6fab",
-    "source/vpp2/src/vpp2/learned_width_kv/lifecycle.py": "798afa47ec5dcc8749982da2d7853eff184e450a32d51bcf70eb4578f2bb0b09",
-    "source/vpp2/src/vpp2/learned_width_kv/masking.py": "3272ea09271130dcb3cab6935f6bb7d61bda66b9982c3baabcd61ee0efd914a0",
-    "source/vpp2/src/vpp2/learned_width_kv/performance.py": "a2fd9597c51bf0cf5c435b48e8f57792df24d2723a6d7b6b841d6fa24f17ab2f",
-    "source/vpp2/src/vpp2/learned_width_kv/reliability.py": "f74399fb67427093b3e4c7df9749379f9291b7ac58bd6632db226b6e78fdbfd0",
-    "source/vpp2/src/vpp2/learned_width_kv/reporting.py": "f22c1d8806a5b7757bb6dd99b15153d0305b2eeb0790bb37763337690d642036",
-    "source/vpp2/src/vpp2/learned_width_kv/runtime.py": "513da1e06ba9509164f017e0b8dbf6b219d7e1cff82edb2d2fd8d2d7839e666f",
-    "source/vpp2/src/vpp2/learned_width_kv/train.py": "5813ebd4fefec6d3a8cead9e0c645d10734f5ec88c4d74a5742587f2eb4de6c3",
-    "source/vpp2/src/vpp2/learned_width_kv/validation.py": "d5594ce875c8011bff12d3c77083dca68878178ed022b8079761267d04a82386",
-    "source/vpp2/src/vpp2/learned_width_kv/verification.py": "1bc43d48ef0ec3056af614adf43ca0744d94b1e77fe970168fef87c7cf2e6fe8",
-    "source/vpp2/src/vpp2/libero/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/libero/action_model.py": "178fbe4cf02b9a2e133ef31389b6bb2f9baf0504bd0818a265b6bff32ebc3914",
-    "source/vpp2/src/vpp2/libero/cli.py": "fcf68a88e2db0a9b786c6199d9b719802fe4b84514741658b49e2b06e2473e45",
-    "source/vpp2/src/vpp2/libero/data.py": "19b699381e7d6fa769b5090c7125ae01b7584576e969468660ee65b1585ef5f1",
-    "source/vpp2/src/vpp2/libero/diagnostics.py": "9540e31dfd2def06bec942ffe73262278700a17a340e14cb5bd41a295706e147",
-    "source/vpp2/src/vpp2/libero/eval_helpers.py": "ac338d4ade195ceb3a87620d162befa9679e2d7ed3d91b09927e5d59d86d97f4",
-    "source/vpp2/src/vpp2/libero/eval_libero_single.py": "87e6b457dc412ef4fe5d01aa48b1a6d20a28dbce3152848042e2103769e4efef",
-    "source/vpp2/src/vpp2/libero/eval_libero_task_manifest.py": "f0c721a7ddc768edddc8b0e065a4d8aeec35e1a546c88deebd9097ad93f6227b",
-    "source/vpp2/src/vpp2/libero/evaluate.py": "24216f9814f8304e077d72ff8d8c778fd588239313b34cd7d23b960c8f924c67",
-    "source/vpp2/src/vpp2/libero/libero_utils.py": "c7cc23976c86eabcada4846d46345d3605fc2ce03929761646486be3901697d4",
-    "source/vpp2/src/vpp2/libero/prepare_pro.py": "0999050352d924c11d29b22f63e902162d8cfb23f94815c93c01295c11ed81ae",
-    "source/vpp2/src/vpp2/libero/pro_alias.py": "90b5c8c8061bc727a6e413a8746fcc98cbd06e1c7c4175fd223b45af4f10b700",
-    "source/vpp2/src/vpp2/libero/pro_runtime.py": "adf94ed64b032aaeb0d88391e7b653f88f67e76b9411419ece25faa31ccc1340",
-    "source/vpp2/src/vpp2/libero/protocols.py": "68225c95eeeee25032966a7f90c1ba18067c53225397f59a463d764030fd0d93",
-    "source/vpp2/src/vpp2/libero/runtime.py": "1e2e2900a604d33cbdf8774cfa33dcee1065f21dc0b59fcf30799b5098c1662a",
-    "source/vpp2/src/vpp2/libero/trainer.py": "ea29b4e000047a6b0c8aee3cb6f32df9187987a657924085dbafd9628315107b",
-    "source/vpp2/src/vpp2/libero/worker.py": "9df6350c4ab92dea70863c08f1bdcebd0d77574a3a2d8d7ac9899a5a825be6b0",
-    "source/vpp2/src/vpp2/models/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/action_dit.py": "9a654edd860ce77c1308923bd338c7f14d09f68686642531d7143c16b9f6f331",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/compact_channels.py": "6f7ad787624844970985117002ab809bd777f7474969daca121e7101381c6d0c",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/gradient.py": "594249b6eb931048e352fd3ffef9b1ac1db9e4fc74659e04216d6b823dd174f3",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/io.py": "1b68cf2f2fd674c7a59140f8632d4464044bffc2e9277ef5f9a975e7c699966a",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/loader.py": "b9180b9c9177af192e5cb92387b4a09bd0dbc08fb75a94e843b67d315c94e339",
-    "source/vpp2/src/vpp2/models/wan21_14b/mot.py": "cc14a909bfb181a441bc9d8d6de12f3c31d89f3dc3415031e34a5f8c21410019",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/scheduler_continuous.py": "15de04574c1d8ccb794ca6560b1e501e5c83cc86104bb80f1eb3cc3b0324ffe0",
-    "source/vpp2/src/vpp2/models/wan21_14b/vpp2.py": "8c419af905a8ffa93d77c827a283b7aeff744f3a461c390783eef8a333c31756",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan21.py": "01b45cc71f69a6914f7b286559317c8e8485d0b0fec93bb28ebc37c183c295f4",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_clip.py": "f65e44c9d1b7811d2996b3fa22e6fe4c2ac5e6edf1fc2c14251cf582bf8d1f05",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_dit.py": "0d1b5837ea923508c05394d56aaa465813e7b4f5aea3b31ceceae39b2de9b0f9",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_text_encoder.py": "902a1e46b21b5c48e166fe03defabb94b3773c6bbe836fc09b9a0d0894eb1f6e",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_vae.py": "33f7d804677b44a6f3ab334684017dcdb1e9dfb748279de76ad94a5bb71a7e51",
-    "source/vpp2/src/vpp2/policy.py": "2a7696cd73ccb99575bf901185cf578c3775945f61f6e67aeb2b4ec6bfc4e30b",
-    "source/vpp2/src/vpp2/preflight.py": "7a91e486675a4cdf38565e0f4b58e60a68bba7de5654cfff736b5df58f5c53a6",
-    "source/vpp2/src/vpp2/robodojo.py": "183db869a622ef17ee92c487949db2e22a7d1365558fe703bfb8dd8fef83e9c6",
-    "source/vpp2/src/vpp2/runtime.py": "e26885a3732f5fdac3d8381da870a1673d0c03354f02f9deee7fe779d645f8a7",
-    "source/vpp2/src/vpp2/trainer.py": "feaa0ac4ed94a376836eb60cfe6f52460392b11f7cafafec9c6d9f151f5b61f1",
-    "source/vpp2/src/vpp2/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/utils/conditioning.py": "43d43f0f3ca0add64498c3d64a3791513a53e3fce0ea53f77ef09580a86b7b2d",
-    "source/vpp2/src/vpp2/utils/fs.py": "69e6df236b43360428a613686c8f37c531c0d16a3a2c5b06a725c46b49d657c5",
-    "source/vpp2/src/vpp2/utils/logging_config.py": "1de4abf66d9d201b49e7bff48291643b9dc5ec9b6e7918ca73e1bd18887b1c43",
-    "source/vpp2/src/vpp2/utils/lr_continuation.py": "72ea9f2af52041db19aab681d7a99fdf6d87d5ed5486f2499383a211c6c1ab76",
-    "source/vpp2/src/vpp2/utils/misc.py": "7e42c330c24604e118f1878473dc804aa8a272944df502ef1414c05be4d5bba2",
-    "source/vpp2/src/vpp2/utils/pytorch_utils.py": "9b0ba19de9a64458b12a0fb24f0811f49d9e9fca5d8b815f5f5ecd3b7886fd63",
-    "source/vpp2/src/vpp2/utils/samplers.py": "4d3bd9b2d498f4ad8f2cf9765d60c4bf5368fa6fa2a7d42d0db6210f66899652",
-    "source/vpp2/src/vpp2/utils/training_profiler.py": "91f6d98577e63d85e474c738a9bc1901edc194bdf318b5387d527c31b6708832",
-    "source/vpp2/src/vpp2/utils/video_caption.py": "d9572deb3690aecb0f16c052016cedc5f94da8ce8f016b4dff1a4e358e18d5e5",
-    "source/vpp2/src/vpp2/utils/video_io.py": "9e9898b021a44e87aa1b4e6d13e3755a9869a2e27a7eb53c72a6802808d5bd80",
-    "source/vpp2/src/vpp2/utils/video_metrics.py": "1840a37685d686336c7326a2a3d4fbe048d1e8e7e431fb14519d360a8b7aaae8",
-    "source/vpp2/src/vpp2/video_training.py": "20f55bf8b0a7dba6422a1115be1c8adf31a0db6a9f53a8f6f53f5d970188bfb2",
-    "source/vpp2/src/vpp2/weight_resize.py": "269edd4b9d32a1c7f363ee082dde7ce57f2176190daab6d5bf4a1c98ab43f0b4",
-    "source/vpp2/src/vpp2/weights.py": "5dbdff347de1b57e30433efca6da28334d0263d27aaa9b2278d95614db1b8641",
-    "source/vpp2/src/vpp2/width_kv/__init__.py": "0183a7106da7e45223550a552db2905d6b59c2a4259f8c86cb98629b03e66b43",
-    "source/vpp2/src/vpp2/width_kv/calibrate.py": "93a06910cb23a9e1dbd961ee9ef9794d70aaaa6186b08483260c8215109598f5",
-    "source/vpp2/src/vpp2/width_kv/common.py": "45f5f7087f5f3dc067358838e3fc738698b796defecb2ea721545d275f767766",
-    "source/vpp2/src/vpp2/width_kv/diagnostics.py": "cf184ae58f5e1c5f4453774eee1e10c85931e618bbbf768ab42f649cf301f5f1",
-    "source/vpp2/src/vpp2/width_kv/reproducibility.py": "89c5377e1314dd6e8de7f67c4385d30f5bd91fa4a5742449f1c1ff8dd578f451",
-    "source/vpp2/src/vpp2/width_kv/scoring.py": "7644358d0b2d29d3b91515fcea56090dbd2a7a0dad2d5e5c88e8d585bff15533",
-    "source/vpp2/src/vpp2/width_kv/slicing.py": "856203c14e0b06d19585a96392d00a00703779ebdb9dbdc9ab1323d50daf8ba4"
-  },
-  "suite": "open",
-  "update": 1000,
-  "validation_gt_action_mse": 0.0005726480308242441,
-  "video_parameters": 12468409664
-}
diff --git a/evaluation/precision-completion.json b/evaluation/precision-completion.json
deleted file mode 100644
index 29d30f84d6a2f9ad4dbf394d530436df5146ce09..0000000000000000000000000000000000000000
--- a/evaluation/precision-completion.json
+++ /dev/null
@@ -1,14 +0,0 @@
-{
-  "checkpoint": 1000,
-  "manifest": "c62f853a713ce9200730a1b937a619358a9c146f2d914ddddfe8988b27ae47d6",
-  "mean_partial_score": 0.40425,
-  "partial_sum": 161.7,
-  "performance_condition": "none",
-  "precision_complete": true,
-  "precision_summary_sha256": "099fa38f19d9f5ff428695ca9999fa5c291081f1519c8320f2101a86c03ff6ef",
-  "rule": "Finish all400 checkpoint1000 Precision episodes, then Memory, Generalization, Long Horizon and Open regardless of success rate or partial score; checkpoint950 cancelled by user.",
-  "scored": 400,
-  "success_rate": 0.315,
-  "successes": 126,
-  "switch": true
-}
diff --git a/evaluation/precision-protocol.json b/evaluation/precision-protocol.json
deleted file mode 100644
index c2b588b2938c7ab0f644c50a8c3038a0dc0cdd3c..0000000000000000000000000000000000000000
--- a/evaluation/precision-protocol.json
+++ /dev/null
@@ -1,699 +0,0 @@
-{
-  "active_policies": [
-    "vpp2"
-  ],
-  "caveat": "Same checkpoint55 policy/simulator code, benchmark layouts, seeds, r16 and new90/old10. Simulator physics are not claimed bitwise deterministic. Offline ranking used replan24 without ensemble.",
-  "checkpoint_files": {
-    "COMPLETE.json": {
-      "mtime_ns": 1793716093276368990,
-      "sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-      "size": 2491
-    },
-    "action.pt": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3",
-      "size": 4057916577
-    },
-    "compact_architecture.json": {
-      "mtime_ns": 1793716074811787315,
-      "sha256": "a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32",
-      "size": 6563270
-    },
-    "dataset_stats.json": {
-      "mtime_ns": 1793716074808787383,
-      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
-      "size": 100256
-    },
-    "validation.json": {
-      "mtime_ns": 1793716074812787292,
-      "sha256": "9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e",
-      "size": 88781
-    },
-    "video.pt": {
-      "mtime_ns": 1793716059522133613,
-      "sha256": "fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb",
-      "size": 24937214042
-    }
-  },
-  "checkpoint_receipt_sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
-  "fresh_policy_per_episode": true,
-  "fresh_simulator_per_episode": true,
-  "gpu_sharing": {
-    "minimum_free_gpu_mib": 61440,
-    "mode": "shared-shard",
-    "one_worker_per_assigned_gpu": true
-  },
-  "id": "266df4b5d9b8eac16145600a8c4f9223d15e03285d61473c3bbd8c43b694f211",
-  "input_hashes": {
-    "inputs/episodes.json": "d33bd02b965fcc9179b4d4a654e82cc4dda85079bb5d560758350d3efb8c98e0",
-    "inputs/master-episodes.json": "336b784b0846bd12ff905bc31c0c204bd0659c7f4e74cee4a85bf24d6a3c1069",
-    "inputs/vpp2-policy.json": "48949df719d515d6d68fe1c657b96129745cdc32a3e9f9542b1adae7c80b2447"
-  },
-  "native_inference": {
-    "horizon": 32,
-    "replan": 16,
-    "seed": 1,
-    "shift": 1,
-    "steps": 10,
-    "temporal_ensemble": {
-      "enabled": true,
-      "new_weight": 0.9,
-      "old_weight": 0.1
-    }
-  },
-  "native_scoring": true,
-  "order": [
-    "vpp2"
-  ],
-  "parent_evaluation": "/data/praveen/vpp2-recovery-best2-insert-tubes-20261102/update_000055",
-  "parent_protocol": "6d165d60796bc93532ded29dfe114a2551f0ee3e2af2adc8b1ddabccb6073f6a",
-  "primary_episodes": 400,
-  "repetitions": 1,
-  "schema": "recovery-precision-sequential-1000-950-v1",
-  "scope": true,
-  "selection_metric": "task_macro_ground_truth_native_action_mse",
-  "source_hashes": {
-    "control/simulator_recovery.py": "fcee066ec9e75ec081b0a955b136e2ab898cd1ee529a2c1e74a453c350c05367",
-    "native_eval/__init__.py": "2deca42c85edcf3d80dba72439ad95641937396e27e1149d77182cfc082267cd",
-    "native_eval/body_coverage.py": "50ce610f86584965281893ff2f98a78643c7e3207014212aad4c13fce4c8a782",
-    "native_eval/camera_group_cleanup.py": "8df69fdba913ec41ff7b1256ab1b43921a0b6fbef7ffbb6dacbf2adaba0f7843",
-    "native_eval/camera_lifecycle.py": "b0952d95d1da8a1bd3e9f544a2eb1eb0f516159b3442666c7acd9bbaee71687a",
-    "native_eval/check.py": "9587d7b200969ccb78ddf750dd3d8616d5e2245e29baa3a0d798e27a55bdaa5e",
-    "native_eval/compact_worker.py": "129a2e26a85e021f7a0282201dec976104a8e1ea1f0a67e6b75ce760257ee46e",
-    "native_eval/completion_integrity.py": "3b737f6dcf9ae9908b1a3602f66fda46c3b8399ee47b1c61f5c0fb9741031029",
-    "native_eval/criteria.py": "97ebc94d91c62d793e614f178d2d120df989faafb54881dd1d3f5d703df8bb76",
-    "native_eval/diagnostics.py": "53b0deb571984fd02ff65c400d5f3b9bc2754b8207ea8edf4e5017c3588ed3cd",
-    "native_eval/ensemble_check.py": "0bef52411b2ec2dd7e1ad21b8690b1f88ed229107b6e46b8dc43c34ca92a3572",
-    "native_eval/evidence.py": "cf7f769cfd838227b26101d7656ebcaede1b1d45a80785a32efb26c9befef0e9",
-    "native_eval/frame_epoch.py": "7b1d3b10f46d5acf82f6e983ecd5f8995310b2fef9fee04bfae168fda02e3efc",
-    "native_eval/hardware.py": "e30ed3359d91595d3053e24124445498a3f1be9d5c0d52f97e1e7d7d88268a5f",
-    "native_eval/hydra_frame.py": "c1aefee86396025b405e050597af7b67bb07c0dd1e65ea4a9be1ed78f8d99072",
-    "native_eval/particle_evidence.py": "378d667b2ab80176045e0bdcdbe3cb39adb612a356b8f6b81d13ae531bde711f",
-    "native_eval/policy.py": "e1cebbfb03c137bd4bf9f054c69be74f9a7f1df5f7ca13c250e606c2346b55b0",
-    "native_eval/prepare.py": "486df6d062098ea07ac63c4d3be51b5f27baf524d0d004411275243783d02ac3",
-    "native_eval/process_lifetime.py": "7dd20bee91c51546b6da067b96012dc5d5a013e8389dafd567b72c1f742cf97d",
-    "native_eval/provenance.py": "8fc2da1678832be9e9b8b778ff0dd1604db5d464cf1a19ff6137acf6b83d3e11",
-    "native_eval/receipts.py": "9ba5800e6e7e796937992117569efe1b3d625900a8e2ebf7955940a4e8cc81e0",
-    "native_eval/replay_receipts.py": "1c60f9bddccf7245e32df97b9e90749bc3ddbe5d65f37ed2d40ba339b458017e",
-    "native_eval/replay_server.py": "1c2c19bae8d1cd3ff564f1775540718c33293663d30161ea7bd293b8f322cab3",
-    "native_eval/report.py": "9d9f0ddc794d2af2ad6e349f991450592a6fe98db292e1ac3ae4fcadea961fc4",
-    "native_eval/repro.py": "f0e4ecd0ba009774105cebeba31f49a70f34c4f88f11855c939368027789a928",
-    "native_eval/reset_coverage.py": "3c6a414878c4cc2ef4114170356db0a6b2193b03ce5c3b4f6293c10c7d3138b0",
-    "native_eval/run.py": "23d2d58fe1160a03bdcf9591b5d5fae671d21832008cc4958ac58e060b88caa1",
-    "native_eval/run_one.py": "dddeea1abb269293bd08c3f5a69ea0a668620b18d07154ca1c0740a0bb213b99",
-    "native_eval/scope.py": "5787bb7f0760bde49bfb54f09d89e5718471286cbce3d768bcca964a487bf594",
-    "native_eval/simulator.py": "d40972be52b7544f62653114217e213d19d4b386bf2991abfbbbc618ab09cf83",
-    "native_eval/status.py": "b49d03681dbc28b80f39f4d6163c3a6e7c5a1ca62f1f0ef502c77dfbf149ca02",
-    "native_eval/topology.py": "062e9246890ad1f2e8bb8b64b3bbe6598f9dce6fcc33ec3eaa106f3cbc0f0ae5",
-    "native_eval/transport.py": "eb6cc295d064a142372dd51859e91233f61f1ad2c8ac8bf3880739f3fe4abec4",
-    "native_eval/vpp2_pilot.py": "571ca3eeba11fbf2afd4a4058629a280bc0ad4bd3eecca178ddde4aa543bd8de",
-    "source/simulator/LICENSE": "7794bb06af8fe5485ca912454ad1665ccd7846c45f9c2b1258658e480948cdbe",
-    "source/simulator/XPolicyLab/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
-    "source/simulator/XPolicyLab/XPolicyLab.py": "a3909558697b73479f8baee03e15b4d74060a7c9ac2f31fcbc6cd659e7364862",
-    "source/simulator/XPolicyLab/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/client_server/__init__.py": "093e47e707f4a5598b0159b5ee371cd9a918b802c70a6e66154a18e9ef8fc5db",
-    "source/simulator/XPolicyLab/client_server/ws/__init__.py": "79905987a12ebf318dc0b2b9ec6f375d9b4561bac29d6c27d04e318d8f6e00b3",
-    "source/simulator/XPolicyLab/client_server/ws/__main__.py": "ee1c5807dcdf1c8f6721c5447463eb51babdb0f9ffbd91d1e7343834ca4aa19a",
-    "source/simulator/XPolicyLab/client_server/ws/model_client.py": "4d6e225ae5f577d862366350dca8742e91dd9527d6e76b8b2a6d48c991212d67",
-    "source/simulator/XPolicyLab/client_server/ws/model_server.py": "4988823b436cd05e6a4d8b13012cc30c7ffd57711b72497b1223a6b10dc0c043",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/__init__.py": "19aeae375e94e58d437d789d0aff07b232c13f375235384d784984631d25d5d9",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/client.py": "b309f58b1bddd99e56578026b61a15e6cfdc07e843e8a082bfd291d0e853ce5a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/codec.py": "8a97ef7e102fd954c266352d81eaceef0f1b0f5504571c03037683efe387865a",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/exceptions.py": "73161f4cdcc7ffc8a87cec94f5459bfcb9113076be272d6cb2a3ff8d69e0c6ce",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/keepalive.py": "7f6b9e636eefcd6c80a4b62e404a6f89e2f7e1a3b6008500312a972d6990a21d",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/messages.py": "d05c8cbc20e0cbe1502551864907f25897035993f920748c9290c944c959eab7",
-    "source/simulator/XPolicyLab/client_server/ws/protocol/schemas.py": "f1ca82eafa8dcad6d968508adb51c25342b291f7b1c553fff305c381a6098bd0",
-    "source/simulator/XPolicyLab/model_template.py": "4f8d1eeb8073aa7ae6093abf0f395cd5fef70ff4a4bd1fa6b2d83b356a90754a",
-    "source/simulator/XPolicyLab/policy/VPP2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.py": "ab94a8569e6168ab7cde6660890035e11b34961ff2648854a7b61d188c555b6e",
-    "source/simulator/XPolicyLab/policy/VPP2/deploy.yml": "d2c3f06eb2ffa2d1866767dc20db8d60e542fc2418dbc4cf86d4a0f86f944c5e",
-    "source/simulator/XPolicyLab/policy/VPP2/model.py": "699acc0c459c5997db4e3155e2b1b7eb64bf750552436074a369c4eb6e9ffef9",
-    "source/simulator/XPolicyLab/policy/VPP2/temporal_ensemble.py": "42be4b09c56377f4686b2e722de760fa3bafd1c6b92605ada269d94e19efc95e",
-    "source/simulator/XPolicyLab/policy/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.py": "42dc75140935046dfe83f19b3834bd7cf8372e1bdbe51f598b1f91b8d557b7fb",
-    "source/simulator/XPolicyLab/policy/internw0/deploy.yml": "e59de6edeee891d2aa24902ae9034e9c10889fb20116bc9263f30d773a4b0d10",
-    "source/simulator/XPolicyLab/utils/get_free_port.sh": "f45bcb327b45c91367e3e5cdff72d3a42bfb7a09f50e4774701637c4c23fc1a4",
-    "source/simulator/XPolicyLab/utils/load_file.py": "488590d76084f419a36c2fb3e658eeb240ae554678899e9f2ce0cc10da5c4b57",
-    "source/simulator/XPolicyLab/utils/process_data.py": "e044e627eb2677907e9227eda8458c857208913bd484c5e922b53025010eef0b",
-    "source/simulator/XPolicyLab/utils/robot/_robot_info.json": "8a5df36e750971003b484ad770a7a243e523e986d2977a4ecdbeecd4a07946bf",
-    "source/simulator/XPolicyLab/utils/wait_for_policy_server.sh": "f75a8360d83f2960f365dae77804c94a28aca332c2fbc62cdb4a39ae220374ad",
-    "source/simulator/env/camera_manager/camera_manager.py": "1a4250cb3bdc7305fe362a7c6b113f57e6f3e9e4983b69bd61e2c84e8e995471",
-    "source/simulator/env/camera_manager/capture/camera_view.py": "be9ec0d4df8a532c35adc7090fbd7d6a76c163f90e8dac30496ee4f875929869",
-    "source/simulator/env/camera_manager/capture/render_sync.py": "48d70e9a95f3445ae464ab7ea00a00c0d4829752eb55c4ce52ef73ab1f453cae",
-    "source/simulator/env/camera_manager/capture/tiled_capture_manager.py": "959b4d87d96e676bc279d01e7755557bb8c6e73920a75944259f72b7ac385f8b",
-    "source/simulator/env/description_manager/desc_manager.py": "bb566028b523a4cc27363cee2e526c2d3e008e671c9773d4f0b8401f58f0388e",
-    "source/simulator/env/environment/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/simulator/env/environment/base_env.py": "ade58ebc19c79bdbf720def30786a0c397b1c7d2141fb6a1bfd63c80f8a3db8e",
-    "source/simulator/env/environment/isaac/__init__.py": "123023ddc75a0f9b1139c7322bee64a27f0e756b6e84242f43a1481428596126",
-    "source/simulator/env/environment/isaac/direct_rl_env.py": "77958f14c465932697efed302c8ef13054f2ec61ee3a4bc5ad66d0390b71df77",
-    "source/simulator/env/environment/isaac/isaac_rl_env.py": "144af40adc2113751bc1bd72caadab145e918b316de53ef3e4e42358c01a1fbb",
-    "source/simulator/env/environment/task_env.py": "ac816cf26efabbb8ac4276d2a2a09fbd557e3db805514db0c3dd7ac22b58e931",
-    "source/simulator/env/global_configs.py": "e96f6141b42e635d69bba5b4727fb1fec54f8965431b3a70d95fe5dcd3d0ee45",
-    "source/simulator/env/observation_manager/obs_manager.py": "771d6cda443b591f3e7850b901a06b4e653a2feefcbabd23608042377fe9be4b",
-    "source/simulator/env/planner_manager/.gitignore": "bdcb46018aafd15f5c2d40adbbccbb06f2dec5b9ec4311cbf776075bdffd4156",
-    "source/simulator/env/planner_manager/curobo_planner.py": "3cd830c002901959409ef6f64d54106d8106b48471b266409abc2199f3cb33e1",
-    "source/simulator/env/reward_manager/func_parser.py": "eb928a05c3eb6bae429a922c59e64272c2c3851c22d779143d45627e47b3a504",
-    "source/simulator/env/reward_manager/reward_manager.py": "bc56efd3aaa861b4b38960dc4fb7778a9b4da9cf131f815ac3170300dd233a41",
-    "source/simulator/env/robot_manager/.gitignore": "fd94daa58076711ae84376170199a63f66177471e587a76252399fe79eb231e2",
-    "source/simulator/env/robot_manager/control_manager.py": "81b7ecc3bed8248c16aa1517e686682d8e28ccfa9e550d45011b579a9219caab",
-    "source/simulator/env/robot_manager/frozen_robot_configs/franka.yml": "a3c944f8cdd07fc1a4eb54cb915f565763292acb24f9c4f725534271449f138c",
-    "source/simulator/env/robot_manager/frozen_robot_configs/x5.yml": "4df90d234656c34132ca573e69a55de5984270e103aac2ea8ab9956a913685b5",
-    "source/simulator/env/robot_manager/robot_class/franka.py": "dc49eb7738a8b58fb731595340c60a1ce32f52ef0dc63a3da84f9c25879af140",
-    "source/simulator/env/robot_manager/robot_class/x5.py": "4748cbe4d273449a48d09817767dec361f53f8e5cdc6e43368fbb8724141b654",
-    "source/simulator/env/robot_manager/robot_config/franka.py": "8bf57a07913573e74dad97aec6c50e58b67746fa90875915d74ce213856ce35a",
-    "source/simulator/env/robot_manager/robot_config/x5.py": "13d35630f833d1b0a1e3df0bea05f843e88aae1248b59af8aadfefa9cad29bbb",
-    "source/simulator/env/robot_manager/robot_manager.py": "19636ff932a0e2192341b57d8695d4102b870bd10c69c9343d712682fcc15082",
-    "source/simulator/env/scene_manager/layout_manager.py": "1b4c940200dd94a1d0300e103d168b9a823e0dc00ce7c921afe16e81d4e01ddc",
-    "source/simulator/env/scene_manager/objects/articulation.py": "0c542a1fc8ba7a61bf71b343745c71949c863291c1b167d4d7048fe2ef9ca4c4",
-    "source/simulator/env/scene_manager/objects/background.py": "04afea9b5c4e21a105ba3b08ce0e5b6692c082cf17ec1053fdffd5dba1c9511b",
-    "source/simulator/env/scene_manager/objects/dynamic.py": "7a4652c649f503681914e37c50c959c7ca8457918bb00628413f4288993f8ea7",
-    "source/simulator/env/scene_manager/objects/fluid.py": "12b183d4d798d6b7188fd42747ee1927747318c386a91135551d1fa0561c41b5",
-    "source/simulator/env/scene_manager/objects/garment.py": "85f38bd85b52da8dd2ce34ce3692d45848016a9f0f14ecf13114705d4a1ff46a",
-    "source/simulator/env/scene_manager/objects/geometry.py": "3100d61b152b894b3b59be0b3be1d2708f2d8f363ddb8311dabfb5411c731f3d",
-    "source/simulator/env/scene_manager/objects/ground.py": "12dd51addae6f5b7c6ad410b4ae7248231d216cf63513c564057d577a09a2972",
-    "source/simulator/env/scene_manager/objects/light.py": "ba8c8529c6608592d4d0eb9638820b70ae6b3c0cd616631bc6dabdb1c20e421a",
-    "source/simulator/env/scene_manager/objects/physics_material.py": "a1ac0dc9928e9253e644ceb9a604aed262514aaad0473aff303bfae448e76c29",
-    "source/simulator/env/scene_manager/objects/primitives.py": "848f4ab08d86307812ebaea18cc86a51638395f0e6ea194b412d45a699ced926",
-    "source/simulator/env/scene_manager/objects/rigid.py": "8260fd2d8350d6a354c8281685bc7e25531923327078752efe0538d9cd78450b",
-    "source/simulator/env/scene_manager/objects/room.py": "946aa1afc23ab9b8fd17394517a5f54605bbbc609582d0854fed2078b3b2d8ee",
-    "source/simulator/env/scene_manager/objects/table.py": "c0e71396b0d53a250f74c75490e91086d67962237400a49f886a13ec07875308",
-    "source/simulator/env/scene_manager/scene_manager.py": "37911b12601038c9ee9bd180cae7a4751ea649bcdaf1ba2628f22c65cfd7dc72",
-    "source/simulator/env/seed_manager/seed_manager.py": "99bcf5a90392c8927aa0dbd799c07e06b9b03befe8b8eabaa8744a3dc9605ba4",
-    "source/simulator/env/seeding.py": "3f88d2a635c528253439cb31a29e5b9d0f47f6aaef0e975d86d9a8cf174be26b",
-    "source/simulator/env_cfg/arx_x5.yml": "7691350159fdeca92fd1b0b2da36913464ccdea3f3251eefb0176aa51db0807f",
-    "source/simulator/env_cfg/camera/camera_config.yml": "c1752170da855e735beba27fddbf78dad16f2395e4ef8b24e9827a768e616a62",
-    "source/simulator/env_cfg/camera/template.py": "68287ed1daec3b8ed7f053bf97a4d48bbddeb8b5757b53e438ac3c4240a9e61b",
-    "source/simulator/env_cfg/robot/_robot_info.json": "4c9de628bc1572502fc48d876c023fc63e02760e1a11a2a5258dea30ffa330e8",
-    "source/simulator/env_cfg/robot/dual_x5.yml": "4eefaccc75c5f83443b7bb85ef1de68ee7a3772e27af2b04a5e8d8769ab9baa1",
-    "source/simulator/env_cfg/robot/dual_x5_and_franka_competition.yml": "ec893f9fde577d0d1209e7dcb85ec2dd74288b636d0c68316cb07bc817b743ab",
-    "source/simulator/env_cfg/scene/conveyor.yml": "c1dfd27faf978f90dfc4825cd83bd203823547c46793f4f9201ce9f4b43beee7",
-    "source/simulator/env_cfg/scene/default.yml": "729df024b9707a55449da94f980c1ca116035e09a1284e2fe071f1b38efd061a",
-    "source/simulator/env_cfg/sim/sim_config.yml": "e442ede1c3d90fcfb50b9461c854d31b22dad9bfbb23397db228b27a6a486963",
-    "source/simulator/scripts/eval_policy.sh": "786966648496b7b68aad81e000a3f92910302eb391dbd07fec0a2dc91107956e",
-    "source/simulator/scripts/internal/summarize_result.py": "9ec361af587e1788d83d3cf8670e05b817ff252d33c33829a6337dc16ae370d2",
-    "source/simulator/scripts/internal/task_inventory.py": "ec82252b0f87fa5a3dbec2f682c170b25cbe47fe181e431a81df2c4b0675084d",
-    "source/simulator/src/__init__.py": "75bfbedb80f2e1421ad787992d78c23349bc8738b9edef7a76a3af83c9586fd6",
-    "source/simulator/src/eval_client/eval_env.py": "c577b27dab3c87588940670da81bcff007cd64b687eee8e0cc6684e5e24966b1",
-    "source/simulator/src/eval_client/main.py": "de2216bb5df707e505f7e2f3d31702907a4c1ab65205c5ea35494021f168c80e",
-    "source/simulator/src/eval_client/physx_warning_monitor.py": "6a9f8e8d75b1031d2cf2feb6dd72d37370d85d5bd80c4bea798622e23ce8dae6",
-    "source/simulator/task/RoboDojo/config/_task.yml": "655a87cd59fa6272bf352d32deb8ade162737b49a9824e16d428fecf0803380b",
-    "source/simulator/task/RoboDojo/config/align_blocks.yml": "082c9d7a32968052b8ce954bee212c7ec0a96b756150701d4b531abdea59be2b",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number.yml": "3bf6b7e286ea328fb1e4a994a8d30af9afb4b0b4dcd7ceafed43cc6160c9978f",
-    "source/simulator/task/RoboDojo/config/arrange_largest_number_random.yml": "145b26eb5c8ed7d084a4ddc6e21694c542f4a86150756188ea38dcf66e565603",
-    "source/simulator/task/RoboDojo/config/build_tower.yml": "e1bab8756bbc5664e3171e2389c060a79a3fbbf92a6af4ef555dcaf59fce275c",
-    "source/simulator/task/RoboDojo/config/classify_objects.yml": "80a754d2aedd6a89f58185eae36f5b2eb1a28250f40a67325f576fde34650384",
-    "source/simulator/task/RoboDojo/config/classify_objects_by_language.yml": "ac6c739ca82a64b53c799af79b44363db1a88246f76d867f8830729e5613507b",
-    "source/simulator/task/RoboDojo/config/cover_blocks.yml": "68f90c85fad2b152cf7853cf48888d09ce74bb27a692ff2feacd11160771c5a7",
-    "source/simulator/task/RoboDojo/config/deposit_coin.yml": "f547b12216fb9a5dfe2e7d7e882d89e1e53ee1a805c943ba09a3c2cdea8c17a3",
-    "source/simulator/task/RoboDojo/config/fasten_screws.yml": "5fbe4cf5b3ee2eafc2771137b13176f01617f8b2131f02503e676a7b62e2eabf",
-    "source/simulator/task/RoboDojo/config/fill_egg_holder.yml": "d79129278c9a5488d63bb1d3865f2c6b39f21f85957ec94ace091af9ae29e31f",
-    "source/simulator/task/RoboDojo/config/fill_pen_holder.yml": "ac72e5582495e8757c547997b317e2fe990370009527b1faa6bfbf4e1275537c",
-    "source/simulator/task/RoboDojo/config/fold_clothes.yml": "56447e3f592e0db388ebb7482ec758ecd5485ec81b9227c2dffe5390bd7d8c8e",
-    "source/simulator/task/RoboDojo/config/fold_clothes_random.yml": "eafe52c0abcc1889069383b008cbe95c6af2d18ceafb1aecc78f0fae3b491b04",
-    "source/simulator/task/RoboDojo/config/general_pickup.yml": "35de68117cb0ed596b30db1c62f1f9090f2cd1237242a03d5aaeb7186e791d93",
-    "source/simulator/task/RoboDojo/config/hang_mugs.yml": "0086d374d3f7e82abb85e43a17cdfa437e1ac645a0171fb47496c2b81427dce9",
-    "source/simulator/task/RoboDojo/config/hang_mugs_random.yml": "847dfd39059275a671b5beff0243bf7bbaeebd9e5c0612a1877cdf12b9007dc6",
-    "source/simulator/task/RoboDojo/config/imitate_sorting_sequence.yml": "d9e194d8c742288601c641ccb75052a67da46e536c0b06c7fb7311d9e2c61a42",
-    "source/simulator/task/RoboDojo/config/insert_key.yml": "c0b409d5c020e0ad706c378b6f63b39dfe7bed4682d978907c40e1cfa7623335",
-    "source/simulator/task/RoboDojo/config/insert_tubes.yml": "5c070e5c764a2ea93f8de0b52e6a664a23c717963fe09fe3af17918a4a1ee9ef",
-    "source/simulator/task/RoboDojo/config/make_kong.yml": "8a33c9e57b6e6c719c4800911c37272ba055ad0f4cda2f1de9d531ee2910351a",
-    "source/simulator/task/RoboDojo/config/make_toast.yml": "d190a26233db174538ff0052b4c2958a3b107f0ccc63d0fdfe28da4c7f5dc57b",
-    "source/simulator/task/RoboDojo/config/make_toast_random.yml": "ca2c1e55a0975c8b019523710e90131504d8d2afeb829be362e4d7927fe97b48",
-    "source/simulator/task/RoboDojo/config/match_and_pick_from_conveyor.yml": "0ae1c2537d9b1880748316928cfd7ba4c6421ecccfedb85c4c5395a7a86c5361",
-    "source/simulator/task/RoboDojo/config/organize_table.yml": "9ffa15dd279060d4f042285a2c53a341c7208e2f89a2f3dd36ee08b22aa453fa",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box.yml": "e232962b8394d08c04c5da14deb88d7b233afa00ab99af8001c084218951d7ca",
-    "source/simulator/task/RoboDojo/config/pack_objects_into_box_random.yml": "7ac7f034fcb169e33a7f0a03b2b77e54d0206ed715409359b579055a701196ef",
-    "source/simulator/task/RoboDojo/config/pick_from_conveyor_by_image.yml": "6b310b70a0ead9d09e7c221c6bbc952140a44b1d43034613bf02e954eebcf9dd",
-    "source/simulator/task/RoboDojo/config/play_Xylophone.yml": "28adedb0f03c49693af32cea37855f3f95ec3f76f031bfd571fb8bfffb2ec59c",
-    "source/simulator/task/RoboDojo/config/play_stacking_toy.yml": "a09afc106161064a2ea6a40882d717f7872bd4875309a97431d6863a9b7ffa1c",
-    "source/simulator/task/RoboDojo/config/play_tic_tac_toe.yml": "63f863a7fb81a040f65803c64f97c4e9577d4fe32cd0f92563bdc72d1aa336d2",
-    "source/simulator/task/RoboDojo/config/plug_in_charger.yml": "2046b47112049e038ecfa24a8c8044fb1db84f629a4a32eeb24822157d2d4dcb",
-    "source/simulator/task/RoboDojo/config/pour_balls_into_vase.yml": "084a707af61d24123912777b417fe0c0b230eb9653e686af42570791e924ff8b",
-    "source/simulator/task/RoboDojo/config/pour_by_language.yml": "efb2e3c3cb08dd6d2d2ae011b3e9a218b5f1f0a92f025b8e3b9e8abd7011c250",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup.yml": "32092b11667b3306579b04225303610c061585ace28531fa9996afc339c5a2ba",
-    "source/simulator/task/RoboDojo/config/pour_liquid_into_cup_random.yml": "1603feeab6f0f7dc9c4897e7f40f322b86112b6ab987dad0598f42ac1b35db69",
-    "source/simulator/task/RoboDojo/config/press_by_number.yml": "b784643495bcbca933f5a5fbc91b7dad3465fe888239b358d516977c78c38f63",
-    "source/simulator/task/RoboDojo/config/push_T.yml": "b55ea11002ff3e5dc63982ca8c1247fd7bee4f4a607f4d674e220b86cf120a25",
-    "source/simulator/task/RoboDojo/config/push_T_random.yml": "07609ec20cea40afa1d5720e895fb1b445411f261b2e5b6850e75d0f1930d32d",
-    "source/simulator/task/RoboDojo/config/put_bottles_into_dustbin.yml": "7eddf43e287b6e8aea686c6d686fb87e13ac1be1cb5aac16a8a2b0d2e9eb976e",
-    "source/simulator/task/RoboDojo/config/solve_equation.yml": "bdaad2ff1cc8b7ec6413a3715d49ea9cefd664e7e204544977832e18785f1e75",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size.yml": "8927cb2ebb18449554d851e629bd7d568e12e46c5c3c636c8e7122f4fab1ed1f",
-    "source/simulator/task/RoboDojo/config/sort_nesting_dolls_by_size_random.yml": "f895d3ce7eaae99454706afccd09011f56d077998bcc1e95840116658b78f139",
-    "source/simulator/task/RoboDojo/config/stack_blocks.yml": "2802e88ebc3e2d3d05ce345d75e9ad6151a6690e4ef1e0021a5223d291a402d2",
-    "source/simulator/task/RoboDojo/config/stack_blocks_by_language.yml": "4fbfed4f7fc49120831087e709a8dbdfdb4504eb143196000d5790b54eb3bc40",
-    "source/simulator/task/RoboDojo/config/stack_blocks_random.yml": "e25d2a096601da30abdcfe1256a464eaf48c0c8f844eb2c737232626d49639d5",
-    "source/simulator/task/RoboDojo/config/stack_bowls.yml": "4e2aaa12b8b0f58ca801f1a93638c46d3b49ebad809c5a8011382ccda846dcb1",
-    "source/simulator/task/RoboDojo/config/stack_bowls_random.yml": "5c08af1bcc925d958d53de51e8ebd336006203f1bc81e4ef68f3719eadfabb49",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones.yml": "0fdaab803a9370a9e2d1fc8d86a5788daa3940c08fd6f767334f625642fca3e2",
-    "source/simulator/task/RoboDojo/config/store_laptop_and_headphones_random.yml": "93f0ca39bc8bb32df94e84a02f22ed4d7860f37cdc0b9dd17e44721445849576",
-    "source/simulator/task/RoboDojo/config/store_tools_in_toolbox.yml": "641c8ef92393bd986e408d26a84b1cfcdac27466a5ac5f0b3ebe4bb84430c527",
-    "source/simulator/task/RoboDojo/config/swap_T.yml": "1daf769911e13714b4aeabd208292931e32f4dd39968893152bb7bb477c46c9b",
-    "source/simulator/task/RoboDojo/config/swap_blocks.yml": "b6bb2a9df494719f26c6b1aab2e7576107ae76e2fbac8b63a5770f21e0e9f05f",
-    "source/simulator/task/RoboDojo/config/sweep_blocks.yml": "98aa25ac39be9d335c71b2b586bbba372789e9d61e652e1147152c87febbf1f5",
-    "source/simulator/task/RoboDojo/config/sweep_blocks_random.yml": "4ebff5e2cad07f958dc4b2793b86b1aa803c0809394818ee3752d300e58b6187",
-    "source/simulator/task/RoboDojo/task_registry.py": "11218028f772d9253ee709a7d85aa72d1e1fd87e4f8d33ee3787d6064438f9ab",
-    "source/simulator/task/RoboDojo/tasks/align_blocks.py": "c8597e96132cd7c9d30f3ae101f7bbe8ac28597b629599d75d442bbea4c6e74c",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number.py": "277acd20ae4345dc71267762e588b7b0343387c71685183890f044a4808b7a20",
-    "source/simulator/task/RoboDojo/tasks/arrange_largest_number_random.py": "01757509978663bf7c0c95059e8ae18d45b8fe0953f6b54540e2bd6bf141182e",
-    "source/simulator/task/RoboDojo/tasks/build_tower.py": "814cf6ffb459871e5cf49f4e127903dbfbc30926f33db71de0b84a71ab235d6e",
-    "source/simulator/task/RoboDojo/tasks/classify_objects.py": "aef4333c088766375e21a928d53a9a93aaee285fecae41e22385acdc74dd11ac",
-    "source/simulator/task/RoboDojo/tasks/classify_objects_by_language.py": "242a8f048f487eb174ef246be78613bbb5f153549ee11a2650412718cf3b36c9",
-    "source/simulator/task/RoboDojo/tasks/cover_blocks.py": "9d14931b16842b93bea8abf9a07991a93a7e663b93c5fd1695108c8b922ab862",
-    "source/simulator/task/RoboDojo/tasks/deposit_coin.py": "f9868ef827ec17b8db6c4650cc5d9a9e593d20d00ee822e2212ce229ec01fb61",
-    "source/simulator/task/RoboDojo/tasks/fasten_screws.py": "2e187a28e33e44efb87baae6463b3ec35264ccd41e35c21a1be44f6361675185",
-    "source/simulator/task/RoboDojo/tasks/fill_egg_holder.py": "506dfb772f8882ae53966e3580affde19c3c0f11591c78cc5f5b1d997ba04f7e",
-    "source/simulator/task/RoboDojo/tasks/fill_pen_holder.py": "8706dce239487e078b21af0b95bcc5cd4eabdcdadcdcbc6400d6b138d982f1e6",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes.py": "ecea9c287d0579605330168e6f36734704fb14a6b1e5dec82b6f2ee3a56ec20a",
-    "source/simulator/task/RoboDojo/tasks/fold_clothes_random.py": "9e40abe3f7201a8c13961edc1ca8075260c315584d3d6b93fa1df966d13d1e88",
-    "source/simulator/task/RoboDojo/tasks/general_pickup.py": "1a9eef7fe518f9d2c6805ad1a8f97c1045fa83726f84b9c63f67ffdfabbc40a7",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs.py": "4bc6586bd87de030b9c580fea2a45e7977c3b6a4d75cf3c43b546d349d42e906",
-    "source/simulator/task/RoboDojo/tasks/hang_mugs_random.py": "ebfa4a95deb954aa7a9063a07fe366cb5ea1835d3f52342b10d531225dfeee92",
-    "source/simulator/task/RoboDojo/tasks/imitate_sorting_sequence.py": "18400cb3c5f5103492819df2dcdf05a77916106e8a503cb62b4ee6cdbaa9305b",
-    "source/simulator/task/RoboDojo/tasks/insert_key.py": "180978a2915744f0596fbca4ea0ad6dcb47dce005d43f06807249bc8b4d1a828",
-    "source/simulator/task/RoboDojo/tasks/insert_tubes.py": "9a68795ddbd0b816ffd7627e673390ecd83594c0e9d374d4e3abf3435ad439e1",
-    "source/simulator/task/RoboDojo/tasks/make_kong.py": "b9266689d65f9213b05219ad952344f9fd63be8892097ec0103af99791b245d5",
-    "source/simulator/task/RoboDojo/tasks/make_toast.py": "1047f484a73d1f92304393879b51f87e33d661b6ad7937198352fe37a0d3233d",
-    "source/simulator/task/RoboDojo/tasks/make_toast_random.py": "69fb879b89e98f7369971834eb6d1b95c5a65f5ffe978b2cffae0b12c6b756ca",
-    "source/simulator/task/RoboDojo/tasks/match_and_pick_from_conveyor.py": "53110b8a0c971bd3ae8878bfb7d0dcd9626bd9b610d4e7fc91772313b435a5c0",
-    "source/simulator/task/RoboDojo/tasks/organize_table.py": "aceb298e78f3f39e4cf9a9f4c5a4f747b2c55f1a6006de2065e110a05f4295d6",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box.py": "84a1cf1dd5e166f36291130f9d8cb5dfe4b3f4732a46674bef6374a7ea2d66be",
-    "source/simulator/task/RoboDojo/tasks/pack_objects_into_box_random.py": "4db259a1b97cdcc39ebc9794244fadc79d31c4ae64069c9b11a233356fcb655c",
-    "source/simulator/task/RoboDojo/tasks/pick_from_conveyor_by_image.py": "a0d29576d0aab8232d2fd3ebcc8c510e81c70b4cc04602a5be3e99ee2275e01d",
-    "source/simulator/task/RoboDojo/tasks/play_Xylophone.py": "b7a4b0bc0457bcf55d714ca1867476fc4aabc53084adfc78d3ff5d1437bf7847",
-    "source/simulator/task/RoboDojo/tasks/play_stacking_toy.py": "23361329ee7efe330da1f1c241415205bafed2b6c082a913659d868a80d338c3",
-    "source/simulator/task/RoboDojo/tasks/play_tic_tac_toe.py": "8f75ebfe55c997ad9bd3b342db46e843e87397db94356f31cadcbe0cb483bca7",
-    "source/simulator/task/RoboDojo/tasks/plug_in_charger.py": "aaa59ade46dc54f430e1148e3f6fae487970cd29eb2c07b5ff7b4045edb831b6",
-    "source/simulator/task/RoboDojo/tasks/pour_balls_into_vase.py": "eae388b93ec196c401f76bbab3b0119cb2164f90bece0225d4fb925d3f2b73ab",
-    "source/simulator/task/RoboDojo/tasks/pour_by_language.py": "f7a15a7fca2c649a7918d68960ede27e667f409a339b668dd9876db890724cfb",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup.py": "73f8c91b57c352fdc1e7ba3fe5354b3c10e4156d8e5ba777762ad0fa1aff5663",
-    "source/simulator/task/RoboDojo/tasks/pour_liquid_into_cup_random.py": "a1c4097b6e34b2c62885b3a9c7d2584a93d9e296dc41d311a80c2e6d541d53af",
-    "source/simulator/task/RoboDojo/tasks/press_by_number.py": "006f969087c00c417a63875a9de153f698c52d1f1ca0caf0aa16076aa15059c5",
-    "source/simulator/task/RoboDojo/tasks/push_T.py": "a01d753bc320ffbf492ccc29e5f7ab3b39e6c7d255cd25804db1934a30fa5b41",
-    "source/simulator/task/RoboDojo/tasks/push_T_random.py": "dc8d9c8e0d2e8dfffa2b7d433a8278c0eaf707cc4861b15035364c2e70061dab",
-    "source/simulator/task/RoboDojo/tasks/put_bottles_into_dustbin.py": "3ba6e19df0c5ff19ba961092f031ce14e91e72a9a4a5934eece6000a1ffa1c48",
-    "source/simulator/task/RoboDojo/tasks/solve_equation.py": "e287a2f36b82b28e8ca7824f63b1b723b02972ea6dfc115ff32b2882729751f7",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size.py": "1838fa6b515bad3796895cb5c1883c6a42789b7dada5434765764600e70beb49",
-    "source/simulator/task/RoboDojo/tasks/sort_nesting_dolls_by_size_random.py": "b793ab5bd32bd2c10bd46874440904837f9c4e290e7508b06426e2e96425709a",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks.py": "50bd388f9e47002966fbb0146742467a820295b8686e0f86379f0efbcfffcc1b",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_by_language.py": "36a83925f35c7a4caac5add328dd35cb1d346bad833d3064f455d88221e7a41e",
-    "source/simulator/task/RoboDojo/tasks/stack_blocks_random.py": "19d4e5cd693e068ad64de75cbcd24977ab8d39e22fa17362a391dcd733bc9bb5",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls.py": "022c129ec3493183b1a26688a0402f5a00a1546ae011999e9c19f4c0029e2b7d",
-    "source/simulator/task/RoboDojo/tasks/stack_bowls_random.py": "a0e20469eb44a9ee98650d09386814c1df5264c68ec139f06aabf0c86f092514",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones.py": "2bee4e31f0594900629fe7f43266e16fecb310b152930ce51524507136809a92",
-    "source/simulator/task/RoboDojo/tasks/store_laptop_and_headphones_random.py": "58c3c5113054a379214d64ae406ed8dde99603cf684ce9cd8c7bcb72ed334382",
-    "source/simulator/task/RoboDojo/tasks/store_tools_in_toolbox.py": "2969fc3eb171560b3c0bcea62b9f38b8b35f3aecc9bcc7085c3b1cec365b48fc",
-    "source/simulator/task/RoboDojo/tasks/swap_T.py": "57fc289b3957caa454e1d8e005e3a57c973d6ab40121845c94799e61e4bfafdf",
-    "source/simulator/task/RoboDojo/tasks/swap_blocks.py": "08ed24ae9dff27f4e5c850161a02f62ce237859c096d9b06a853a7890e0ad4ed",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks.py": "65786217e199c1e2b26ec8e2c232bf98ecb304a20939d0908fe12bc41e990614",
-    "source/simulator/task/RoboDojo/tasks/sweep_blocks_random.py": "68b6127038322cf001d5f614060b8a1d96e64a49efddfcdcae91127d5c206bc5",
-    "source/simulator/utils/cluttered_generator.py": "f59cfcfed2b4cce5ae9aeaf3339c86d17e237874e3526bded3cb05136dbcbd50",
-    "source/simulator/utils/ensure_usd_path.py": "7c8e71cf91636e630b1b174098fe0b38740857daeaf207e2fae772c4727b11bf",
-    "source/simulator/utils/load_file.py": "ef3a6dc96aa1a7442cfd24f7be9671052a98d2f20783c3e71032332704132943",
-    "source/simulator/utils/path.py": "1d427c7cf33e748e01d0d083c9432d8be8b5c3a1f35c5fb0f6ebb0714098f1d1",
-    "source/simulator/utils/pipeline_utils.py": "d5d8da29e6ca84ab04323a2f79130f930572785b4399634e06a42834ed4087ae",
-    "source/simulator/utils/rotations.py": "8e814f79938bdb9e85f1c024825617cc08d281b61728c6410b23ef1bd852c340",
-    "source/simulator/utils/save_file.py": "6cff5950f73bf7a1ed404f804d6f08af712974f22c3955de4cdf142f211aaf00",
-    "source/simulator/utils/transformer.py": "73de4f5e1e0ff60d4f2e1aaf6104368e8f919269facb68b4fa0344d35bc08cbe",
-    "source/simulator/utils/usd_schema.py": "260d30b0ada2976136cad3e49e810b4409304eb4a2bc60bd9ed710864d3953ac",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/METADATA": "614985b278f601b8ef863f82783b991cc3ada4b89f592eb71c7c48c5f2ef2298",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/RECORD": "c7660eb223c8080f75c49690eaf0a9c9c00b6195511791b33b15e6b24e42edd7",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/WHEEL": "942926c567d0b1273d09f20295879abd37507845f6b683eab9f87d808df91100",
-    "source/transport_deps/annotated_types-0.8.0.dist-info/licenses/LICENSE": "fe1049884b1a0d9342901e88e07f32925d24b3121d9972b6a6805fb9824b095d",
-    "source/transport_deps/annotated_types/__init__.py": "a7104a4d439b27a9f74fc0be236b9ba1b7831e6044026802a205abc1298a9bc8",
-    "source/transport_deps/annotated_types/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/annotated_types/test_cases.py": "d867472acb57b81ce97f78972828e6bc5669ca4db33fd1e6f645c6183846068e",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/AUTHORS.md": "5fe0446ddd6d1c70fa298747368582dc1d91c23d62c8fb7228595bc87fd53232",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/LICENSE.md": "509aa7af19ae2b5b1681b22b969627ee9e1fd15ff7ea5bf9f6e301fc44d5a6c2",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/METADATA": "b7ee8fa1bf018747082cd6d2fcaa14e5d0f5956b5477ab5810eb61ac4cc7bcf9",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/RECORD": "51331727bf8d9ff60510de5e7756c5dd72aba7663f3673e2827bf1ca5c09c7e3",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/WHEEL": "1d7faff7e9e8524c94a31c99d4f312b92edab94c43011e1505da182ea0f4c70b",
-    "source/transport_deps/msgpack_numpy-0.4.8.dist-info/top_level.txt": "19da698ad1490bdde1c7f18fabaceb2625aaf48b364d4dd7ad404c7e5eed75b9",
-    "source/transport_deps/msgpack_numpy.py": "0a45a7430e2cefa2e74235c4c6a34a22ead2fb57a7de8b2631f45df977701d63",
-    "source/transport_deps/pydantic-2.12.3.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic-2.12.3.dist-info/METADATA": "348336c208743aafeabfa5537c29f429035a3107b7a78bbb6a3bf7b3a42fe4aa",
-    "source/transport_deps/pydantic-2.12.3.dist-info/RECORD": "089c7ceb033486c4e6106749bb1a38c80b63905da74d842d067735e8b9c86e4d",
-    "source/transport_deps/pydantic-2.12.3.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic-2.12.3.dist-info/WHEEL": "aad0b0a12256807936d52d4a6f88a1773236ae527564a688bab4e3fe780e8724",
-    "source/transport_deps/pydantic-2.12.3.dist-info/licenses/LICENSE": "a9e186f3ca16b5eef84318e7a701721351a00cb7b8ae3a4394b67b49e3529ef3",
-    "source/transport_deps/pydantic/__init__.py": "e62127278c07bf5384cdd2903f368a69929f3b8a524000bae4e0eb608ebf4bc6",
-    "source/transport_deps/pydantic/_internal/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/_internal/_config.py": "4d66708377346d91e24f76e8479f9862a928b877308d175e9e6c861e87d5ec4d",
-    "source/transport_deps/pydantic/_internal/_core_metadata.py": "63f836b778bbba5b8afb05c266fcc97d114c3d4336dda0582df69ee05cc13f2d",
-    "source/transport_deps/pydantic/_internal/_core_utils.py": "d63aeee156c9d31eb747a76d55cb8e23e74a4130bf77f9529c9584050cc63444",
-    "source/transport_deps/pydantic/_internal/_dataclasses.py": "4e4d6611a7e169dd6457b2b9b4f5f9070c5649763b0be30ac1fd0e2c58089440",
-    "source/transport_deps/pydantic/_internal/_decorators.py": "3e7c80a0a4a0dc135b0954999f0ab0f67684835503b58bc34fd2e5ba280f88ef",
-    "source/transport_deps/pydantic/_internal/_decorators_v1.py": "b5f75f76940a6384765c23b0a876d97a840cbabe95362811adf86e757047c77f",
-    "source/transport_deps/pydantic/_internal/_discriminated_union.py": "68c9744914b24321df5b8fa49673131cdbf0491a2a1371f73d157fd39bd1b138",
-    "source/transport_deps/pydantic/_internal/_docs_extraction.py": "7f2ce74801e1e40ce88675d94ad5741ef1fe9d16dabcd1cfca0fa49f1f92fc41",
-    "source/transport_deps/pydantic/_internal/_fields.py": "6127c42aadb51608cb27a62a6172a1d1e104b399f130fbd0ea1a7da40f0dcdfc",
-    "source/transport_deps/pydantic/_internal/_forward_ref.py": "e67dd8efedc028b9fcfc54b761cecabad2e23d48725e602d90464e6859e89f03",
-    "source/transport_deps/pydantic/_internal/_generate_schema.py": "4d3e3dbf3633a87f74ad6aefe69b4da19823cceb11d0a3e54a4a8f545ae7ac1c",
-    "source/transport_deps/pydantic/_internal/_generics.py": "10baa34fa2cccd0cd600ad04079ffdaa47bf8806b3cf4390e20ba7a7fb8ab986",
-    "source/transport_deps/pydantic/_internal/_git.py": "2303e1dc33df6b65eadeb06e07d371f25b91d80d62ebd41d7937d659722e0958",
-    "source/transport_deps/pydantic/_internal/_import_utils.py": "4d18710f93ae63a094a2c8a805d05c254b34a26ec820e36265d60057bcd0f233",
-    "source/transport_deps/pydantic/_internal/_internal_dataclass.py": "fdb79d7355dbbaeca044688b66a914930c05a506a847584a2e5479d359f2c926",
-    "source/transport_deps/pydantic/_internal/_known_annotated_metadata.py": "25ceca4cd159a01ddffb489b3ff36024834e795bd8137ab73930500e354c9375",
-    "source/transport_deps/pydantic/_internal/_mock_val_ser.py": "c2645115206fa9f70b6c8e353ec165881e2ed80677989a597a21116c3df1293a",
-    "source/transport_deps/pydantic/_internal/_model_construction.py": "c24f9b3460c026f76e686be7d14d3ff33125d06111bb4b2126f37cfd97e461ac",
-    "source/transport_deps/pydantic/_internal/_namespace_utils.py": "865dfe4d102bf365368d3c8f3f7b7e42ab2f28b8abc6d90b7cd7eb37e7e9d31f",
-    "source/transport_deps/pydantic/_internal/_repr.py": "8d07e726ec83c50a5244d846dbd208f4f5fc7b836fdaa599ac4c3696a8a1dd41",
-    "source/transport_deps/pydantic/_internal/_schema_gather.py": "54b12fe754d811e78d0f6733b32ae626ad4c56726a4ce9a72da9fb546e3873c0",
-    "source/transport_deps/pydantic/_internal/_schema_generation_shared.py": "17fadb41bae4a289a0c6cb24747a4fd23509ed309f7b40400c0124aba089e273",
-    "source/transport_deps/pydantic/_internal/_serializers.py": "6085af4a6011e5f9db19258239076e5add72078650638d9e02bb9cf9e9eb6fa7",
-    "source/transport_deps/pydantic/_internal/_signature.py": "f049633c97b8a529daa6e8ab1b90e4040803d618201f1100cb314f1fef47d331",
-    "source/transport_deps/pydantic/_internal/_typing_extra.py": "755c481a1b0be37cda79b3423108752b94235fa2386bd3959cd01c2f6a486bfc",
-    "source/transport_deps/pydantic/_internal/_utils.py": "0878c7fb4ce75235fd479b144a33d3f23f3efbb3f37bcca36a27ac7130152e24",
-    "source/transport_deps/pydantic/_internal/_validate_call.py": "3df7559d2ce15ceac436d683a03c373c558f5580f95bf80d60f7bca77520e8b8",
-    "source/transport_deps/pydantic/_internal/_validators.py": "76fd1ad8d91ce33718aafdf51a1fd021dda571ff96d24429574a1237380475f8",
-    "source/transport_deps/pydantic/_migration.py": "545ef72d1094cf722b6f9c55b75de36f7340717567105e93d7e27438b7de6790",
-    "source/transport_deps/pydantic/alias_generators.py": "28cd67deee097cb48197552e620de1a181f35c90fecaf82b9eaf2ed5c73087f0",
-    "source/transport_deps/pydantic/aliases.py": "be1087ca84969d7f8427ec166f9aa3e31c91b2c8065a74d07f3429e06499f6e8",
-    "source/transport_deps/pydantic/annotated_handlers.py": "59fc854aac28108157061ed3ef73f270a9682350e25f8e465a2d3e24eb024786",
-    "source/transport_deps/pydantic/class_validators.py": "8bf5778fe3d87462d22e68ff20965e913463ba73bc488573f0b325aae3f23fb1",
-    "source/transport_deps/pydantic/color.py": "033a867d540717ddbf65cb4372e7b40cce324e9d8fead7a49304483535ab2c8a",
-    "source/transport_deps/pydantic/config.py": "b83676c4df89ec255035e9c269529c62c244cc83a92aa9607d0bc3774d35c229",
-    "source/transport_deps/pydantic/dataclasses.py": "dbf8cdfb9d595abbb222c7a78cb235f9585de2a206345ddeb9ef6319c769a872",
-    "source/transport_deps/pydantic/datetime_parse.py": "402f9680cc4caffc10fe6357512ec055ffb684b121befb0f6353d0ca148639d9",
-    "source/transport_deps/pydantic/decorator.py": "617fa3500a6ee4029a55628fa1a57e9fee25ed46d2ebd184b7d45ade1b3398a2",
-    "source/transport_deps/pydantic/deprecated/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/deprecated/class_validators.py": "10071a550339ce9db0066974c9b37ad827d07f226f0cbc7941df4f938fed6f85",
-    "source/transport_deps/pydantic/deprecated/config.py": "93f96c564e7ba5ac4b24e701b9e1f4edcbb53a012059d541c66ea57da0b70825",
-    "source/transport_deps/pydantic/deprecated/copy_internals.py": "1a177ebe431de446020a0c821ad3ca3b9f27a7d704281402eaa9017881052368",
-    "source/transport_deps/pydantic/deprecated/decorator.py": "4c19ba6c9ef026c36287ff16ab92330dcc0fdf69bdfef7f1b3de9d7ac2ee934d",
-    "source/transport_deps/pydantic/deprecated/json.py": "1e55821b7e5146bc72cee4d2e8b4d089907046198365699ea901fcacb5bac00f",
-    "source/transport_deps/pydantic/deprecated/parse.py": "1b377a6ff83ccc95dcb84ed046ae5a761c7f10c25a8577dca57085d1182baaa2",
-    "source/transport_deps/pydantic/deprecated/tools.py": "36b9bda054595a9f3e8e57ef3e02482c4b32c29e18cd90f9d972063c32f11dc2",
-    "source/transport_deps/pydantic/env_settings.py": "e881de79612a9543d152fdd5f805e217f5bdd5f836270fccdced25df827f7b20",
-    "source/transport_deps/pydantic/error_wrappers.py": "44aea6a804dcf72303f8a043f48252f47a4a0a9ad61ddf30a3ce019e6fb77d1f",
-    "source/transport_deps/pydantic/errors.py": "edcb41342b6de7b919171ef52ecd87f3a22e7d0011bf8c0f29ff0385db159f9c",
-    "source/transport_deps/pydantic/experimental/__init__.py": "413eeb298743b0289327d1848e6b107561d2730a4aaeb3641aaeafa8e34fe914",
-    "source/transport_deps/pydantic/experimental/arguments_schema.py": "1059e35ffba5a7eb4fc948d05f9a6642dba0d4e14bfc071cf1b70c6cb77e7d56",
-    "source/transport_deps/pydantic/experimental/missing_sentinel.py": "8507a382d174d3052e40c9e2f78dbd7af83e797c88c292af8ec0fca1fa9f8fec",
-    "source/transport_deps/pydantic/experimental/pipeline.py": "2affddbdc7b12ae99acdf44bd32ebd01ac9e03a1f7ed2ae6b19dd2525fe7d2a6",
-    "source/transport_deps/pydantic/fields.py": "5ae0c63af076d8a5aeb96ddf5e74b85af838a97fed769f17ec1ac09736b8b30f",
-    "source/transport_deps/pydantic/functional_serializers.py": "ac4cc7dfdd73ab2de8fdb5a4d9012ebf24a67103599b05e62502c2dd9185ed00",
-    "source/transport_deps/pydantic/functional_validators.py": "73ffbbc1e5a918d70e61f45f55416ed758ebc4c54c75f63f73ee22b2dc24f796",
-    "source/transport_deps/pydantic/generics.py": "d19a99f4ef5a9e7223ff7986051aa9b3886d7b2ddbe65575f9ddad5313cc3270",
-    "source/transport_deps/pydantic/json.py": "647f11908ee1e01cfece9f0e7530316c952856f72853e8e131d459d01fa4d317",
-    "source/transport_deps/pydantic/json_schema.py": "677786d509be226f8c417ebac28c963fd5b2c055272d116d68e0cbced559f937",
-    "source/transport_deps/pydantic/main.py": "efb414c13b5041d51a7e8af78abe70d820b5b34be2805b460747d2953bc5dbb2",
-    "source/transport_deps/pydantic/mypy.py": "a7a294d46c0f1dacc5ec4e6f26ad6e2dde2d1dde8313a6eb7b8fa6e4b9f6de6b",
-    "source/transport_deps/pydantic/networks.py": "2c995c70c79df7b1007a9293e8e98f23d73a8e1e62c1b063935fe35f5449c37d",
-    "source/transport_deps/pydantic/parse.py": "c2477cd9d82dbd6b43f3de54fc8e84d61b6a3251a1052604577f5cb8149ea370",
-    "source/transport_deps/pydantic/plugin/__init__.py": "6bb4f0dfae94e8ade496d08235963bea773db2cfbbb86190e344d769df4eca94",
-    "source/transport_deps/pydantic/plugin/_loader.py": "f502d79de2c49afca15e46bff63e0bae46e67b8a8fbfaa98a6196c8c5d8c1ac0",
-    "source/transport_deps/pydantic/plugin/_schema_validator.py": "41b9aab06df730199fb4d4369cd8ae376d8b85badec5403b8a90d5bf7274d815",
-    "source/transport_deps/pydantic/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/root_model.py": "06f98bb56e22d7574993e74b38cdeb97e8e725d40679e41314172610eabaa4c8",
-    "source/transport_deps/pydantic/schema.py": "56aaa3beafcb9daa5592779b51ddc1a7f275a768176599d94602f8f1b5441bba",
-    "source/transport_deps/pydantic/tools.py": "88742977c489e43093b4f5796ad015d3a4fcf5b8d268c15e659d8b5fd95ab196",
-    "source/transport_deps/pydantic/type_adapter.py": "553fbeca0e1adbbb21481cd61c13cacf8f777f7890f686dd9049218d92a513b4",
-    "source/transport_deps/pydantic/types.py": "9ea752f89d995ea4e1daa7b224ecc14c1b475b267961115ef2068c579f5df471",
-    "source/transport_deps/pydantic/typing.py": "3fb7de037e4cc1370bb11d6e2fb75bd12fa8c9d071a1b9976b9e580e80606a34",
-    "source/transport_deps/pydantic/utils.py": "d799d1d90a6a4c1165415e1336d4c8b4cc93271fdf6f257e80f988118d46a287",
-    "source/transport_deps/pydantic/v1/__init__.py": "14b43c212a7a3156517e39d2edf4389b51714311421764158a41380cafb83e11",
-    "source/transport_deps/pydantic/v1/_hypothesis_plugin.py": "e444b9c56bb0d45400b3298b7b3cbc4209d5cf466955f5378e4993ef81f6ed54",
-    "source/transport_deps/pydantic/v1/annotated_types.py": "ba4d8d000c6a88d10b2a388787287129a20e87c1759585bf2f3ad6dd7ee86417",
-    "source/transport_deps/pydantic/v1/class_validators.py": "50b39a21481850306c1cbec411511aadc33e52e6ca52082837c8526c3a27b051",
-    "source/transport_deps/pydantic/v1/color.py": "8990012d8a7a395a28d801643fd229ae2ff049cebe2a496ef18ba149aaed7798",
-    "source/transport_deps/pydantic/v1/config.py": "6ba3f459eafdc7871bc0a5bb5eff29a1252a33858ff912d6c17e9832962af400",
-    "source/transport_deps/pydantic/v1/dataclasses.py": "efce1caaf2276f020f5abf6eb1fa57ddc87bcf8b77276b532ba374ebbff0935a",
-    "source/transport_deps/pydantic/v1/datetime_parse.py": "e10cb5910a6adeb3556497881de48f0e9b92d81be1a7528fb73246d71bbe1f4d",
-    "source/transport_deps/pydantic/v1/decorator.py": "cda6b1c71a163c29bcd7c0f56ecd3285aa518d79b7d95f06d191d609d335b970",
-    "source/transport_deps/pydantic/v1/env_settings.py": "03d557c2d465d36018fa31f40b4a2ecb954dc377e2e85fe9933b870e380000e3",
-    "source/transport_deps/pydantic/v1/error_wrappers.py": "ebadb931fc3daa40b63708ad07f2450167bc07e5efeb3054eeb2fd936f2d7f2a",
-    "source/transport_deps/pydantic/v1/errors.py": "988c0f103e6f18ce50e6fe02e19d493e5753447fa89af132947ea4b0c84e98fc",
-    "source/transport_deps/pydantic/v1/fields.py": "56a5890ab894362132a6d5eba03549e74d49a550347a7d9500d724b2a5cbd9bf",
-    "source/transport_deps/pydantic/v1/generics.py": "5730bd61457e11b3e94376807e4d5c3457a3efdfc8cf39e443b5ab9934d94bd1",
-    "source/transport_deps/pydantic/v1/json.py": "590e47cbf848a5f751dd84bc93a37613a28c273b1d6c18bf95d58e3c9695f353",
-    "source/transport_deps/pydantic/v1/main.py": "cee36974de50d15d301b65144cab741d4cb75c9e0e02f3d26437625d8f85233b",
-    "source/transport_deps/pydantic/v1/mypy.py": "0a5f1745f0a621c544de3e40114e760bc88387c95c5fffc3de1cf68c85b1300b",
-    "source/transport_deps/pydantic/v1/networks.py": "1d836d2807ce98e9ca269b0383583a488923f563e153ffa2f25e5ed89281a46e",
-    "source/transport_deps/pydantic/v1/parse.py": "049b5daa2651b5abfd55114298ec6863e2889909a33f2f80fcda288e21546716",
-    "source/transport_deps/pydantic/v1/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic/v1/schema.py": "6aa06e03ef9cabc8005648a6e4124f140487cce67c745b66157fdf346afa8a9e",
-    "source/transport_deps/pydantic/v1/tools.py": "d650dd5c79348cbe6e3f7bb9442600bd402518296000efb8e6592af63edfc810",
-    "source/transport_deps/pydantic/v1/types.py": "07397e45c9e2b4f0479eac308fd8afec98c7b8dd46a67587db874a905de5f5ef",
-    "source/transport_deps/pydantic/v1/typing.py": "ec67418356131d42d4f35b6107ff5c8d13437d97e7e248685fb9c6b70fe47821",
-    "source/transport_deps/pydantic/v1/utils.py": "339151c9f3546f50369a4f6568680255d7c71dbdc0b5082b8cee6a7f205fe310",
-    "source/transport_deps/pydantic/v1/validators.py": "9725249f5316847c65097e597c48058ff0801e88e888f71a41e31d10cf57be25",
-    "source/transport_deps/pydantic/v1/version.py": "1d79d75bed5b316e6a2a196be5180e10fa21ad90c20d2bb2cbcfa08bc182819a",
-    "source/transport_deps/pydantic/validate_call_decorator.py": "f23a8b9605d38d6123e1d5c3834c08dc41902a46f42670ac2ff2525236d4e528",
-    "source/transport_deps/pydantic/validators.py": "a706c825755bd42576980138c3f1067cd8fb0f0cec29a5b0fed4cbe9cbe24eeb",
-    "source/transport_deps/pydantic/version.py": "4e6124eaa03e540d4f56c737f5b66584f010032bda71de812b3c57007c96d395",
-    "source/transport_deps/pydantic/warnings.py": "5aed551b3aef159c384fac8220a1e31fb2d263ae878dbb720856e7e6e5a8309e",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/METADATA": "3434fb886fe254ed414541f305a86e43ff575093f8f6d15c89bd11144ed8e131",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/RECORD": "7f3a82de0054511dbd2f7309c5e5b6ddb3d96c6230bfa83fd73bbdffc51cd0f1",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/WHEEL": "7f32b0ec3e21425972f904759f2efc26a495ce223c174ce187bb87ac82f5240a",
-    "source/transport_deps/pydantic_core-2.41.4.dist-info/licenses/LICENSE": "2afdd30d54b4d62b6f488a6bcc1546e84ec5061f13f4209c03d012348783795a",
-    "source/transport_deps/pydantic_core/__init__.py": "9cad6292b75254af606a970aae9bff6e5ae9f0b08089cd63acafa391b60794d2",
-    "source/transport_deps/pydantic_core/_pydantic_core.cpython-310-x86_64-linux-gnu.so": "cd9bb47485cdffc363010507a214155966ca860b0b93bda39583492c6764e631",
-    "source/transport_deps/pydantic_core/_pydantic_core.pyi": "3ea1dbd4182f08cf9341f24b3c5cf7db77a05b3535ffe9e23525127a3617a11f",
-    "source/transport_deps/pydantic_core/core_schema.py": "bbdc850b72d685133a0e250fed263b33691dcdf3813492f3c0e31001e3d46005",
-    "source/transport_deps/pydantic_core/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/INSTALLER": "ceebae7b8927a3227e5303cf5e0f1f7b34bb542ad7250ac03fbcde36ec2f1508",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/METADATA": "bad3d9e6342ef6750cf5128ec942327b8bebdce0b6e4d6f67cb15b27ae5118da",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/RECORD": "cf5e3948866214f2e9ab0c1d43c157e9f9c14f59cf8442936c6cb97d5f2b6a5d",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/REQUESTED": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/WHEEL": "ccec20e2307acd7da453dd7437e70c6b08e2bc3ead3bc3445af135da37b56d59",
-    "source/transport_deps/typing_inspection-0.4.4.dist-info/licenses/LICENSE": "804b59b25f2c31bd278f9202a19ae49a3945aa2664387e2d0a128c7cacc61ec3",
-    "source/transport_deps/typing_inspection/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/introspection.py": "5888c45bd627da692d9527699e3f549adf35651596b1656772310e12181b4396",
-    "source/transport_deps/typing_inspection/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/transport_deps/typing_inspection/typing_objects.py": "a778d6290d51581c9c9a23e0263af2a4604c76def5c4d40a7cd1a375b0046ff6",
-    "source/transport_deps/typing_inspection/typing_objects.pyi": "3bd24dd92d0270d31d64fd40abc1ef8693f5322aab9a78aeb3c5b62f992f84be",
-    "source/vpp2/src/vpp2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/adaptive_mask/__init__.py": "a73f5515ffb77d323f9c324776e9c166d13817c470506cfd5db4ef0546cbbdda",
-    "source/vpp2/src/vpp2/adaptive_mask/audit.py": "c949909a2e60c3c4389c0221c79459961f2c4925baa07f7985774a52257a1d17",
-    "source/vpp2/src/vpp2/adaptive_mask/budget.py": "74e32285903dbf673ecadb6a60d917a990b9c6619beac9f926c1b3cbdae3887c",
-    "source/vpp2/src/vpp2/adaptive_mask/controller.py": "328f636201a845ebdc211340a9014a5076fa895e10cf9c2e1f91b96761a2514b",
-    "source/vpp2/src/vpp2/adaptive_mask/data.py": "e3b633fedaa503b101c80610cbed0933e9e245acf56a6c0692c4d99620faf84c",
-    "source/vpp2/src/vpp2/adaptive_mask/equivalence.py": "8658cb69d549fa0575570364fe3aa1e0b121fcf8b9965b17cd174d459dddf5ab",
-    "source/vpp2/src/vpp2/adaptive_mask/execution.py": "efec52c9b2eabf0300c72edea18f9cd4cc16a0889496793abb6a50161f2328ef",
-    "source/vpp2/src/vpp2/adaptive_mask/export.py": "0ce51e74787d28bbe9f4e73a3774297b424aa131fa78968d24154ff56db98b0f",
-    "source/vpp2/src/vpp2/adaptive_mask/feedback.py": "519fa4e6416723fc000754815a2f05d3415299ae35724f37bb806ebc22393d38",
-    "source/vpp2/src/vpp2/adaptive_mask/geometry.py": "f1ce4110c3a3a893f4f240c49e2ccc4e93215ac4ba9e447de8042f9b19afab98",
-    "source/vpp2/src/vpp2/adaptive_mask/hidden.py": "968560bd6dae0fce59bd28084091b10d559cf165c81764f6ad24b7492101810d",
-    "source/vpp2/src/vpp2/adaptive_mask/losses.py": "4a7dea6f87ef0fa38a7202d85d98d37460f8f4dc13ca5336f32b1da8b84a4322",
-    "source/vpp2/src/vpp2/adaptive_mask/masking.py": "31b595916410c60f772fff1ab63e92c854c8fd0d2261e6b865fd05ce2c61b9c6",
-    "source/vpp2/src/vpp2/adaptive_mask/monitor.py": "48b71f499dd5a0a79557eeb918f5fb1a641dc8654e5d412aa628ca9c57b5dcf8",
-    "source/vpp2/src/vpp2/adaptive_mask/physical.py": "f54e8b4fae8026cc89f296d74bf4cbea06fa4e126905b572a22c7d8226f0238c",
-    "source/vpp2/src/vpp2/adaptive_mask/preflight.py": "b646e7ca785afc87356e3e1184e48f535792ebafc84936450fc83bfd3f9f5ea8",
-    "source/vpp2/src/vpp2/adaptive_mask/priority.py": "68ccff1a384081f0c9a82e8f54ad2c720293f8500e7440a469b5e8fac1ee608a",
-    "source/vpp2/src/vpp2/adaptive_mask/quality.py": "c3b3a2264ca0b4773805f4485ebfe91bb4ab861e226ff4bd27bd667ab372b59b",
-    "source/vpp2/src/vpp2/adaptive_mask/quality_policy.py": "f100b7d4a412592777d936a6cf3c73a9f787cc4a62d5f1b8a369fed74acf897f",
-    "source/vpp2/src/vpp2/adaptive_mask/reliability.py": "306b41720154810bd972fe4d9c4c388e6967539d076ddbedd3a432ac62d13104",
-    "source/vpp2/src/vpp2/adaptive_mask/report.py": "f71401e09297daed8bbb72cb3af3b887e5fdf5ec282f47637fbb01392d280171",
-    "source/vpp2/src/vpp2/adaptive_mask/runtime.py": "ed74bfedd147389ea01f3be8c2979fa06c7152326ccc29ee347175690cb9c6cc",
-    "source/vpp2/src/vpp2/adaptive_mask/state.py": "5e8f95021eeb35511f5481e4e91ae687fe605f9e88d17b3b98d377cc967c17e1",
-    "source/vpp2/src/vpp2/adaptive_mask/train.py": "ad95349bfb242bcecd7ddf82fc58b31d63f2938a2101f9560bbc9e305f5f5b20",
-    "source/vpp2/src/vpp2/benchmark/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/benchmark/audit.py": "8587be371653c38b9926aa0ace006a244f270e0ef0020147a7bf5b0fc85d4538",
-    "source/vpp2/src/vpp2/benchmark/packed.py": "5047fd02f3fcc68e41f543fe83bc9202769a05794d898524d6d25e308f2b5125",
-    "source/vpp2/src/vpp2/benchmark/scheduler_io.py": "d6df06259549faed2da9f2888b540eac8e1ff2895c339eea58b4ea6a7e3b3cb0",
-    "source/vpp2/src/vpp2/checkpoint_compat.py": "462c5f93812483b5871633b200bc2ffecb9c201a2caf6b14f8f36dbe95fba50a",
-    "source/vpp2/src/vpp2/cli.py": "54c19a0f7e25b53bbf56ec13b0f537c9d7b81282e1e80575bddfb470172164e3",
-    "source/vpp2/src/vpp2/compact_kv/__init__.py": "f34f2f77f68399f45f6656b562a5c331f2504672dc145df368e086532191f766",
-    "source/vpp2/src/vpp2/compact_kv/blocks.py": "662c1f35055c2be2e2a14f2d92615d046f5e454879e34af39e3c7aec3af34098",
-    "source/vpp2/src/vpp2/compact_kv/checkpointing.py": "3c33c156e96f26ceb22eee61cfb8b18c3cb9560db0ab405cfd4c46efe5aba353",
-    "source/vpp2/src/vpp2/compact_kv/eval_prepare.py": "bdb33444056245e3cc773b0849b4db608b454cd24a7ac5b752f1ba26ff978865",
-    "source/vpp2/src/vpp2/compact_kv/export_check.py": "e6cba70b47027882a34d587aba0da5f9371128ec2db2282b91ad2d49e40065b9",
-    "source/vpp2/src/vpp2/compact_kv/gpu_check.py": "fedb6a9ab69b51080161ff2a4feae3f85262cc50f932f48e8decf04c1d147a20",
-    "source/vpp2/src/vpp2/compact_kv/losses.py": "914437ebf3ad91fd2cffa893a65ecb04ab01f2b337658a781a472feeb50c0a06",
-    "source/vpp2/src/vpp2/compact_kv/model.py": "332ebaebf543d40a34d188ebbe81055ec3e4bcff164bfe90d15aa6cef0748b8d",
-    "source/vpp2/src/vpp2/compact_kv/prepare.py": "7b01764cad0fafc0cc9441952b2e928048b0a43a6105dbd0eb64c349f4b9bf1f",
-    "source/vpp2/src/vpp2/compact_kv/provenance.py": "103e0f7100be56233aa86041ba9bc427ccb0db4fe045c4e6b8d3481ab6bb42da",
-    "source/vpp2/src/vpp2/compact_kv/reporting.py": "0b9ff602fc3425d1b9c122a77da3a426814e11c9cbdcb6b9a75a2a48130087a7",
-    "source/vpp2/src/vpp2/compact_kv/sampling.py": "2e8d82704f07853a7b487d5aaedef1b2f1fcfc5247e9100a56ed21d69262eae2",
-    "source/vpp2/src/vpp2/compact_kv/spec.py": "9bc920fc7d4398e424a0bde38ce97f87473aa417cf68b48e2c292a59491bc7d7",
-    "source/vpp2/src/vpp2/compact_kv/train.py": "a9e4410c5cde5df002d9808554538c29b77ed9e2d2bedc9ac5f8465eed70cc3b",
-    "source/vpp2/src/vpp2/compact_kv/validation.py": "73b62d02ce1d518cba7b9462e85c3c8d272ae9974e1161b1b6adea6f0040e2b5",
-    "source/vpp2/src/vpp2/convert_robodojo.py": "7ec2d6bfffbb4f5891ee33a4bf630cf27b17628506aa3cee1c46b3050027b290",
-    "source/vpp2/src/vpp2/critical_windows/__init__.py": "22721b77e25ec4874c4853d2f2991a14dff6aaa665da500f2830826b06545478",
-    "source/vpp2/src/vpp2/critical_windows/contracts.py": "ded4a239f7599785d74ff57ff892c5aa74f0280d6759460c93942e91f505d47a",
-    "source/vpp2/src/vpp2/critical_windows/coverage.py": "6c18bf995aacf413bbe4536eaeed92ab4b2f15f9eaaaa03a2730256ea7922ce4",
-    "source/vpp2/src/vpp2/critical_windows/inference.py": "df301bd187b27d20b6623febcb5e85815e1b7d4a1e9524a8a65bdf5904bf7232",
-    "source/vpp2/src/vpp2/critical_windows/metrics.py": "ae2bab145555d55696ef479758a5bae6780527fd3085a4f46d035b667ff32736",
-    "source/vpp2/src/vpp2/critical_windows/packets.py": "8f7d2c411407d95b1ff4d0e5b9055589f6114b1bb1020eeb7e1b106f330b853a",
-    "source/vpp2/src/vpp2/critical_windows/prepare_staged.py": "694cb5699de2d9f5ce1bc4667627b47d6bf0698715b4aad86907139859caffe5",
-    "source/vpp2/src/vpp2/critical_windows/reselect.py": "8d0e31ca1c17906d74de6439276e49ff7ea78abfc66fa556d8a306070b3dbf4c",
-    "source/vpp2/src/vpp2/critical_windows/score.py": "32daceb9a3fd957d9be75823b4d519629bc60a64d1da62f465e172cb1ec98d55",
-    "source/vpp2/src/vpp2/critical_windows/selection.py": "a7c35f6719d3308bfb738965d17544a15985c6d684bb47eac7272d6155d96786",
-    "source/vpp2/src/vpp2/critical_windows/staged_review.py": "d5cfb3f49ec5acc6914da98f5413d1549b2695e7c85c9aed9c0db83dee6be754",
-    "source/vpp2/src/vpp2/data.py": "8574e1ae3c6ca6e4b84633ee1815f2b3f2ea17c1e55f12fda0b00c5c5a6a7b53",
-    "source/vpp2/src/vpp2/datasets/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/dataset_utils.py": "e1a1471b0aa70917eaf699f85a9acf3ac70d9fb6eb9ebaf02d2bf2f218d6ab2f",
-    "source/vpp2/src/vpp2/datasets/lerobot/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/base_lerobot_dataset.py": "729fb3fd4ca531212eadb4a4e9c7e9f97ad34cca091d66380acd6b179f0c29c6",
-    "source/vpp2/src/vpp2/datasets/lerobot/constants.py": "11ac9260341e1560c5eb0dd7acb79ed63c74180b8ab4113c27622ff213db417f",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_dataset.py": "bfe75bb9593ec646236777d90d9fa309fa0e1188db07e07a3be0810edbe3da56",
-    "source/vpp2/src/vpp2/datasets/lerobot/critical_phase_sampling.py": "266b0b7b023f014f4ccd935b6015b0a05b4659f372264fdefd997ea9692b244a",
-    "source/vpp2/src/vpp2/datasets/lerobot/episode_instruction.py": "74767839a813c7902861f4561101541c282b72e261df1fd28fe8d06c91231b69",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/compute_stats.py": "7d7daab219332b5abdfd5ec67ccaf3502e8eb583a17a4e482c3f3d4f7a691e89",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/utils.py": "4fdb4ac0653201e6743f2161c485213429948c1e1d5b068d03b4d97803254006",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/datasets/video_utils.py": "4f66c410ae6aa5a98acfc3320f3dd7b017d0285117054f14226f9e5d82395087",
-    "source/vpp2/src/vpp2/datasets/lerobot/lerobot/lerobot_dataset.py": "5659e992010bb6674ec50ab0d33b28a8e9dab0ecc895c199662af4c40fe56629",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/base_processor.py": "d3327cbe398d77761df960014d41e70d349bb0fa42bca874333b9ee970222447",
-    "source/vpp2/src/vpp2/datasets/lerobot/processors/vpp2_processor.py": "43c5604cbfc315b7377316cb97c9a6f3322124404397abfd01c6c3d0090cd414",
-    "source/vpp2/src/vpp2/datasets/lerobot/robot_video_dataset.py": "8eabfbd649116b0b4d6c9704c0df34febbf8875f6ab863dd75f4107dbbcdef73",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/action_state_merger.py": "39fd44f254bd7fb71b9269d75d8f9b5a5bbbe167ea438dda92d40eda1d256734",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/image.py": "f494d262ce29770e959bdff60b997f7a07e5f00d10ef515eee7be813d1db038f",
-    "source/vpp2/src/vpp2/datasets/lerobot/transforms/rotation.py": "feb59d26539be6cfdb86488f0c10c11488ed73901658bc657e67a755ddb9e67e",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/normalizer.py": "8207d813b698a91dfdbe6910a4aa1fe0e775a2a8edf5c05b2d7a42591e9f8adb",
-    "source/vpp2/src/vpp2/datasets/lerobot/utils/rotation.py": "8ea2c72565707f5979967a1abeb5edf1eaa2c3fec55a7b50e640de869fc9fbcf",
-    "source/vpp2/src/vpp2/datasets/lerobot/video_event_dataset.py": "c20d3f5e2bc05e1b7ed94ae38872883a976a179e09b7fa051e94f94e2836c02d",
-    "source/vpp2/src/vpp2/enhance_prompt.py": "f31b3190c24bad073c5e3121ded4f564232729566e8c1fcc3e76fe75b28faf45",
-    "source/vpp2/src/vpp2/evaluation.py": "9b76882149e3c3a8c417a0e23514af1bb305c01bf0b6d21fff1f70bc9843f55a",
-    "source/vpp2/src/vpp2/flexible_mask/__init__.py": "64bed1808976d1f3c708154c36a97288a75abc6463d49d7f0065cb8360a2fb16",
-    "source/vpp2/src/vpp2/flexible_mask/calibration_batch.py": "dc5d83fc9e2bb595b9e3aeb1923115ebb9345283290a7c3f95c79fdeef7d3e76",
-    "source/vpp2/src/vpp2/flexible_mask/checkpoint.py": "c8531727e2388ec3d7a9770a504ceda119afcf03d64b2462befcd45a48949f9d",
-    "source/vpp2/src/vpp2/flexible_mask/coefficient.py": "07a3cc8d8e6c8b60119b9f94ae85074da017b08158fa0a0912cc9051b6df8ef7",
-    "source/vpp2/src/vpp2/flexible_mask/config.py": "4e1e9b289e293145b4e895c8ef7347bc837d01c929e05fd5cbf65bdb3c20be5b",
-    "source/vpp2/src/vpp2/flexible_mask/continuation.py": "2016463b4efe01b5d0a9ee84f03c3d95ee16f8e24f230faa465598cc978da29b",
-    "source/vpp2/src/vpp2/flexible_mask/controller.py": "263f1c9091e0c4318aeefb61e91549ba905ff046a625e35e1b18c94da4b4668d",
-    "source/vpp2/src/vpp2/flexible_mask/data.py": "d245bafd6f04f33ae7cef7c01d5db876a1a2a1a138c4de8a139d3ea78095521c",
-    "source/vpp2/src/vpp2/flexible_mask/diagnose.py": "faa29a0fa57c66af3b2e817640fdedfddc998bf878357f484415b244555ae4a5",
-    "source/vpp2/src/vpp2/flexible_mask/export.py": "3064dfd82bb55a145c16c77bbe17244970d07d18e944b793f4b7368dbcede764",
-    "source/vpp2/src/vpp2/flexible_mask/gradient_probe.py": "6fe3f13540cb9d8dd20c0b6429268ca74625dfad1e41806412c0f980fac76509",
-    "source/vpp2/src/vpp2/flexible_mask/losses.py": "df264b71365885ea90329c0b0facc92db3d2eb6e04324ea90a4701a1cd457abf",
-    "source/vpp2/src/vpp2/flexible_mask/metrics.py": "e976dd066e4b5e7e7ae79db77b9d1be64d17c212c61d0bc113b87a048e8f2b6f",
-    "source/vpp2/src/vpp2/flexible_mask/performance.py": "95b3977fd739cc98f7de94b14fb490a7c132021e742e7179dc5b5dddd6557c68",
-    "source/vpp2/src/vpp2/flexible_mask/preflight.py": "f794b997a492911c4d77656310c89d3b85e54983b3b889ba9904be64fb961e43",
-    "source/vpp2/src/vpp2/flexible_mask/prepare.py": "d6cf82e27555ecc13973727f4b71ce7e40558fef7d95805665fc7e458a42fad4",
-    "source/vpp2/src/vpp2/flexible_mask/pressure.py": "6cf74edcf3625fce9eec689d70841c6c1481f7c1abd87e5bf70bb38eb34632a2",
-    "source/vpp2/src/vpp2/flexible_mask/recover.py": "418b2d5ecdd08b08686b2a07bbeb0f969471eb73c04ac6952dcfbcda3b123f30",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_best.py": "917d924c5708584a0192050a6d8210809e905938aaf877dd81cff3be287ea289",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_checkpointing.py": "6f6a61c2ee7076c6b0977d1acb489fee7df04322d7f6b13bee9bb8229be319df",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_export.py": "c50433aa1b7e528fb23c514dddb5b9197f44c662829838ecda106980ab0dd54d",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_handoff.py": "aa590f707738e4b7c7250f901e29795c7441430f961bfa764437cf34bdc514b1",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_kv_decay.py": "bf83761798deea4f89c32d14f43bbfb0f16f751b3230962456c073b843dcd08a",
-    "source/vpp2/src/vpp2/flexible_mask/recovery_subset.py": "7f7154cab6d910b747a0189742ee625983d08f13de50ca60bd4154f61df2b0d3",
-    "source/vpp2/src/vpp2/flexible_mask/report.py": "5b49283d472bace3a6ad56ce568af04f2c3d04add1a3920ad5006fe98374bce6",
-    "source/vpp2/src/vpp2/flexible_mask/rng.py": "01680b5b7e8a29ce92e96d756e12ceb80bb0e0ce1ebb642a14fa2a3a34776daf",
-    "source/vpp2/src/vpp2/flexible_mask/runtime.py": "373a3ecbc2ab6592a359daf6adfd8c958cb657f393e698ca2cb834f06750be4e",
-    "source/vpp2/src/vpp2/flexible_mask/select.py": "b547857bad61ae7ebd25b888fba73c2ca8f50a8eb31551ca54a909c75c8984be",
-    "source/vpp2/src/vpp2/flexible_mask/selection.py": "b9453c8735051b54ae7dc8610d2ca332eae8ca1a3e099aaadaf5ab75bfb0f8b0",
-    "source/vpp2/src/vpp2/flexible_mask/speed_review.py": "9d7099d1e173026274eb439169df2ad4812ec1d56a5f0b21242f972cd835ffd2",
-    "source/vpp2/src/vpp2/flexible_mask/throughput.py": "ce21bdfda51ae10d1470ac537d910d9f46f60a38910888808ffb3f5a80174d58",
-    "source/vpp2/src/vpp2/flexible_mask/train.py": "86625613bbc692722c6d2d24278a586f88ce0d76263a9daf92562a9002a8e037",
-    "source/vpp2/src/vpp2/flexible_mask/verify_export.py": "fb68e62b0da98e6f96cfa694bd061162bbeec1763e5787e057a1cbac5b2ce8f6",
-    "source/vpp2/src/vpp2/full_mask/__init__.py": "27c9ede277380e8d64586371f2a03334e68ab2d3bfe516d95e8cb96099827929",
-    "source/vpp2/src/vpp2/full_mask/calibration.py": "5d97e0385eba82252e12cc1562d1a77da2f2bce29332d758c1b5b6b84c5bb47a",
-    "source/vpp2/src/vpp2/full_mask/conditioning_cache.py": "07adf95165db4c3c1c469695d6b7b6fe4cf4fa9aff0a3d360f4229bffe9be17e",
-    "source/vpp2/src/vpp2/full_mask/controller.py": "0e6a969b89a4275352cc5a1e00bdfe8e3f8754a6ebd7f9266b92179cbe6bdd10",
-    "source/vpp2/src/vpp2/full_mask/cpu_contracts.py": "1a46a53a68d2d8a202c5f98d9a78f6d5f5041dada0c2239ed81e7916a28f0758",
-    "source/vpp2/src/vpp2/full_mask/curation.py": "4ffd7b0212bdbca1daf73ba7c24423523437012150fc044a43bbb33cce747545",
-    "source/vpp2/src/vpp2/full_mask/curation_v3.py": "9dba386d66869de0fba321ade79b2692d47f111ccfebc7f367194cef7a4b4e8c",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_checks.py": "688ea4e7e44c4ecc1a3cd81cf682e2cbf611d9ad2d52744ba1308324da70febc",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_features.py": "11011a089f1586045e2972cf267773aaf4602cf08c8b0e755bbf017ede1e6e43",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_finalize.py": "a2a7812eb6c670686a7a4f4adff0a2c6363dbc078c71d948537f1c5fd546254f",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_probe.py": "2755e9cabd0fe02e70cb8330c083af881a5e4696697d58cf7aa01bd00ef57f6d",
-    "source/vpp2/src/vpp2/full_mask/curation_v3_quality.py": "b9170398762112a23bb133dde53c6a49193c671c767647239c8b0ff15159530c",
-    "source/vpp2/src/vpp2/full_mask/data.py": "668a7c5f88a3ab17854eb0df240c974082f4aa64604cf11d3143b18337f73181",
-    "source/vpp2/src/vpp2/full_mask/distributed_smoke.py": "f699d4f330309f29a537fbabb61c60f61eb3322d7304605351dba3d2acb48e8e",
-    "source/vpp2/src/vpp2/full_mask/gpu_prepare.py": "92ce96dc4c1b6d6ebc22d99935472c15d86ba2a7c050e31640d2c32e08748ec5",
-    "source/vpp2/src/vpp2/full_mask/monitor.py": "2724a934d1f96444227c854f75ef84df6025c02485aa68eca34f3b28d509fe19",
-    "source/vpp2/src/vpp2/full_mask/native_monitor.py": "5572712a340f17692a669d2a4c41d1d691164925eb7bd06a56fff1dc2afd69fa",
-    "source/vpp2/src/vpp2/full_mask/panel.py": "2844b663340cbfb54f7622f3447a7831e2809050739a52dc076fd6ae32ed9e46",
-    "source/vpp2/src/vpp2/full_mask/review_decisions.py": "ed1184bcda7b5bb3fb11bb4262c4461c408e3d3ac0d23e80d2f8cafcf39c7e18",
-    "source/vpp2/src/vpp2/full_mask/review_report.py": "752ba9e8a3caf5b8cefa2d18e6fe14860e6f998f57c03af0940af0943b09a9b1",
-    "source/vpp2/src/vpp2/full_mask/runtime.py": "ae30b4d3fd3c888bef17672376dc3b34491cb296a1883959674ad465579efe7b",
-    "source/vpp2/src/vpp2/full_mask/search_quality.py": "ee3238d3366799e97d3f9a90c11f579ff8f51d5bd0642f52c37d9dcdbcf02e6e",
-    "source/vpp2/src/vpp2/full_mask/train.py": "f4127e6580522d4d9e5a50db10d201526e948d4d3479ddfbbc12e60ec61c7501",
-    "source/vpp2/src/vpp2/full_mask/verify_data.py": "72ee60ba2016d5f73eb2ccb01b666a5dac37138620efc4a79a20607df3658fb3",
-    "source/vpp2/src/vpp2/inference.py": "42b3057a85aaacbb87440bc740f13b09b9c82cd51d8c5af4e080716aabdcf38e",
-    "source/vpp2/src/vpp2/learned_width_kv/__init__.py": "0c59cb8424b6cd64508a005625e9edf63c84d24618a5d2caf6e6fa5d63cc859e",
-    "source/vpp2/src/vpp2/learned_width_kv/calibrate.py": "b89bbf103b49fd85f85f8c300a7868bb42ae7a13f29bd9878dbd5a11061ae51d",
-    "source/vpp2/src/vpp2/learned_width_kv/checkpointing.py": "7d30e46b95d91df8aff088346ae75809958622b3a8e8d953db4f30adfbaebadd",
-    "source/vpp2/src/vpp2/learned_width_kv/common.py": "52dff98c6317b34df2a13313a641998acf4f9e18ac31ea4e8e25227afa14620d",
-    "source/vpp2/src/vpp2/learned_width_kv/controller.py": "7b5a8793a540e1e7464185229189f4f1f1acadcd7cf3c2bf477bda5de7d18787",
-    "source/vpp2/src/vpp2/learned_width_kv/data_scope.py": "0ddde6805f646afbf5f5966c53530f829becd083802f0e0a8697dc1004e2003e",
-    "source/vpp2/src/vpp2/learned_width_kv/distributed_check.py": "f89f153a20730a6bc3c70133479f34918d30c5a48775bb3f6c14bb25701801ce",
-    "source/vpp2/src/vpp2/learned_width_kv/gradient_probe.py": "3bed964f6488e95b386cc58058e7f28aff50c9a98dd526f3365891a9049a6fab",
-    "source/vpp2/src/vpp2/learned_width_kv/lifecycle.py": "798afa47ec5dcc8749982da2d7853eff184e450a32d51bcf70eb4578f2bb0b09",
-    "source/vpp2/src/vpp2/learned_width_kv/masking.py": "3272ea09271130dcb3cab6935f6bb7d61bda66b9982c3baabcd61ee0efd914a0",
-    "source/vpp2/src/vpp2/learned_width_kv/performance.py": "a2fd9597c51bf0cf5c435b48e8f57792df24d2723a6d7b6b841d6fa24f17ab2f",
-    "source/vpp2/src/vpp2/learned_width_kv/reliability.py": "f74399fb67427093b3e4c7df9749379f9291b7ac58bd6632db226b6e78fdbfd0",
-    "source/vpp2/src/vpp2/learned_width_kv/reporting.py": "f22c1d8806a5b7757bb6dd99b15153d0305b2eeb0790bb37763337690d642036",
-    "source/vpp2/src/vpp2/learned_width_kv/runtime.py": "513da1e06ba9509164f017e0b8dbf6b219d7e1cff82edb2d2fd8d2d7839e666f",
-    "source/vpp2/src/vpp2/learned_width_kv/train.py": "5813ebd4fefec6d3a8cead9e0c645d10734f5ec88c4d74a5742587f2eb4de6c3",
-    "source/vpp2/src/vpp2/learned_width_kv/validation.py": "d5594ce875c8011bff12d3c77083dca68878178ed022b8079761267d04a82386",
-    "source/vpp2/src/vpp2/learned_width_kv/verification.py": "1bc43d48ef0ec3056af614adf43ca0744d94b1e77fe970168fef87c7cf2e6fe8",
-    "source/vpp2/src/vpp2/libero/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/libero/action_model.py": "178fbe4cf02b9a2e133ef31389b6bb2f9baf0504bd0818a265b6bff32ebc3914",
-    "source/vpp2/src/vpp2/libero/cli.py": "fcf68a88e2db0a9b786c6199d9b719802fe4b84514741658b49e2b06e2473e45",
-    "source/vpp2/src/vpp2/libero/data.py": "19b699381e7d6fa769b5090c7125ae01b7584576e969468660ee65b1585ef5f1",
-    "source/vpp2/src/vpp2/libero/diagnostics.py": "9540e31dfd2def06bec942ffe73262278700a17a340e14cb5bd41a295706e147",
-    "source/vpp2/src/vpp2/libero/eval_helpers.py": "ac338d4ade195ceb3a87620d162befa9679e2d7ed3d91b09927e5d59d86d97f4",
-    "source/vpp2/src/vpp2/libero/eval_libero_single.py": "87e6b457dc412ef4fe5d01aa48b1a6d20a28dbce3152848042e2103769e4efef",
-    "source/vpp2/src/vpp2/libero/eval_libero_task_manifest.py": "f0c721a7ddc768edddc8b0e065a4d8aeec35e1a546c88deebd9097ad93f6227b",
-    "source/vpp2/src/vpp2/libero/evaluate.py": "24216f9814f8304e077d72ff8d8c778fd588239313b34cd7d23b960c8f924c67",
-    "source/vpp2/src/vpp2/libero/libero_utils.py": "c7cc23976c86eabcada4846d46345d3605fc2ce03929761646486be3901697d4",
-    "source/vpp2/src/vpp2/libero/prepare_pro.py": "0999050352d924c11d29b22f63e902162d8cfb23f94815c93c01295c11ed81ae",
-    "source/vpp2/src/vpp2/libero/pro_alias.py": "90b5c8c8061bc727a6e413a8746fcc98cbd06e1c7c4175fd223b45af4f10b700",
-    "source/vpp2/src/vpp2/libero/pro_runtime.py": "adf94ed64b032aaeb0d88391e7b653f88f67e76b9411419ece25faa31ccc1340",
-    "source/vpp2/src/vpp2/libero/protocols.py": "68225c95eeeee25032966a7f90c1ba18067c53225397f59a463d764030fd0d93",
-    "source/vpp2/src/vpp2/libero/runtime.py": "1e2e2900a604d33cbdf8774cfa33dcee1065f21dc0b59fcf30799b5098c1662a",
-    "source/vpp2/src/vpp2/libero/trainer.py": "ea29b4e000047a6b0c8aee3cb6f32df9187987a657924085dbafd9628315107b",
-    "source/vpp2/src/vpp2/libero/worker.py": "9df6350c4ab92dea70863c08f1bdcebd0d77574a3a2d8d7ac9899a5a825be6b0",
-    "source/vpp2/src/vpp2/models/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/action_dit.py": "9a654edd860ce77c1308923bd338c7f14d09f68686642531d7143c16b9f6f331",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/compact_channels.py": "6f7ad787624844970985117002ab809bd777f7474969daca121e7101381c6d0c",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/gradient.py": "594249b6eb931048e352fd3ffef9b1ac1db9e4fc74659e04216d6b823dd174f3",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/io.py": "1b68cf2f2fd674c7a59140f8632d4464044bffc2e9277ef5f9a975e7c699966a",
-    "source/vpp2/src/vpp2/models/wan21_14b/helpers/loader.py": "b9180b9c9177af192e5cb92387b4a09bd0dbc08fb75a94e843b67d315c94e339",
-    "source/vpp2/src/vpp2/models/wan21_14b/mot.py": "cc14a909bfb181a441bc9d8d6de12f3c31d89f3dc3415031e34a5f8c21410019",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/models/wan21_14b/schedulers/scheduler_continuous.py": "15de04574c1d8ccb794ca6560b1e501e5c83cc86104bb80f1eb3cc3b0324ffe0",
-    "source/vpp2/src/vpp2/models/wan21_14b/vpp2.py": "8c419af905a8ffa93d77c827a283b7aeff744f3a461c390783eef8a333c31756",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan21.py": "01b45cc71f69a6914f7b286559317c8e8485d0b0fec93bb28ebc37c183c295f4",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_clip.py": "f65e44c9d1b7811d2996b3fa22e6fe4c2ac5e6edf1fc2c14251cf582bf8d1f05",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_dit.py": "0d1b5837ea923508c05394d56aaa465813e7b4f5aea3b31ceceae39b2de9b0f9",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_text_encoder.py": "902a1e46b21b5c48e166fe03defabb94b3773c6bbe836fc09b9a0d0894eb1f6e",
-    "source/vpp2/src/vpp2/models/wan21_14b/wan_video_vae.py": "33f7d804677b44a6f3ab334684017dcdb1e9dfb748279de76ad94a5bb71a7e51",
-    "source/vpp2/src/vpp2/policy.py": "2a7696cd73ccb99575bf901185cf578c3775945f61f6e67aeb2b4ec6bfc4e30b",
-    "source/vpp2/src/vpp2/preflight.py": "7a91e486675a4cdf38565e0f4b58e60a68bba7de5654cfff736b5df58f5c53a6",
-    "source/vpp2/src/vpp2/robodojo.py": "183db869a622ef17ee92c487949db2e22a7d1365558fe703bfb8dd8fef83e9c6",
-    "source/vpp2/src/vpp2/runtime.py": "e26885a3732f5fdac3d8381da870a1673d0c03354f02f9deee7fe779d645f8a7",
-    "source/vpp2/src/vpp2/trainer.py": "feaa0ac4ed94a376836eb60cfe6f52460392b11f7cafafec9c6d9f151f5b61f1",
-    "source/vpp2/src/vpp2/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "source/vpp2/src/vpp2/utils/conditioning.py": "43d43f0f3ca0add64498c3d64a3791513a53e3fce0ea53f77ef09580a86b7b2d",
-    "source/vpp2/src/vpp2/utils/fs.py": "69e6df236b43360428a613686c8f37c531c0d16a3a2c5b06a725c46b49d657c5",
-    "source/vpp2/src/vpp2/utils/logging_config.py": "1de4abf66d9d201b49e7bff48291643b9dc5ec9b6e7918ca73e1bd18887b1c43",
-    "source/vpp2/src/vpp2/utils/lr_continuation.py": "72ea9f2af52041db19aab681d7a99fdf6d87d5ed5486f2499383a211c6c1ab76",
-    "source/vpp2/src/vpp2/utils/misc.py": "7e42c330c24604e118f1878473dc804aa8a272944df502ef1414c05be4d5bba2",
-    "source/vpp2/src/vpp2/utils/pytorch_utils.py": "9b0ba19de9a64458b12a0fb24f0811f49d9e9fca5d8b815f5f5ecd3b7886fd63",
-    "source/vpp2/src/vpp2/utils/samplers.py": "4d3bd9b2d498f4ad8f2cf9765d60c4bf5368fa6fa2a7d42d0db6210f66899652",
-    "source/vpp2/src/vpp2/utils/training_profiler.py": "91f6d98577e63d85e474c738a9bc1901edc194bdf318b5387d527c31b6708832",
-    "source/vpp2/src/vpp2/utils/video_caption.py": "d9572deb3690aecb0f16c052016cedc5f94da8ce8f016b4dff1a4e358e18d5e5",
-    "source/vpp2/src/vpp2/utils/video_io.py": "9e9898b021a44e87aa1b4e6d13e3755a9869a2e27a7eb53c72a6802808d5bd80",
-    "source/vpp2/src/vpp2/utils/video_metrics.py": "1840a37685d686336c7326a2a3d4fbe048d1e8e7e431fb14519d360a8b7aaae8",
-    "source/vpp2/src/vpp2/video_training.py": "20f55bf8b0a7dba6422a1115be1c8adf31a0db6a9f53a8f6f53f5d970188bfb2",
-    "source/vpp2/src/vpp2/weight_resize.py": "269edd4b9d32a1c7f363ee082dde7ce57f2176190daab6d5bf4a1c98ab43f0b4",
-    "source/vpp2/src/vpp2/weights.py": "5dbdff347de1b57e30433efca6da28334d0263d27aaa9b2278d95614db1b8641",
-    "source/vpp2/src/vpp2/width_kv/__init__.py": "0183a7106da7e45223550a552db2905d6b59c2a4259f8c86cb98629b03e66b43",
-    "source/vpp2/src/vpp2/width_kv/calibrate.py": "93a06910cb23a9e1dbd961ee9ef9794d70aaaa6186b08483260c8215109598f5",
-    "source/vpp2/src/vpp2/width_kv/common.py": "45f5f7087f5f3dc067358838e3fc738698b796defecb2ea721545d275f767766",
-    "source/vpp2/src/vpp2/width_kv/diagnostics.py": "cf184ae58f5e1c5f4453774eee1e10c85931e618bbbf768ab42f649cf301f5f1",
-    "source/vpp2/src/vpp2/width_kv/reproducibility.py": "89c5377e1314dd6e8de7f67c4385d30f5bd91fa4a5742449f1c1ff8dd578f451",
-    "source/vpp2/src/vpp2/width_kv/scoring.py": "7644358d0b2d29d3b91515fcea56090dbd2a7a0dad2d5e5c88e8d585bff15533",
-    "source/vpp2/src/vpp2/width_kv/slicing.py": "856203c14e0b06d19585a96392d00a00703779ebdb9dbdc9ab1323d50daf8ba4"
-  },
-  "update": 1000,
-  "validation_gt_action_mse": 0.0005726480308242441,
-  "video_parameters": 12468409664
-}
diff --git a/inference/README.md b/inference/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..38e99e161105c1710074b500ade351715c7e08dd
--- /dev/null
+++ b/inference/README.md
@@ -0,0 +1,13 @@
+# Bagel-Robotics-1 public inference adapter
+
+This folder is the standalone XPolicyLab adapter for AdaPrune checkpoint 1000. It contains inference code and evaluation instructions, with no training repository dependency. Model tensors download separately from `bageldotcom/adaprune`, pinned to `2651631c5d08819088475e8778676f5cf6d9afcf` and verified against `policy/bagel_robotics_1/artifact-manifest.json`.
+
+Install this adapter into an existing XPolicyLab checkout (the installer refuses to overwrite an existing policy):
+
+```bash
+python install_adapter.py /absolute/path/to/XPolicyLab
+```
+
+Then follow [Installation](policy/bagel_robotics_1/README.md#installation) for the Python 3.10 environment, exact checkpoint and encoder download, and debug checks. [REEVALUATION.md](policy/bagel_robotics_1/REEVALUATION.md) gives native simulator task/suite/three-seed commands. RoboDojo and its assets are installed separately.
+
+The adapter is prepared against XPolicyLab `6baccd57ff958570d0db1588370289ee36c8bb70`; source and dependency pins are preserved in the policy folder. This public release is not an XPolicyLab PR or an official leaderboard result. Training/pruning/export source and the technical report are planned for November 20, 2026.
diff --git a/inference/SHA256SUMS b/inference/SHA256SUMS
new file mode 100644
index 0000000000000000000000000000000000000000..655413b1e67d62beb5f63b559849a56550afa1ad
--- /dev/null
+++ b/inference/SHA256SUMS
@@ -0,0 +1,81 @@
+9ee395e4c4603776300a8cf2b8e1b986c47df4f2aa00bccaef7d32fe1a302724  README.md
+56f26e5b6ba87080c82703e102240fdfbd955c731fb2a044609ffff099795b3f  install_adapter.py
+6c9437916a0dc5b6ed5fd56b361a4a1894be910e6965c9570e8683cc74a40813  policy/bagel_robotics_1/.gitignore
+24cb153ee8bababbfc20fe7acd36768923fbcf2a07d0aa653f460282761ccf8f  policy/bagel_robotics_1/EVALUATION_RESULTS.md
+8379dddfbbb1154ad9cc57f008f1f8d4a8d26c3de99bd6b36aa4d8067888aa92  policy/bagel_robotics_1/LICENSE
+2b2e02ed1e08922744249a51678dea3b765dd0d10d9d908e93b3b9b53531775a  policy/bagel_robotics_1/README.md
+4fa25718ab410d9926541d5c682693a0772f2d7df2f1fd9c1c4a205d1e0b2521  policy/bagel_robotics_1/REEVALUATION.md
+2f051521079a9de8815faafcb20098f1fe7fa167425e8c8088e7a42c2d50bb8e  policy/bagel_robotics_1/THIRD_PARTY_NOTICES.md
+1d22bc8164fcca87f47b003406878112087118c4a5e23e6019ae1998e2b53418  policy/bagel_robotics_1/VALIDATION.md
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/__init__.py
+1d89d4f59bf0d9c5d7cdf29de28baec2280c99c6a28b9176365781f4f506fd30  policy/bagel_robotics_1/artifact-manifest.json
+76fec0d61c9344d8d2c5f61853a93269c8b19ae16d08de673f2ced5bdfe08b79  policy/bagel_robotics_1/assets.py
+2a16b03892ede7aa0a93cfd65e01a239498aa992bc4f1affb9cdde54253dae08  policy/bagel_robotics_1/backend.py
+68f337b21b1e517e1df27550e64fbc542b644528adc8f3f47fb8e44849185af6  policy/bagel_robotics_1/deploy.py
+4c315728926a12e96cd2aee3ff2f66d2f8ee7b175f331745f6aec01771c88a9a  policy/bagel_robotics_1/deploy.yml
+e8ce9e58f22e4330b2f1cca1f45e304d9cd466941ec795ebcf46782d1299fe93  policy/bagel_robotics_1/download_checkpoint.py
+467e70c368f1574f6a39afdb3c84d6b10e8430bdbf638d9742a58585c7e60bc3  policy/bagel_robotics_1/download_checkpoint.sh
+daae036dc45b13f90f1b5b5c9e813262d62585aba467534e2402d4933e8dd97c  policy/bagel_robotics_1/eval.sh
+6110ad701ed4c7dd9768b61add6df64ace15c7109ccafdace4ea8ab8934b23cb  policy/bagel_robotics_1/evidence/action-comparison.json
+2c8b3a55f75117533bf7ec0cfa8cfa7c42cd9b51d2517161fc022e8e15c5b9d3  policy/bagel_robotics_1/evidence/five-suite-results.json
+3187c8c75bb85c393b65f23838e8163fd201081b82b492601b2a6c7bb552d43a  policy/bagel_robotics_1/evidence/gpu-validation.json
+f73d49a6b9c96824eddb76ae146b720a1d2feafba69c2ab7e9222c9167f3985f  policy/bagel_robotics_1/evidence/local-results.json
+a299b9f60b3820af5292ef8b410d73aaa67a97f41e44c8415e9ed743d67f4f7b  policy/bagel_robotics_1/evidence/pr-audit.json
+d842f9639df62df7fddc84f095771b438f02f955605e221e4f7b160893592454  policy/bagel_robotics_1/install.sh
+c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4  policy/bagel_robotics_1/licenses/Apache-2.0.txt
+7794bb06af8fe5485ca912454ad1665ccd7846c45f9c2b1258658e480948cdbe  policy/bagel_robotics_1/licenses/RoboDojo-MIT.txt
+c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4  policy/bagel_robotics_1/licenses/XPolicyLab.txt
+56418464b2a88e7aea31d436ee42851dfababad97db9178abdce318b5a7e621b  policy/bagel_robotics_1/model.py
+2d1698608d28de6395f8c8794460b292f0233d2a4a5ef76a5d83152631ffe389  policy/bagel_robotics_1/requirements-transport.txt
+8379dddfbbb1154ad9cc57f008f1f8d4a8d26c3de99bd6b36aa4d8067888aa92  policy/bagel_robotics_1/runtime/LICENSE
+bc40a4d1ce1487f1f0bb4b272af7abe6db345ec5fa3cc1373afc7ad36fa7743c  policy/bagel_robotics_1/runtime/THIRD_PARTY_NOTICES.md
+c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4  policy/bagel_robotics_1/runtime/licenses/Apache-2.0.txt
+2258964574a8c567f3f6f871e54ef3664218bed4c40ee73eee4ed8df304fab0b  policy/bagel_robotics_1/runtime/pyproject.toml
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/__init__.py
+70e67d196cb46263677ebe873c5837657d6b8f90ca40cb6845f4782f5377ba2d  policy/bagel_robotics_1/runtime/src/bagel_cp1000/checkpoint_compat.py
+8729d8e1386c74a0e3bd1c800a7c883725c9d00441f19bcbe7e40c2f8ff7af84  policy/bagel_robotics_1/runtime/src/bagel_cp1000/data.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/__init__.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/__init__.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/__init__.py
+df79b044f93bcc9f4e93a2575ed5a7bc191595e3916753fcc26d3dd00db892c3  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/base_processor.py
+b00acaa43c3d18323f56ec9899b2de3ee5a7df8c2289ceea283725ca62aae773  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/vpp2_processor.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/__init__.py
+39fd44f254bd7fb71b9269d75d8f9b5a5bbbe167ea438dda92d40eda1d256734  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/action_state_merger.py
+f494d262ce29770e959bdff60b997f7a07e5f00d10ef515eee7be813d1db038f  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/image.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/utils/__init__.py
+28cedb2f4d69257a5f4da10e406dddba78ddd2a95408a9a55d6f74be5899d8dd  policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/utils/normalizer.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/__init__.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/__init__.py
+780171144ff43f9282875af6c42a16db96b1f9ff56d01742a1024f7888fbb0fb  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/action_dit.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/__init__.py
+6f7ad787624844970985117002ab809bd777f7474969daca121e7101381c6d0c  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/compact_channels.py
+594249b6eb931048e352fd3ffef9b1ac1db9e4fc74659e04216d6b823dd174f3  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/gradient.py
+39cb43aafb25c9046d19921cdddc6df2bae2b3eb267f5f1c271254dd216908e5  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/io.py
+ce632db536b479c96a7160968eb314deb4e2dd1ab1b986e68f5dd8dea0ed6468  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/loader.py
+5f42547a8571a2758ea6d4285834d5eaafc67285c6db42d44ec3c317e8402cc2  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/mot.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/schedulers/__init__.py
+15de04574c1d8ccb794ca6560b1e501e5c83cc86104bb80f1eb3cc3b0324ffe0  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/schedulers/scheduler_continuous.py
+41e8632ad7898bf3a63cf7c874179adae609583ceb1b4dadf2b4c22669865323  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/vpp2.py
+d203d5ec0f6c913b052e22202eff0fe238e087040ae0de9843b7a22a5533b5f3  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan21.py
+f65e44c9d1b7811d2996b3fa22e6fe4c2ac5e6edf1fc2c14251cf582bf8d1f05  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_clip.py
+516b8dd09d576898c8760968ada4b617f5fe6a7c3fd945787898ae326df8e9a1  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_dit.py
+902a1e46b21b5c48e166fe03defabb94b3773c6bbe836fc09b9a0d0894eb1f6e  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_text_encoder.py
+33f7d804677b44a6f3ab334684017dcdb1e9dfb748279de76ad94a5bb71a7e51  policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_vae.py
+35ca975c19a1290531c0027e5cf849ceabee724ff771ed4b49dd653c125c8deb  policy/bagel_robotics_1/runtime/src/bagel_cp1000/policy.py
+dcf2942f3b9a02590902212b3c4f24b297c485ed0be65702ee540529dd4f55c0  policy/bagel_robotics_1/runtime/src/bagel_cp1000/runtime.py
+e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855  policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/__init__.py
+dd3afea396e0ba3e619a1a66121f3c41dd0ca4174e1cf1f9f459b200f23e2690  policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/logging_config.py
+9b0ba19de9a64458b12a0fb24f0811f49d9e9fca5d8b815f5f5ecd3b7886fd63  policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/pytorch_utils.py
+113091a0c7fa38ecda3aae470bb6f151d0d034cfc187d4d752350f8cb9d71191  policy/bagel_robotics_1/setup_eval_env_client.sh
+428ada84e7b5d2c131118e7c8c47a65fb60bc34ce49db94b2e33608c354b93b9  policy/bagel_robotics_1/setup_eval_policy_server.sh
+6597accf2e5ebc6ac46cdc9cfefc9a1622f24fecc9059516dab4fc3abd4530c6  policy/bagel_robotics_1/source-provenance.json
+42be4b09c56377f4686b2e722de760fa3bafd1c6b92605ada269d94e19efc95e  policy/bagel_robotics_1/temporal_ensemble.py
+d5a7f0219ee4fc94c32add51121b83bf2138950a39de67cd42d9a76dddf391eb  policy/bagel_robotics_1/tests/fixtures/checkpoint_processor.json
+add0937c0762cba1395426f8c46fc3b5b9bc2d6da63a51202b7b781a53732db9  policy/bagel_robotics_1/tests/test_backend.py
+97b8ff038365aed20925627fd6f5ed589b9a4b74a3795742fe381f925746259e  policy/bagel_robotics_1/tests/test_checkpoint_processor.py
+59ba14286d9b0ab3d2d29d59fa8b04bb3cdc51dcfadeb6f4b8df7a0df417cb11  policy/bagel_robotics_1/tests/test_contract.py
+a602f672afb21b6662e18882a104e1fc16e46a31a3aa73d9598a1eee715045b4  policy/bagel_robotics_1/tests/test_validation.py
+df0f4687516d3392bfee26e747352fa6510c33ff5102da5dbe830dfb11235c48  policy/bagel_robotics_1/validation.py
+34ca874def38c294228fb542fde970ba1bc3ccc9ff390b23ebe1fc40e3a37459  policy/bagel_robotics_1/validation_compare.py
+9fb0ffad9e69f97c34a90428831ef7eed219fe62c989b0e855dc0d5f6fb2e3ec  policy/bagel_robotics_1/validation_replay.py
+45a6a985ffc7e4c64df13e37fb6afffc649f0675cdfb7bbdd64aeb37b0feee7d  policy/bagel_robotics_1/validation_server.py
diff --git a/inference/install_adapter.py b/inference/install_adapter.py
new file mode 100644
index 0000000000000000000000000000000000000000..bda9a874c1a5159fb1ea78ba6c2cf2a7c82f3d5e
--- /dev/null
+++ b/inference/install_adapter.py
@@ -0,0 +1,40 @@
+"""Install this standalone policy into an existing XPolicyLab checkout.
+
+Refuses to overwrite any existing adapter. Does not publish, submit, or load GPUs.
+"""
+import argparse
+import json
+from pathlib import Path
+import shutil
+import subprocess
+
+ROOT = Path(__file__).resolve().parent
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("xpolicylab", type=Path)
+    args = parser.parse_args()
+    xpl = args.xpolicylab.expanduser().resolve()
+    for rel in ["model_template.py", "setup_policy_server.py", "utils/checkpoint_resolver.py"]:
+        if not (xpl / rel).is_file():
+            parser.error(f"Not a compatible XPolicyLab checkout: missing {rel}")
+    source = ROOT / "policy/bagel_robotics_1"
+    destination = xpl / "policy/bagel_robotics_1"
+    if destination.exists():
+        parser.error(f"Refusing to overwrite {destination}; choose a clean checkout")
+    shutil.copytree(source, destination,
+                    ignore=shutil.ignore_patterns("__pycache__", "*.pyc", "*.egg-info",
+                                                  "checkpoints", ".venv", "build", "dist", ".pytest_cache", "validation-output"))
+    expected = json.loads((source / "source-provenance.json").read_text())["xpolicylab_base"]
+    version = subprocess.run(["git", "-C", str(xpl), "rev-parse", "HEAD"],
+                             capture_output=True, text=True)
+    print(f"Installed: {destination}")
+    print(f"Prepared against XPolicyLab commit: {expected}")
+    if version.returncode == 0 and version.stdout.strip() != expected:
+        print("This checkout differs from the pinned reference; rerun adapter checks.")
+    print(f"Next: bash {destination}/install.sh /absolute/path/to/new-policy-venv")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/inference/policy/bagel_robotics_1/.gitignore b/inference/policy/bagel_robotics_1/.gitignore
new file mode 100644
index 0000000000000000000000000000000000000000..92e78f945da7430756a7592348acf89c52d93516
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/.gitignore
@@ -0,0 +1,9 @@
+checkpoints/
+.venv/
+__pycache__/
+*.pyc
+*.egg-info/
+.pytest_cache/
+build/
+dist/
+validation-output/
diff --git a/inference/policy/bagel_robotics_1/EVALUATION_RESULTS.md b/inference/policy/bagel_robotics_1/EVALUATION_RESULTS.md
new file mode 100644
index 0000000000000000000000000000000000000000..d2b0231c15d11a9c682f6c39d9378fa9d4cba4ac
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/EVALUATION_RESULTS.md
@@ -0,0 +1,103 @@
+# Checkpoint-1000 research evaluation results
+
+Bagel-Robotics-1 / AdaPrune Final, checkpoint 1000. Success means full task completion; partial scores average credited progress over all reported outcomes, including failures. Partial values here are out of 100. Generalization combines standard and randomized episodes; other suites use their recorded standard scene selections.
+
+| Suite | Successful / reported outcomes | Success rate | Mean partial /100 |
+|---|---:|---:|---:|
+| Precision | 126/400 | 31.50% | 40.43 |
+| Memory† | 173/300 | 57.67% | 58.00 |
+| Long Horizon‡ | 137/400 | 34.25% | 53.00 |
+| Generalization | 183/600 | 30.50% | 36.82 |
+| Open§ | 16/400 | 4.00% | 4.48 |
+| **Equal-weight five-suite average** | **635/2,100 total; not the averaging formula** | **31.58%** | **38.55** |
+
+## Original scenes, repairs and replacements
+
+† Memory has 299 valid original results: 173 successes, **57.86% success and 58.19/100 partial**. Original sorting-sequence episode 47 remains unscored. Its separately repaired variant is a valid failure with score zero; the headline includes that variant, giving 300 reported outcomes. The repaired variant is not an unchanged-original result.
+
+‡ Long Horizon includes 398 original results plus egg-holder replacements 50 and 51 for invalid original layouts 15 and 20. Both replacements failed the task, with partial scores 0 and 0.25. Original-only totals: 137/398 successes, 53.2035/100 partial.
+
+§ Open includes 399 original results plus pickup replacement 50 for invalid original layout 5. That replacement failed with partial zero. Original-only totals: 16/399 successes, 4.4862/100 partial.
+
+Generalization combines standard and randomized episodes over all 12 tasks (600 results). The nine user-supplied completed tasks contribute 131/450 successes; the three node-2 tasks contribute 52/150. The original stack-block layout 6 and laptop/headphones layout 22 succeeded on retry; these are original scenes, not replacements. Cancellation of duplicate queue entries does not discard the separately supplied completed results.
+
+These totals contain **2,096 original results, three replacements and one repaired variant**. The overall score gives each suite equal weight, not each episode. Organizer acceptance of local substitutions or repairs is not assumed.
+
+## Sources and scope
+
+Precision, Open and Long Horizon retain the original receipt hashes and replacement evidence in `evidence/local-results.json`. The completed five-suite calculation is in `evidence/five-suite-results.json`.
+
+Memory and nine Generalization tasks come from the user's supplied completed task summaries. Their raw episode logs were not independently re-audited during this packaging update. Stack blocks, store laptop/headphones and sort nesting dolls use node-2 receipts verified in the preceding results review, including successful original-layout retries. No new multi-seed estimate is implied.
+
+These results belong to the frozen research runner. Packaged-adapter debug, smoke and matched-action validation are reported separately in `VALIDATION.md`; research scores do not substitute for those tests. This is not an official leaderboard entry.
+
+## Task-level breakdown
+
+### Precision
+
+| Task | Successful / reported outcomes | Success rate | Mean partial /100 |
+|---|---:|---:|---:|
+| Build tower | 32/50 | 64.00% | 73.60 |
+| Deposit coin | 11/50 | 22.00% | 26.80 |
+| Fasten screws | 18/50 | 36.00% | 66.80 |
+| Insert key | 0/50 | 0.00% | 15.00 |
+| Insert tubes | 32/50 | 64.00% | 75.20 |
+| Play xylophone | 0/50 | 0.00% | 0.00 |
+| Plug in charger | 2/50 | 4.00% | 4.00 |
+| Pour balls into vase | 31/50 | 62.00% | 62.00 |
+
+### Memory
+
+| Task | Successful / reported outcomes | Success rate | Mean partial /100 |
+|---|---:|---:|---:|
+| Cover blocks | 50/50 | 100.00% | 100.00 |
+| Imitate sorting sequence | 0/50 | 0.00% | 2.00 |
+| Match and pick from conveyor | 46/50 | 92.00% | 92.00 |
+| Press by number | 50/50 | 100.00% | 100.00 |
+| Swap T | 11/50 | 22.00% | 22.00 |
+| Swap blocks | 16/50 | 32.00% | 32.00 |
+
+### Long Horizon
+
+| Task | Successful / reported outcomes | Success rate | Mean partial /100 |
+|---|---:|---:|---:|
+| Classify objects | 16/50 | 32.00% | 44.50 |
+| Fill egg holder | 2/50 | 4.00% | 13.80 |
+| Fill pen holder | 11/50 | 22.00% | 43.90 |
+| Make kong | 43/50 | 86.00% | 86.00 |
+| Organize table | 9/50 | 18.00% | 64.50 |
+| Play stacking toy | 6/50 | 12.00% | 32.00 |
+| Play tic tac toe | 1/50 | 2.00% | 40.50 |
+| Put bottles into dustbin | 49/50 | 98.00% | 98.80 |
+
+### Generalization
+
+| Task | Successful / reported outcomes | Success rate | Mean partial /100 |
+|---|---:|---:|---:|
+| Arrange largest number | 2/50 | 4.00% | 11.00 |
+| Fold clothes | 36/50 | 72.00% | 74.40 |
+| Hang mugs | 14/50 | 28.00% | 38.50 |
+| Make toast | 22/50 | 44.00% | 52.50 |
+| Pack objects into box | 6/50 | 12.00% | 35.40 |
+| Pour liquid into cup | 22/50 | 44.00% | 44.00 |
+| Push T | 1/50 | 2.00% | 2.00 |
+| Sort nesting dolls by size | 10/50 | 20.00% | 20.00 |
+| Stack blocks | 22/50 | 44.00% | 50.00 |
+| Stack bowls | 26/50 | 52.00% | 58.90 |
+| Store laptop and headphones | 20/50 | 40.00% | 51.20 |
+| Sweep blocks | 2/50 | 4.00% | 4.00 |
+
+### Open
+
+| Task | Successful / reported outcomes | Success rate | Mean partial /100 |
+|---|---:|---:|---:|
+| Align blocks | 0/50 | 0.00% | 0.00 |
+| General pickup | 12/50 | 24.00% | 24.00 |
+| Stack blocks by language | 4/50 | 8.00% | 9.20 |
+| Solve equation | 0/50 | 0.00% | 0.00 |
+| Classify objects by language | 0/50 | 0.00% | 2.60 |
+| Pick from conveyor by image | 0/50 | 0.00% | 0.00 |
+| Store tools in toolbox | 0/50 | 0.00% | 0.00 |
+| Pour by language | 0/50 | 0.00% | 0.00 |
+
+Checkpoint identity: `790bffc9464fa7a026aa43b217aec2033d186895cad643166daa0bcc01914d5c`.
diff --git a/inference/policy/bagel_robotics_1/LICENSE b/inference/policy/bagel_robotics_1/LICENSE
new file mode 100644
index 0000000000000000000000000000000000000000..90124d129334cd5de11aa0fd606182b3ffd442cb
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/LICENSE
@@ -0,0 +1,24 @@
+MIT License
+
+Copyright (c) 2026 The FastWAM Authors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+Except for third-party code and files that retain their original license
+headers, this repository is released under the MIT License.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
diff --git a/inference/policy/bagel_robotics_1/README.md b/inference/policy/bagel_robotics_1/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..07a9b607c8b6040dcfb467228021e6ef0b90f324
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/README.md
@@ -0,0 +1,219 @@
+# Bagel-Robotics-1
+
+**Contributor:** Bagel Labs | **Paper:** Technical report in preparation | **arXiv:** Not yet available | **Original code:** [VPP2](https://github.com/roboterax/video-prediction-policy-2)
+
+Evaluation-only adapter for the physically compact VPP2 recovery checkpoint 1000.
+The video expert has 12,468,409,664 parameters, 23.95% fewer than the recorded
+16,394,878,784-parameter parent video expert. Action2B remains unchanged. The
+`runtime/` directory vendors a minimal isolated inference runtime and compact
+loader; this is a different artifact from Bagel-Robotics-1-Preview.
+
+Shared conventions — argument meanings, checkpoint naming, split-machine deployment, `EVAL_ENV_TYPE` — are documented in the [XPolicyLab README](https://github.com/XPolicyLab/XPolicyLab/blob/main/README.md). Official results: [RoboDojo LeaderBoard](https://robodojo-benchmark.com/LeaderBoard).
+
+## Installation
+
+Install this directory under `XPolicyLab/policy/bagel_robotics_1` in a RoboDojo
+workspace. `env_cfg/` belongs next to XPolicyLab, not inside the policy. Install
+RoboDojo and its simulator/assets separately using the
+[official installation guide](https://robodojo-benchmark.com/doc/usage/install-and-download/).
+
+**1. Set paths and install the policy environment.** Replace the example checkout
+path below. Run the following sections in the same shell so these variables remain
+available. Use **Python 3.10**, the version used for the GPU qualification; set
+`PYTHON_BIN` to its executable if it is not named `python3.10`.
+
+```bash
+BAGEL_XPL_ROOT=/absolute/path/to/XPolicyLab
+BAGEL_POLICY_DIR="$BAGEL_XPL_ROOT/policy/bagel_robotics_1"
+BAGEL_POLICY_ENV="$BAGEL_POLICY_DIR/.venv"
+cd "$BAGEL_POLICY_DIR"
+PYTHON_BIN=python3.10 bash install.sh "$BAGEL_POLICY_ENV"
+```
+
+The installer creates a venv, installs PyTorch 2.11 / torchvision 0.26 CUDA 13.0
+wheels and installs the isolated runtime plus XPolicyLab. `TORCH_INDEX_URL` changes
+the PyTorch wheel index. GPU qualification used Python 3.10.21 and PyTorch
+2.11.0+cu130 on an RTX PRO 6000 Blackwell; installation on another stack does not
+establish numerical equivalence. The recorded GPU has 96 GB VRAM; a lower
+minimum has not been qualified. The isolated validation jobs reserved 128 GiB
+host RAM and eight CPUs; these are tested allocations, not measured minima.
+
+**2. Download and verify before evaluation.** The checkpoint is public at
+[bageldotcom/adaprune](https://huggingface.co/bageldotcom/adaprune). No login, token
+or access request is required.
+
+```bash
+PYTHON_BIN="$BAGEL_POLICY_ENV/bin/python" bash download_checkpoint.sh
+```
+
+The download is approximately **45.7 GB** before cache/installation overhead.
+To check pinned metadata before downloading tensors, run
+`PYTHON_BIN="$BAGEL_POLICY_ENV/bin/python" bash download_checkpoint.sh --metadata-only`
+first. This checks four small JSON files only; run the full downloader above
+before evaluation. The downloader verifies every checkpoint and auxiliary-asset hash. For existing
+files, follow [Model Assets](#model-assets) instead. Continue to
+[Evaluation](#evaluation) only after the download or verification succeeds.
+
+## Data Processing
+
+No conversion is needed for inference. The research recovery run used a curated
+30,000-window RoboDojo EE16 export; training reproduction is outside this initial
+adapter. `process_data.sh` is intentionally omitted under the proposed eval-only
+exception. Inputs are decoded RGB arrays from XPolicyLab's shared server.
+
+## Training
+
+`train.sh` is intentionally omitted. Training, mask learning and export are not included in this adapter; the
+standalone review package contains a method note. The technical report and training,
+pruning, and export code are planned for release on **November 20, 2026**.
+The proposed eval-only integration and release schedule remain subject to maintainer
+coordination; this preparation does not claim that the exception is approved.
+
+## Evaluation
+
+Supported: `bench_name=RoboDojo`, `env_cfg_type=arx_x5`, `action_type=ee`.
+Use the same checkpoint and settings for every task. Complete Installation first.
+
+**3. Check both debug transports.** These load the real model but use synthetic
+observations. Run from the policy directory with the environment defined above.
+On a dedicated workstation, choose a GPU with sufficient free memory:
+
+```bash
+cd "$BAGEL_POLICY_DIR"
+BAGEL_GPU_ID=0
+```
+
+On a scheduled cluster, run inside an allocation and set `BAGEL_GPU_ID` to its
+assigned single GPU instead; preserve the scheduler's `CUDA_VISIBLE_DEVICES`.
+For a one-GPU allocation that exports this variable:
+
+```bash
+BAGEL_GPU_ID="${CUDA_VISIBLE_DEVICES:?Run inside a one-GPU allocation}"
+```
+
+Run both commands sequentially:
+
+```bash
+EVAL_ENV_TYPE=debug DEBUG_OBS_ENCODED=0 DEBUG_EVAL_EPISODES=1 bash eval.sh \
+  RoboDojo stack_blocks cp1000 arx_x5 ee 1 \
+  "$BAGEL_GPU_ID" "$BAGEL_GPU_ID" "$BAGEL_POLICY_ENV" "$BAGEL_POLICY_ENV"
+
+EVAL_ENV_TYPE=debug DEBUG_OBS_ENCODED=1 DEBUG_EVAL_EPISODES=1 bash eval.sh \
+  RoboDojo stack_blocks cp1000 arx_x5 ee 1 \
+  "$BAGEL_GPU_ID" "$BAGEL_GPU_ID" "$BAGEL_POLICY_ENV" "$BAGEL_POLICY_ENV"
+```
+
+Each must finish with `[TestEnv] Episode finished` and `[MAIN] eval finished`,
+exit zero, and contain no traceback. Debug completion is an interface check,
+not a task success result. A venv path in argument 10 is supported for debug
+only; simulation uses the official conda client launcher.
+
+**4. Run a small native simulator check.** Set the installed RoboDojo path and
+simulator conda environment name. The following command matches the tested
+`insert_tubes` smoke: seed 1, two episodes, original layouts 0 and 1. It uses
+the same GPU selected above for the policy and simulator.
+
+```bash
+BAGEL_SIM_ROOT=/absolute/path/to/RoboDojo
+BAGEL_SIM_ENV=your_simulator_conda_env
+cd "$BAGEL_SIM_ROOT"
+EVAL_ENV_TYPE=sim bash scripts/robodojo.sh eval \
+  --policy-dir "$BAGEL_POLICY_DIR" \
+  --task insert_tubes --ckpt cp1000 --env-cfg arx_x5 --action-type ee \
+  --seed 1 --eval-num 2 --policy-env "$BAGEL_POLICY_ENV" \
+  --eval-env "$BAGEL_SIM_ENV" \
+  --policy-gpu "$BAGEL_GPU_ID" --env-gpu "$BAGEL_GPU_ID"
+```
+
+For the second tested task, repeat with `--task store_laptop_and_headphones`.
+Preserve native `_result.json` files under RoboDojo's `eval_result/` and record
+source revisions and any startup errors. See [VALIDATION.md](VALIDATION.md) for
+measured outcomes and the shared-installation startup workaround used in our
+qualification. Keep these two-episode checks separate from full benchmark runs.
+
+For a full task, suite, or three-seed run, follow [Reevaluation](REEVALUATION.md).
+It uses the official runner, native episode counts and original layouts.
+
+The direct launcher retains XPolicyLab's ten-argument contract (run from the
+policy directory):
+
+```text
+bash eval.sh <bench_name> <task_name> <ckpt_name> <env_cfg_type> <action_type> <seed> \
+  <policy_gpu_id> <env_gpu_id> <policy_env_or_venv_path> <simulator_conda_env>
+```
+
+The custom `deploy.py` records every executed observation, calls reset at episode
+boundaries and stops sending actions when the environment terminates. This is
+necessary for the checkpoint's history conditioning. Batched execution is disabled;
+the batch methods only support a single environment and reject larger batches.
+
+## Model Assets
+
+The Installation quickstart authenticates and downloads before evaluation.
+The script downloads the six exact bundle files into `checkpoints/cp1000` and
+the pinned public VAE/CLIP/UMT5/tokenizer assets into
+`checkpoints/Wan2.1-I2V-14B-480P`. Bundle repository:
+[bageldotcom/adaprune](https://huggingface.co/bageldotcom/adaprune),
+revision `2651631c5d08819088475e8778676f5cf6d9afcf` (**public; no login required**).
+Access must be arranged for evaluators. No token is included.
+
+For existing artifacts, set `BAGEL_CP1000_BUNDLE` and `BAGEL_CP1000_WAN_ROOT`.
+`bundle_path` and `wan_model_dir` in `deploy.yml` are alternatives. The environment
+variables take precedence; otherwise the shared XPolicyLab checkpoint resolver
+uses `ckpt_name`. `artifact-manifest.json` pins every file and encoder revision.
+
+```bash
+export BAGEL_CP1000_BUNDLE=/absolute/path/to/update_001000
+export BAGEL_CP1000_WAN_ROOT=/absolute/path/to/Wan2.1-I2V-14B-480P
+"$BAGEL_POLICY_ENV/bin/python" "$BAGEL_POLICY_DIR/download_checkpoint.py" \
+  --verify-only --bundle "$BAGEL_CP1000_BUNDLE" --wan-root "$BAGEL_CP1000_WAN_ROOT"
+```
+
+Full file hashes are checked before loading. `--metadata-only` downloads only
+small checkpoint metadata for CPU architecture checks and does not qualify weights.
+
+## Configuration
+
+BF16, action horizon 32, execute/replan 16, ten Euler steps, sigma shift 1,
+90/10 temporal blending, and `eval_batch: false` are fixed and validated.
+`seed` is supplied by the evaluator; task names never select alternate models.
+The server launcher also preserves the recorded deterministic backend controls
+(`PYTHONHASHSEED=1`, `CUBLAS_WORKSPACE_CONFIG=:4096:8`, deterministic algorithms,
+and the recorded TF32/matmul settings). Direct API users must set those startup
+environment variables before launching Python. The checkpoint loader accepts only the compact channel architecture and loads
+weights strictly. There is no teacher, mask-training module or dense fallback.
+
+Optional validation: setting `BAGEL_VALIDATION_TRACE` to a new absolute output
+directory enables lossless capture around the official policy server. It records
+the first 209 observations per episode and fingerprints preprocessing, history
+and actions. It changes no predictions or scoring and is disabled by default.
+See [VALIDATION.md](VALIDATION.md) for capture/replay scope.
+
+## Notes
+
+Mask-based structured pruning and recovery produce physically smaller,
+nonuniform FFN and cross-attention layers. The method is intended for compatible
+WAM backbones; the evidence here is limited to VPP2. The technical report will
+provide the detailed algorithm and recovery recipe.
+
+
+The recorded research results cover all five suites:
+
+| Suite | Successful / reported outcomes | Success rate | Mean partial /100 |
+|---|---:|---:|---:|
+| Precision | 126/400 | 31.50% | 40.43 |
+| Memory† | 173/300 | 57.67% | 58.00 |
+| Long Horizon‡ | 137/400 | 34.25% | 53.00 |
+| Generalization | 183/600 | 30.50% | 36.82 |
+| Open§ | 16/400 | 4.00% | 4.48 |
+| **Equal-weight five-suite average** | **635/2,100 total; not the averaging formula** | **31.58%** | **38.55** |
+
+Memory includes one repaired variant, Long Horizon two labeled replacements, and Open one labeled replacement. Generalization combines standard and randomized episodes. Memory and nine Generalization tasks use user-supplied summaries; see [EVALUATION_RESULTS.md](EVALUATION_RESULTS.md) for complete counts, task results, source limitations and original-only figures. These are research-runner results, not official leaderboard scores or proof of packaged-adapter equivalence.
+
+The learned masks have been exported into nonuniform per-layer FFN/cross-attention
+reductions. Action2B and all video self-attention heads are retained. Eligibility
+as a new model and the evaluation-only exception remain subject to committee review.
+
+Package validation so far: 15 CPU tests, real compact-video strict tensor loading, all checkpoint/encoder hashes, static checks, isolated wheel processor loading, and clean-checkout installation. Both real-checkpoint GPU debug transports passed in job 53609. Four native simulator episodes completed with valid results, and all eight action/history/reset/backend comparison checks passed with zero recorded differences. See [VALIDATION.md](VALIDATION.md) for the measured task scores and test scope.
+Reviewer access remains pending. The technical report and training/pruning/export code are planned for **November 20, 2026**. Full-suite
+coverage is a separate leaderboard-verification milestone. See [validation evidence](VALIDATION.md) and [historical task results](EVALUATION_RESULTS.md). The standalone review package contains the fuller internal method note.
diff --git a/inference/policy/bagel_robotics_1/REEVALUATION.md b/inference/policy/bagel_robotics_1/REEVALUATION.md
new file mode 100644
index 0000000000000000000000000000000000000000..ee620a70099602a593c6c8565747bbcc08fe9009
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/REEVALUATION.md
@@ -0,0 +1,131 @@
+# Reevaluate Bagel-Robotics-1 checkpoint 1000
+
+Use the [README quickstart](README.md#installation) to install Python 3.10 policy
+dependencies, download and verify the pinned files, and pass both
+debug checks. The standalone inference runtime is included; the private research
+training repository is not needed. Install RoboDojo and its assets separately.
+
+The only policy checkpoint is `cp1000`, from
+[bageldotcom/adaprune](https://huggingface.co/bageldotcom/adaprune),
+revision `2651631c5d08819088475e8778676f5cf6d9afcf`. The downloader and loader reject
+files that do not match `artifact-manifest.json`. The VAE, image/text encoders and
+tokenizer are downloaded automatically at their separately pinned public revision.
+The checkpoint and all required public encoder assets download without authentication.
+
+## Set the local paths
+
+Replace the example paths and simulator conda environment. Reuse the policy
+venv created in the README. GPU commands on a cluster belong inside an allocation;
+set `BAGEL_GPU_ID` to its assigned single GPU. On a dedicated workstation, the
+example selects GPU 0. Do not overwrite a scheduler's GPU visibility assignment.
+
+```bash
+BAGEL_XPL_ROOT=/absolute/path/to/XPolicyLab
+BAGEL_POLICY_DIR="$BAGEL_XPL_ROOT/policy/bagel_robotics_1"
+BAGEL_POLICY_ENV="$BAGEL_POLICY_DIR/.venv"
+BAGEL_SIM_ROOT=/absolute/path/to/RoboDojo
+BAGEL_SIM_ENV=your_simulator_conda_env
+BAGEL_GPU_ID=0
+cd "$BAGEL_SIM_ROOT"
+```
+
+Qualification used XPolicyLab `6baccd57ff958570d0db1588370289ee36c8bb70` and
+RoboDojo `f57e1633ffed59761bc8e568cfe7530c4b33381c`. Record the revisions used for
+reevaluation. A different simulator or numerical backend requires fresh checking.
+The [official installation guide](https://robodojo-benchmark.com/doc/usage/install-and-download/)
+provides the simulator workspace layout, assets and conda setup.
+
+## Inspect the command, then check two episodes
+
+The dry run does not load the model or simulator:
+
+```bash
+EVAL_ENV_TYPE=sim bash scripts/robodojo.sh eval \
+  --policy-dir "$BAGEL_POLICY_DIR" --task insert_tubes --ckpt cp1000 \
+  --env-cfg arx_x5 --action-type ee --seed 1 --eval-num 2 \
+  --policy-env "$BAGEL_POLICY_ENV" --eval-env "$BAGEL_SIM_ENV" \
+  --policy-gpu "$BAGEL_GPU_ID" --env-gpu "$BAGEL_GPU_ID" --dry-run
+```
+
+Remove `--dry-run` to execute the check. The recorded qualification used original
+layouts 0/1 and produced 2/2 successes for `insert_tubes`. Repeating with
+`--task store_laptop_and_headphones` produced 1/2 successes and 60/100 mean partial
+score. These are four integration episodes, not full-suite performance guarantees.
+[Validation evidence](VALIDATION.md) records their scope and backend.
+
+## Evaluate complete tasks or suites
+
+Use the official `benchmark` entry point with `--eval-num native`: the benchmark
+then selects its 25/50 episode counts. The following runs the entire Memory suite
+sequentially with seed 1. Append `--dry-run` first to inspect all resolved commands.
+
+```bash
+EVAL_ENV_TYPE=sim bash scripts/robodojo.sh benchmark \
+  --policy-dir "$BAGEL_POLICY_DIR" --ckpt cp1000 \
+  --env-cfg arx_x5 --action-type ee --seed 1 --eval-num native \
+  --policy-env "$BAGEL_POLICY_ENV" --eval-env "$BAGEL_SIM_ENV" \
+  --policy-gpu "$BAGEL_GPU_ID" --env-gpu "$BAGEL_GPU_ID" \
+  --dimension memory
+```
+
+Use `--dimension generalization`, `precision`, `long-horizon`, `open`, or `all`.
+For one complete task, use `--only insert_tubes` with `--dimension precision`.
+Generalization includes each standard task and its `_random` sibling; each has
+25 episodes per seed. The five suites contain 54 runnable configurations, merged
+into 42 task rows. Preserve canonical names such as `play_Xylophone`, `swap_T`,
+`push_T` and `push_T_random`.
+
+## Run the three-seed verification scope
+
+Use a dedicated result workspace for a complete run, separate from the smoke
+checks. This command is a full benchmark, not a quick installation test. Run it
+only with the intended evaluation allocation and disk space. It runs tasks and
+seeds sequentially; it never changes the checkpoint or inference settings.
+
+```bash
+set -euo pipefail
+for BAGEL_EVAL_SEED in 0 1 2; do
+  EVAL_ENV_TYPE=sim bash scripts/robodojo.sh benchmark \
+    --policy-dir "$BAGEL_POLICY_DIR" --ckpt cp1000 \
+    --env-cfg arx_x5 --action-type ee --seed "$BAGEL_EVAL_SEED" \
+    --eval-num native --dimension all \
+    --policy-env "$BAGEL_POLICY_ENV" --eval-env "$BAGEL_SIM_ENV" \
+    --policy-gpu "$BAGEL_GPU_ID" --env-gpu "$BAGEL_GPU_ID"
+done
+bash scripts/robodojo.sh summarize
+```
+
+The official [Quick Evaluation guide](https://robodojo-benchmark.com/doc/usage/quick-evaluation/)
+also documents multi-GPU task sharding. Use one unbatched policy process per
+worker; `eval_batch: false` is required. Shared state across independent episodes
+or changing inference settings is not part of this submission.
+
+## Preserve and report the results
+
+Keep each run's native `_result.json`, logs, source revisions and videos. Run
+`bash scripts/robodojo.sh summarize` from the matching RoboDojo checkout and retain
+`eval_result/RoboDojo/_summary.md`. The native summary distinguishes incomplete
+coverage; its across-seed statistics must not be replaced with statistics over
+episodes or combined with the two-episode smoke results.
+
+The default reevaluation uses original supplied layouts. If a scene fails before
+any robot action, preserve its startup error and retry history; do not silently
+replace the scene or score an infrastructure failure as a policy failure. Seek
+the committee's handling decision. The historical research run used three
+labeled replacements and one repaired Memory variant, described in
+[EVALUATION_RESULTS.md](EVALUATION_RESULTS.md); those exceptions are not activated
+by these commands. Official verification and acceptance remain with RoboDojo.
+
+## Common setup problems
+
+| Symptom | Check |
+|---|---|
+| HTTP 401/403 or repository not found | Use the canonical public repository and pinned revision; retry without a stale cached credential (`HF_HUB_DISABLE_IMPLICIT_TOKEN=1`). |
+| Hash/size mismatch | Check the pinned revision and local file completeness; do not disable verification. |
+| Missing encoder or tokenizer | Run the full downloader; `--metadata-only` does not fetch tensors or encoders. |
+| Missing `env_cfg` or simulator | Complete the official workspace setup; `env_cfg/` belongs beside XPolicyLab. |
+| Shared Isaac/Kit cache permissions | Use a writable installation/cache; the recorded local workaround is documented in VALIDATION.md. No task or scoring patches are supplied. |
+| Insufficient GPU memory | Qualification used a 96 GB RTX PRO 6000 Blackwell; a smaller minimum has not been qualified. |
+
+The policy logs the checkpoint identity after file verification. Keep that log
+with the run so the evaluated checkpoint is unambiguous.
diff --git a/inference/policy/bagel_robotics_1/THIRD_PARTY_NOTICES.md b/inference/policy/bagel_robotics_1/THIRD_PARTY_NOTICES.md
new file mode 100644
index 0000000000000000000000000000000000000000..7c7140daa83b46bf89a8ef0a227cc2585c1d8932
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/THIRD_PARTY_NOTICES.md
@@ -0,0 +1,13 @@
+# Source and artifact notices
+
+- The retained VPP2/FastWAM-derived runtime carries the original MIT notice in
+  `LICENSE`, copyright 2026 The FastWAM Authors. Bagel-specific packaging does not
+  remove that attribution. See `source-provenance.json` for frozen-source hashes.
+- Wan2.1/DiffSynth-derived model components retain Apache-2.0 terms; see
+  `licenses/Apache-2.0.txt` and upstream Wan2.1/DiffSynth source notices.
+- The XPolicyLab interface follows its upstream license, included in
+  `licenses/XPolicyLab.txt`. RoboDojo and simulator assets are not bundled.
+- Encoder weights, tokenizer files and model weights have their own upstream
+  terms. The code license does not assign a new license to those artifacts.
+- The existing preview PR by Soroush Mehraban was consulted as a reference; its
+  alternate checkpoint, code and release promises are not this submission.
diff --git a/inference/policy/bagel_robotics_1/VALIDATION.md b/inference/policy/bagel_robotics_1/VALIDATION.md
new file mode 100644
index 0000000000000000000000000000000000000000..4c9981011f1581f8fa35bee61f600f06b59d9135
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/VALIDATION.md
@@ -0,0 +1,45 @@
+# Adapter validation
+
+Status: **Technical qualification passed: 15 CPU tests, both real-checkpoint debug transports, four valid native simulator episodes, and all eight exact replay/backend checks. The technical report and training/pruning/export code are planned for November 20, 2026; evaluator access and user review remain before submission.**
+
+| Check | Evidence and scope |
+|---|---|
+| Static/template audit | 50 Python files compiled, five shell scripts checked, clean diff, exact PR-template headings |
+| CPU regression | 15 tests passed: compact architecture metadata, corrupt artifacts, blending, observation timing, reset, backend settings, lossless array integrity and comparison failures |
+| Real video tensors | Strict CPU loading passed; 12,468,409,664 compact video parameters |
+| Artifact integrity | All six bundle files and nine auxiliary assets matched pinned sizes/hashes |
+| Fresh policy installation | Python 3.10.21, PyTorch 2.11.0+cu130; dependency consistency check passed |
+| Source preservation | 30 numerical modules and 16 policy functions match the frozen reference after namespace normalization; factory and prompt match |
+| Simulator preflight | Official doctor: 12 passes, zero failures; Isaac and conda checks intentionally skipped at this non-GPU stage |
+| Plain and encoded debug | Passed with real checkpoint: job 53609, exit 0, both evaluation-finished markers |
+| Simulator smoke | 4/4 valid, original layouts 0/1, seed 1: Insert tubes 2/2 successes, 100/100 partial; Store laptop/headphones 1/2 successes, 60/100 partial |
+| Numerical action comparison | Passed: all eight checks, zero recorded differences; job 53612 completed with exit 0 |
+
+XPolicyLab base: `6baccd57ff958570d0db1588370289ee36c8bb70`.
+Simulator validation checkout: `f57e1633ffed59761bc8e568cfe7530c4b33381c`.
+
+Two real recordings of 209 observations / 14 action chunks each were replayed in A/B/A order, giving 42 chunks per implementation. Reference-versus-package, original-capture-versus-replay, reset and backend checks all passed with zero recorded differences. Evidence: `evidence/gpu-validation.json` and `evidence/action-comparison.json`. Equality is established for these fixtures on PyTorch 2.11.0+cu130 and NVIDIA RTX PRO 6000 Blackwell, not every input or backend.
+
+The original Store laptop/headphones layout 1 failed with 0.2/1 partial and is retained. Debug completion is an interface check; four simulator smoke episodes do not establish full-suite performance. Historical research scores are separately documented in EVALUATION_RESULTS.md.
+
+## Reproduce the transport checks
+
+Use the policy README's debug command with the actual checkpoint. Repeat it with `DEBUG_OBS_ENCODED=1`. Both runs must finish normally and print the evaluation-finished marker.
+
+## Optional lossless capture
+
+`BAGEL_VALIDATION_TRACE=/absolute/path/to/new-capture-directory` enables `validation_server.py` through the normal server launcher. The wrapper uses the official server and records already-decoded observations, inference inputs, history fingerprints, raw predictions, blending and commands. It does not modify actions, seeds, task termination or scoring.
+
+Capture is limited to the first 209 observations per episode. The default server path does not import or enable the recorder. Array blobs are local test artifacts and are excluded from the PR.
+
+`validation_replay.py` replays two fixtures in A/B/A order. `validation_compare.py` checks exact equality of reference versus packaged outputs, capture versus replay, history and resets, and verifies matched GPU/backend settings. The original frozen research implementation is needed for the reference comparison and is not redistributed in this inference package.
+
+All planned qualification stages completed. Failed attempts remain preserved in the outer verification packet; full camera fixtures and videos remain on the validation machine and are excluded from the PR.
+
+## Packaging repair
+
+The first debug attempt (53541) failed before opening the server because serialized processor targets still referred to the original `vpp2` namespace and the processor modules were missing. Its dependent simulator job 53542 was cancelled before execution. The repair vendors the evaluated processor/transform modules unchanged except import namespace, and recursively maps Hydra targets into `bagel_cp1000`. Four new CPU cases test the real processor metadata and legacy target formats while rejecting research-package imports. Failed evidence is preserved in the outer review packet.
+
+## Shared simulator installation
+
+The local qualification environment reuses an existing shared Isaac Sim installation. Attempt 53610 stopped at its noninteractive startup prompt; attempt 53611 then stopped on read-only shared cache permissions. Neither reached a scored episode. Attempt 53612 restores the same noninteractive flag and user-owned portable Kit directory used by the frozen evaluation runner. A recorded one-line change to the validation checkout's launch script permits the portable-root argument; task, physics, layout and scoring code remain unchanged. This launch-only patch is in the outer verification packet and is not part of the policy PR. See [Kit portable configuration](https://docs.omniverse.nvidia.com/kit/docs/kit-manual/latest/guide/configuring.html).
diff --git a/inference/policy/bagel_robotics_1/__init__.py b/inference/policy/bagel_robotics_1/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/artifact-manifest.json b/inference/policy/bagel_robotics_1/artifact-manifest.json
new file mode 100644
index 0000000000000000000000000000000000000000..fcc1c37a1eaf549c4b77576d0dcbcfcc197c44aa
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/artifact-manifest.json
@@ -0,0 +1,112 @@
+{
+  "schema": "bagel-cp1000-submission-v1",
+  "display_name": "Bagel-Robotics-1",
+  "checkpoint_identity": "790bffc9464fa7a026aa43b217aec2033d186895cad643166daa0bcc01914d5c",
+  "update": 1000,
+  "video_parameters": 12468409664,
+  "repo_id": "bageldotcom/adaprune",
+  "revision": "2651631c5d08819088475e8778676f5cf6d9afcf",
+  "private": false,
+  "files": {
+    "COMPLETE.json": {
+      "sha256": "eee062fbcb8c92a881117aa7f06b5cd039dfb9912d85571b59a20aa30837b0e1",
+      "size": 2491
+    },
+    "action.pt": {
+      "sha256": "9b04836abe5a30972e69ee84e5f5e39c09bf53dd45865551e1d04b42ead90ec3",
+      "size": 4057916577
+    },
+    "compact_architecture.json": {
+      "sha256": "a7f6696ab9a9beca785531c2402b81ccd1550745cb6e9ccaad67f658630f6b32",
+      "size": 6563270
+    },
+    "dataset_stats.json": {
+      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
+      "size": 100256
+    },
+    "validation.json": {
+      "sha256": "9806ad2b0b58f879351da58eede46507a79433ec9e4eb8f37f415cf58f2ef70e",
+      "size": 88781
+    },
+    "video.pt": {
+      "sha256": "fb62db0b3a74a53aeab60f4b1acc8a5be47274f5c9db166684043521fe2a40bb",
+      "size": 24937214042
+    }
+  },
+  "wan": {
+    "repo": "Haodong082399/VPP2",
+    "revision": "04824ddb3aae11f13db50984ed7d540c3b77dd9f",
+    "files": [
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/LICENSE.txt",
+        "blobId": "261eeb9e9f8b2b4b0d119366dda99c6fd7d35c64",
+        "size": 11357
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth",
+        "blobId": "5897fba405232a6b07a947d6188d19a8e050ccfb",
+        "size": 507609880,
+        "lfs": {
+          "sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+          "size": 507609880,
+          "pointerSize": 134
+        }
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/config.json",
+        "blobId": "f5a12b285ec9c2497cafc4a4443fc6ba61b1a30b",
+        "size": 250
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/google/umt5-xxl/special_tokens_map.json",
+        "blobId": "14855e7052ffbb595057dfd791d293c1c940db2c",
+        "size": 6623
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/google/umt5-xxl/spiece.model",
+        "blobId": "2fe5f347e9f9367585589ae89e997dfbd5cf802c",
+        "size": 4548313,
+        "lfs": {
+          "sha256": "e3909a67b780650b35cf529ac782ad2b6b26e6d1f849d3fbb6a872905f452458",
+          "size": 4548313,
+          "pointerSize": 132
+        }
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/google/umt5-xxl/tokenizer.json",
+        "blobId": "7bd4ee3cf5633e49c0e06a3aac2a621559e6222a",
+        "size": 16837417,
+        "lfs": {
+          "sha256": "6e197b4d3dbd71da14b4eb255f4fa91c9c1f2068b20a2de2472967ca3d22602b",
+          "size": 16837417,
+          "pointerSize": 133
+        }
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/google/umt5-xxl/tokenizer_config.json",
+        "blobId": "4e1cc1cd85599ce0b47fd0a746af188fe4043ff2",
+        "size": 61728
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth",
+        "blobId": "cbdd1a5bd944b2b4c47e21435106316971ac76d8",
+        "size": 4772359047,
+        "lfs": {
+          "sha256": "628c9998b613391f193eb67ff68da9667d75f492911e4eb3decf23460a158c38",
+          "size": 4772359047,
+          "pointerSize": 135
+        }
+      },
+      {
+        "rfilename": "checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth",
+        "blobId": "d5dad910304ab4b909a2c8a225a71840606e6de4",
+        "size": 11361920418,
+        "lfs": {
+          "sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+          "size": 11361920418,
+          "pointerSize": 136
+        }
+      }
+    ]
+  }
+}
diff --git a/inference/policy/bagel_robotics_1/assets.py b/inference/policy/bagel_robotics_1/assets.py
new file mode 100644
index 0000000000000000000000000000000000000000..041f23bb350b59af8c4e9f45c96a32200a33e9b5
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/assets.py
@@ -0,0 +1,104 @@
+"""Pinned CP1000 artifact and inference contract; safe to import without torch."""
+from __future__ import annotations
+
+import hashlib
+import json
+import os
+from pathlib import Path
+
+POLICY_DIR = Path(__file__).resolve().parent
+MANIFEST = json.loads((POLICY_DIR / "artifact-manifest.json").read_text())
+FIXED_SETTINGS = {
+    "bench_name": "RoboDojo", "env_cfg_type": "arx_x5", "action_type": "ee",
+    "protocol": "ws", "eval_batch": False, "mixed_precision": "bf16",
+    "action_horizon": 32, "replan_steps": 16, "num_inference_steps": 10,
+    "sigma_shift": 1.0, "deployment_adapter": "robodojo_ee16",
+    "temporal_ensemble": {"enabled": True, "new_weight": 0.9, "old_weight": 0.1},
+}
+
+
+def file_digest(path, algorithm="sha256", git_blob=False):
+    digest = hashlib.new(algorithm)
+    if git_blob:
+        digest.update(f"blob {path.stat().st_size}\0".encode())
+    with path.open("rb") as handle:
+        for chunk in iter(lambda: handle.read(8 * 1024 * 1024), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def verify_bundle(bundle, *, full_hash=True):
+    bundle = Path(bundle).expanduser().resolve()
+    for name, expected in MANIFEST["files"].items():
+        path = bundle / name
+        if not path.is_file():
+            raise FileNotFoundError(f"Missing checkpoint artifact: {path}")
+        if path.stat().st_size != expected["size"]:
+            raise ValueError(f"Checkpoint size mismatch: {name}")
+        if full_hash or name not in {"action.pt", "video.pt"}:
+            if file_digest(path) != expected["sha256"]:
+                raise ValueError(f"Checkpoint SHA-256 mismatch: {name}")
+    receipt = json.loads((bundle / "COMPLETE.json").read_text())
+    if receipt.get("identity") != MANIFEST["checkpoint_identity"] or receipt.get("update") != 1000:
+        raise ValueError("Not the approved recovery checkpoint 1000")
+    architecture = json.loads((bundle / "compact_architecture.json").read_text())
+    if architecture.get("schema") != "vpp2-compact-channels-v1" or architecture.get("parameters") != MANIFEST["video_parameters"]:
+        raise ValueError("Unexpected compact architecture")
+    return {"checkpoint_identity": receipt["identity"], "update": 1000,
+            "files": len(MANIFEST["files"]), "full_hash": full_hash}
+
+
+def verify_wan(wan_root, *, full_hash=True):
+    wan_root = Path(wan_root).expanduser().resolve()
+    prefix = "checkpoints/Wan2.1-I2V-14B-480P/"
+    for entry in MANIFEST["wan"]["files"]:
+        path = wan_root / entry["rfilename"].removeprefix(prefix)
+        if not path.is_file():
+            raise FileNotFoundError(f"Missing encoder asset: {path}")
+        if path.stat().st_size != entry["size"]:
+            raise ValueError(f"Encoder asset size mismatch: {path.name}")
+        if full_hash:
+            lfs = entry.get("lfs")
+            actual = file_digest(path, "sha256" if lfs else "sha1", git_blob=not bool(lfs))
+            expected = lfs["sha256"] if lfs else entry["blobId"]
+            if actual != expected:
+                raise ValueError(f"Encoder asset checksum mismatch: {path.name}")
+    return {"repo": MANIFEST["wan"]["repo"], "revision": MANIFEST["wan"]["revision"],
+            "files": len(MANIFEST["wan"]["files"]), "full_hash": full_hash}
+
+
+def validate_settings(model_cfg):
+    cfg = dict(model_cfg)
+    for key, expected in FIXED_SETTINGS.items():
+        actual = cfg.get(key, expected)
+        if actual != expected:
+            raise ValueError(f"CP1000 requires {key}={expected!r}; got {actual!r}")
+        cfg[key] = expected
+    seed = cfg.get("seed", 1)
+    if isinstance(seed, bool) or not isinstance(seed, int) or seed < 0:
+        raise ValueError("seed must be a nonnegative integer supplied by the evaluator")
+    cfg["seed"] = seed
+    return cfg
+
+
+def prepare_model_config(model_cfg, policy_dir=POLICY_DIR):
+    from XPolicyLab.utils.checkpoint_resolver import resolve_checkpoint_root
+
+    cfg = validate_settings(model_cfg)
+    explicit = os.environ.get("BAGEL_CP1000_BUNDLE") or cfg.get("bundle_path")
+    lookup = dict(cfg)
+    if explicit:
+        lookup["model_path"] = explicit
+    bundle = resolve_checkpoint_root(lookup, Path(policy_dir) / "checkpoints",
+                                     policy_dir=policy_dir)
+    wan_root = os.environ.get("BAGEL_CP1000_WAN_ROOT") or cfg.get("wan_model_dir")
+    if not wan_root:
+        wan_root = Path(policy_dir) / "checkpoints/Wan2.1-I2V-14B-480P"
+    # Verify before loading any tensors. No silently accepted alternate weights.
+    verification = {"checkpoint": verify_bundle(bundle), "encoders": verify_wan(wan_root)}
+    print("[Bagel-Robotics-1] " + json.dumps(verification), flush=True)
+    cfg.update(checkpoint_path=str(bundle / "action.pt"),
+               video_checkpoint_path=str(bundle / "video.pt"),
+               dataset_stats_path=str(bundle / "dataset_stats.json"),
+               wan_model_dir=str(Path(wan_root).expanduser().resolve()))
+    return cfg
diff --git a/inference/policy/bagel_robotics_1/backend.py b/inference/policy/bagel_robotics_1/backend.py
new file mode 100644
index 0000000000000000000000000000000000000000..85667024949973c25f9ecc2dbc3f94a67d70d52f
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/backend.py
@@ -0,0 +1,26 @@
+"""Numerical controls recorded by the checkpoint-1000 node-2 evaluation."""
+import os
+import random
+
+
+def configure_backend(seed):
+    if os.environ.get("PYTHONHASHSEED") != "1":
+        raise RuntimeError("Launch Python with PYTHONHASHSEED=1 (see setup_eval_policy_server.sh)")
+    if os.environ.get("CUBLAS_WORKSPACE_CONFIG") != ":4096:8":
+        raise RuntimeError("Launch Python with CUBLAS_WORKSPACE_CONFIG=:4096:8")
+    import numpy as np
+    import torch
+    random.seed(seed)
+    np.random.seed(seed % (2 ** 32))
+    torch.manual_seed(seed)
+    if torch.cuda.is_available():
+        torch.cuda.manual_seed_all(seed)
+    torch.backends.cudnn.benchmark = False
+    torch.backends.cudnn.deterministic = True
+    torch.backends.cudnn.allow_tf32 = True
+    torch.backends.cuda.matmul.allow_tf32 = False
+    torch.set_float32_matmul_precision("highest")
+    torch.use_deterministic_algorithms(True, warn_only=False)
+    return {"torch": torch.__version__, "cuda": torch.version.cuda,
+            "deterministic_algorithms": torch.are_deterministic_algorithms_enabled(),
+            "float32_matmul_precision": torch.get_float32_matmul_precision()}
diff --git a/inference/policy/bagel_robotics_1/deploy.py b/inference/policy/bagel_robotics_1/deploy.py
new file mode 100644
index 0000000000000000000000000000000000000000..9562d8adb50a27422059d4d0ffeb12dd04a74fa6
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/deploy.py
@@ -0,0 +1,36 @@
+import os
+import socket
+import uuid
+
+# VPP2 keeps per-episode observation history on the policy server.  Every
+# executed action records exactly one observation through update_obs, and each
+# call carries this process's client id so one server can serve several
+# simulator clients without mixing their histories.  A single-client official
+# evaluation behaves exactly as without the id.
+_CLIENT_ID_FIELD = "_vpp2_client_id"
+_CLIENT_ID = f"{socket.gethostname()}:{os.getpid()}:{uuid.uuid4().hex[:8]}"
+
+
+def _with_client_id(obs):
+    tagged = dict(obs)
+    tagged[_CLIENT_ID_FIELD] = _CLIENT_ID
+    return tagged
+
+
+def eval_one_episode(TASK_ENV, model_client):
+    model_client.call(func_name="reset")
+    model_client.call(func_name="reset_client", obs={_CLIENT_ID_FIELD: _CLIENT_ID})
+
+    while not TASK_ENV.is_episode_end():  # Check whether the episode ends
+        obs = TASK_ENV.get_obs()  # Get Observation
+        model_client.call(func_name="update_obs", obs=_with_client_id(obs))
+
+        actions = model_client.call(func_name="get_action", obs={_CLIENT_ID_FIELD: _CLIENT_ID})
+        for action_idx, action in enumerate(actions):
+            if TASK_ENV.is_episode_end():
+                break
+            TASK_ENV.take_action(action)
+
+            if action_idx != len(actions) - 1 and not TASK_ENV.is_episode_end():
+                obs = TASK_ENV.get_obs()  # Get Observation
+                model_client.call(func_name="update_obs", obs=_with_client_id(obs))
diff --git a/inference/policy/bagel_robotics_1/deploy.yml b/inference/policy/bagel_robotics_1/deploy.yml
new file mode 100644
index 0000000000000000000000000000000000000000..68075ab87f920a6585db75ab488ef83d2e2707a0
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/deploy.yml
@@ -0,0 +1,26 @@
+policy_name: bagel_robotics_1
+protocol: ws
+host: localhost
+port: null
+bench_name: RoboDojo
+task_name: null
+ckpt_name: cp1000
+env_cfg_type: arx_x5
+action_type: ee
+seed: 1
+gpu_id: null
+eval_batch: false
+bundle_path: null
+wan_model_dir: null
+deployment_adapter: robodojo_ee16
+device: cuda
+mixed_precision: bf16
+action_horizon: 32
+replan_steps: 16
+num_inference_steps: 10
+sigma_shift: 1.0
+default_instruction: follow the instruction
+temporal_ensemble:
+  enabled: true
+  new_weight: 0.9
+  old_weight: 0.1
diff --git a/inference/policy/bagel_robotics_1/download_checkpoint.py b/inference/policy/bagel_robotics_1/download_checkpoint.py
new file mode 100644
index 0000000000000000000000000000000000000000..ccd940f8a86ec002daad1fb83061793a262025e1
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/download_checkpoint.py
@@ -0,0 +1,48 @@
+"""Download the pinned existing private CP1000 artifact and public encoder assets."""
+from pathlib import Path
+import argparse
+import json
+from assets import MANIFEST, POLICY_DIR, verify_bundle, verify_wan
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--bundle", type=Path, default=POLICY_DIR / "checkpoints/cp1000")
+    parser.add_argument("--wan-root", type=Path, default=POLICY_DIR / "checkpoints/Wan2.1-I2V-14B-480P")
+    parser.add_argument("--verify-only", action="store_true")
+    parser.add_argument("--metadata-only", action="store_true",
+                        help="Download/check small checkpoint metadata; does not qualify tensor weights")
+    args = parser.parse_args()
+    if args.metadata_only and args.verify_only:
+        parser.error("Choose metadata-only or verify-only")
+    if not args.verify_only:
+        from huggingface_hub import hf_hub_download
+        for filename in MANIFEST["files"]:
+            if args.metadata_only and filename in {"action.pt", "video.pt"}:
+                continue
+            hf_hub_download(MANIFEST["repo_id"], filename, revision=MANIFEST["revision"],
+                            local_dir=args.bundle)
+        if args.metadata_only:
+            from assets import file_digest
+            for name, expected in MANIFEST["files"].items():
+                if name not in {"action.pt", "video.pt"}:
+                    if file_digest(args.bundle / name) != expected["sha256"]:
+                        raise ValueError(f"Metadata mismatch: {name}")
+            print("Pinned metadata verified. Tensor weights and encoders NOT checked.")
+            return
+        prefix = "checkpoints/Wan2.1-I2V-14B-480P/"
+        # Download just the encoders/tokenizer, never the original dense video model.
+        for entry in MANIFEST["wan"]["files"]:
+            source = hf_hub_download(MANIFEST["wan"]["repo"], entry["rfilename"],
+                                     revision=MANIFEST["wan"]["revision"])
+            import shutil
+            destination = args.wan_root / entry["rfilename"].removeprefix(prefix)
+            destination.parent.mkdir(parents=True, exist_ok=True)
+            if not destination.exists():
+                shutil.copyfile(source, destination)
+    print(json.dumps({"checkpoint": verify_bundle(args.bundle),
+                      "encoders": verify_wan(args.wan_root)}, indent=2))
+
+
+if __name__ == "__main__":
+    main()
diff --git a/inference/policy/bagel_robotics_1/download_checkpoint.sh b/inference/policy/bagel_robotics_1/download_checkpoint.sh
new file mode 100644
index 0000000000000000000000000000000000000000..66f21f7224226622aa25037911c5671141ec4873
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/download_checkpoint.sh
@@ -0,0 +1,4 @@
+#!/usr/bin/env bash
+set -euo pipefail
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+exec "${PYTHON_BIN:-python}" "${SCRIPT_DIR}/download_checkpoint.py" "$@"
diff --git a/inference/policy/bagel_robotics_1/eval.sh b/inference/policy/bagel_robotics_1/eval.sh
new file mode 100644
index 0000000000000000000000000000000000000000..99c05abed3ca49e8d1605f946bd5b0b00f9168bc
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/eval.sh
@@ -0,0 +1,68 @@
+#!/bin/bash
+set -euo pipefail
+
+bench_name=${1}
+task_name=${2}
+ckpt_name=${3}
+env_cfg_type=${4}
+action_type=${5}
+seed=${6}
+policy_gpu_id=${7}
+env_gpu_id=${8}
+policy_conda_env=${9}
+eval_env_conda_env=${10}
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+XPL_ROOT="$(cd "${SCRIPT_DIR}/../.." && pwd)"
+BENCH_ROOT="$(cd "${XPL_ROOT}/.." && pwd)"
+UTILS_DIR="${XPL_ROOT}/utils"
+
+SERVER_SCRIPT="${SCRIPT_DIR}/setup_eval_policy_server.sh"
+CLIENT_SCRIPT="${SCRIPT_DIR}/setup_eval_env_client.sh"
+
+policy_server_port=$(bash "${UTILS_DIR}/get_free_port.sh")
+policy_server_ip="localhost"
+
+additional_info="ckpt_name=${ckpt_name},action_type=${action_type}"
+
+cleanup() {
+    if [[ -n "${SERVER_PID:-}" ]]; then
+        echo "[MAIN] kill server ${SERVER_PID}"
+        kill "${SERVER_PID}" 2>/dev/null || true
+    fi
+}
+trap cleanup EXIT
+
+echo "[MAIN] start server, policy_server_port=${policy_server_port}"
+
+bash "${SERVER_SCRIPT}" \
+    "${bench_name}" \
+    "${task_name}" \
+    "${ckpt_name}" \
+    "${env_cfg_type}" \
+    "${action_type}" \
+    "${seed}" \
+    "${policy_gpu_id}" \
+    "${policy_conda_env}" \
+    "${policy_server_port}" &
+
+SERVER_PID=$!
+
+bash "${UTILS_DIR}/wait_for_policy_server.sh" "${policy_server_ip}" "${policy_server_port}" "${SERVER_PID}" "Policy server" 1200
+
+echo "[MAIN] start client, server=${policy_server_ip}:${policy_server_port}"
+
+bash "${CLIENT_SCRIPT}" \
+    "${bench_name}" \
+    "${task_name}" \
+    "${ckpt_name}" \
+    "${env_cfg_type}" \
+    "${action_type}" \
+    "${seed}" \
+    "${env_gpu_id}" \
+    "${eval_env_conda_env}" \
+    "${additional_info}" \
+    "${policy_server_port}" \
+    "${policy_server_ip}"
+
+echo "[MAIN] eval finished"
diff --git a/inference/policy/bagel_robotics_1/evidence/action-comparison.json b/inference/policy/bagel_robotics_1/evidence/action-comparison.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca39bce72c422493afb69a1abaebc7e32e6527d0
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/evidence/action-comparison.json
@@ -0,0 +1,105 @@
+{
+  "passed": true,
+  "checks": {
+    "reference_vs_packaged_0": {
+      "passed": true,
+      "events": [
+        280,
+        280
+      ],
+      "max_abs_stored_array_difference": 0.0,
+      "difference_count": 0,
+      "differences": []
+    },
+    "reference_vs_packaged_1": {
+      "passed": true,
+      "events": [
+        280,
+        280
+      ],
+      "max_abs_stored_array_difference": 0.0,
+      "difference_count": 0,
+      "differences": []
+    },
+    "reference_vs_packaged_2": {
+      "passed": true,
+      "events": [
+        280,
+        280
+      ],
+      "max_abs_stored_array_difference": 0.0,
+      "difference_count": 0,
+      "differences": []
+    },
+    "captured_vs_packaged_0": {
+      "passed": true,
+      "events": [
+        280,
+        280
+      ],
+      "max_abs_stored_array_difference": 0.0,
+      "difference_count": 0,
+      "differences": []
+    },
+    "captured_vs_packaged_1": {
+      "passed": true,
+      "events": [
+        280,
+        280
+      ],
+      "max_abs_stored_array_difference": 0.0,
+      "difference_count": 0,
+      "differences": []
+    },
+    "reference_reset_A_to_B_to_A": {
+      "passed": true,
+      "events": [
+        280,
+        280
+      ],
+      "max_abs_stored_array_difference": 0.0,
+      "difference_count": 0,
+      "differences": []
+    },
+    "packaged_reset_A_to_B_to_A": {
+      "passed": true,
+      "events": [
+        280,
+        280
+      ],
+      "max_abs_stored_array_difference": 0.0,
+      "difference_count": 0,
+      "differences": []
+    },
+    "matching_backend": {
+      "passed": true,
+      "reference": {
+        "torch": "2.11.0+cu130",
+        "cuda": "13.0",
+        "cudnn": 91900,
+        "gpu": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
+        "deterministic": true,
+        "warn_only": false,
+        "cudnn_benchmark": false,
+        "cudnn_deterministic": true,
+        "cudnn_tf32": true,
+        "matmul_tf32": false,
+        "matmul_precision": "highest"
+      },
+      "packaged": {
+        "torch": "2.11.0+cu130",
+        "cuda": "13.0",
+        "cudnn": 91900,
+        "gpu": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
+        "deterministic": true,
+        "warn_only": false,
+        "cudnn_benchmark": false,
+        "cudnn_deterministic": true,
+        "cudnn_tf32": true,
+        "matmul_tf32": false,
+        "matmul_precision": "highest"
+      }
+    }
+  },
+  "scope": "Same observations, seed 1, hardware and backend; not a benchmark score."
+}
diff --git a/inference/policy/bagel_robotics_1/evidence/five-suite-results.json b/inference/policy/bagel_robotics_1/evidence/five-suite-results.json
new file mode 100644
index 0000000000000000000000000000000000000000..32e5a88090efe96f45983aea48a35d7a77121733
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/evidence/five-suite-results.json
@@ -0,0 +1,448 @@
+{
+  "schema": "bagel-cp1000-five-suite-report-v1",
+  "checkpoint": 1000,
+  "display_name": "Bagel-Robotics-1",
+  "official_leaderboard_result": false,
+  "scope": "Research-runner reported results; not packaged-adapter validation",
+  "aggregation": "Equal-weight mean of five suites; Generalization combines standard and randomized episodes",
+  "suites": {
+    "precision": {
+      "scored": 400,
+      "successes": 126,
+      "success_rate": 0.315,
+      "mean_partial_score": 0.40425,
+      "tasks": {
+        "build_tower": {
+          "name": "Build tower",
+          "scored": 50,
+          "successes": 32,
+          "success_rate": 0.64,
+          "mean_partial_score": 0.736,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "deposit_coin": {
+          "name": "Deposit coin",
+          "scored": 50,
+          "successes": 11,
+          "success_rate": 0.22,
+          "mean_partial_score": 0.268,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "fasten_screws": {
+          "name": "Fasten screws",
+          "scored": 50,
+          "successes": 18,
+          "success_rate": 0.36,
+          "mean_partial_score": 0.6679999999999999,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "insert_key": {
+          "name": "Insert key",
+          "scored": 50,
+          "successes": 0,
+          "success_rate": 0,
+          "mean_partial_score": 0.15,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "insert_tubes": {
+          "name": "Insert tubes",
+          "scored": 50,
+          "successes": 32,
+          "success_rate": 0.64,
+          "mean_partial_score": 0.752,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "play_Xylophone": {
+          "name": "Play xylophone",
+          "scored": 50,
+          "successes": 0,
+          "success_rate": 0,
+          "mean_partial_score": 0,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "plug_in_charger": {
+          "name": "Plug in charger",
+          "scored": 50,
+          "successes": 2,
+          "success_rate": 0.04,
+          "mean_partial_score": 0.04,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "pour_balls_into_vase": {
+          "name": "Pour balls into vase",
+          "scored": 50,
+          "successes": 31,
+          "success_rate": 0.62,
+          "mean_partial_score": 0.62,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        }
+      }
+    },
+    "memory": {
+      "scored": 300,
+      "successes": 173,
+      "success_rate": 0.5766666666666667,
+      "mean_partial_score": 0.58,
+      "tasks": {
+        "cover_blocks": {
+          "name": "Cover blocks",
+          "scored": 50,
+          "successes": 50,
+          "success_rate": 1,
+          "mean_partial_score": 1,
+          "source": "user-supplied completed task summary"
+        },
+        "imitate_sorting_sequence": {
+          "name": "Imitate sorting sequence",
+          "scored": 50,
+          "successes": 0,
+          "success_rate": 0,
+          "mean_partial_score": 0.02,
+          "source": "user-supplied completed task summary"
+        },
+        "match_and_pick_from_conveyor": {
+          "name": "Match and pick from conveyor",
+          "scored": 50,
+          "successes": 46,
+          "success_rate": 0.92,
+          "mean_partial_score": 0.92,
+          "source": "user-supplied completed task summary"
+        },
+        "press_by_number": {
+          "name": "Press by number",
+          "scored": 50,
+          "successes": 50,
+          "success_rate": 1,
+          "mean_partial_score": 1,
+          "source": "user-supplied completed task summary"
+        },
+        "swap_T": {
+          "name": "Swap T",
+          "scored": 50,
+          "successes": 11,
+          "success_rate": 0.22,
+          "mean_partial_score": 0.22,
+          "source": "user-supplied completed task summary"
+        },
+        "swap_blocks": {
+          "name": "Swap blocks",
+          "scored": 50,
+          "successes": 16,
+          "success_rate": 0.32,
+          "mean_partial_score": 0.32,
+          "source": "user-supplied completed task summary"
+        }
+      }
+    },
+    "long_horizon": {
+      "scored": 400,
+      "successes": 137,
+      "success_rate": 0.3425,
+      "mean_partial_score": 0.53,
+      "tasks": {
+        "classify_objects": {
+          "name": "Classify objects",
+          "scored": 50,
+          "successes": 16,
+          "success_rate": 0.32,
+          "mean_partial_score": 0.445,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "fill_egg_holder": {
+          "name": "Fill egg holder",
+          "scored": 50,
+          "successes": 2,
+          "success_rate": 0.04,
+          "mean_partial_score": 0.138,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "fill_pen_holder": {
+          "name": "Fill pen holder",
+          "scored": 50,
+          "successes": 11,
+          "success_rate": 0.22,
+          "mean_partial_score": 0.439,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "make_kong": {
+          "name": "Make kong",
+          "scored": 50,
+          "successes": 43,
+          "success_rate": 0.86,
+          "mean_partial_score": 0.86,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "organize_table": {
+          "name": "Organize table",
+          "scored": 50,
+          "successes": 9,
+          "success_rate": 0.18,
+          "mean_partial_score": 0.645,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "play_stacking_toy": {
+          "name": "Play stacking toy",
+          "scored": 50,
+          "successes": 6,
+          "success_rate": 0.12,
+          "mean_partial_score": 0.32,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "play_tic_tac_toe": {
+          "name": "Play tic tac toe",
+          "scored": 50,
+          "successes": 1,
+          "success_rate": 0.02,
+          "mean_partial_score": 0.405,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "put_bottles_into_dustbin": {
+          "name": "Put bottles into dustbin",
+          "scored": 50,
+          "successes": 49,
+          "success_rate": 0.98,
+          "mean_partial_score": 0.988,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        }
+      }
+    },
+    "generalization": {
+      "scored": 600,
+      "successes": 183,
+      "success_rate": 0.305,
+      "mean_partial_score": 0.36824999999999997,
+      "tasks": {
+        "arrange_largest_number": {
+          "name": "Arrange largest number",
+          "scored": 50,
+          "successes": 2,
+          "success_rate": 0.04,
+          "mean_partial_score": 0.11,
+          "source": "user-supplied completed task summary"
+        },
+        "fold_clothes": {
+          "name": "Fold clothes",
+          "scored": 50,
+          "successes": 36,
+          "success_rate": 0.72,
+          "mean_partial_score": 0.7440000000000001,
+          "source": "user-supplied completed task summary"
+        },
+        "hang_mugs": {
+          "name": "Hang mugs",
+          "scored": 50,
+          "successes": 14,
+          "success_rate": 0.28,
+          "mean_partial_score": 0.385,
+          "source": "user-supplied completed task summary"
+        },
+        "make_toast": {
+          "name": "Make toast",
+          "scored": 50,
+          "successes": 22,
+          "success_rate": 0.44,
+          "mean_partial_score": 0.525,
+          "source": "user-supplied completed task summary"
+        },
+        "pack_objects_into_box": {
+          "name": "Pack objects into box",
+          "scored": 50,
+          "successes": 6,
+          "success_rate": 0.12,
+          "mean_partial_score": 0.354,
+          "source": "user-supplied completed task summary"
+        },
+        "pour_liquid_into_cup": {
+          "name": "Pour liquid into cup",
+          "scored": 50,
+          "successes": 22,
+          "success_rate": 0.44,
+          "mean_partial_score": 0.44,
+          "source": "user-supplied completed task summary"
+        },
+        "push_T": {
+          "name": "Push T",
+          "scored": 50,
+          "successes": 1,
+          "success_rate": 0.02,
+          "mean_partial_score": 0.02,
+          "source": "user-supplied completed task summary"
+        },
+        "sort_nesting_dolls_by_size": {
+          "name": "Sort nesting dolls by size",
+          "scored": 50,
+          "successes": 10,
+          "success_rate": 0.2,
+          "mean_partial_score": 0.2,
+          "source": "node-2 receipts verified in prior results review"
+        },
+        "stack_blocks": {
+          "name": "Stack blocks",
+          "scored": 50,
+          "successes": 22,
+          "success_rate": 0.44,
+          "mean_partial_score": 0.5,
+          "source": "node-2 receipts verified in prior results review"
+        },
+        "stack_bowls": {
+          "name": "Stack bowls",
+          "scored": 50,
+          "successes": 26,
+          "success_rate": 0.52,
+          "mean_partial_score": 0.589,
+          "source": "user-supplied completed task summary"
+        },
+        "store_laptop_and_headphones": {
+          "name": "Store laptop and headphones",
+          "scored": 50,
+          "successes": 20,
+          "success_rate": 0.4,
+          "mean_partial_score": 0.512,
+          "source": "node-2 receipts verified in prior results review"
+        },
+        "sweep_blocks": {
+          "name": "Sweep blocks",
+          "scored": 50,
+          "successes": 2,
+          "success_rate": 0.04,
+          "mean_partial_score": 0.04,
+          "source": "user-supplied completed task summary"
+        }
+      }
+    },
+    "open": {
+      "scored": 400,
+      "successes": 16,
+      "success_rate": 0.04,
+      "mean_partial_score": 0.044750000000000005,
+      "tasks": {
+        "align_blocks": {
+          "name": "Align blocks",
+          "scored": 50,
+          "successes": 0,
+          "success_rate": 0,
+          "mean_partial_score": 0,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "general_pickup": {
+          "name": "General pickup",
+          "scored": 50,
+          "successes": 12,
+          "success_rate": 0.24,
+          "mean_partial_score": 0.24,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "stack_blocks_by_language": {
+          "name": "Stack blocks by language",
+          "scored": 50,
+          "successes": 4,
+          "success_rate": 0.08,
+          "mean_partial_score": 0.092,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "solve_equation": {
+          "name": "Solve equation",
+          "scored": 50,
+          "successes": 0,
+          "success_rate": 0,
+          "mean_partial_score": 0,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "classify_objects_by_language": {
+          "name": "Classify objects by language",
+          "scored": 50,
+          "successes": 0,
+          "success_rate": 0,
+          "mean_partial_score": 0.026000000000000002,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "pick_from_conveyor_by_image": {
+          "name": "Pick from conveyor by image",
+          "scored": 50,
+          "successes": 0,
+          "success_rate": 0,
+          "mean_partial_score": 0,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "store_tools_in_toolbox": {
+          "name": "Store tools in toolbox",
+          "scored": 50,
+          "successes": 0,
+          "success_rate": 0,
+          "mean_partial_score": 0,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        },
+        "pour_by_language": {
+          "name": "Pour by language",
+          "scored": 50,
+          "successes": 0,
+          "success_rate": 0,
+          "mean_partial_score": 0,
+          "source": "saved checkpoint-1000 research evidence with receipt hashes"
+        }
+      }
+    }
+  },
+  "provenance_note": "Memory and nine Generalization tasks use user-supplied summaries; raw episode records for those tasks were not independently re-audited. The other three Generalization tasks were verified against node-2 receipts in the preceding results review.",
+  "exceptions": [
+    {
+      "suite": "memory",
+      "type": "repaired_variant",
+      "original_layout": 47,
+      "original_status": "unscored",
+      "variant_success": false,
+      "variant_partial_score": 0,
+      "reported_scored": 300,
+      "original_scored": 299,
+      "original_successes": 173,
+      "original_mean_partial_score": 0.5819397993311036
+    },
+    {
+      "suite": "open",
+      "type": "replacement",
+      "task": "general_pickup",
+      "original_layouts": [
+        5
+      ],
+      "replacement_layouts": [
+        50
+      ],
+      "replacement_successes": [
+        false
+      ],
+      "replacement_partial_scores": [
+        0
+      ]
+    },
+    {
+      "suite": "long_horizon",
+      "type": "replacement",
+      "task": "fill_egg_holder",
+      "original_layouts": [
+        15,
+        20
+      ],
+      "replacement_layouts": [
+        50,
+        51
+      ],
+      "replacement_successes": [
+        false,
+        false
+      ],
+      "replacement_partial_scores": [
+        0,
+        0.25
+      ]
+    }
+  ],
+  "overall": {
+    "success_rate": 0.3158333333333333,
+    "mean_partial_score": 0.38545,
+    "reported_scored": 2100,
+    "successes": 635,
+    "original_results": 2096,
+    "replacements": 3,
+    "repaired_variants": 1
+  }
+}
diff --git a/inference/policy/bagel_robotics_1/evidence/gpu-validation.json b/inference/policy/bagel_robotics_1/evidence/gpu-validation.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce1d916a4950c7f6906fd7073e3d4b2d24e29eaa
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/evidence/gpu-validation.json
@@ -0,0 +1,168 @@
+{
+  "passed": true,
+  "date_utc": "2026-11-06",
+  "public_model_name": "Bagel-Robotics-1",
+  "checkpoint": {
+    "update": 1000,
+    "identity": "790bffc9464fa7a026aa43b217aec2033d186895cad643166daa0bcc01914d5c",
+    "revision": "2651631c5d08819088475e8778676f5cf6d9afcf"
+  },
+  "source_revisions": {
+    "simulator": "f57e1633ffed59761bc8e568cfe7530c4b33381c",
+    "xpolicylab": "6baccd57ff958570d0db1588370289ee36c8bb70"
+  },
+  "adapter_code_integrity": {
+    "passed": true,
+    "files_checked": 57,
+    "manifest_sha256": "a97b59051c7c5c1652a21c2f0e60b745114551004234f46015ab1c1fb4aa32ad"
+  },
+  "debug": {
+    "passed": true,
+    "plain": true,
+    "encoded": true,
+    "job_id": "53609"
+  },
+  "simulator": {
+    "passed": true,
+    "seed": 1,
+    "errors": [],
+    "scope": "Adapter smoke test; not a complete suite or official score."
+  },
+  "simulator_tasks": {
+    "insert_tubes": {
+      "episodes": 2,
+      "successes": 2,
+      "success_rate_percent": 100,
+      "partial_score_100": 100.0,
+      "layouts": [
+        0,
+        1
+      ],
+      "details": [
+        {
+          "layout_id": 0,
+          "success": true,
+          "score": 1.0
+        },
+        {
+          "layout_id": 1,
+          "success": true,
+          "score": 1.0
+        }
+      ],
+      "result_sha256": "52f265125beb3fa9d1c040451392dd0762a9561d1105825712fda3fc84e10243",
+      "video_files": 6
+    },
+    "store_laptop_and_headphones": {
+      "episodes": 2,
+      "successes": 1,
+      "success_rate_percent": 50,
+      "partial_score_100": 60.0,
+      "layouts": [
+        0,
+        1
+      ],
+      "details": [
+        {
+          "layout_id": 0,
+          "success": true,
+          "score": 1.0
+        },
+        {
+          "layout_id": 1,
+          "success": false,
+          "score": 0.2
+        }
+      ],
+      "result_sha256": "95e73d2f069cf0f3d9418e654206afda6a6670ec8a9b8661d232299349614f55",
+      "video_files": 6
+    }
+  },
+  "replay": {
+    "passed": true,
+    "fixtures": [
+      {
+        "task": "insert_tubes",
+        "coverage": {
+          "observations": 209,
+          "action_chunks": 14,
+          "history_mature": true,
+          "capture_limit": 209
+        },
+        "trace_sha256": "e4956cf7959ea7fa0f261d8395e6c6387b1329dcbcfe99e2e3dd200b5632217f"
+      },
+      {
+        "task": "store_laptop_and_headphones",
+        "coverage": {
+          "observations": 209,
+          "action_chunks": 14,
+          "history_mature": true,
+          "capture_limit": 209
+        },
+        "trace_sha256": "5174d8b3ac7565a38e913813c4c9d3f4a58077ac2dffd0276d1383781ca011e1"
+      }
+    ],
+    "order": [
+      "A",
+      "B",
+      "A"
+    ],
+    "observations_per_run": 209,
+    "chunks_per_run": 14,
+    "runs_per_implementation": 3,
+    "checks": 8,
+    "recorded_differences": 0,
+    "max_abs_stored_array_difference": 0.0
+  },
+  "jobs": [
+    {
+      "JobID": "53541",
+      "State": "FAILED",
+      "ExitCode": "1:0",
+      "Elapsed": "00:01:13"
+    },
+    {
+      "JobID": "53542",
+      "State": "CANCELLED",
+      "ExitCode": "0:0",
+      "Elapsed": "00:00:00"
+    },
+    {
+      "JobID": "53609",
+      "State": "COMPLETED",
+      "ExitCode": "0:0",
+      "Elapsed": "00:02:20"
+    },
+    {
+      "JobID": "53610",
+      "State": "FAILED",
+      "ExitCode": "1:0",
+      "Elapsed": "00:01:08"
+    },
+    {
+      "JobID": "53611",
+      "State": "FAILED",
+      "ExitCode": "55:0",
+      "Elapsed": "00:01:09"
+    },
+    {
+      "JobID": "53612",
+      "State": "COMPLETED",
+      "ExitCode": "0:0",
+      "Elapsed": "00:15:45"
+    }
+  ],
+  "scope": "Technical package qualification on the recorded backend. Four smoke episodes are separate from historical suite results; this is not official leaderboard evaluation.",
+  "simulator_launch_adjustments": [
+    "Established noninteractive Kit setting",
+    "User-owned portable Kit cache via a recorded launcher-only patch",
+    "Read-path bridge to native eval_result; no result contents changed"
+  ],
+  "receipts": {
+    "simulator-result.json": "360a928cbffca156e08f461a9b5a810cc648a9c4de72121089563e21b60937c3",
+    "action-comparison.json": "6110ad701ed4c7dd9768b61add6df64ace15c7109ccafdace4ea8ab8934b23cb",
+    "gpu-validation-result.json": "46b138b64700468bdef996abd972395502cddaf2ca25eb0dccd39b710345b070",
+    "adapter-integrity-after.json": "4b463b710cd3f1724789121005167f8b22a369ddd7c151f9a5501e448fb2cbae",
+    "slurm-final.json": "766906b88c3153fb1abb641579302b9639234ccac995015bcb58ead33622344d"
+  }
+}
diff --git a/inference/policy/bagel_robotics_1/evidence/local-results.json b/inference/policy/bagel_robotics_1/evidence/local-results.json
new file mode 100644
index 0000000000000000000000000000000000000000..63b90a262e4205cacdbbab94f95ca135e4e66d45
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/evidence/local-results.json
@@ -0,0 +1,569 @@
+{
+  "checkpoint": 1000,
+  "seed": 1,
+  "official_leaderboard_result": false,
+  "suites": {
+    "precision": {
+      "source_file": "progress-001000.json",
+      "source_sha256": "099fa38f19d9f5ff428695ca9999fa5c291081f1519c8320f2101a86c03ff6ef",
+      "original": {
+        "complete": true,
+        "expected": 400,
+        "mean_partial_score": 0.40425000000000033,
+        "protocol": "266df4b5d9b8eac16145600a8c4f9223d15e03285d61473c3bbd8c43b694f211",
+        "scored": 400,
+        "success_rate": 0.315,
+        "successes": 126,
+        "tasks": {
+          "build_tower": {
+            "complete": true,
+            "expected": 50,
+            "mean_partial_score": 0.7360000000000002,
+            "scored": 50,
+            "success_rate": 0.64,
+            "successes": 32
+          },
+          "deposit_coin": {
+            "complete": true,
+            "expected": 50,
+            "mean_partial_score": 0.2679999999999999,
+            "scored": 50,
+            "success_rate": 0.22,
+            "successes": 11
+          },
+          "fasten_screws": {
+            "complete": true,
+            "expected": 50,
+            "mean_partial_score": 0.6679999999999999,
+            "scored": 50,
+            "success_rate": 0.36,
+            "successes": 18
+          },
+          "insert_key": {
+            "complete": true,
+            "expected": 50,
+            "mean_partial_score": 0.15000000000000013,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          },
+          "insert_tubes": {
+            "complete": true,
+            "expected": 50,
+            "mean_partial_score": 0.7519999999999999,
+            "scored": 50,
+            "success_rate": 0.64,
+            "successes": 32
+          },
+          "play_Xylophone": {
+            "complete": true,
+            "expected": 50,
+            "mean_partial_score": 0.0,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          },
+          "plug_in_charger": {
+            "complete": true,
+            "expected": 50,
+            "mean_partial_score": 0.04,
+            "scored": 50,
+            "success_rate": 0.04,
+            "successes": 2
+          },
+          "pour_balls_into_vase": {
+            "complete": true,
+            "expected": 50,
+            "mean_partial_score": 0.62,
+            "scored": 50,
+            "success_rate": 0.62,
+            "successes": 31
+          }
+        },
+        "update": 1000
+      },
+      "reported_with_labeled_replacements": {
+        "scored": 400,
+        "successes": 126,
+        "success_rate": 0.315,
+        "mean_partial_score": 0.40425000000000033,
+        "replacement_count": 0,
+        "tasks": {
+          "build_tower": {
+            "mean_partial_score": 0.7360000000000002,
+            "scored": 50,
+            "success_rate": 0.64,
+            "successes": 32
+          },
+          "deposit_coin": {
+            "mean_partial_score": 0.2679999999999999,
+            "scored": 50,
+            "success_rate": 0.22,
+            "successes": 11
+          },
+          "fasten_screws": {
+            "mean_partial_score": 0.6679999999999999,
+            "scored": 50,
+            "success_rate": 0.36,
+            "successes": 18
+          },
+          "insert_key": {
+            "mean_partial_score": 0.15000000000000013,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          },
+          "insert_tubes": {
+            "mean_partial_score": 0.7519999999999999,
+            "scored": 50,
+            "success_rate": 0.64,
+            "successes": 32
+          },
+          "play_Xylophone": {
+            "mean_partial_score": 0.0,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          },
+          "plug_in_charger": {
+            "mean_partial_score": 0.04,
+            "scored": 50,
+            "success_rate": 0.04,
+            "successes": 2
+          },
+          "pour_balls_into_vase": {
+            "mean_partial_score": 0.62,
+            "scored": 50,
+            "success_rate": 0.62,
+            "successes": 31
+          }
+        }
+      }
+    },
+    "open": {
+      "source_file": "progress.json",
+      "source_sha256": "32e307e43f14291eb654f11d457be3bd29c28d0e7e6d3bc132de8d26eec8f3c3",
+      "original": {
+        "accounted": 399,
+        "expected": 400,
+        "mean_partial_score": 0.04486215538847117,
+        "native_complete": false,
+        "reporting_complete": false,
+        "scored": 399,
+        "success_rate": 0.040100250626566414,
+        "successes": 16,
+        "suite": "open",
+        "tasks": {
+          "align_blocks": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.0,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0,
+            "unscored": 0,
+            "valid_failures": 50
+          },
+          "classify_objects_by_language": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.026000000000000002,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0,
+            "unscored": 0,
+            "valid_failures": 50
+          },
+          "general_pickup": {
+            "accounted": 49,
+            "expected": 50,
+            "mean_partial_score": 0.24489795918367346,
+            "native_complete": false,
+            "reporting_complete": false,
+            "scored": 49,
+            "success_rate": 0.24489795918367346,
+            "successes": 12,
+            "unscored": 0,
+            "valid_failures": 37
+          },
+          "pick_from_conveyor_by_image": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.0,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0,
+            "unscored": 0,
+            "valid_failures": 50
+          },
+          "pour_by_language": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.0,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0,
+            "unscored": 0,
+            "valid_failures": 50
+          },
+          "solve_equation": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.0,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0,
+            "unscored": 0,
+            "valid_failures": 50
+          },
+          "stack_blocks_by_language": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.092,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.08,
+            "successes": 4,
+            "unscored": 0,
+            "valid_failures": 46
+          },
+          "store_tools_in_toolbox": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.0,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0,
+            "unscored": 0,
+            "valid_failures": 50
+          }
+        },
+        "unscored": 0,
+        "update": 1000,
+        "valid_failures": 383,
+        "variants": {
+          "standard": {
+            "accounted": 399,
+            "expected": 400,
+            "mean_partial_score": 0.04486215538847117,
+            "native_complete": false,
+            "reporting_complete": false,
+            "scored": 399,
+            "success_rate": 0.040100250626566414,
+            "successes": 16,
+            "unscored": 0,
+            "valid_failures": 383
+          }
+        }
+      },
+      "reported_with_labeled_replacements": {
+        "scored": 400,
+        "successes": 16,
+        "success_rate": 0.04,
+        "mean_partial_score": 0.04475,
+        "replacement_count": 1,
+        "tasks": {
+          "align_blocks": {
+            "mean_partial_score": 0.0,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          },
+          "classify_objects_by_language": {
+            "mean_partial_score": 0.026000000000000002,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          },
+          "general_pickup": {
+            "mean_partial_score": 0.24,
+            "scored": 50,
+            "success_rate": 0.24,
+            "successes": 12
+          },
+          "pick_from_conveyor_by_image": {
+            "mean_partial_score": 0.0,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          },
+          "pour_by_language": {
+            "mean_partial_score": 0.0,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          },
+          "solve_equation": {
+            "mean_partial_score": 0.0,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          },
+          "stack_blocks_by_language": {
+            "mean_partial_score": 0.092,
+            "scored": 50,
+            "success_rate": 0.08,
+            "successes": 4
+          },
+          "store_tools_in_toolbox": {
+            "mean_partial_score": 0.0,
+            "scored": 50,
+            "success_rate": 0.0,
+            "successes": 0
+          }
+        }
+      }
+    },
+    "long_horizon": {
+      "source_file": "progress.json",
+      "source_sha256": "9bc9c571ee6099957f954be58e5cf7654a1439975389277b8c29d036b3ad1e48",
+      "original": {
+        "accounted": 398,
+        "expected": 400,
+        "mean_partial_score": 0.532035175879397,
+        "native_complete": false,
+        "reporting_complete": false,
+        "scored": 398,
+        "success_rate": 0.3442211055276382,
+        "successes": 137,
+        "suite": "long_horizon",
+        "tasks": {
+          "classify_objects": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.445,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.32,
+            "successes": 16,
+            "unscored": 0,
+            "valid_failures": 34
+          },
+          "fill_egg_holder": {
+            "accounted": 48,
+            "expected": 50,
+            "mean_partial_score": 0.13854166666666667,
+            "native_complete": false,
+            "reporting_complete": false,
+            "scored": 48,
+            "success_rate": 0.041666666666666664,
+            "successes": 2,
+            "unscored": 0,
+            "valid_failures": 46
+          },
+          "fill_pen_holder": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.439,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.22,
+            "successes": 11,
+            "unscored": 0,
+            "valid_failures": 39
+          },
+          "make_kong": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.86,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.86,
+            "successes": 43,
+            "unscored": 0,
+            "valid_failures": 7
+          },
+          "organize_table": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.645,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.18,
+            "successes": 9,
+            "unscored": 0,
+            "valid_failures": 41
+          },
+          "play_stacking_toy": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.32,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.12,
+            "successes": 6,
+            "unscored": 0,
+            "valid_failures": 44
+          },
+          "play_tic_tac_toe": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.405,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.02,
+            "successes": 1,
+            "unscored": 0,
+            "valid_failures": 49
+          },
+          "put_bottles_into_dustbin": {
+            "accounted": 50,
+            "expected": 50,
+            "mean_partial_score": 0.988,
+            "native_complete": true,
+            "reporting_complete": true,
+            "scored": 50,
+            "success_rate": 0.98,
+            "successes": 49,
+            "unscored": 0,
+            "valid_failures": 1
+          }
+        },
+        "unscored": 0,
+        "update": 1000,
+        "valid_failures": 261,
+        "variants": {
+          "standard": {
+            "accounted": 398,
+            "expected": 400,
+            "mean_partial_score": 0.532035175879397,
+            "native_complete": false,
+            "reporting_complete": false,
+            "scored": 398,
+            "success_rate": 0.3442211055276382,
+            "successes": 137,
+            "unscored": 0,
+            "valid_failures": 261
+          }
+        }
+      },
+      "reported_with_labeled_replacements": {
+        "scored": 400,
+        "successes": 137,
+        "success_rate": 0.3425,
+        "mean_partial_score": 0.53,
+        "replacement_count": 2,
+        "tasks": {
+          "classify_objects": {
+            "mean_partial_score": 0.445,
+            "scored": 50,
+            "success_rate": 0.32,
+            "successes": 16
+          },
+          "fill_egg_holder": {
+            "mean_partial_score": 0.138,
+            "scored": 50,
+            "success_rate": 0.04,
+            "successes": 2
+          },
+          "fill_pen_holder": {
+            "mean_partial_score": 0.439,
+            "scored": 50,
+            "success_rate": 0.22,
+            "successes": 11
+          },
+          "make_kong": {
+            "mean_partial_score": 0.86,
+            "scored": 50,
+            "success_rate": 0.86,
+            "successes": 43
+          },
+          "organize_table": {
+            "mean_partial_score": 0.645,
+            "scored": 50,
+            "success_rate": 0.18,
+            "successes": 9
+          },
+          "play_stacking_toy": {
+            "mean_partial_score": 0.32,
+            "scored": 50,
+            "success_rate": 0.12,
+            "successes": 6
+          },
+          "play_tic_tac_toe": {
+            "mean_partial_score": 0.405,
+            "scored": 50,
+            "success_rate": 0.02,
+            "successes": 1
+          },
+          "put_bottles_into_dustbin": {
+            "mean_partial_score": 0.988,
+            "scored": 50,
+            "success_rate": 0.98,
+            "successes": 49
+          }
+        }
+      }
+    }
+  },
+  "replacements": [
+    {
+      "suite": "open",
+      "episode": "general_pickup--0050",
+      "outcome": {
+        "executed_steps": 200,
+        "layout_id": 50,
+        "score": 0.0,
+        "step_budget": 200,
+        "success": false,
+        "termination_reason": "step_budget"
+      },
+      "receipt_sha256": "4f2a6cebf43dc12ea30ee63981e720e7afcec93b03ad3f6a4ba460c59f2d9cf4",
+      "protocol": "e32aa0e95d1bdfe638c65520e8512c1e0da1cab16744ac560ba2c8ad59dcbc5d",
+      "source_sha256": "a9c67e0796355fb49b13de78212bb5ef0983d3e2d18bb0480d0c38f2bcbe1eae",
+      "source_file": "replacement-result.json"
+    },
+    {
+      "suite": "long_horizon",
+      "episode": "fill_egg_holder--0050",
+      "outcome": {
+        "executed_steps": 700,
+        "layout_id": 50,
+        "score": 0.0,
+        "step_budget": 700,
+        "success": false,
+        "termination_reason": "step_budget"
+      },
+      "receipt_sha256": "e997fd981a5855be673538943400865788b20f4850d0da684d9752ff4778bafa",
+      "protocol": "a876503909848ac16a62806bed0177445f8ee0681970f82f1d49150a1ab5b5b6",
+      "source_sha256": "af24963ead6a0db6175e26b51df2a551e3a3b7c4854659c8b7995ff24a4fb927",
+      "source_file": "fill_egg_holder--0050-result.json"
+    },
+    {
+      "suite": "long_horizon",
+      "episode": "fill_egg_holder--0051",
+      "outcome": {
+        "executed_steps": 700,
+        "layout_id": 51,
+        "score": 0.25,
+        "step_budget": 700,
+        "success": false,
+        "termination_reason": "step_budget"
+      },
+      "receipt_sha256": "8a9be90e270a00edd75615061418570fe83af25e28eb742ce2548d10b792713f",
+      "protocol": "a876503909848ac16a62806bed0177445f8ee0681970f82f1d49150a1ab5b5b6",
+      "source_sha256": "48566afa4d9dd6f1c53c601974cc583fc4c519a57c96c94515e361ba1daf03b0",
+      "source_file": "fill_egg_holder--0051-result.json"
+    }
+  ],
+  "note": "Local standard-layout results. Open/Long Horizon include labeled startup-invalid replacements. No five-suite benchmark aggregate or packaged-adapter rollout claim."
+}
diff --git a/inference/policy/bagel_robotics_1/evidence/pr-audit.json b/inference/policy/bagel_robotics_1/evidence/pr-audit.json
new file mode 100644
index 0000000000000000000000000000000000000000..ad9d686daefda3c4c8575af5ee9a28326e2557a0
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/evidence/pr-audit.json
@@ -0,0 +1,20 @@
+{
+  "passed": true,
+  "errors": [],
+  "python_files": 50,
+  "shell_scripts": 5,
+  "cpu_tests": {
+    "tests": 15,
+    "failures": 0,
+    "errors": 0,
+    "skipped": 0
+  },
+  "pr_template_headings_exact": true,
+  "gpu_qualification": "passed",
+  "ready_for_pr": false,
+  "remaining_gates": [
+    "dedicated private reviewer access",
+    "maintainer agreement on proposed eval-only integration and release timeline",
+    "user review before PR creation"
+  ]
+}
diff --git a/inference/policy/bagel_robotics_1/install.sh b/inference/policy/bagel_robotics_1/install.sh
new file mode 100644
index 0000000000000000000000000000000000000000..655eecb24e443f96698966843a7a2dbc3c70adc1
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/install.sh
@@ -0,0 +1,11 @@
+#!/usr/bin/env bash
+set -euo pipefail
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+XPL_ROOT="$(cd "${SCRIPT_DIR}/../.." && pwd)"
+ENV_DIR="${1:-${SCRIPT_DIR}/.venv}"
+"${PYTHON_BIN:-python3}" -m venv "${ENV_DIR}"
+"${ENV_DIR}/bin/python" -m pip install --upgrade pip
+"${ENV_DIR}/bin/python" -m pip install torch==2.11.0 torchvision==0.26.0 \
+  --index-url "${TORCH_INDEX_URL:-https://download.pytorch.org/whl/cu130}"
+"${ENV_DIR}/bin/python" -m pip install -c "${SCRIPT_DIR}/requirements-transport.txt" -e "${SCRIPT_DIR}/runtime" -e "${XPL_ROOT}"
+echo "Policy environment: ${ENV_DIR}"
diff --git a/inference/policy/bagel_robotics_1/licenses/Apache-2.0.txt b/inference/policy/bagel_robotics_1/licenses/Apache-2.0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..261eeb9e9f8b2b4b0d119366dda99c6fd7d35c64
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/licenses/Apache-2.0.txt
@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
diff --git a/inference/policy/bagel_robotics_1/licenses/RoboDojo-MIT.txt b/inference/policy/bagel_robotics_1/licenses/RoboDojo-MIT.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2d51370cc56c4c2b931be99bc3ced43566393741
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/licenses/RoboDojo-MIT.txt
@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Yue Chen <yuechen020614@gmail.com>
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
diff --git a/inference/policy/bagel_robotics_1/licenses/XPolicyLab.txt b/inference/policy/bagel_robotics_1/licenses/XPolicyLab.txt
new file mode 100644
index 0000000000000000000000000000000000000000..261eeb9e9f8b2b4b0d119366dda99c6fd7d35c64
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/licenses/XPolicyLab.txt
@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
diff --git a/inference/policy/bagel_robotics_1/model.py b/inference/policy/bagel_robotics_1/model.py
new file mode 100644
index 0000000000000000000000000000000000000000..9d6c28561a98a68d8485a70f6052a2e01288631c
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/model.py
@@ -0,0 +1,361 @@
+import os
+from pathlib import Path
+from typing import Any
+import cv2
+import numpy as np
+from .temporal_ensemble import TemporalEnsemble
+from XPolicyLab.model_template import ModelTemplate
+from XPolicyLab.utils.process_data import get_robot_action_dim_info
+from .assets import prepare_model_config
+from .backend import configure_backend
+
+POLICY_DIR = Path(__file__).resolve().parent
+
+
+def _is_none_like(value: Any) -> bool:
+    if value is None:
+        return True
+    return isinstance(value, str) and value.strip().lower() in {"", "none", "null"}
+
+
+def _standardize_rgb(image: np.ndarray, *, resize: bool = True) -> np.ndarray:
+    image = np.asarray(image)
+    if image.ndim != 3 or image.shape[-1] != 3:
+        raise ValueError(f"Expected HWC image with 3 channels, got {image.shape}")
+    if image.dtype != np.uint8:
+        image = np.clip(image, 0, 255).astype(np.uint8)
+    if resize and image.shape[:2] != (240, 320):
+        image = cv2.resize(image, (320, 240), interpolation=cv2.INTER_AREA)
+    if resize and image.shape != (240, 320, 3):
+        raise ValueError(f"Expected standardized RGB shape (240, 320, 3), got {image.shape}")
+    return image
+
+
+def _get_instruction(obs: dict, fallback: str) -> str:
+    value = obs.get("task_instruction")
+    if value is None:
+        value = obs.get("instruction", obs.get("instructions"))
+    if isinstance(value, (list, tuple)):
+        return str(value[0]) if value else fallback
+    if value is None:
+        return fallback
+    if hasattr(value, "item"):
+        value = value.item()
+    text = str(value).strip()
+    return text if text else fallback
+
+
+class Model(ModelTemplate):
+    def __init__(self, model_cfg):
+        self.model_cfg = prepare_model_config(model_cfg, POLICY_DIR)
+        self.backend_settings = configure_backend(self.model_cfg["seed"])
+        info = get_robot_action_dim_info(self.model_cfg["env_cfg_type"])
+        if len(info["arm_dim"]) != 2 or list(info["ee_dim"]) != [1, 1]:
+            raise ValueError("The checkpoint requires two arms and one gripper value per arm")
+        self.robot_action_dim_info = info
+        self.action_dim = sum(7 + int(d) for d in info["ee_dim"])
+        if self.action_dim != 16:
+            raise ValueError("Checkpoint action width differs from the registered robot")
+        self.action_type = self.model_cfg["action_type"]
+        if (
+            self.action_type != "ee"
+            or self.model_cfg.get("deployment_adapter", "robodojo_ee16") != "robodojo_ee16"
+        ):
+            raise ValueError("This adapter only supports RoboDojo EE16")
+        self.env_cfg_type = self.model_cfg["env_cfg_type"]
+        self.action_horizon = 1
+        self.replan_steps = int(self.model_cfg.get("replan_steps") or 16)
+        self.default_instruction = str(
+            self.model_cfg.get("default_instruction")
+            or self.model_cfg.get("prompt")
+            or "follow the instruction"
+        )
+        self.deployment_adapter = (
+            str(self.model_cfg.get("deployment_adapter") or "robodojo_ee16").strip().lower()
+        )
+        self.last_obs = None
+        self.last_instruction = self.default_instruction
+        self.model = None
+        self._request_context = {}
+        self._virtual_env_ids = {}
+        self._next_virtual_env_id = 0
+        self._client_last_use = {}
+        self._client_use_counter = 0
+        self._max_retained_clients = int(self.model_cfg.get("max_retained_clients") or 64)
+        self._last_infer_step = {}
+        self.history_guard_duplicates = 0
+        self.history_guard_mismatches = 0
+        self._active_client = None
+        self._seen_client_ids = False
+        self._last_obs_by_client = {}
+        self._last_instruction_by_client = {}
+        self._last_virtual_by_client = {}
+        self._infer_count_by_virtual = {}
+        self._ensemble_by_virtual = {}
+        self._last_ensemble = None
+        checkpoint_path = self.model_cfg.get("checkpoint_path") or self.model_cfg.get(
+            "ckpt_setting"
+        )
+        dataset_stats_path = self.model_cfg.get("dataset_stats_path")
+        if _is_none_like(checkpoint_path):
+            raise FileNotFoundError(
+                "VPP2 requires checkpoint_path/ckpt_setting for real deployment."
+            )
+        if _is_none_like(dataset_stats_path):
+            raise FileNotFoundError("VPP2 requires dataset_stats_path for real deployment.")
+        from bagel_cp1000.policy import get_model
+
+        upstream_cfg = dict(self.model_cfg)
+        upstream_cfg["ckpt_setting"] = str(Path(checkpoint_path).expanduser().resolve())
+        upstream_cfg["dataset_stats_path"] = str(Path(dataset_stats_path).expanduser().resolve())
+        self.model = get_model(upstream_cfg)
+        video = self.model.model.video_expert
+        if sum(p.numel() for p in video.parameters()) != 12468409664:
+            raise ValueError("Loaded video expert is not the pinned compact CP1000 model")
+        if any("mask_logits" in name for name, _ in video.named_parameters()):
+            raise ValueError("Training mask parameters must not be present at inference")
+        self.action_horizon = int(self.model.action_horizon)
+        self.replan_steps = int(self.model.replan_steps)
+        if self.model_cfg.get("temporal_ensemble") != {"enabled": True, "new_weight": 0.9, "old_weight": 0.1}:
+            raise ValueError("Expected the authorized 90/10 temporal ensemble")
+        if (self.action_horizon, self.replan_steps) != (32, 16):
+            raise ValueError("Expected horizon 32, execute 16")
+
+    @staticmethod
+    def _pack_robodojo_ee16_state(obs: dict) -> np.ndarray:
+        state = obs["state"]
+        parts = []
+        for side in ("left", "right"):
+            pose = np.asarray(state[f"{side}_ee_pose"], dtype=np.float32).reshape(7)
+            gripper = np.asarray(state[f"{side}_ee_joint_state"], dtype=np.float32).reshape(1)
+            parts.extend((pose, gripper))
+        packed = np.concatenate(parts, axis=0)
+        if packed.shape != (16,):
+            raise ValueError(f"Expected RoboDojo EE state [16], got {packed.shape}")
+        return packed
+
+    @staticmethod
+    def _unpack_robodojo_ee16_actions(actions: np.ndarray):
+        packed = np.asarray(actions, dtype=np.float32)
+        if packed.ndim == 1:
+            packed = packed[None, :]
+        if packed.ndim != 2 or packed.shape[1] != 16:
+            raise ValueError(f"Expected RoboDojo EE actions [T,16], got {packed.shape}")
+        result = []
+        for action in packed:
+            result.append(
+                {
+                    "left_ee_pose": action[0:7],
+                    "left_ee_joint_state": action[7:8],
+                    "right_ee_pose": action[8:15],
+                    "right_ee_joint_state": action[15:16],
+                }
+            )
+        return result
+
+    def _encode_observation(self, obs: dict) -> dict:
+        vision = obs["vision"]
+        adapted = {
+            "env_idx": self._virtual_env_idx(int(obs.get("env_idx", 0))),
+            "observation": {
+                "head_camera": {
+                    "rgb": _standardize_rgb(vision["cam_head"]["color"], resize=False)
+                },
+                "left_camera": {
+                    "rgb": _standardize_rgb(vision["cam_left_wrist"]["color"], resize=False)
+                },
+                "right_camera": {
+                    "rgb": _standardize_rgb(vision["cam_right_wrist"]["color"], resize=False)
+                },
+            },
+            "joint_action": {"vector": self._pack_robodojo_ee16_state(obs)},
+        }
+        return adapted
+
+    CLIENT_ID_FIELD = "_vpp2_client_id"
+
+    def _bind_client(self, obs) -> None:
+        client_id = obs.get(self.CLIENT_ID_FIELD) if isinstance(obs, dict) else None
+        if client_id is None:
+            self._active_client = None
+            return
+        self._seen_client_ids = True
+        self._active_client = ("client", str(client_id))
+
+    def _client_key(self):
+        if getattr(self, "_active_client", None) is not None:
+            return self._active_client
+        context = getattr(self, "_request_context", None) or {}
+        return (context.get("evaluation_id"), context.get("trial_id"))
+
+    def _forget_virtual(self, virtual: int) -> None:
+        self._last_infer_step.pop(virtual, None)
+        self._infer_count_by_virtual.pop(virtual, None)
+        self._ensemble_by_virtual.pop(virtual, None)
+        self._last_ensemble = None
+        if self.model is not None and hasattr(self.model, "reset_env"):
+            self.model.reset_env(virtual)
+
+    def _virtual_env_idx(self, env_idx: int) -> int:
+        client = self._client_key()
+        self._client_use_counter += 1
+        self._client_last_use[client] = self._client_use_counter
+        key = (client, int(env_idx))
+        virtual = self._virtual_env_ids.get(key)
+        if virtual is None:
+            virtual = self._next_virtual_env_id
+            self._next_virtual_env_id += 1
+            self._virtual_env_ids[key] = virtual
+            self._evict_stale_clients()
+        return virtual
+
+    def _evict_stale_clients(self) -> None:
+        clients = {key[0] for key in self._virtual_env_ids}
+        excess = len(clients) - self._max_retained_clients
+        if excess <= 0:
+            return
+        current = self._client_key()
+        stale = sorted(
+            (c for c in clients if c != current), key=lambda c: self._client_last_use.get(c, 0)
+        )[:excess]
+        for client in stale:
+            for key in [k for k in self._virtual_env_ids if k[0] == client]:
+                self._forget_virtual(self._virtual_env_ids.pop(key))
+            self._client_last_use.pop(client, None)
+            self._last_obs_by_client.pop(client, None)
+            self._last_instruction_by_client.pop(client, None)
+            self._last_virtual_by_client.pop(client, None)
+
+    def set_request_context(self, context):
+        self._request_context = dict(context or {})
+        if self.model is not None and hasattr(self.model, "set_request_context"):
+            self.model.set_request_context(self._request_context)
+
+    def _remember(self, encoded, raw_obs) -> None:
+        client = self._client_key()
+        self.last_obs = encoded
+        self.last_instruction = _get_instruction(raw_obs, self.default_instruction)
+        self._last_obs_by_client[client] = self.last_obs
+        self._last_instruction_by_client[client] = self.last_instruction
+        self._last_virtual_by_client[client] = encoded["env_idx"]
+
+    def update_obs(self, obs):
+        self._bind_client(obs)
+        observation_history = obs.get("_vpp2_observation_history")
+        if observation_history:
+            step = (self._request_context or {}).get("step")
+            virtual = self._virtual_env_idx(int(obs.get("env_idx", 0)))
+            if step is not None:
+                last = self._last_infer_step.get(virtual)
+                if last is not None and int(step) <= last:
+                    self.history_guard_duplicates += 1
+                    print(
+                        f"[VPP2][history-guard] duplicate infer step={step} last={last} client={self._client_key()} total_duplicates={self.history_guard_duplicates}",
+                        flush=True,
+                    )
+                    current = observation_history[-1]
+                    self._remember(self._encode_observation(current), current)
+                    self._infer_count_by_virtual[virtual] = max(
+                        0, self._infer_count_by_virtual.get(virtual, 0) - 1
+                    )
+                    return
+                self._last_infer_step[virtual] = int(step)
+            for history_obs in observation_history:
+                if not isinstance(history_obs, dict):
+                    continue
+                encoded = self._encode_observation(history_obs)
+                self._remember(encoded, history_obs)
+                if self.model is not None and hasattr(self.model, "observe"):
+                    self.model.observe(encoded)
+            return
+        encoded = self._encode_observation(obs)
+        self._remember(encoded, obs)
+        if self.model is not None and hasattr(self.model, "observe"):
+            self.model.observe(encoded)
+
+    def _check_history_alignment(self, virtual: int) -> None:
+        """Before the k-th inference of an episode exactly replan_steps*k
+        observations must have been recorded after the anchor (step 0)."""
+        count = self._infer_count_by_virtual.get(virtual, 0)
+        self._infer_count_by_virtual[virtual] = count + 1
+        model = self.model
+        if model is None or not getattr(model, "history_condition_frames", 0):
+            return
+        recorded = getattr(model, "_history_step_by_env", {}).get(virtual)
+        expected = int(self.replan_steps) * count
+        if recorded != expected:
+            self.history_guard_mismatches += 1
+            print(
+                f"[VPP2][history-guard] MISMATCH inference={count} recorded_obs_step={recorded} expected={expected} client={self._client_key()} total_mismatches={self.history_guard_mismatches}",
+                flush=True,
+            )
+
+    def _infer_actions(self, obs, instruction):
+        if obs is None:
+            raise ValueError("No observation is available. Call update_obs() before get_action().")
+        action_chunk = self.model._infer_action_chunk(obs, instruction)
+        action_chunk = np.asarray(action_chunk, dtype=np.float32)
+        if action_chunk.ndim == 1:
+            action_chunk = action_chunk[None, :]
+        virtual = int(obs["env_idx"])
+        ensemble = self._ensemble_by_virtual.setdefault(virtual, TemporalEnsemble())
+        action_chunk = ensemble.apply(action_chunk)
+        self._last_ensemble = ensemble.last_info
+        return self._unpack_robodojo_ee16_actions(action_chunk)
+
+    def get_action(self, obs=None):
+        self._bind_client(obs)
+        client = self._client_key()
+        last_obs = self._last_obs_by_client.get(client, self.last_obs)
+        instruction = self._last_instruction_by_client.get(client, self.last_instruction)
+        virtual = self._last_virtual_by_client.get(client)
+        if virtual is not None:
+            self._check_history_alignment(virtual)
+        return self._infer_actions(last_obs, instruction)
+
+    def _reset_client_state(self, client) -> None:
+        for key in [k for k in self._virtual_env_ids if k[0] == client]:
+            self._forget_virtual(self._virtual_env_ids.pop(key))
+        self._last_obs_by_client.pop(client, None)
+        self._last_instruction_by_client.pop(client, None)
+        self._last_virtual_by_client.pop(client, None)
+
+    def reset_client(self, obs=None):
+        """Episode reset for one identified client (upstream per-call protocol)."""
+        self._bind_client(obs)
+        client = self._client_key()
+        self.last_obs = None
+        self.last_instruction = self.default_instruction
+        if self.model is not None and hasattr(self.model, "reset_env"):
+            self._reset_client_state(client)
+        elif self.model is not None:
+            self.model.reset()
+
+    def reset(self):
+        self._active_client = None
+        self._request_context = {}
+        self._seen_client_ids = False
+        self._next_virtual_env_id = 0
+        self._client_use_counter = 0
+        self.last_obs = None
+        self.last_instruction = self.default_instruction
+        for state in (self._virtual_env_ids, self._client_last_use,
+                      self._last_infer_step, self._infer_count_by_virtual,
+                      self._ensemble_by_virtual, self._last_obs_by_client,
+                      self._last_instruction_by_client, self._last_virtual_by_client):
+            state.clear()
+        self._last_ensemble = None
+        self.history_guard_duplicates = 0
+        self.history_guard_mismatches = 0
+        if self.model is not None:
+            self.model.reset()
+
+    def update_obs_batch(self, obs_list):
+        if len(obs_list) != 1:
+            raise NotImplementedError("CP1000 is qualified for one environment per instance")
+        self.update_obs(obs_list[0])
+
+    def get_action_batch(self, env_idx_list=None):
+        if env_idx_list is not None and len(env_idx_list) != 1:
+            raise NotImplementedError("CP1000 is qualified for one environment per instance")
+        return [self.get_action()]
diff --git a/inference/policy/bagel_robotics_1/requirements-transport.txt b/inference/policy/bagel_robotics_1/requirements-transport.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a81fad56bb51b81de53dbd66e77efc599a2d6d98
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/requirements-transport.txt
@@ -0,0 +1,5 @@
+opencv-python-headless==4.11.0.86
+websockets==16.1.1
+h5py==3.16.0
+PyYAML==6.0.3
+msgpack==1.2.3
diff --git a/inference/policy/bagel_robotics_1/runtime/LICENSE b/inference/policy/bagel_robotics_1/runtime/LICENSE
new file mode 100644
index 0000000000000000000000000000000000000000..90124d129334cd5de11aa0fd606182b3ffd442cb
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/LICENSE
@@ -0,0 +1,24 @@
+MIT License
+
+Copyright (c) 2026 The FastWAM Authors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+Except for third-party code and files that retain their original license
+headers, this repository is released under the MIT License.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
diff --git a/inference/policy/bagel_robotics_1/runtime/THIRD_PARTY_NOTICES.md b/inference/policy/bagel_robotics_1/runtime/THIRD_PARTY_NOTICES.md
new file mode 100644
index 0000000000000000000000000000000000000000..89c5a601d8405bd6c0b3f316168a442e00a5b41c
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/THIRD_PARTY_NOTICES.md
@@ -0,0 +1,8 @@
+# Runtime source notices
+
+The VPP2/FastWAM-derived source retains the MIT notice in LICENSE.
+Wan2.1/DiffSynth-derived components retain Apache-2.0 terms, included in
+licenses/Apache-2.0.txt. Source attribution is retained in individual modules
+where present. Source hashes and packaging changes accompany the adapter in
+source-provenance.json. Model weights and datasets are separate artifacts with
+their own upstream terms; this code license does not relicense those artifacts.
diff --git a/inference/policy/bagel_robotics_1/runtime/licenses/Apache-2.0.txt b/inference/policy/bagel_robotics_1/runtime/licenses/Apache-2.0.txt
new file mode 100644
index 0000000000000000000000000000000000000000..261eeb9e9f8b2b4b0d119366dda99c6fd7d35c64
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/licenses/Apache-2.0.txt
@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
diff --git a/inference/policy/bagel_robotics_1/runtime/pyproject.toml b/inference/policy/bagel_robotics_1/runtime/pyproject.toml
new file mode 100644
index 0000000000000000000000000000000000000000..562c61ea0b38693486cacec458a729fbe508d1c2
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/pyproject.toml
@@ -0,0 +1,20 @@
+[build-system]
+requires = ["setuptools>=77", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "bagel-cp1000-inference"
+version = "0.1.0"
+description = "Isolated inference runtime for Bagel-Robotics-1 recovery checkpoint 1000"
+requires-python = ">=3.10"
+license = "MIT AND Apache-2.0"
+license-files = ["LICENSE", "THIRD_PARTY_NOTICES.md", "licenses/*.txt"]
+dependencies = [
+  "torch==2.11.0", "torchvision==0.26.0", "numpy==1.26.4",
+  "transformers==4.49.0", "huggingface-hub==0.36.0", "hydra-core==1.3.2",
+  "omegaconf==2.3.0", "einops==0.8.1", "Pillow==12.0.0", "rich==14.2.0",
+  "safetensors==0.5.3", "ftfy==6.3.1", "regex>=2024.11", "sentencepiece>=0.2"
+]
+
+[tool.setuptools.packages.find]
+where = ["src"]
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/checkpoint_compat.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/checkpoint_compat.py
new file mode 100644
index 0000000000000000000000000000000000000000..a4da5e8c1d0162debd960c9e03ac94c097beaf4d
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/checkpoint_compat.py
@@ -0,0 +1,42 @@
+"""Read historical checkpoints using the isolated inference namespace.
+
+Legacy identifiers below describe serialized input formats, not current methods.
+No legacy package is imported and tensor keys are unchanged.
+"""
+
+ACTION_FORMAT = "vpp2_wan21_action_v1"
+_LEGACY_ACTION_FORMATS = {"fastwam_wan21_action_v1"}
+_LEGACY_NAMESPACES = ("fastwam.", "robodojo_joint2b.")
+_LEGACY_TARGET_PARTS = {
+    "fastwam": "vpp2",
+    "FastWAM": "VPP2",
+    "fastwam_processor": "vpp2_processor",
+    "FastWAMProcessor": "VPP2Processor",
+    "create_fastwam_wan21_14b": "create_vpp2_wan21_14b",
+}
+
+
+def canonicalize_config(value):
+    """Convert Hydra targets recursively, preserving paths and other values."""
+    if isinstance(value, list):
+        return [canonicalize_config(item) for item in value]
+    if not isinstance(value, dict):
+        return value
+    result = {key: canonicalize_config(item) for key, item in value.items()}
+    target = result.get("_target_")
+    if isinstance(target, str):
+        for prefix in _LEGACY_NAMESPACES:
+            if target.startswith(prefix):
+                target = "vpp2." + target[len(prefix) :]
+                break
+        if target.startswith("vpp2."):
+            target = ".".join(_LEGACY_TARGET_PARTS.get(part, part) for part in target.split("."))
+            target = "bagel_cp1000." + target[len("vpp2.") :]
+        result["_target_"] = target
+    return result
+
+
+def action_config(payload):
+    if payload.get("format") not in {ACTION_FORMAT, *_LEGACY_ACTION_FORMATS}:
+        raise ValueError("Expected an exported VPP2 action checkpoint")
+    return canonicalize_config(payload["resolved_config"])
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/data.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/data.py
new file mode 100644
index 0000000000000000000000000000000000000000..b49a92a1018f69871953bcb448b5c21fb6ae8870
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/data.py
@@ -0,0 +1 @@
+DEFAULT_PROMPT = "A video recorded from a robot's point of view executing the following instruction: {task}"
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/base_processor.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/base_processor.py
new file mode 100644
index 0000000000000000000000000000000000000000..a457628009c319e8b86e4a2c4744759f129d472c
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/base_processor.py
@@ -0,0 +1,270 @@
+from abc import ABC, abstractmethod
+from typing import Dict, Any, Optional, List, Literal
+
+import torch
+import numpy as np
+from copy import deepcopy
+from ..utils.normalizer import LinearNormalizer, NormMode
+from bagel_cp1000.utils.pytorch_utils import dict_apply
+
+
+class BaseProcessor(ABC):
+    def __init__(
+        self,
+        # keys
+        shape_meta: Dict[str, Any],
+        num_obs_steps: int,
+        num_output_cameras: int,
+        action_output_dim: int,
+        proprio_output_dim: int,
+        action_state_transforms: Optional[List[Any]],
+        # action & state normalization
+        use_stepwise_action_norm: bool,
+        norm_default_mode: NormMode,
+        norm_exception_mode: Dict[str, Dict[str, NormMode]],
+        action_state_merger,
+        # image transform
+        train_transforms: Dict[str, List[Any]] | None,
+        val_transforms: Dict[str, List[Any]] | None,
+        # instruction transform
+        drop_high_level_prob: float,
+        use_zh_instruction: bool,
+        tokenizer: Any,
+    ):
+        self.shape_meta = shape_meta
+        self.num_obs_steps = num_obs_steps
+        self.num_output_cameras = num_output_cameras
+        self.action_output_dim = action_output_dim
+        self.proprio_output_dim = proprio_output_dim
+
+        self.drop_high_level_prob = drop_high_level_prob
+        self.use_zh_instruction = use_zh_instruction
+
+        # image
+        self.train_transforms = train_transforms
+        self.val_transforms = val_transforms
+
+        self._is_train = None
+
+        self.action_state_transforms = action_state_transforms
+        self.action_state_merger = action_state_merger
+        self.action_state_merger.set_shape_meta(self.shape_meta)
+
+        self.use_stepwise_action_norm = use_stepwise_action_norm
+        self.norm_default_mode = norm_default_mode
+        self.norm_exception_mode = norm_exception_mode
+        self._normalizer = None
+
+        self.tokenizer = tokenizer
+
+    @property
+    def is_train(self):
+        if self._is_train is None:
+            raise ValueError("is_train has not been set. Please call train() and eval() first.")
+        return self._is_train
+
+    @property
+    def normalizer(self) -> LinearNormalizer:
+        if self._normalizer is None:
+            raise ValueError(
+                "normalizer has not been set. Please call set_normalizer_from_stats() first."
+            )
+        return self._normalizer
+
+    def train(self):
+        self._is_train = True
+        return self
+
+    def eval(self):
+        self._is_train = False
+        return self
+
+    def set_normalizer_from_stats(self, dataset_stats: Dict[str, Any] = None):
+        self._normalizer = LinearNormalizer(
+            use_stepwise_action_norm=self.use_stepwise_action_norm,
+            shape_meta=self.shape_meta,
+            default_mode=self.norm_default_mode,
+            exception_mode=self.norm_exception_mode,
+            stats=dataset_stats,
+        )
+
+    def augment_instruction(self, data: Dict[str, str] | List[str]) -> List[str]:
+        """
+        Args:
+            data: Dict[str, str] | List[str], lerobot sample in raw mcap
+
+        Returns:
+            List[str], processed instructions
+        """
+        # if single instruction, convert to list
+        if "coarse_task" in data:
+            high_level_instruction = data["coarse_task"]
+        else:
+            high_level_instruction = ""
+        if "task" not in data:
+            return f"[high] {high_level_instruction}"
+
+        low_level_instruction = data["task"]
+        # Galaxea lerobot use @ to split Chinese and English instruction
+        if "@" in low_level_instruction:
+            zh, eng = low_level_instruction.split("@")
+            low_level_instruction = zh if self.use_zh_instruction else eng
+
+        if np.random.rand() < self.drop_high_level_prob:
+            instruction = f"[Low]: {low_level_instruction}"
+        else:
+            instruction = f"[High]: {high_level_instruction}, [Low]: {low_level_instruction}"
+
+        return instruction
+
+    def action_state_transform(self, batch):
+        if "action" in batch:
+            for meta in self.shape_meta["action"]:
+                k, meta_shape = meta["key"], meta["raw_shape"]
+                actual_shape = batch["action"][k].shape[-1]
+                assert actual_shape == meta_shape, (
+                    f"Action key {k} actual raw shape {actual_shape} mismatch with meta raw shape {meta_shape}."
+                )
+
+        for meta in self.shape_meta["state"]:
+            k, meta_shape = meta["key"], meta["raw_shape"]
+            actual_shape = batch["state"][k].shape[-1]
+            assert actual_shape == meta_shape, (
+                f"State key {k} actual raw shape {actual_shape} mismatch with meta raw shape{meta_shape}."
+            )
+
+        if self.action_state_transforms is not None:
+            for trans in self.action_state_transforms:
+                batch = trans.forward(batch)
+
+        if "action" in batch:
+            for meta in self.shape_meta["action"]:
+                k, meta_shape = meta["key"], meta["shape"]
+                actual_shape = batch["action"][k].shape[-1]
+                assert actual_shape == meta_shape, (
+                    f"Action key {k} actual transformed shape {actual_shape} mismatch with meta shape {meta_shape}."
+                )
+
+        for meta in self.shape_meta["state"]:
+            k, meta_shape = meta["key"], meta["shape"]
+            actual_shape = batch["state"][k].shape[-1]
+            assert actual_shape == meta_shape, (
+                f"State key {k} actual transformed shape {actual_shape} mismatch with meta raw shape {meta_shape}."
+            )
+
+        return batch
+
+    def preprocess(self, data: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        Preprocess the data for the policy model.
+
+        Args:
+            Data: Dict[str, Any], lerobot sample in raw mcap obtained from dataset __getitem__:
+                - "action": Optional, Dict[str, torch.Tensor] -> [action_horizon, action_dim]
+                - "state": Dict[str, torch.Tensor] -> [num_obs_steps, state_dim]
+                - "images": Dict[str, torch.Tensor] -> [num_obs_steps, C, H, W]
+                - "action_is_pad": Optional, torch.Tensor -> [action_horizon,]
+                - "state_is_pad": torch.Tensor -> [num_obs_steps,]
+                - "image_is_pad": torch.Tensor -> [num_obs_steps,]
+                - "idx": int, sample index
+
+        Returns:
+            Sample: Dict[str, Any], which can collated:
+                - "input_ids": torch.Tensor -> [max_image_text_tokens,]
+                - "attention_mask": torch.Tensor -> [max_image_text_tokens,]
+                - "pixel_values": torch.Tensor -> [num_input_cameras, C, H, W]
+                - "image_is_pad": torch.Tensor -> [num_obs_steps,]
+                - "proprio": torch.Tensor -> [num_obs_steps, proprio_dim]
+                - "state_is_pad": torch.Tensor -> [num_obs_steps,]
+                - "action": Optional, torch.Tensor -> [action_horizon, action_dim]
+                - "action_is_pad": Optional, torch.Tensor -> [action_horizon,]
+                - "gt_action: Optional, deepcopy of input action for open loop eval, which is left untouched
+                - "idx": int, sample index
+        """
+        sample = {}
+        # 1. instruction
+        sample["instruction"] = self.augment_instruction(data)
+        sample["image_is_pad"] = data["image_is_pad"]
+
+        # 2. image
+        processed_images = []
+        for meta in self.shape_meta["images"]:
+            key, shape = meta["key"], meta["shape"]
+            image = data["images"][key]  # [num_obs_steps, C, H, W]
+            assert image.ndim == 4, (
+                f"Expected 4 dimensions (num_obs_steps, C, H, W), got shape {image.shape}"
+            )
+
+            # Apply transforms efficiently on the merged batch
+            transforms = self.train_transforms if self.is_train else self.val_transforms
+            for trans in transforms[key]:
+                image = trans(image)
+
+            meta_shape = [self.num_obs_steps] + shape
+            assert image.shape == meta_shape, (
+                f"Expected shape {meta_shape}, got {image.shape} after transforms for key {key}"
+            )
+
+            processed_images.append(image)
+
+        pixel_values = torch.cat(processed_images, dim=0)  # [num_input_cameras, C, H, W]
+        if self.num_output_cameras > pixel_values.shape[0]:
+            out = torch.zeros(
+                (self.num_output_cameras,) + pixel_values.shape[1:],
+                device=pixel_values.device,
+                dtype=pixel_values.dtype,
+            )
+            out[0 : pixel_values.shape[0]] = pixel_values
+            sample["pixel_values"] = out
+        else:
+            sample["pixel_values"] = pixel_values
+
+        # Copy action before transform for open-loop evaluation,
+        # disabled for training dataset as it may cause collating key problem.
+        if not self.is_train and "action" in data:
+            sample["gt_action"] = deepcopy(data["action"])
+
+        # 3. action & state
+        data = self.action_state_transform(data)
+        data = self.normalizer.forward(data)
+        data = self.action_state_merger.forward(data)
+
+        if "action" in data:
+            sample["action"] = data["action"]  # [action_horizon, action_dim]
+            sample["action_is_pad"] = data["action_is_pad"]  # [action_horizon,]
+            sample["action_dim_is_pad"] = data["action_dim_is_pad"]  # [action_dim,]
+            assert sample["action"].shape[-1] == self.action_output_dim
+
+        # TODO: rename all "state" into "proprio"
+        sample["proprio"] = data["state"]  # [num_obs_steps, proprio_dim]
+        sample["proprio_is_pad"] = data["state_is_pad"]  # [num_obs_steps,]
+        sample["proprio_dim_is_pad"] = data["state_dim_is_pad"]  # [proprio_dim,]
+        assert sample["proprio"].shape[-1] == self.proprio_output_dim
+
+        sample["idx"] = data["idx"]
+
+        sample = self.tokenizer(sample)
+
+        return sample
+
+    def postprocess(self, data: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        Postprocess the data for the policy model.
+
+        Args:
+            data: Dict[str, Any], lerobot sample in raw mcap
+
+        Returns:
+            data: Dict[str, Any], processed data including unnormalized action
+        """
+        assert "action" in data, "Action is required in postprocess"
+        data["state"] = data.pop("proprio")
+        data = self.action_state_merger.backward(data)
+        data = self.normalizer.backward(data)
+        if self.action_state_transforms is not None:
+            for trans in reversed(self.action_state_transforms):
+                data = trans.backward(data)
+
+        start_obs_step = self.num_obs_steps - 1
+        data["action"] = dict_apply(data["action"], lambda x: x[:, start_obs_step:, :])
+        return data
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/vpp2_processor.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/vpp2_processor.py
new file mode 100644
index 0000000000000000000000000000000000000000..d0479c35ef858f054b0db2b45902132398db47b6
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/processors/vpp2_processor.py
@@ -0,0 +1,324 @@
+from abc import ABC, abstractmethod
+from typing import Dict, Any, Optional, List, Literal
+
+import torch
+import numpy as np
+from copy import deepcopy
+from ..utils.normalizer import LinearNormalizer, NormMode
+from bagel_cp1000.utils.pytorch_utils import dict_apply
+from bagel_cp1000.utils.logging_config import get_logger
+from .base_processor import BaseProcessor
+
+logger = get_logger(__name__)
+
+
+class VPP2Processor(BaseProcessor):
+    def __init__(
+        self,
+        # keys
+        shape_meta: Dict[str, Any],
+        num_obs_steps: int,
+        num_output_cameras: int,
+        action_output_dim: int,
+        proprio_output_dim: int,
+        action_state_transforms: Optional[List[Any]],
+        # action & state normalization
+        use_stepwise_action_norm: bool,
+        norm_default_mode: NormMode,
+        norm_exception_mode: Dict[str, Dict[str, NormMode]],
+        action_state_merger,
+        # image transform
+        train_transforms: Dict[str, List[Any]] | None,
+        val_transforms: Dict[str, List[Any]] | None,
+        # instruction transform
+        drop_high_level_prob: float = 1.0,
+        use_zh_instruction: bool = False,
+        tokenizer: Optional[Any] = None,
+        delta_action_dim_mask: Optional[Dict[str, List[bool]]] = None,
+        num_image_obs_steps: Optional[int] = None,
+    ):
+        self.shape_meta = shape_meta
+        self.num_obs_steps = num_obs_steps
+        self.num_image_obs_steps = (
+            num_obs_steps if num_image_obs_steps is None else num_image_obs_steps
+        )
+        if self.num_image_obs_steps <= 0:
+            raise ValueError(
+                f"`num_image_obs_steps` must be positive, got {self.num_image_obs_steps}"
+            )
+        self.num_output_cameras = num_output_cameras
+        self.action_output_dim = action_output_dim
+        self.proprio_output_dim = proprio_output_dim
+
+        self.drop_high_level_prob = drop_high_level_prob
+        self.use_zh_instruction = use_zh_instruction
+
+        # image
+        self.train_transforms = train_transforms
+        self.val_transforms = val_transforms
+
+        self._is_train = None
+
+        self.action_state_transforms = action_state_transforms
+        self.action_state_merger = action_state_merger
+        self.action_state_merger.set_shape_meta(self.shape_meta)
+
+        self.use_stepwise_action_norm = use_stepwise_action_norm
+        self.norm_default_mode = norm_default_mode
+        self.norm_exception_mode = norm_exception_mode
+        self._normalizer = None
+
+        self.tokenizer = tokenizer
+        if delta_action_dim_mask is None:
+            self.delta_action_dim_mask = None
+        else:
+            action_meta = self.shape_meta["action"]
+            expected_keys = [m["key"] for m in action_meta]
+            provided_keys = list(delta_action_dim_mask.keys())
+            if set(provided_keys) != set(expected_keys):
+                raise ValueError(
+                    f"`delta_action_dim_mask` keys mismatch. Expected {expected_keys}, got {provided_keys}."
+                )
+
+            self.delta_action_dim_mask = {}
+            for meta in action_meta:
+                key = meta["key"]
+                expected_dim = meta["shape"]
+                mask = delta_action_dim_mask[key]
+                if len(mask) != expected_dim:
+                    raise ValueError(
+                        f"`delta_action_dim_mask[{key}]` length must be {expected_dim}, got {len(mask)}."
+                    )
+                self.delta_action_dim_mask[key] = torch.as_tensor(mask, dtype=torch.bool)
+
+    @property
+    def is_train(self):
+        if self._is_train is None:
+            raise ValueError("is_train has not been set. Please call train() and eval() first.")
+        return self._is_train
+
+    @property
+    def normalizer(self) -> LinearNormalizer:
+        if self._normalizer is None:
+            raise ValueError(
+                "normalizer has not been set. Please call set_normalizer_from_stats() first."
+            )
+        return self._normalizer
+
+    def train(self):
+        self._is_train = True
+        return self
+
+    def eval(self):
+        self._is_train = False
+        return self
+
+    def set_normalizer_from_stats(self, dataset_stats: Dict[str, Any] = None):
+        self._normalizer = LinearNormalizer(
+            use_stepwise_action_norm=self.use_stepwise_action_norm,
+            shape_meta=self.shape_meta,
+            default_mode=self.norm_default_mode,
+            exception_mode=self.norm_exception_mode,
+            stats=dataset_stats,
+        )
+
+    def augment_instruction(self, data: Dict[str, str] | List[str]) -> List[str]:
+        """
+        Args:
+            data: Dict[str, str] | List[str], lerobot sample in raw mcap
+
+        Returns:
+            List[str], processed instructions
+        """
+        # if single instruction, convert to list
+        if "coarse_task" in data:
+            high_level_instruction = data["coarse_task"]
+        else:
+            high_level_instruction = ""
+        if "task" not in data:
+            return f"[high] {high_level_instruction}"
+
+        low_level_instruction = data["task"]
+        # Galaxea lerobot use @ to split Chinese and English instruction
+        if "@" in low_level_instruction:
+            zh, eng = low_level_instruction.split("@")
+            low_level_instruction = zh if self.use_zh_instruction else eng
+
+        if np.random.rand() < self.drop_high_level_prob:
+            instruction = f"{low_level_instruction}"
+        else:
+            instruction = f"[High]: {high_level_instruction}, [Low]: {low_level_instruction}"
+
+        return instruction
+
+    def action_state_transform(self, batch):
+        if "action" in batch:
+            for meta in self.shape_meta["action"]:
+                k, meta_shape = meta["key"], meta["raw_shape"]
+                actual_shape = batch["action"][k].shape[-1]
+                assert actual_shape == meta_shape, (
+                    f"Action key {k} actual raw shape {actual_shape} mismatch with meta raw shape {meta_shape}."
+                )
+
+        for meta in self.shape_meta["state"]:
+            k, meta_shape = meta["key"], meta["raw_shape"]
+            actual_shape = batch["state"][k].shape[-1]
+            assert actual_shape == meta_shape, (
+                f"State key {k} actual raw shape {actual_shape} mismatch with meta raw shape {meta_shape}."
+            )
+
+        if self.action_state_transforms is not None:
+            for trans in self.action_state_transforms:
+                batch = trans.forward(batch)
+
+        if "action" in batch:
+            for meta in self.shape_meta["action"]:
+                k, meta_shape = meta["key"], meta["shape"]
+                actual_shape = batch["action"][k].shape[-1]
+                assert actual_shape == meta_shape, (
+                    f"Action key {k} actual transformed shape {actual_shape} mismatch with meta shape {meta_shape}."
+                )
+
+        for meta in self.shape_meta["state"]:
+            k, meta_shape = meta["key"], meta["shape"]
+            actual_shape = batch["state"][k].shape[-1]
+            assert actual_shape == meta_shape, (
+                f"State key {k} actual transformed shape {actual_shape} mismatch with meta raw shape {meta_shape}."
+            )
+
+        return batch
+
+    def preprocess(self, data: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        Preprocess the data for the policy model.
+
+        Args:
+            Data: Dict[str, Any], lerobot sample in raw mcap obtained from dataset __getitem__:
+                - "action": Optional, Dict[str, torch.Tensor] -> [action_horizon, action_dim]
+                - "state": Dict[str, torch.Tensor] -> [num_obs_steps, state_dim]
+                - "images": Dict[str, torch.Tensor] -> [num_obs_steps, C, H, W]
+                - "action_is_pad": Optional, torch.Tensor -> [action_horizon,]
+                - "state_is_pad": torch.Tensor -> [num_obs_steps,]
+                - "image_is_pad": torch.Tensor -> [num_obs_steps,]
+                - "idx": int, sample index
+
+        Returns:
+            Sample: Dict[str, Any], which can collated:
+                - "input_ids": torch.Tensor -> [max_image_text_tokens,]
+                - "attention_mask": torch.Tensor -> [max_image_text_tokens,]
+                - "pixel_values": torch.Tensor -> [num_input_cameras, C, H, W]
+                - "image_is_pad": torch.Tensor -> [num_obs_steps,]
+                - "proprio": torch.Tensor -> [num_obs_steps, proprio_dim]
+                - "state_is_pad": torch.Tensor -> [num_obs_steps,]
+                - "action": Optional, torch.Tensor -> [action_horizon, action_dim]
+                - "action_is_pad": Optional, torch.Tensor -> [action_horizon,]
+                - "gt_action: Optional, deepcopy of input action for open loop eval, which is left untouched
+                - "idx": int, sample index
+        """
+        sample = {}
+        # 1. instruction
+        sample["instruction"] = self.augment_instruction(data)
+        sample["image_is_pad"] = data["image_is_pad"]
+
+        # 2. image
+        processed_images = []
+        for meta in self.shape_meta["images"]:
+            key, shape = meta["key"], meta["shape"]
+            image = data["images"][key]  # [num_obs_steps, C, H, W]
+            assert image.ndim == 4, (
+                f"Expected 4 dimensions (num_obs_steps, C, H, W), got shape {image.shape}"
+            )
+
+            # Apply transforms efficiently on the merged batch
+            transforms = self.train_transforms if self.is_train else self.val_transforms
+            current_transforms = transforms[key] if isinstance(transforms, dict) else transforms
+            for trans in current_transforms:
+                image = trans(image)
+
+            meta_shape = [self.num_image_obs_steps] + shape
+            assert image.shape == meta_shape, (
+                f"Expected shape {meta_shape}, got {image.shape} after transforms for key {key}"
+            )
+
+            processed_images.append(image)
+        pixel_values = torch.stack(processed_images, dim=0)  # [num_input_cameras, T, C, H, W]
+
+        if self.num_output_cameras > pixel_values.shape[0]:
+            out = torch.zeros(
+                (self.num_output_cameras,) + pixel_values.shape[1:],
+                device=pixel_values.device,
+                dtype=pixel_values.dtype,
+            )
+            out[0 : pixel_values.shape[0]] = pixel_values
+            sample["pixel_values"] = out
+        elif self.num_output_cameras < pixel_values.shape[0]:
+            logger.warning(
+                f"num_output_cameras {self.num_output_cameras} is less than the number of cameras in data {pixel_values.shape[0]}, "
+                f"truncating the input to the first {self.num_output_cameras} cameras."
+            )
+            sample["pixel_values"] = pixel_values[: self.num_output_cameras]
+        else:
+            sample["pixel_values"] = pixel_values
+
+        # Copy action before transform for open-loop evaluation,
+        # disabled for training dataset as it may cause collating key problem.
+        if not self.is_train and "action" in data:
+            sample["gt_action"] = deepcopy(data["action"])
+
+        # 3. action & state.  Convert absolute labels to their configured
+        # representation before zeroing padded delta dimensions.  Doing this
+        # in the opposite order would turn a padded zero into ``-state`` for
+        # an absolute-to-delta transform.
+        data = self.action_state_transform(data)
+        if "action" in data and self.delta_action_dim_mask is not None:
+            action_is_pad = torch.as_tensor(data["action_is_pad"], dtype=torch.bool)
+            if bool(action_is_pad.any().item()):
+                for key, dim_mask in self.delta_action_dim_mask.items():
+                    cur_action = data["action"][key]
+                    cur_action_is_pad = action_is_pad.to(device=cur_action.device)
+                    cur_dim_mask = dim_mask.to(device=cur_action.device)
+                    pad_delta_mask = cur_action_is_pad.unsqueeze(1) & cur_dim_mask.unsqueeze(0)
+                    cur_action[pad_delta_mask] = 0.0
+        data = self.normalizer.forward(data)
+        data = self.action_state_merger.forward(data)
+
+        if "action" in data:
+            sample["action"] = data["action"]  # [action_horizon, action_dim]
+            sample["action_is_pad"] = data["action_is_pad"]  # [action_horizon,]
+            sample["action_dim_is_pad"] = data["action_dim_is_pad"]  # [action_dim,]
+            assert sample["action"].shape[-1] == self.action_output_dim
+            # sample["action"][sample["action_is_pad"], :-1] = 0.0 # NOTE: we assume use delta_eef_pose + gripper， so pad action is 0
+
+        # TODO: rename all "state" into "proprio"
+        sample["proprio"] = data["state"]  # [num_obs_steps, proprio_dim]
+        sample["proprio_is_pad"] = data["state_is_pad"]  # [num_obs_steps,]
+        sample["proprio_dim_is_pad"] = data["state_dim_is_pad"]  # [proprio_dim,]
+        assert sample["proprio"].shape[-1] == self.proprio_output_dim
+
+        sample["idx"] = data["idx"]
+
+        # sample = self.tokenizer(sample)
+
+        return sample
+
+    def postprocess(self, data: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        Postprocess the data for the policy model.
+
+        Args:
+            data: Dict[str, Any], lerobot sample in raw mcap
+
+        Returns:
+            data: Dict[str, Any], processed data including unnormalized action
+        """
+        assert "action" in data, "Action is required in postprocess"
+        data["state"] = data.pop("proprio")
+        data = self.action_state_merger.backward(data)
+        data = self.normalizer.backward(data)
+        if self.action_state_transforms is not None:
+            for trans in reversed(self.action_state_transforms):
+                data = trans.backward(data)
+
+        start_obs_step = self.num_obs_steps - 1
+        data["action"] = dict_apply(data["action"], lambda x: x[:, start_obs_step:, :])
+        return data
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/action_state_merger.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/action_state_merger.py
new file mode 100644
index 0000000000000000000000000000000000000000..f1b259785b0eb7151109fb087ea6489a9b40e9af
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/action_state_merger.py
@@ -0,0 +1,76 @@
+from typing import Dict
+
+import torch
+from torch.nn.functional import pad
+
+
+class ConcatLeftAlign:
+    def __init__(self, action_target_dim: int | None = None, state_target_dim: int | None = None):
+        self.action_target_dim = action_target_dim
+        self.state_target_dim = state_target_dim
+
+    def set_shape_meta(self, shape_meta):
+        self.action_meta = shape_meta["action"]
+        self.state_meta = shape_meta["state"]
+
+    def forward(self, batch):
+        if "action" in batch:
+            batch["action"] = self._concat(batch["action"], self.action_meta)
+            batch["action"], batch["action_dim_is_pad"] = self._pad(
+                batch["action"], self.action_target_dim
+            )
+
+        batch["state"] = self._concat(batch["state"], self.state_meta)
+        batch["state"], batch["state_dim_is_pad"] = self._pad(batch["state"], self.state_target_dim)
+
+        return batch
+
+    def backward(self, batch):
+        if self.state_target_dim is not None:
+            assert batch["state"].shape[-1] == self.state_target_dim
+        batch["state"] = self._crop(batch["state"], self.state_meta)
+        batch["state"] = self._split(batch["state"], self.state_meta)
+
+        if self.action_target_dim is not None:
+            assert batch["action"].shape[-1] == self.action_target_dim
+        batch["action"] = self._crop(batch["action"], self.action_meta)
+        batch["action"] = self._split(batch["action"], self.action_meta)
+
+        return batch
+
+    @staticmethod
+    def _pad(x: torch.Tensor, dim: int):
+        if dim is None:
+            dim = x.shape[-1]
+
+        assert x.ndim == 2 and x.shape[-1] <= dim
+        pad_dim = dim - x.shape[-1]
+        x_padded = pad(x, (0, pad_dim))
+        mask = torch.zeros_like(x[0]).bool()
+        mask = pad(mask, (0, pad_dim), value=True)
+        return x_padded, mask
+
+    @staticmethod
+    def _crop(x: torch.Tensor, meta: int):
+        assert x.ndim == 3
+        dim = sum([m["shape"] for m in meta])
+        x = x[:, :, :dim]
+        return x
+
+    @staticmethod
+    def _concat(x: Dict[str, torch.Tensor], meta: Dict[str, Dict]):
+        x = torch.cat([x[m["key"]] for m in meta], dim=-1)
+        assert x.ndim == 2
+        return x
+
+    @staticmethod
+    def _split(x: torch.Tensor, meta: Dict[str, Dict]):
+        assert x.ndim == 3
+        y = {}
+        idx = 0
+        for m in meta:
+            key, dim = m["key"], m["shape"]
+            y[key] = x[:, :, idx : idx + dim]
+            idx += dim
+
+        return y
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/image.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/image.py
new file mode 100644
index 0000000000000000000000000000000000000000..7ea98e3fe0b396ea734ea5eb302ef68c4dc9b14f
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/transforms/image.py
@@ -0,0 +1,68 @@
+import torch
+import torch.nn as nn
+import torchvision.transforms as TF
+import torchvision.transforms.functional as TF_F
+
+
+class ToTensor(nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    def forward(self, x: torch.Tensor):
+        assert x.dtype == torch.uint8
+        x = x.to(torch.float32) / 255.0
+        return x
+
+
+class Pad(nn.Module):
+    def __init__(self, padding, fill=0, padding_mode="constant"):
+        super().__init__()
+        self.padding = padding
+        self.fill = fill
+        self.padding_mode = padding_mode
+        self.pad = TF.Pad(padding=tuple(padding), fill=fill, padding_mode=padding_mode)
+
+    def forward(self, x: torch.Tensor):
+        assert x.ndim == 4, "Can only pad tensor of 4 dims."
+        return self.pad(x)
+
+
+class CenterCropAspectResize(nn.Module):
+    """Center-crop a video tensor to an aspect ratio, then resize it.
+
+    The LeRobot loader supplies ``[T,C,H,W]`` tensors.  RoboColiseum's head
+    and wrist cameras have different native geometries, so a direct resize
+    would make training and benchmark inference disagree geometrically.  This
+    transform establishes a common 4:3 camera contract before the views are
+    composed into the model's T-shaped canvas.
+    """
+
+    def __init__(self, size, aspect_ratio: float = 4.0 / 3.0):
+        super().__init__()
+        if len(size) != 2 or int(size[0]) <= 0 or int(size[1]) <= 0:
+            raise ValueError(f"size must be [height,width], got {size!r}")
+        if float(aspect_ratio) <= 0:
+            raise ValueError(f"aspect_ratio must be positive, got {aspect_ratio!r}")
+        self.size = [int(size[0]), int(size[1])]
+        self.aspect_ratio = float(aspect_ratio)
+
+    def forward(self, x: torch.Tensor):
+        if x.ndim != 4:
+            raise ValueError(f"CenterCropAspectResize expects [T,C,H,W], got {tuple(x.shape)}")
+        height, width = (int(value) for value in x.shape[-2:])
+        current_ratio = width / height
+        if current_ratio > self.aspect_ratio:
+            crop_height = height
+            crop_width = max(1, int(round(height * self.aspect_ratio)))
+        else:
+            crop_width = width
+            crop_height = max(1, int(round(width / self.aspect_ratio)))
+        top = (height - crop_height) // 2
+        left = (width - crop_width) // 2
+        x = TF_F.crop(x, top, left, crop_height, crop_width)
+        return TF_F.resize(
+            x,
+            size=self.size,
+            interpolation=TF_F.InterpolationMode.BILINEAR,
+            antialias=True,
+        )
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/utils/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/utils/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/utils/normalizer.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/utils/normalizer.py
new file mode 100644
index 0000000000000000000000000000000000000000..f979a7fa2305007e4538bc51454a0360bafe3c39
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/datasets/lerobot/utils/normalizer.py
@@ -0,0 +1,221 @@
+from typing import Literal, Dict, Annotated, Union, Any, List, Tuple, Optional
+import torch
+import json
+from collections import defaultdict
+import numpy as np
+from omegaconf import DictConfig, OmegaConf
+from pathlib import Path
+from bagel_cp1000.utils.logging_config import get_logger
+
+from bagel_cp1000.utils.pytorch_utils import dict_apply
+
+logger = get_logger(__name__)
+
+ConstConstStr = Annotated[
+    str, "format: 'const_min/const_max', where const_min and const_max give the constant range"
+]
+NormMode = Union[Literal["min/max", "q01/q99", "z-score"], ConstConstStr]
+
+
+class LinearNormalizer:
+    def __init__(
+        self,
+        shape_meta,
+        use_stepwise_action_norm,
+        default_mode: NormMode,
+        exception_mode: Dict[str, Dict[str, NormMode]],
+        stats: Dict[str, Dict[str, Dict[str, torch.Tensor]]],
+    ):
+        super().__init__()
+        self.normalizers = {"action": {}, "state": {}}
+        self.stats = stats
+
+        for meta in shape_meta["action"]:
+            key = meta["key"]
+
+            if use_stepwise_action_norm:
+                cur_stats = {
+                    k.removeprefix("stepwise_"): v
+                    for k, v in stats["action"][key].items()
+                    if k.startswith("stepwise_")
+                }
+            else:
+                cur_stats = {
+                    k.removeprefix("global_"): v
+                    for k, v in stats["action"][key].items()
+                    if k.startswith("global_")
+                }
+
+            if (
+                exception_mode is not None
+                and "action" in exception_mode
+                and key in exception_mode["action"]
+            ):
+                cur_mode = exception_mode["action"][key]
+            else:
+                cur_mode = default_mode
+
+            self.normalizers["action"][key] = SingleFieldLinearNormalizer(
+                stats=cur_stats,
+                mode=cur_mode,
+            )
+
+        for meta in shape_meta["state"]:
+            key = meta["key"]
+            cur_stats = {
+                k.removeprefix("global_"): v
+                for k, v in stats["state"][key].items()
+                if k.startswith("global_")
+            }
+
+            if (
+                exception_mode is not None
+                and "state" in exception_mode
+                and key in exception_mode["state"]
+            ):
+                cur_mode = exception_mode["state"][key]
+            else:
+                cur_mode = default_mode
+
+            self.normalizers["state"][key] = SingleFieldLinearNormalizer(
+                stats=cur_stats,
+                mode=cur_mode,
+            )
+
+    def get_stats(self):
+        stats = {
+            "action": {key: norm.get_stats() for key, norm in self.normalizers["action"].items()},
+            "state": {key: norm.get_stats() for key, norm in self.normalizers["state"].items()},
+        }
+        return stats
+
+    def forward(self, batch: Dict[str, Dict[str, torch.Tensor]]) -> torch.Tensor:
+        if "action" in batch:
+            for key, norm in self.normalizers["action"].items():
+                batch["action"][key] = norm.forward(batch["action"][key])
+
+        for key, norm in self.normalizers["state"].items():
+            batch["state"][key] = norm.forward(batch["state"][key])
+
+        return batch
+
+    def backward(self, batch: Dict[str, Dict[str, torch.Tensor]]) -> torch.Tensor:
+        for key, norm in self.normalizers["action"].items():
+            batch["action"][key] = norm.backward(batch["action"][key])
+
+        for key, norm in self.normalizers["state"].items():
+            batch["state"][key] = norm.backward(batch["state"][key])
+
+        return batch
+
+
+class SingleFieldLinearNormalizer:
+    std_reg = 1e-8
+    range_tol = 1e-4
+    output_max = 1.0
+    output_min = -1.0
+
+    def __init__(self, stats, mode: NormMode = "min/max"):
+        self.stats = stats
+        self.mode = mode
+
+        if mode == "z-score":
+            input_mean, input_std = stats["mean"], stats["std"]
+            scale = 1.0 / (input_std + self.std_reg)
+            offset = -input_mean / (input_std + self.std_reg)
+        else:
+            if mode == "min/max":
+                input_min, input_max = stats["min"], stats["max"]
+            elif mode == "q01/q99":
+                input_min, input_max = stats["q01"], stats["q99"]
+            else:
+                # parse const_min/const_max
+                input_min, input_max = map(float, mode.split("/"))
+                input_min = torch.full_like(stats["min"], input_min)
+                input_max = torch.full_like(stats["max"], input_max)
+
+            input_range = input_max - input_min
+            ignore_dim = input_range < self.range_tol
+            input_range[ignore_dim] = self.output_max - self.output_min
+            scale = (self.output_max - self.output_min) / input_range
+            offset = self.output_min - scale * input_min
+            offset[ignore_dim] = (self.output_max + self.output_min) / 2 - input_min[ignore_dim]
+
+        self.scale = scale
+        self.offset = offset
+
+    def get_stats(self):
+        return self.stats
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        x = x * self.scale + self.offset
+        x = torch.clamp(x, -5.0, 5.0)
+        return x
+
+    def backward(self, x: torch.Tensor) -> torch.Tensor:
+        x = (x - self.offset) / self.scale
+        return x
+
+
+def save_dataset_stats_to_json(dataset_stats: dict, file_path: str):
+
+    def convert_tensor(obj):
+        if isinstance(obj, torch.Tensor):
+            return obj.detach().cpu().numpy().tolist()
+        elif isinstance(obj, (defaultdict, dict)):
+            return {k: convert_tensor(v) for k, v in dict(obj).items()}
+        elif isinstance(obj, (list, tuple)):
+            return [convert_tensor(item) for item in obj]
+        elif isinstance(obj, (int, float, str, bool, type(None))):
+            return obj
+        else:
+            return str(obj)
+
+    serializable_stats = convert_tensor(dataset_stats)
+
+    with open(file_path, "w", encoding="utf-8") as f:
+        json.dump(serializable_stats, f, ensure_ascii=False, indent=2)
+
+
+def load_dataset_stats_from_json(file_path: str, try_convert_tensor: bool = True) -> Dict[str, Any]:
+
+    def is_numeric_list(obj):
+        if isinstance(obj, list):
+            if not obj:
+                return True
+            first = obj[0]
+            if isinstance(first, (int, float)):
+                return all(isinstance(x, (int, float)) for x in obj)
+            elif isinstance(first, list):
+                return all(is_numeric_list(item) for item in obj)
+            else:
+                return False
+        return False
+
+    def convert_back_to_tensor(obj):
+        if isinstance(obj, dict):
+            return {k: convert_back_to_tensor(v) for k, v in obj.items()}
+        elif isinstance(obj, list):
+            if is_numeric_list(obj):
+                try:
+                    arr = np.array(obj)
+                    return torch.from_numpy(arr)
+                except Exception:
+                    return [convert_back_to_tensor(item) for item in obj]
+            else:
+                return [convert_back_to_tensor(item) for item in obj]
+        else:
+            return obj
+
+    with open(file_path, "r", encoding="utf-8") as f:
+        data = json.load(f)
+
+    if try_convert_tensor:
+        data = convert_back_to_tensor(data)
+
+    data = dict_apply(
+        data,
+        lambda x: x.to(torch.float32) if isinstance(x, torch.Tensor) else x,
+    )
+
+    return data
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/action_dit.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/action_dit.py
new file mode 100644
index 0000000000000000000000000000000000000000..a75f31c59bc88dcfb013c18d769c15f9a53fb7a7
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/action_dit.py
@@ -0,0 +1,458 @@
+import os
+import torch
+import torch.nn as nn
+from typing import Any, Dict, Optional
+
+from bagel_cp1000.utils.logging_config import get_logger
+
+from .helpers.gradient import gradient_checkpoint_forward
+from .wan_video_dit import (
+    DiTBlock,
+    WanRMSNorm,
+    rope_apply,
+    sinusoidal_embedding_1d,
+    precompute_freqs_cis,
+)
+
+logger = get_logger(__name__)
+
+
+class ActionHead(nn.Module):
+    def __init__(self, hidden_dim: int, out_dim: int, eps: float):
+        super().__init__()
+        self.norm = nn.LayerNorm(hidden_dim, eps=eps, elementwise_affine=False)
+        self.proj = nn.Linear(hidden_dim, out_dim)
+        self.modulation = nn.Parameter(torch.randn(1, 2, hidden_dim) / hidden_dim**0.5)
+
+    def forward(self, x: torch.Tensor, t: torch.Tensor) -> torch.Tensor:
+        shift, scale = (self.modulation.to(dtype=t.dtype, device=t.device) + t.unsqueeze(1)).chunk(
+            2, dim=1
+        )
+        shift = shift.squeeze(1)
+        scale = scale.squeeze(1)
+        return self.proj(self.norm(x) * (1 + scale.unsqueeze(1)) + shift.unsqueeze(1))
+
+
+class ActionDiT(nn.Module):
+    ACTION_BACKBONE_SKIP_PREFIXES = (
+        "action_encoder.",
+        "head.",
+        "video_kv_projections.",
+        "video_k_norms.",
+        "video_token_encoder.",
+    )
+    ACTION_BACKBONE_META_KEYS = (
+        "hidden_dim",
+        "ffn_dim",
+        "num_layers",
+        "num_heads",
+        "attn_head_dim",
+        "text_dim",
+        "freq_dim",
+        "eps",
+    )
+
+    def __init__(
+        self,
+        hidden_dim: int,
+        action_dim: int,
+        ffn_dim: int,
+        text_dim: int,
+        freq_dim: int,
+        eps: float,
+        num_heads: int,
+        attn_head_dim: int,
+        num_layers: int,
+        use_gradient_checkpointing: bool = False,
+        use_video_tokens: bool = False,
+        video_token_dim: Optional[int] = None,
+        video_token_latent_dim: int = 2560,
+    ):
+        super().__init__()
+        self.hidden_dim = hidden_dim
+        self.action_dim = action_dim
+        self.ffn_dim = ffn_dim
+        self.text_dim = text_dim
+        self.freq_dim = freq_dim
+        self.num_heads = num_heads
+        self.attn_head_dim = attn_head_dim
+        self.attn_hidden_dim = num_heads * attn_head_dim
+        self.use_video_tokens = bool(use_video_tokens)
+        self.video_token_dim = None if video_token_dim is None else int(video_token_dim)
+        self.video_token_latent_dim = int(video_token_latent_dim)
+
+        if num_heads <= 0:
+            raise ValueError(f"`num_heads` must be > 0, got {num_heads}")
+        if attn_head_dim <= 0:
+            raise ValueError(f"`attn_head_dim` must be > 0, got {attn_head_dim}")
+        if attn_head_dim % 2 != 0:
+            raise ValueError(f"`attn_head_dim` must be even for RoPE, got {attn_head_dim}")
+        if self.use_video_tokens and (self.video_token_dim is None or self.video_token_dim <= 0):
+            raise ValueError(
+                "`use_video_tokens=true` requires a positive `video_token_dim`, "
+                f"got {video_token_dim}."
+            )
+        if self.use_video_tokens and self.video_token_latent_dim <= 0:
+            raise ValueError(
+                f"`video_token_latent_dim` must be positive, got {self.video_token_latent_dim}."
+            )
+
+        self.action_encoder = nn.Linear(action_dim, hidden_dim)
+        self.text_embedding = nn.Sequential(
+            nn.Linear(text_dim, hidden_dim),
+            nn.GELU(approximate="tanh"),
+            nn.Linear(hidden_dim, hidden_dim),
+        )
+        self.time_embedding = nn.Sequential(
+            nn.Linear(freq_dim, hidden_dim),
+            nn.SiLU(),
+            nn.Linear(hidden_dim, hidden_dim),
+        )
+        self.time_projection = nn.Sequential(nn.SiLU(), nn.Linear(hidden_dim, hidden_dim * 6))
+        self.blocks = nn.ModuleList(
+            [
+                DiTBlock(
+                    hidden_dim=hidden_dim,
+                    attn_head_dim=attn_head_dim,
+                    num_heads=num_heads,
+                    ffn_dim=ffn_dim,
+                    eps=eps,
+                )
+                for _ in range(num_layers)
+            ]
+        )
+        self.video_kv_projections = nn.ModuleList()
+        self.video_k_norms = nn.ModuleList()
+        self.video_token_encoder: Optional[nn.Linear] = None
+        if self.use_video_tokens:
+            self.video_token_encoder = nn.Linear(
+                self.video_token_dim,
+                self.video_token_latent_dim,
+            )
+            self.video_kv_projections.extend(
+                nn.ModuleDict(
+                    {
+                        "k": nn.Linear(
+                            self.video_token_latent_dim,
+                            self.attn_hidden_dim,
+                        ),
+                        "v": nn.Linear(
+                            self.video_token_latent_dim,
+                            self.attn_hidden_dim,
+                        ),
+                    }
+                )
+                for _ in range(num_layers)
+            )
+            self.video_k_norms.extend(
+                WanRMSNorm(self.attn_hidden_dim, eps=eps) for _ in range(num_layers)
+            )
+        self.head = nn.Linear(hidden_dim, action_dim)
+        self.freqs = precompute_freqs_cis(attn_head_dim, end=1024)
+
+        self.use_gradient_checkpointing = use_gradient_checkpointing
+
+    def encode_video_tokens(self, video_tokens: torch.Tensor) -> torch.Tensor:
+        """Compress frozen Wan video features before action-owned K/V projections."""
+        if self.video_token_encoder is None:
+            raise RuntimeError("Video-token encoding is disabled for this ActionDiT.")
+        if video_tokens.ndim != 3 or video_tokens.shape[-1] != self.video_token_dim:
+            raise ValueError(
+                "`video_tokens` must have shape [B,Sv,video_token_dim], got "
+                f"{tuple(video_tokens.shape)} with video_token_dim={self.video_token_dim}."
+            )
+        return self.video_token_encoder(video_tokens)
+
+    def project_video_tokens(
+        self,
+        layer_idx: int,
+        video_tokens: torch.Tensor,
+        video_freqs: torch.Tensor,
+    ) -> tuple[torch.Tensor, torch.Tensor]:
+        """Build one action layer's K/V from the shared compressed video tokens."""
+        if not self.use_video_tokens:
+            raise RuntimeError("Video-token K/V projection is disabled for this ActionDiT.")
+        if video_tokens.ndim != 3:
+            raise ValueError(f"`video_tokens` must be [B,Sv,Dv], got {tuple(video_tokens.shape)}.")
+        if video_tokens.shape[-1] != self.video_token_latent_dim:
+            raise ValueError(
+                "Encoded video token dim must be "
+                f"{self.video_token_latent_dim}, got {video_tokens.shape[-1]}."
+            )
+        if not 0 <= int(layer_idx) < len(self.video_kv_projections):
+            raise ValueError(
+                f"Action layer index must be in [0, {len(self.video_kv_projections) - 1}], "
+                f"got {layer_idx}."
+            )
+
+        layer_idx = int(layer_idx)
+        projection = self.video_kv_projections[layer_idx]
+        k = projection["k"](video_tokens)
+        k = self.video_k_norms[layer_idx](k)
+        k = rope_apply(k, video_freqs, self.num_heads)
+        v = projection["v"](video_tokens)
+        return k, v
+
+    @classmethod
+    def backbone_key_set(cls, keys) -> set[str]:
+        return {
+            key
+            for key in keys
+            if not any(key.startswith(prefix) for prefix in cls.ACTION_BACKBONE_SKIP_PREFIXES)
+        }
+
+    @classmethod
+    def from_pretrained(
+        cls,
+        action_dit_config: dict[str, Any],
+        action_dit_pretrained_path: str | None = None,
+        skip_dit_load_from_pretrain: bool = False,
+        device: str = "cuda",
+        torch_dtype: torch.dtype = torch.bfloat16,
+    ) -> "ActionDiT":
+        if action_dit_config is None:
+            raise ValueError("`action_dit_config` is required for ActionDiT.from_pretrained().")
+        if skip_dit_load_from_pretrain:
+            logger.info(
+                "Skipping ActionDiT pretrained load (`skip_dit_load_from_pretrain=True`); "
+                "initializing action expert randomly and expecting checkpoint override."
+            )
+            return cls(**action_dit_config).to(device=device, dtype=torch_dtype)
+        if not action_dit_pretrained_path:
+            logger.info(
+                "No `action_dit_pretrained_path` provided, initializing ActionDiT with random weights."
+            )
+            return cls(**action_dit_config).to(device=device, dtype=torch_dtype)
+        from pathlib import Path
+
+        p = Path(action_dit_pretrained_path)
+        if not p.is_absolute():
+            p = Path(__file__).resolve().parents[4] / p
+        action_dit_pretrained_path = str(p)
+        if not os.path.isfile(action_dit_pretrained_path):
+            raise FileNotFoundError(
+                f"`action_dit_pretrained_path` does not exist: {action_dit_pretrained_path}"
+            )
+
+        action_cfg = dict(action_dit_config)
+        action_expert = cls(**action_cfg).to(device=device, dtype=torch_dtype)
+        action_state = action_expert.state_dict()
+        expected_backbone_keys = cls.backbone_key_set(action_state.keys())
+
+        payload = torch.load(action_dit_pretrained_path, map_location="cpu")
+        if not isinstance(payload, dict):
+            raise ValueError(
+                f"Invalid action backbone payload type from {action_dit_pretrained_path}: {type(payload)}"
+            )
+
+        policy = payload.get("policy", {})
+        if policy:
+            logger.info(f"ActionDiT backbone payload policy: {policy}")
+
+        meta = payload.get("meta")
+        expected_meta = {
+            "hidden_dim": int(action_cfg["hidden_dim"]),
+            "ffn_dim": int(action_cfg["ffn_dim"]),
+            "num_layers": int(action_cfg["num_layers"]),
+            "num_heads": int(action_cfg["num_heads"]),
+            "attn_head_dim": int(action_cfg["attn_head_dim"]),
+            "text_dim": int(action_cfg["text_dim"]),
+            "freq_dim": int(action_cfg["freq_dim"]),
+            "eps": float(action_cfg["eps"]),
+        }
+        for key in cls.ACTION_BACKBONE_META_KEYS:
+            if key not in meta:
+                raise ValueError(f"`meta.{key}` missing in {action_dit_pretrained_path}")
+            expected_value = expected_meta[key]
+            got_value = meta[key]
+            if key == "eps":
+                if abs(float(got_value) - float(expected_value)) > 1e-12:
+                    raise ValueError(
+                        f"`meta.{key}` mismatch in {action_dit_pretrained_path}: "
+                        f"expected {expected_value}, got {got_value}"
+                    )
+            elif int(got_value) != int(expected_value):
+                if key == "num_heads" and bool(action_cfg.get("use_video_tokens", False)):
+                    logger.warning(
+                        "Action backbone num_heads changed from %s to %s for video-token mode; "
+                        "shape-incompatible attention tensors will remain randomly initialized.",
+                        got_value,
+                        expected_value,
+                    )
+                    continue
+                raise ValueError(
+                    f"`meta.{key}` mismatch in {action_dit_pretrained_path}: "
+                    f"expected {expected_value}, got {got_value}"
+                )
+
+        backbone_state_dict = payload.get("backbone_state_dict")
+        if not isinstance(backbone_state_dict, dict):
+            raise ValueError(
+                f"`backbone_state_dict` must be a dict in {action_dit_pretrained_path}, "
+                f"got {type(backbone_state_dict)}"
+            )
+
+        provided_keys = set(backbone_state_dict.keys())
+        missing_keys = sorted(expected_backbone_keys - provided_keys)
+        unexpected_keys = sorted(provided_keys - expected_backbone_keys)
+        if missing_keys or unexpected_keys:
+            raise ValueError(
+                "Action backbone key mismatch in preprocessed payload. "
+                f"missing={missing_keys[:10]}{'...' if len(missing_keys) > 10 else ''}, "
+                f"unexpected={unexpected_keys[:10]}{'...' if len(unexpected_keys) > 10 else ''}"
+            )
+
+        merged_state = dict(action_state)
+        skipped_shape_keys: list[str] = []
+        for key in expected_backbone_keys:
+            value = backbone_state_dict[key]
+            if not isinstance(value, torch.Tensor):
+                raise ValueError(
+                    f"`backbone_state_dict[{key}]` must be torch.Tensor in {action_dit_pretrained_path}, "
+                    f"got {type(value)}"
+                )
+            target = merged_state[key]
+            if tuple(value.shape) != tuple(target.shape):
+                if bool(action_cfg.get("use_video_tokens", False)):
+                    skipped_shape_keys.append(key)
+                    continue
+                raise ValueError(
+                    f"Shape mismatch for `{key}` in {action_dit_pretrained_path}: "
+                    f"expected {tuple(target.shape)}, got {tuple(value.shape)}"
+                )
+            merged_state[key] = value.to(device=target.device, dtype=target.dtype)
+
+        if skipped_shape_keys:
+            logger.warning(
+                "Kept random initialization for %d shape-incompatible ActionDiT tensors "
+                "(examples: %s).",
+                len(skipped_shape_keys),
+                skipped_shape_keys[:8],
+            )
+
+        action_expert.load_state_dict(merged_state, strict=True)
+        logger.info(
+            "Loaded ActionDiT backbone from %s (keys=%d; random_kept_prefixes=%s).",
+            action_dit_pretrained_path,
+            len(expected_backbone_keys),
+            list(cls.ACTION_BACKBONE_SKIP_PREFIXES),
+        )
+        return action_expert.to(device=device, dtype=torch_dtype)
+
+    def pre_dit(
+        self,
+        action_tokens: torch.Tensor,
+        timestep: torch.Tensor,
+        context: torch.Tensor,
+        context_mask: Optional[torch.Tensor] = None,
+    ) -> Dict[str, Any]:
+        if action_tokens.ndim != 3:
+            raise ValueError(
+                f"`action_tokens` must be 3D [B, T, action_dim], got shape {tuple(action_tokens.shape)}"
+            )
+        if action_tokens.shape[2] != self.action_dim:
+            raise ValueError(
+                f"`action_tokens` last dim must be {self.action_dim}, got {action_tokens.shape[2]}"
+            )
+        if timestep.ndim != 1:
+            raise ValueError(f"`timestep` must be 1D [B] or [1], got shape {tuple(timestep.shape)}")
+        if context.ndim != 3:
+            raise ValueError(f"`context` must be 3D [B, L, D], got shape {tuple(context.shape)}")
+
+        batch_size = action_tokens.shape[0]
+        if context.shape[0] != batch_size:
+            raise ValueError(
+                f"Batch mismatch between action tokens and text context: {batch_size} vs {context.shape[0]}"
+            )
+        if timestep.shape[0] not in (1, batch_size):
+            raise ValueError(
+                f"`timestep` length must be 1 or batch_size({batch_size}), got {timestep.shape[0]}"
+            )
+        if timestep.shape[0] == 1 and batch_size > 1:
+            if self.training:
+                raise ValueError("During training, action timestep length must match batch_size.")
+            timestep = timestep.expand(batch_size)
+
+        if context_mask is None:
+            context_mask = torch.ones(
+                (batch_size, context.shape[1]), dtype=torch.bool, device=context.device
+            )
+        else:
+            if context_mask.ndim != 2:
+                raise ValueError(
+                    f"`context_mask` must be 2D [B, L], got shape {tuple(context_mask.shape)}"
+                )
+            if context_mask.shape[0] != batch_size or context_mask.shape[1] != context.shape[1]:
+                raise ValueError(
+                    f"`context_mask` shape must match `context` shape [B, L], got {tuple(context_mask.shape)} vs {tuple(context.shape)}"
+                )
+
+        seq_len = action_tokens.shape[1]
+        if seq_len > self.freqs.shape[0]:
+            raise ValueError(
+                f"Action token length {seq_len} exceeds RoPE cache {self.freqs.shape[0]}."
+            )
+
+        time_weight = self.time_embedding[0].weight
+        time_input = sinusoidal_embedding_1d(self.freq_dim, timestep).to(
+            device=time_weight.device,
+            dtype=time_weight.dtype,
+        )
+        t = self.time_embedding(time_input)
+        t_mod = self.time_projection(t).unflatten(1, (6, self.hidden_dim))
+
+        tokens = self.action_encoder(action_tokens)
+        context_emb = self.text_embedding(context)
+        context_attn_mask = context_mask.unsqueeze(1).expand(-1, seq_len, -1)
+        freqs = self.freqs[:seq_len].view(seq_len, 1, -1).to(tokens.device)
+
+        return {
+            "tokens": tokens,
+            "freqs": freqs,
+            "t": t,
+            "t_mod": t_mod,
+            "context": context_emb,
+            "context_mask": context_attn_mask,
+            "meta": {
+                "batch_size": batch_size,
+                "seq_len": seq_len,
+            },
+        }
+
+    def post_dit(self, tokens: torch.Tensor, pre_state: Dict[str, Any]) -> torch.Tensor:
+        return self.head(tokens)
+
+    def forward(
+        self,
+        action_tokens: torch.Tensor,
+        timestep: torch.Tensor,
+        context: torch.Tensor,
+        context_mask: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        pre_state = self.pre_dit(
+            action_tokens=action_tokens,
+            timestep=timestep,
+            context=context,
+            context_mask=context_mask,
+        )
+        x = pre_state["tokens"]
+        context = pre_state["context"]
+        t_mod = pre_state["t_mod"]
+        freqs = pre_state["freqs"]
+        context_mask = pre_state["context_mask"]
+
+        for block in self.blocks:
+            if self.use_gradient_checkpointing:
+                x = gradient_checkpoint_forward(
+                    block,
+                    self.use_gradient_checkpointing,
+                    x,
+                    context,
+                    t_mod,
+                    freqs,
+                    context_mask=context_mask,
+                )
+            else:
+                x = block(x, context, t_mod, freqs, context_mask=context_mask)
+
+        return self.post_dit(x, pre_state)
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/compact_channels.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/compact_channels.py
new file mode 100644
index 0000000000000000000000000000000000000000..bb445b22096660f99b076e159d1143270c64beb3
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/compact_channels.py
@@ -0,0 +1,27 @@
+"""Native heterogeneous channel architecture; no training/mask dependencies."""
+from torch import nn
+from ..wan_video_dit import WanVideoDiT,I2VCrossAttention,precompute_freqs_cis_3d
+
+
+def build(spec,device='cpu'):
+    if spec.get('schema')!='vpp2-compact-channels-v1':raise ValueError('Unknown channel architecture')
+    g=spec['geometry'];selection=spec['selection'];h=len(selection['hidden_indices'])
+    def valid(ids,maximum):
+        return ids==sorted(set(ids)) and len(ids)>0 and all(type(i) is int and 0<=i<maximum for i in ids)
+    if not valid(selection['hidden_indices'],g['hidden']):raise ValueError('Invalid hidden coordinate identities')
+    for key,total in (('ffn_indices',g['ffn']),('cross_head_indices',g['heads'])):
+        if len(selection[key])!=g['layers'] or any(not valid(x,total) for x in selection[key]):
+            raise ValueError('Invalid per-layer identities')
+        if selection[key][-1]!=list(range(total)):raise ValueError('Terminal internals must remain complete')
+    args=dict(spec['model_kwargs']);args.update(dim=h,hidden_dim=h,ffn_dim=g['ffn'],num_layers=g['layers'],
+                                               num_heads=g['heads'],attn_head_dim=g['head_dim'])
+    import torch
+    with torch.device(device):
+        model=WanVideoDiT(**args)
+        for i,b in enumerate(model.blocks):
+            f=len(selection['ffn_indices'][i]);heads=len(selection['cross_head_indices'][i])
+            b.ffn[0]=nn.Linear(h,f);b.ffn[2]=nn.Linear(f,h);b.ffn_dim=f
+            b.cross_attn=I2VCrossAttention(h,g['head_dim'],heads,b.cross_attn.norm_q.eps,True)
+    model.freqs=precompute_freqs_cis_3d(g['head_dim'])
+    if sum(p.numel() for p in model.parameters())!=spec['parameters']:raise ValueError('Physical channel parameter count mismatch')
+    return model
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/gradient.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/gradient.py
new file mode 100644
index 0000000000000000000000000000000000000000..0b0051cee98064f6ce22e13f263d464fe11261fb
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/gradient.py
@@ -0,0 +1,26 @@
+import torch
+
+
+def create_custom_forward(module):
+    def custom_forward(*inputs, **kwargs):
+        return module(*inputs, **kwargs)
+
+    return custom_forward
+
+
+def gradient_checkpoint_forward(
+    model,
+    use_gradient_checkpointing,
+    *args,
+    **kwargs,
+):
+    if use_gradient_checkpointing:
+        model_output = torch.utils.checkpoint.checkpoint(
+            create_custom_forward(model),
+            *args,
+            **kwargs,
+            use_reentrant=False,
+        )
+    else:
+        model_output = model(*args, **kwargs)
+    return model_output
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/io.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/io.py
new file mode 100644
index 0000000000000000000000000000000000000000..06349566aa60dce485b299aa5ce6c7eccc979daa
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/io.py
@@ -0,0 +1,58 @@
+import os
+import torch
+from safetensors import safe_open
+from bagel_cp1000.utils.logging_config import get_logger
+
+logger = get_logger(__name__)
+
+
+def load_state_dict(file_path, torch_dtype=None, device="cpu"):
+    if isinstance(file_path, list):
+        state_dict = {}
+        for file_path_ in file_path:
+            state_dict.update(load_state_dict(file_path_, torch_dtype=torch_dtype, device=device))
+        return state_dict
+    if file_path.endswith(".safetensors"):
+        return load_state_dict_from_safetensors(file_path, torch_dtype=torch_dtype, device=device)
+    return load_state_dict_from_bin(file_path, torch_dtype=torch_dtype, device=device)
+
+
+def load_state_dict_from_safetensors(file_path, torch_dtype=None, device="cpu"):
+    state_dict = {}
+    with safe_open(file_path, framework="pt", device=str(device)) as f:
+        for key in f.keys():
+            value = f.get_tensor(key)
+            if torch_dtype is not None:
+                value = value.to(torch_dtype)
+            state_dict[key] = value
+    return state_dict
+
+
+def load_state_dict_from_bin(file_path, torch_dtype=None, device="cpu"):
+    mmap_env = os.environ.get("VPP2_TORCH_LOAD_MMAP", "").strip().lower()
+    mmap = False
+    if mmap_env in {"1", "true", "yes", "on"}:
+        mmap = True
+    elif mmap_env in {"0", "false", "no", "off"}:
+        mmap = False
+    try:
+        state_dict = torch.load(file_path, map_location=device, weights_only=True, mmap=mmap)
+    except RuntimeError:
+        if mmap_env:
+            raise
+        logger.warning(
+            "mmap load failed for %s; retrying with default torch.load behavior.", file_path
+        )
+        state_dict = torch.load(file_path, map_location=device, weights_only=True)
+    if len(state_dict) == 1:
+        if "state_dict" in state_dict:
+            state_dict = state_dict["state_dict"]
+        elif "module" in state_dict:
+            state_dict = state_dict["module"]
+        elif "model_state" in state_dict:
+            state_dict = state_dict["model_state"]
+    if torch_dtype is not None:
+        for key in state_dict:
+            if isinstance(state_dict[key], torch.Tensor):
+                state_dict[key] = state_dict[key].to(torch_dtype)
+    return state_dict
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/loader.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/loader.py
new file mode 100644
index 0000000000000000000000000000000000000000..0dc7f1322c9a3b625bde2263f6613e02b51dd94d
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/helpers/loader.py
@@ -0,0 +1,230 @@
+from __future__ import annotations
+
+import glob
+import inspect
+import json
+import os
+import time
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+import torch
+
+from .io import load_state_dict
+from ..wan_video_clip import WanImageEncoderCLIP
+from ..wan_video_dit import WanVideoDiT
+from ..wan_video_text_encoder import HuggingfaceTokenizer, WanTextEncoder
+from ..wan_video_vae import WanVideoVAE
+from bagel_cp1000.utils.logging_config import get_logger
+
+logger = get_logger(__name__)
+SKIPPED_PRETRAIN_SENTINEL = "SKIPPED_PRETRAIN"
+
+
+@dataclass
+class Wan21I2V14BLoadedComponents:
+    dit: WanVideoDiT
+    vae: WanVideoVAE
+    text_encoder: WanTextEncoder | None
+    tokenizer: HuggingfaceTokenizer | None
+    clip_encoder: WanImageEncoderCLIP | None
+    dit_path: str
+    vae_path: str
+    text_encoder_path: str | None
+    tokenizer_path: str | None
+    clip_encoder_path: str | None
+
+
+def _resolve_model_dir(model_id: str | os.PathLike[str]) -> Path:
+    path = Path(model_id)
+    if path.exists():
+        return path
+    raise FileNotFoundError(f"Local Wan model directory does not exist: {path}")
+
+
+def _validate_dit_config(dit_config: dict[str, Any]) -> dict[str, Any]:
+    if not isinstance(dit_config, dict):
+        raise ValueError(f"`dit_config` must be a dict, got {type(dit_config)}")
+    validated = dict(dit_config)
+    sig = inspect.signature(WanVideoDiT.__init__)
+    allowed = {k for k in sig.parameters if k != "self"}
+    unknown = sorted(set(validated) - allowed)
+    if unknown:
+        raise ValueError(
+            f"Unknown keys in `dit_config`: {unknown}. Allowed keys: {sorted(allowed)}"
+        )
+    return validated
+
+
+def _load_json_config(model_dir: Path) -> dict[str, Any]:
+    config_path = model_dir / "config.json"
+    if not config_path.is_file():
+        return {}
+    with config_path.open("r", encoding="utf-8") as f:
+        config = json.load(f)
+    return {k: v for k, v in config.items() if not k.startswith("_")}
+
+
+def _convert_vae_state_dict(state_dict: dict[str, torch.Tensor]) -> dict[str, torch.Tensor]:
+    if "model_state" in state_dict:
+        state_dict = state_dict["model_state"]
+    if any(k.startswith("model.") for k in state_dict):
+        return state_dict
+    return {f"model.{k}": v for k, v in state_dict.items()}
+
+
+def _load_state_into_model(
+    model: torch.nn.Module, path, torch_dtype: torch.dtype, strict: bool = True
+):
+    state_dict = load_state_dict(path, torch_dtype=torch_dtype, device="cpu")
+    if isinstance(state_dict, dict):
+        for key in ("dit", "state_dict", "module", "model_state"):
+            nested = state_dict.get(key)
+            if isinstance(nested, dict):
+                state_dict = nested
+                break
+    missing, unexpected = _load_state_dict_assign(model, state_dict, strict=strict)
+    if missing or unexpected:
+        logger.warning(
+            "Loaded %s with missing=%d unexpected=%d", path, len(missing), len(unexpected)
+        )
+    return model
+
+
+def _load_state_dict_assign(
+    model: torch.nn.Module, state_dict: dict[str, torch.Tensor], strict: bool = True
+):
+    use_assign = os.environ.get("VPP2_LOAD_ASSIGN", "1").strip().lower() not in {
+        "0",
+        "false",
+        "no",
+        "off",
+    }
+    if use_assign and "assign" in inspect.signature(model.load_state_dict).parameters:
+        return model.load_state_dict(state_dict, strict=strict, assign=True)
+    return model.load_state_dict(state_dict, strict=strict)
+
+
+def _maybe_stagger_model_load():
+    delay_s = float(os.environ.get("VPP2_MODEL_LOAD_STAGGER_SECONDS", "0") or 0)
+    if delay_s <= 0:
+        return
+    rank = int(os.environ.get("LOCAL_RANK", os.environ.get("RANK", "0")) or 0)
+    sleep_s = delay_s * rank
+    if sleep_s > 0:
+        logger.info("Staggering Wan2.1-I2V-14B load for rank=%d by %.1f seconds.", rank, sleep_s)
+        time.sleep(sleep_s)
+
+
+def load_wan21_i2v_14b_components(
+    device: str = "cuda",
+    torch_dtype: torch.dtype = torch.bfloat16,
+    model_id: str = "weights/Wan2.1-I2V-14B-480P",
+    dit_checkpoint_path: str | os.PathLike[str] | None = None,
+    tokenizer_model_id: str | None = None,
+    tokenizer_max_len: int = 512,
+    dit_config: dict[str, Any] | None = None,
+    skip_dit_load_from_pretrain: bool = False,
+    load_text_encoder: bool = True,
+    load_clip_encoder: bool = True,
+):
+    logger.info("Loading Wan2.1-I2V-14B components...")
+    _maybe_stagger_model_load()
+    start = time.time()
+    model_dir = _resolve_model_dir(model_id)
+    if not model_dir.is_dir():
+        raise FileNotFoundError(f"Wan2.1-I2V-14B model directory does not exist: {model_dir}")
+
+    config = _load_json_config(model_dir)
+    if dit_config is not None:
+        config.update(dict(dit_config))
+    validated_dit_config = _validate_dit_config(config)
+
+    if dit_checkpoint_path:
+        dit_checkpoint = Path(dit_checkpoint_path)
+        if not dit_checkpoint.is_absolute():
+            dit_checkpoint = model_dir / dit_checkpoint
+        if not dit_checkpoint.is_file() and not skip_dit_load_from_pretrain:
+            raise FileNotFoundError(f"DiT checkpoint does not exist: {dit_checkpoint}")
+        dit_paths = [str(dit_checkpoint)]
+    else:
+        dit_paths = sorted(glob.glob(str(model_dir / "diffusion_pytorch_model-*.safetensors")))
+        if not dit_paths:
+            single = model_dir / "diffusion_pytorch_model.safetensors"
+            if single.is_file():
+                dit_paths = [str(single)]
+        if not dit_paths and not skip_dit_load_from_pretrain:
+            raise FileNotFoundError(f"No DiT checkpoint found under {model_dir}")
+
+    vae_path = model_dir / "Wan2.1_VAE.pth"
+    text_path = model_dir / "models_t5_umt5-xxl-enc-bf16.pth"
+    clip_path = model_dir / "models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"
+    tokenizer_path = (
+        Path(tokenizer_model_id) if tokenizer_model_id else model_dir / "google" / "umt5-xxl"
+    )
+
+    # This release accepts only the physically compact CP1000 architecture.
+    # It never constructs/loads the dense teacher or imports mask-training code.
+    if skip_dit_load_from_pretrain or len(dit_paths) != 1:
+        raise ValueError("CP1000 requires one explicit compact video checkpoint")
+    compact_manifest = Path(dit_paths[0]).parent / "compact_architecture.json"
+    spec = json.loads(compact_manifest.read_text())
+    from .compact_channels import build
+    dit = build(spec, device="meta")
+    if os.environ.get("VPP2_LOAD_ASSIGN", "1").lower() in {"0", "false", "no", "off"}:
+        raise ValueError("Compact meta construction requires assign loading")
+    _load_state_into_model(dit, dit_paths, torch_dtype=torch_dtype, strict=True)
+    dit = dit.to(device=device, dtype=torch_dtype)
+    dit_path_out = dit_paths[0]
+
+    vae = WanVideoVAE(z_dim=16)
+    vae_state = load_state_dict(str(vae_path), torch_dtype=torch.float32, device="cpu")
+    _load_state_dict_assign(vae, _convert_vae_state_dict(vae_state), strict=True)
+    vae = vae.eval().requires_grad_(False).to(device=device, dtype=torch.float32)
+
+    text_encoder = None
+    tokenizer = None
+    text_encoder_path = None
+    tokenizer_path_out = None
+    if load_text_encoder:
+        text_encoder = WanTextEncoder()
+        _load_state_into_model(text_encoder, str(text_path), torch_dtype=torch_dtype, strict=True)
+        text_encoder = text_encoder.to(device=device, dtype=torch_dtype)
+        tokenizer = HuggingfaceTokenizer(
+            name=str(tokenizer_path), seq_len=int(tokenizer_max_len), clean="whitespace"
+        )
+        text_encoder_path = str(text_path)
+        tokenizer_path_out = str(tokenizer_path)
+
+    clip_encoder = None
+    clip_encoder_path = None
+    if load_clip_encoder:
+        clip_encoder = WanImageEncoderCLIP()
+        clip_state = load_state_dict(str(clip_path), torch_dtype=torch.float16, device="cpu")
+        clip_state = {
+            k: v for k, v in clip_state.items() if k.startswith("visual.") or k == "log_scale"
+        }
+        missing, unexpected = _load_state_dict_assign(clip_encoder, clip_state, strict=True)
+        if missing or unexpected:
+            raise ValueError(
+                f"Unexpected CLIP visual state mismatch: missing={missing[:10]}, unexpected={unexpected[:10]}"
+            )
+        clip_encoder = (
+            clip_encoder.eval().requires_grad_(False).to(device=device, dtype=torch.float16)
+        )
+        clip_encoder_path = str(clip_path)
+
+    logger.info("Finished loading Wan2.1-I2V-14B components in %.2f seconds.", time.time() - start)
+    return Wan21I2V14BLoadedComponents(
+        dit=dit,
+        vae=vae,
+        text_encoder=text_encoder,
+        tokenizer=tokenizer,
+        clip_encoder=clip_encoder,
+        dit_path=dit_path_out,
+        vae_path=str(vae_path),
+        text_encoder_path=text_encoder_path,
+        tokenizer_path=tokenizer_path_out,
+        clip_encoder_path=clip_encoder_path,
+    )
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/mot.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/mot.py
new file mode 100644
index 0000000000000000000000000000000000000000..50bcf6d5c8d001b64de71c32b939937b9a391ef3
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/mot.py
@@ -0,0 +1,971 @@
+from __future__ import annotations
+
+from typing import Callable, Dict, Optional
+
+import torch
+import torch.nn as nn
+
+from .wan_video_dit import flash_attention, modulate, rope_apply
+from bagel_cp1000.utils.logging_config import get_logger
+
+logger = get_logger(__name__)
+
+
+class MoT(nn.Module):
+    def __init__(
+        self,
+        mixtures: Dict[str, nn.Module],
+        mot_checkpoint_mixed_attn: bool = True,
+        use_fixed_video_layers: Optional[int] = None,
+        use_video_tokens: bool = False,
+    ):
+        super().__init__()
+        if not mixtures:
+            raise ValueError("`mixtures` cannot be empty.")
+        if "video" not in mixtures or "action" not in mixtures:
+            raise ValueError("`mixtures` must include both 'video' and 'action' experts.")
+
+        self.mixtures = nn.ModuleDict(mixtures)
+        self.expert_order = list(self.mixtures.keys())
+        self.mot_checkpoint_mixed_attn = mot_checkpoint_mixed_attn
+        self.use_fixed_video_layers = (
+            None if use_fixed_video_layers is None else int(use_fixed_video_layers)
+        )
+        self.use_video_tokens = bool(use_video_tokens)
+        if mot_checkpoint_mixed_attn:
+            logger.info(
+                "Using gradient checkpointing for mixture attention. This will save memory but use more computation."
+            )
+
+        first_expert = self.mixtures[self.expert_order[0]]
+        self.num_layers = len(first_expert.blocks)
+        self.num_heads = first_expert.num_heads
+        self.attn_head_dim = first_expert.attn_head_dim
+
+        for name in self.expert_order[1:]:
+            expert = self.mixtures[name]
+            if len(expert.blocks) != self.num_layers:
+                raise ValueError(
+                    f"All experts must have same number of layers; got {self.num_layers} and {len(expert.blocks)}"
+                )
+            if expert.num_heads != self.num_heads and not self.use_video_tokens:
+                raise ValueError(
+                    f"All experts must have same num_heads; got {self.num_heads} and {expert.num_heads}"
+                )
+            if expert.attn_head_dim != self.attn_head_dim:
+                raise ValueError(
+                    "All experts must have same attn_head_dim; "
+                    f"got {self.attn_head_dim} and {expert.attn_head_dim}"
+                )
+
+        if self.use_fixed_video_layers is not None and not (
+            0 <= self.use_fixed_video_layers < self.num_layers
+        ):
+            raise ValueError(
+                "`use_fixed_video_layers` is a zero-based video layer index and must be in "
+                f"[0, {self.num_layers - 1}], got {self.use_fixed_video_layers}."
+            )
+        action_uses_video_tokens = bool(getattr(self.mixtures["action"], "use_video_tokens", False))
+        if action_uses_video_tokens != self.use_video_tokens:
+            raise ValueError(
+                "MoT and ActionDiT disagree on `use_video_tokens`: "
+                f"{self.use_video_tokens} vs {action_uses_video_tokens}."
+            )
+
+        logger.info(
+            f"Initialized MoT with experts: {self.expert_order}, num_layers={self.num_layers}"
+        )
+        for name in self.expert_order:
+            expert = self.mixtures[name]
+            logger.info(
+                f"  Expert '{name}': num_params={sum(p.numel() for p in expert.parameters()) / 1e9:.2f} B"
+            )
+
+    def _video_source_layer(self, action_layer_idx: int) -> int:
+        if self.use_fixed_video_layers is not None:
+            return self.use_fixed_video_layers
+        return int(action_layer_idx)
+
+    def materialize_action_video_kv_cache(
+        self,
+        video_cache: list[dict[str, torch.Tensor]],
+    ) -> list[dict[str, torch.Tensor]]:
+        """Project raw video tokens once for reuse across action denoising steps."""
+        if not self.use_video_tokens:
+            return video_cache
+        if len(video_cache) != self.num_layers:
+            raise ValueError(
+                f"`video_cache` must contain {self.num_layers} layers, got {len(video_cache)}."
+            )
+
+        expert = self.mixtures["action"]
+        encoded_by_source: dict[int, torch.Tensor] = {}
+        action_cache: list[dict[str, torch.Tensor]] = []
+        for layer_idx in range(self.num_layers):
+            source_layer_idx = self._video_source_layer(layer_idx)
+            source_cache = video_cache[source_layer_idx]
+            if "tokens" not in source_cache or "freqs" not in source_cache:
+                raise ValueError(
+                    f"`video_cache[{source_layer_idx}]` must contain `tokens` and `freqs`."
+                )
+            if source_layer_idx not in encoded_by_source:
+                encoded_by_source[source_layer_idx] = expert.encode_video_tokens(
+                    source_cache["tokens"]
+                )
+            k, v = expert.project_video_tokens(
+                layer_idx=layer_idx,
+                video_tokens=encoded_by_source[source_layer_idx],
+                video_freqs=source_cache["freqs"],
+            )
+            action_cache.append({"action_k": k, "action_v": v})
+        return action_cache
+
+    @staticmethod
+    def _split_modulation(block, t_mod: torch.Tensor):
+        has_seq = len(t_mod.shape) == 4
+        chunk_dim = 2 if has_seq else 1
+
+        base_mod = block.modulation.to(dtype=t_mod.dtype, device=t_mod.device)
+        shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = (base_mod + t_mod).chunk(
+            6, dim=chunk_dim
+        )
+        if has_seq:
+            # means t_mod has separate modulation for each token, otherwise same modulation for all tokens in the block
+            shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = (
+                shift_msa.squeeze(2),
+                scale_msa.squeeze(2),
+                gate_msa.squeeze(2),
+                shift_mlp.squeeze(2),
+                scale_mlp.squeeze(2),
+                gate_mlp.squeeze(2),
+            )
+        return shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp
+
+    def _mixed_attention(
+        self,
+        q_cat: torch.Tensor,
+        k_cat: torch.Tensor,
+        v_cat: torch.Tensor,
+        attention_mask: torch.Tensor,
+        num_heads: Optional[int] = None,
+    ) -> torch.Tensor:
+        attn_mask = attention_mask.to(device=q_cat.device)
+        attention_heads = self.num_heads if num_heads is None else int(num_heads)
+
+        def _forward(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor) -> torch.Tensor:
+            return flash_attention(q=q, k=k, v=v, num_heads=attention_heads, ctx_mask=attn_mask)
+
+        if self.mot_checkpoint_mixed_attn and self.training and (not getattr(self,'skip_nograd_checkpointing',False) or torch.is_grad_enabled()):
+            return torch.utils.checkpoint.checkpoint(
+                _forward,
+                q_cat,
+                k_cat,
+                v_cat,
+                use_reentrant=False,
+            )
+        return _forward(q_cat, k_cat, v_cat)
+
+    @staticmethod
+    def _apply_expert_post_block(
+        block,
+        residual_x: torch.Tensor,
+        mixed_attn_out: torch.Tensor,
+        gate_msa: torch.Tensor,
+        shift_mlp: torch.Tensor,
+        scale_mlp: torch.Tensor,
+        gate_mlp: torch.Tensor,
+        context_payload: Optional[dict],
+    ) -> torch.Tensor:
+        x = block.gate(residual_x, gate_msa, block.self_attn.o(mixed_attn_out))
+
+        if context_payload is not None:
+            context = context_payload.get("context")
+            if context is not None:
+                context_mask = context_payload.get("mask")
+                if context_mask is not None and context_mask.dim() == 3:
+                    context_mask = context_mask.unsqueeze(1)
+                x = x + block.cross_attn(block.norm3(x), context, ctx_mask=context_mask)
+
+        mlp_input = modulate(block.norm2(x), shift_mlp, scale_mlp)
+        x = block.gate(x, gate_mlp, block.ffn(mlp_input))
+        return x
+
+    def _build_expert_attention_io(
+        self,
+        expert,
+        block,
+        x: torch.Tensor,
+        freqs: torch.Tensor,
+        t_mod: torch.Tensor,
+    ) -> tuple[
+        torch.Tensor,
+        torch.Tensor,
+        torch.Tensor,
+        torch.Tensor,
+        torch.Tensor,
+        torch.Tensor,
+        torch.Tensor,
+        torch.Tensor,
+        bool,
+    ]:
+        """Build per-expert attention tensors and post-block states.
+
+        Args:
+            expert: Expert module that owns this `block`; only used to read
+                `use_gradient_checkpointing`.
+            block: Transformer block for current layer (`expert.blocks[layer_idx]`).
+            x: Current expert tokens, shape [B, S, D].
+            freqs: RoPE frequencies aligned with token sequence, shape [S, 1, rope_dim].
+            t_mod: Time modulation tensor for this expert/layer.
+
+        Returns:
+            q: Query after q-proj, RMSNorm, and RoPE, shape [B, S, H*Dh].
+            k: Key after k-proj, RMSNorm, and RoPE, shape [B, S, H*Dh].
+            v: Value after v-proj, shape [B, S, H*Dh].
+            residual_x: Original input `x` for residual path in post block.
+            gate_msa: Gating tensor for self-attention residual branch.
+            shift_mlp: Shift tensor for MLP modulation.
+            scale_mlp: Scale tensor for MLP modulation.
+            gate_mlp: Gating tensor for MLP residual branch.
+            use_gradient_checkpointing: Whether this expert enables checkpointing.
+        """
+        shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = self._split_modulation(
+            block, t_mod
+        )
+        attn_input = modulate(block.norm1(x), shift_msa, scale_msa)
+
+        q = block.self_attn.norm_q(block.self_attn.q(attn_input))
+        k = block.self_attn.norm_k(block.self_attn.k(attn_input))
+        v = block.self_attn.v(attn_input)
+
+        q = rope_apply(q, freqs, block.num_heads)
+        k = rope_apply(k, freqs, block.num_heads)
+
+        use_gradient_checkpointing = bool(getattr(expert, "use_gradient_checkpointing", False))
+        return (
+            q,
+            k,
+            v,
+            x,
+            gate_msa,
+            shift_mlp,
+            scale_mlp,
+            gate_mlp,
+            use_gradient_checkpointing,
+        )
+
+    def _apply_post_with_optional_checkpoint(
+        self,
+        block,
+        residual_x: torch.Tensor,
+        gate_msa: torch.Tensor,
+        shift_mlp: torch.Tensor,
+        scale_mlp: torch.Tensor,
+        gate_mlp: torch.Tensor,
+        use_gradient_checkpointing: bool,
+        mixed_slice: torch.Tensor,
+        context_payload: Optional[dict],
+    ) -> torch.Tensor:
+        """Apply post-attention computations, with optional checkpointing.
+
+        Args:
+            block: Transformer block for current layer.
+            residual_x: Residual input tokens before attention update, shape [B, S, D].
+            gate_msa: Gating tensor used after mixed self-attention.
+            shift_mlp: Shift tensor for MLP input modulation.
+            scale_mlp: Scale tensor for MLP input modulation.
+            gate_mlp: Gating tensor used after MLP.
+            use_gradient_checkpointing: If True and training, checkpoint this post block.
+            mixed_slice: Mixed-attention output for this expert, shape [B, S, H*Dh].
+            context_payload: Optional dict for cross-attention.
+                - `context`: encoder states [B, L, D]
+                - `mask`: attention mask [B, S, L] or [B, 1, S, L]
+
+        Returns:
+            Updated expert tokens after self-attn residual, optional cross-attn, and MLP.
+        """
+
+        def _post_fn(
+            _mixed_slice: torch.Tensor,
+            _x: torch.Tensor,
+            _gate_msa: torch.Tensor,
+            _shift_mlp: torch.Tensor,
+            _scale_mlp: torch.Tensor,
+            _gate_mlp: torch.Tensor,
+            _block=block,
+            _context_payload=context_payload,
+        ) -> torch.Tensor:
+            return self._apply_expert_post_block(
+                block=_block,
+                residual_x=_x,
+                mixed_attn_out=_mixed_slice,
+                gate_msa=_gate_msa,
+                shift_mlp=_shift_mlp,
+                scale_mlp=_scale_mlp,
+                gate_mlp=_gate_mlp,
+                context_payload=_context_payload,
+            )
+
+        if use_gradient_checkpointing and self.training and (not getattr(self,'skip_nograd_checkpointing',False) or torch.is_grad_enabled()):
+            return torch.utils.checkpoint.checkpoint(
+                _post_fn,
+                mixed_slice,
+                residual_x,
+                gate_msa,
+                shift_mlp,
+                scale_mlp,
+                gate_mlp,
+                use_reentrant=False,
+            )
+        return _post_fn(
+            mixed_slice,
+            residual_x,
+            gate_msa,
+            shift_mlp,
+            scale_mlp,
+            gate_mlp,
+        )
+
+    def prefill_video_cache(
+        self,
+        video_tokens: torch.Tensor,
+        video_freqs: torch.Tensor,
+        video_t_mod: torch.Tensor,
+        video_context_payload: Optional[dict],
+        video_attention_mask: torch.Tensor,
+        return_final_tokens: bool = False,
+        max_layers: Optional[int] = None,
+        video_feature_callback: Optional[Callable[[int, torch.Tensor], None]] = None,
+    ) -> list[dict[str, torch.Tensor]] | tuple[list[dict[str, torch.Tensor]], torch.Tensor]:
+        """Prefill video branch once and cache K/V or raw tokens for action.
+
+        Args:
+            video_tokens: Video tokens before layer 0, shape [B, Sv, D].
+            video_freqs: Video RoPE frequencies, shape [Sv, 1, rope_dim].
+            video_t_mod: Video time modulation tensor.
+            video_context_payload: Optional dict for video cross-attention.
+                - `context`: encoder states [B, L, D]
+                - `mask`: attention mask [B, Sv, L] or [B, 1, Sv, L]
+            video_attention_mask: Video self-attention mask, shape [Sv, Sv].
+            video_feature_callback: Optional read-only collector called with
+                (zero-based layer index, post-layer tokens), outside activation
+                checkpointing. Must not modify the tokens in place.
+
+        Returns:
+            Layer-wise cache list with length `num_layers`, or `max_layers`
+            when a bounded prefill is requested.
+            In direct mode, required entries contain video `k`/`v`. With
+            `use_video_tokens=true`, required entries contain pre-layer video
+            `tokens` and their RoPE `freqs` instead.
+        """
+        if "video" not in self.mixtures:
+            raise ValueError("MoT requires `video` expert for `prefill_video_cache`.")
+        if video_attention_mask.ndim != 2:
+            raise ValueError(
+                f"`video_attention_mask` must be 2D [S,S], got shape {tuple(video_attention_mask.shape)}"
+            )
+        if video_attention_mask.shape[0] != video_attention_mask.shape[1]:
+            raise ValueError(
+                f"`video_attention_mask` must be square, got shape {tuple(video_attention_mask.shape)}"
+            )
+        if video_attention_mask.shape[0] != video_tokens.shape[1]:
+            raise ValueError(
+                "`video_attention_mask` seq length mismatch: "
+                f"mask={video_attention_mask.shape[0]} vs tokens={video_tokens.shape[1]}"
+            )
+
+        layer_count = self.num_layers if max_layers is None else int(max_layers)
+        if not 1 <= layer_count <= self.num_layers:
+            raise ValueError(f"`max_layers` must be in [1, {self.num_layers}], got {layer_count}.")
+        if self.use_fixed_video_layers is not None and self.use_fixed_video_layers >= layer_count:
+            raise ValueError(
+                "`max_layers` must include `use_fixed_video_layers`; got "
+                f"max_layers={layer_count}, fixed={self.use_fixed_video_layers}."
+            )
+
+        expert = self.mixtures["video"]
+        x = video_tokens
+        kv_cache: list[dict[str, torch.Tensor]] = []
+        for layer_idx in range(layer_count):
+            block = expert.blocks[layer_idx]
+            if getattr(block, "compact_cache_block", False):
+                if self.use_video_tokens or self.use_fixed_video_layers is not None:
+                    raise ValueError("Compact K/V blocks require native direct per-layer K/V")
+                payload = video_context_payload or {}
+                args = (x, video_t_mod, video_freqs, payload.get("context"),
+                        payload.get("mask"), video_attention_mask)
+                if self.training and bool(getattr(expert, "use_gradient_checkpointing", False)):
+                    x, k, v = torch.utils.checkpoint.checkpoint(block, *args, use_reentrant=False)
+                else:
+                    x, k, v = block(*args)
+                kv_cache.append({"k": k, "v": v})
+                if video_feature_callback is not None:
+                    video_feature_callback(layer_idx, x)
+                continue
+            # Build video Q/K/V from current layer input tokens.
+            (
+                q,
+                k,
+                v,
+                residual_x,
+                gate_msa,
+                shift_mlp,
+                scale_mlp,
+                gate_mlp,
+                use_gradient_checkpointing,
+            ) = self._build_expert_attention_io(
+                expert=expert,
+                block=block,
+                x=x,
+                freqs=video_freqs,
+                t_mod=video_t_mod,
+            )
+            # Video prefill uses only video self-attention mask.
+            mixed = self._mixed_attention(
+                q_cat=q,
+                k_cat=k,
+                v_cat=v,
+                attention_mask=video_attention_mask,
+                num_heads=expert.num_heads,
+            )
+            cache_entry: dict[str, torch.Tensor] = {}
+            if self.use_video_tokens:
+                if self.use_fixed_video_layers is None or layer_idx == self.use_fixed_video_layers:
+                    cache_entry.update({"tokens": x, "freqs": video_freqs})
+            elif self.use_fixed_video_layers is None or layer_idx == self.use_fixed_video_layers:
+                cache_entry.update({"k": k, "v": v})
+
+            # Update video tokens for the next layer and persist its action input.
+            x = self._apply_post_with_optional_checkpoint(
+                block=block,
+                residual_x=residual_x,
+                gate_msa=gate_msa,
+                shift_mlp=shift_mlp,
+                scale_mlp=scale_mlp,
+                gate_mlp=gate_mlp,
+                use_gradient_checkpointing=use_gradient_checkpointing,
+                mixed_slice=mixed,
+                context_payload=video_context_payload,
+            )
+            kv_cache.append(cache_entry)
+            if video_feature_callback is not None:
+                # Outside checkpointed blocks: retain the original graph, and
+                # do not collect again during backward recomputation.
+                video_feature_callback(layer_idx, x)
+        if return_final_tokens:
+            return kv_cache, x
+        return kv_cache
+
+    def forward_action_with_video_cache(
+        self,
+        action_tokens: torch.Tensor,
+        action_freqs: torch.Tensor,
+        action_t_mod: torch.Tensor,
+        action_context_payload: Optional[dict],
+        video_kv_cache: list[dict[str, torch.Tensor]],
+        attention_mask: torch.Tensor,
+        video_seq_len: int,
+        clean_first_frame_kv_cache: Optional[list[dict[str, torch.Tensor]]] = None,
+        clean_first_frame_attention_mask: Optional[torch.Tensor] = None,
+        clean_first_frame_video_seq_len: Optional[int] = None,
+        clean_first_frame_start_layer: Optional[int] = None,
+        action_video_conditioning_mask: Optional[torch.Tensor] = None,
+        action_clean_first_frame_conditioning_mask: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        """Run action branch with cached video K/V instead of recomputing video tokens.
+
+        Args:
+            action_tokens: Action tokens before layer 0, shape [B, Sa, D].
+            action_freqs: Action RoPE frequencies, shape [Sa, 1, rope_dim].
+            action_t_mod: Action time modulation tensor.
+            action_context_payload: Optional dict for action cross-attention.
+                - `context`: encoder states [B, L, D]
+                - `mask`: attention mask [B, Sa, L] or [B, 1, Sa, L]
+            video_kv_cache: Layer-wise cached video K/V from `prefill_video_cache`.
+            attention_mask: Joint [video+action] mask, shape [Sv+Sa, Sv+Sa].
+            video_seq_len: Video token count `Sv` in the joint sequence prefix.
+            clean_first_frame_*: Optional first-frame-only cache/mask/prefix
+                used from `clean_first_frame_start_layer` through the final
+                ActionDiT layer. The cache may come from an independent clean
+                prefill or from frame 0 of the noisy cache; this mode is
+                restricted to direct per-layer video K/V.
+            action_video_conditioning_mask: Optional per-sample boolean vector
+                [B]. False removes every cached video key from that sample's
+                action attention while preserving action self-attention.
+            action_clean_first_frame_conditioning_mask: Optional per-sample
+                boolean vector [B]. True replaces the noisy-video prefix with
+                the supplied first-frame-only cache for every ActionDiT layer.
+                Mixed batches retain one common padded attention width.
+
+        Returns:
+            Updated action tokens after all layers, shape [B, Sa, D].
+        """
+        if "action" not in self.mixtures:
+            raise ValueError("MoT requires `action` expert for `forward_action_with_video_cache`.")
+        action_seq_len = int(action_tokens.shape[1])
+        split_enabled = clean_first_frame_start_layer is not None
+        timestep_clean_enabled = action_clean_first_frame_conditioning_mask is not None
+        if split_enabled and timestep_clean_enabled:
+            raise ValueError(
+                "Layer-split and timestep-routed clean-first-frame conditioning "
+                "cannot be enabled together."
+            )
+        if timestep_clean_enabled and action_video_conditioning_mask is not None:
+            raise ValueError(
+                "Clean-first-frame timestep routing and no-video timestep gating "
+                "cannot be enabled together."
+            )
+        clean_cache_enabled = split_enabled or timestep_clean_enabled
+        if clean_cache_enabled:
+            if self.use_fixed_video_layers is not None or self.use_video_tokens:
+                raise ValueError("Clean-first-frame routing requires direct per-layer video K/V.")
+            if clean_first_frame_kv_cache is None:
+                raise ValueError("Clean-first-frame routing requires `clean_first_frame_kv_cache`.")
+            if clean_first_frame_attention_mask is None:
+                raise ValueError(
+                    "Clean-first-frame routing requires `clean_first_frame_attention_mask`."
+                )
+            if clean_first_frame_video_seq_len is None:
+                raise ValueError(
+                    "Clean-first-frame routing requires `clean_first_frame_video_seq_len`."
+                )
+            if len(clean_first_frame_kv_cache) != self.num_layers:
+                raise ValueError(
+                    "`clean_first_frame_kv_cache` must contain "
+                    f"{self.num_layers} layers, got {len(clean_first_frame_kv_cache)}."
+                )
+        if split_enabled:
+            split_layer = int(clean_first_frame_start_layer)
+            if not 0 < split_layer < self.num_layers:
+                raise ValueError(
+                    "`clean_first_frame_start_layer` must be in "
+                    f"[1, {self.num_layers - 1}], got {split_layer}."
+                )
+            if len(video_kv_cache) < split_layer:
+                raise ValueError(
+                    "`video_kv_cache` must cover every noisy-video ActionDiT layer; "
+                    f"need at least {split_layer}, got {len(video_kv_cache)}."
+                )
+        else:
+            split_layer = self.num_layers
+            if len(video_kv_cache) != self.num_layers:
+                raise ValueError(
+                    f"`video_kv_cache` must contain {self.num_layers} layers, got {len(video_kv_cache)}."
+                )
+
+        def _action_mask_rows(
+            mask: torch.Tensor,
+            prefix_seq_len: int,
+            name: str,
+        ) -> torch.Tensor:
+            if mask.ndim != 2:
+                raise ValueError(f"`{name}` must be 2D [S,S], got shape {tuple(mask.shape)}")
+            if mask.shape[0] != mask.shape[1]:
+                raise ValueError(f"`{name}` must be square, got shape {tuple(mask.shape)}")
+            total = int(prefix_seq_len) + action_seq_len
+            if mask.shape[0] != total:
+                raise ValueError(
+                    f"`{name}` seq length mismatch: mask={mask.shape[0]} vs expected_total={total}"
+                )
+            return mask[prefix_seq_len:total, :total]
+
+        action_attention_mask = _action_mask_rows(
+            attention_mask,
+            int(video_seq_len),
+            "attention_mask",
+        )
+        clean_action_attention_mask = None
+        if clean_cache_enabled:
+            clean_action_attention_mask = _action_mask_rows(
+                clean_first_frame_attention_mask,
+                int(clean_first_frame_video_seq_len),
+                "clean_first_frame_attention_mask",
+            )
+
+        clean_conditioning = None
+        any_clean_first_frame = False
+        all_clean_first_frame = False
+        if timestep_clean_enabled:
+            clean_conditioning = action_clean_first_frame_conditioning_mask.to(
+                device=action_tokens.device,
+                dtype=torch.bool,
+            )
+            if clean_conditioning.ndim != 1:
+                raise ValueError(
+                    "`action_clean_first_frame_conditioning_mask` must be 1D "
+                    f"[B], got shape {tuple(clean_conditioning.shape)}"
+                )
+            batch_size = int(action_tokens.shape[0])
+            if clean_conditioning.shape[0] != batch_size:
+                raise ValueError(
+                    "`action_clean_first_frame_conditioning_mask` batch mismatch: "
+                    f"mask={clean_conditioning.shape[0]} vs action={batch_size}"
+                )
+            if int(clean_first_frame_video_seq_len) > int(video_seq_len):
+                raise ValueError(
+                    "Clean-first-frame prefix cannot exceed noisy-video prefix: "
+                    f"clean={clean_first_frame_video_seq_len}, noisy={video_seq_len}."
+                )
+            any_clean_first_frame = bool(clean_conditioning.any().item())
+            all_clean_first_frame = bool(clean_conditioning.all().item())
+            if all_clean_first_frame:
+                action_attention_mask = clean_action_attention_mask
+            elif any_clean_first_frame:
+                clean_prefix_len = int(clean_first_frame_video_seq_len)
+                noisy_prefix_len = int(video_seq_len)
+                clean_prefix_mask = clean_action_attention_mask[:, :clean_prefix_len]
+                clean_padding_mask = torch.zeros(
+                    (
+                        action_seq_len,
+                        noisy_prefix_len - clean_prefix_len,
+                    ),
+                    device=action_attention_mask.device,
+                    dtype=torch.bool,
+                )
+                action_self_mask = action_attention_mask[:, noisy_prefix_len:]
+                padded_clean_mask = torch.cat(
+                    [clean_prefix_mask, clean_padding_mask, action_self_mask],
+                    dim=-1,
+                )
+                action_attention_mask = torch.where(
+                    clean_conditioning[:, None, None],
+                    padded_clean_mask.unsqueeze(0),
+                    action_attention_mask.unsqueeze(0),
+                )
+
+        def _apply_video_conditioning_gate(
+            mask: torch.Tensor,
+            prefix_seq_len: int,
+        ) -> torch.Tensor:
+            if action_video_conditioning_mask is None:
+                return mask
+            conditioning = action_video_conditioning_mask.to(
+                device=mask.device,
+                dtype=torch.bool,
+            )
+            if conditioning.ndim != 1:
+                raise ValueError(
+                    "`action_video_conditioning_mask` must be 1D [B], got "
+                    f"shape {tuple(conditioning.shape)}"
+                )
+            batch_size = int(action_tokens.shape[0])
+            if conditioning.shape[0] != batch_size:
+                raise ValueError(
+                    "`action_video_conditioning_mask` batch mismatch: "
+                    f"mask={conditioning.shape[0]} vs action={batch_size}"
+                )
+            gated = mask.unsqueeze(0).expand(batch_size, -1, -1).clone()
+            gated[:, :, :prefix_seq_len] &= conditioning[:, None, None]
+            return gated
+
+        action_attention_mask = _apply_video_conditioning_gate(
+            action_attention_mask,
+            int(video_seq_len),
+        )
+        if clean_action_attention_mask is not None:
+            clean_action_attention_mask = _apply_video_conditioning_gate(
+                clean_action_attention_mask,
+                int(clean_first_frame_video_seq_len),
+            )
+        all_video_disabled = action_video_conditioning_mask is not None and not bool(
+            action_video_conditioning_mask.to(dtype=torch.bool).any().item()
+        )
+
+        expert = self.mixtures["action"]
+        x = action_tokens
+        has_materialized_action_kv = self.use_video_tokens and all(
+            "action_k" in layer_cache and "action_v" in layer_cache
+            for layer_cache in video_kv_cache
+        )
+        encoded_video_tokens: dict[int, torch.Tensor] = {}
+        if self.use_video_tokens and not has_materialized_action_kv and not all_video_disabled:
+            for source_layer_idx in {
+                self._video_source_layer(layer_idx) for layer_idx in range(self.num_layers)
+            }:
+                source_cache = video_kv_cache[source_layer_idx]
+                if "tokens" not in source_cache:
+                    raise ValueError(
+                        f"`video_kv_cache[{source_layer_idx}]` must contain `tokens` "
+                        "when `use_video_tokens=true`."
+                    )
+                encoded_video_tokens[source_layer_idx] = expert.encode_video_tokens(
+                    source_cache["tokens"]
+                )
+        for layer_idx in range(self.num_layers):
+            block = expert.blocks[layer_idx]
+            # Action query/key/value are still step-dependent and must be recomputed each step.
+            (
+                q_action,
+                k_action,
+                v_action,
+                residual_x,
+                gate_msa,
+                shift_mlp,
+                scale_mlp,
+                gate_mlp,
+                use_gradient_checkpointing,
+            ) = self._build_expert_attention_io(
+                expert=expert,
+                block=block,
+                x=x,
+                freqs=action_freqs,
+                t_mod=action_t_mod,
+            )
+            use_clean_first_frame = (split_enabled and layer_idx >= split_layer) or (
+                timestep_clean_enabled and all_clean_first_frame
+            )
+            if use_clean_first_frame:
+                source_layer_idx = layer_idx
+                layer_cache = clean_first_frame_kv_cache[source_layer_idx]
+                selected_video_seq_len = int(clean_first_frame_video_seq_len)
+                selected_action_attention_mask = clean_action_attention_mask
+                cache_name = "clean_first_frame_kv_cache"
+            else:
+                source_layer_idx = self._video_source_layer(layer_idx)
+                layer_cache = video_kv_cache[source_layer_idx]
+                selected_video_seq_len = int(video_seq_len)
+                selected_action_attention_mask = action_attention_mask
+                cache_name = "video_kv_cache"
+            if all_video_disabled:
+                # The late denoising path has no video-visible samples. Avoid
+                # projecting or concatenating masked video K/V, so inference
+                # refinement also gets the intended attention-width reduction.
+                k_cat = k_action
+                v_cat = v_action
+                selected_action_attention_mask = selected_action_attention_mask[
+                    ..., selected_video_seq_len:
+                ]
+            elif self.use_video_tokens:
+                if has_materialized_action_kv:
+                    k_video = video_kv_cache[layer_idx]["action_k"]
+                    v_video = video_kv_cache[layer_idx]["action_v"]
+                else:
+                    if "freqs" not in layer_cache:
+                        raise ValueError(
+                            f"`video_kv_cache[{source_layer_idx}]` must contain `freqs`."
+                        )
+                    k_video, v_video = expert.project_video_tokens(
+                        layer_idx=layer_idx,
+                        video_tokens=encoded_video_tokens[source_layer_idx],
+                        video_freqs=layer_cache["freqs"],
+                    )
+            else:
+                if "k" not in layer_cache or "v" not in layer_cache:
+                    raise ValueError(
+                        f"`{cache_name}[{source_layer_idx}]` must contain `k` and `v`."
+                    )
+                k_video = layer_cache["k"]
+                v_video = layer_cache["v"]
+            if timestep_clean_enabled and any_clean_first_frame and not all_clean_first_frame:
+                clean_layer_cache = clean_first_frame_kv_cache[layer_idx]
+                if "k" not in clean_layer_cache or "v" not in clean_layer_cache:
+                    raise ValueError(
+                        "`clean_first_frame_kv_cache` entries must contain `k` and `v`."
+                    )
+                clean_k = clean_layer_cache["k"]
+                clean_v = clean_layer_cache["v"]
+                clean_prefix_len = int(clean_first_frame_video_seq_len)
+                if clean_k.shape[1] != clean_prefix_len or clean_v.shape[1] != clean_prefix_len:
+                    raise ValueError(
+                        "`clean_first_frame_kv_cache` seq len mismatch, expected "
+                        f"{clean_prefix_len}."
+                    )
+                clean_k_padded = torch.cat(
+                    [clean_k, k_video[:, clean_prefix_len:]],
+                    dim=1,
+                )
+                clean_v_padded = torch.cat(
+                    [clean_v, v_video[:, clean_prefix_len:]],
+                    dim=1,
+                )
+                k_video = torch.where(
+                    clean_conditioning[:, None, None],
+                    clean_k_padded,
+                    k_video,
+                )
+                v_video = torch.where(
+                    clean_conditioning[:, None, None],
+                    clean_v_padded,
+                    v_video,
+                )
+            if not all_video_disabled:
+                if (
+                    k_video.shape[1] != selected_video_seq_len
+                    or v_video.shape[1] != selected_video_seq_len
+                ):
+                    raise ValueError(
+                        f"`{cache_name}[{source_layer_idx}]` seq len mismatch, "
+                        f"expected {selected_video_seq_len}."
+                    )
+
+                # Mixed attention: action queries attend to cached video K/V
+                # plus current action K/V.
+                k_cat = torch.cat([k_video, k_action], dim=1)
+                v_cat = torch.cat([v_video, v_action], dim=1)
+            mixed = self._mixed_attention(
+                q_cat=q_action,
+                k_cat=k_cat,
+                v_cat=v_cat,
+                attention_mask=selected_action_attention_mask,
+                num_heads=expert.num_heads,
+            )
+            x = self._apply_post_with_optional_checkpoint(
+                block=block,
+                residual_x=residual_x,
+                gate_msa=gate_msa,
+                shift_mlp=shift_mlp,
+                scale_mlp=scale_mlp,
+                gate_mlp=gate_mlp,
+                use_gradient_checkpointing=use_gradient_checkpointing,
+                mixed_slice=mixed,
+                context_payload=action_context_payload,
+            )
+        return x
+
+    def forward(
+        self,
+        embeds_all: Dict[str, torch.Tensor],
+        attention_mask: torch.Tensor,
+        freqs_all: Dict[str, torch.Tensor],
+        context_all: Dict[str, Optional[dict]],
+        t_mod_all: Dict[str, torch.Tensor],
+        video_feature_callback: Optional[Callable[[int, torch.Tensor], None]] = None,
+    ):
+        missing = [k for k in self.expert_order if k not in embeds_all]
+        if missing:
+            raise ValueError(f"Missing expert tokens for {missing}")
+        missing = [k for k in self.expert_order if k not in freqs_all]
+        if missing:
+            raise ValueError(f"Missing expert freqs for {missing}")
+        missing = [k for k in self.expert_order if k not in t_mod_all]
+        if missing:
+            raise ValueError(f"Missing expert t_mod for {missing}")
+
+        if attention_mask.ndim != 2:
+            raise ValueError(
+                f"`attention_mask` must be 2D [S, S], got shape {tuple(attention_mask.shape)}"
+            )
+        if attention_mask.shape[0] != attention_mask.shape[1]:
+            raise ValueError(
+                f"`attention_mask` must be square, got shape {tuple(attention_mask.shape)}"
+            )
+
+        if self.use_fixed_video_layers is not None or self.use_video_tokens:
+            video_seq_len = int(embeds_all["video"].shape[1])
+            action_seq_len = int(embeds_all["action"].shape[1])
+            if attention_mask.shape[0] != video_seq_len + action_seq_len:
+                raise ValueError(
+                    "Attention mask seq length mismatch: "
+                    f"mask={attention_mask.shape[0]} vs tokens={video_seq_len + action_seq_len}"
+                )
+            video_cache, final_video_tokens = self.prefill_video_cache(
+                video_tokens=embeds_all["video"],
+                video_freqs=freqs_all["video"],
+                video_t_mod=t_mod_all["video"],
+                video_context_payload=context_all.get("video"),
+                video_attention_mask=attention_mask[:video_seq_len, :video_seq_len],
+                return_final_tokens=True,
+                video_feature_callback=video_feature_callback,
+            )
+            final_action_tokens = self.forward_action_with_video_cache(
+                action_tokens=embeds_all["action"],
+                action_freqs=freqs_all["action"],
+                action_t_mod=t_mod_all["action"],
+                action_context_payload=context_all.get("action"),
+                video_kv_cache=video_cache,
+                attention_mask=attention_mask,
+                video_seq_len=video_seq_len,
+            )
+            return {"video": final_video_tokens, "action": final_action_tokens}
+
+        tokens_all = {k: v for k, v in embeds_all.items()}
+
+        for layer_idx in range(self.num_layers):
+            q_chunks = []
+            k_chunks = []
+            v_chunks = []
+            cached = {}
+            seq_lens = []
+
+            for name in self.expert_order:
+                expert = self.mixtures[name]
+                block = expert.blocks[layer_idx]
+                x = tokens_all[name]
+                freqs = freqs_all[name]
+                t_mod = t_mod_all[name]
+
+                (
+                    q,
+                    k,
+                    v,
+                    residual_x,
+                    gate_msa,
+                    shift_mlp,
+                    scale_mlp,
+                    gate_mlp,
+                    use_gradient_checkpointing,
+                ) = self._build_expert_attention_io(
+                    expert=expert,
+                    block=block,
+                    x=x,
+                    freqs=freqs,
+                    t_mod=t_mod,
+                )
+
+                q_chunks.append(q)
+                k_chunks.append(k)
+                v_chunks.append(v)
+                seq_lens.append(x.shape[1])
+                cached[name] = {
+                    "block": block,
+                    "residual_x": residual_x,
+                    "gate_msa": gate_msa,
+                    "shift_mlp": shift_mlp,
+                    "scale_mlp": scale_mlp,
+                    "gate_mlp": gate_mlp,
+                    "use_gradient_checkpointing": use_gradient_checkpointing,
+                }
+
+            # 3. concat all tokens for mixed attention
+            q_cat = torch.cat(q_chunks, dim=1)
+            k_cat = torch.cat(k_chunks, dim=1)
+            v_cat = torch.cat(v_chunks, dim=1)
+
+            total_seq = q_cat.shape[1]
+            if attention_mask.shape[0] != total_seq:
+                raise ValueError(
+                    "Attention mask seq length mismatch: "
+                    f"mask={attention_mask.shape[0]} vs tokens={total_seq}"
+                )
+
+            mixed = self._mixed_attention(
+                q_cat=q_cat, k_cat=k_cat, v_cat=v_cat, attention_mask=attention_mask
+            )
+
+            start = 0
+            for name, seq_len in zip(self.expert_order, seq_lens):
+                # 4. split mixed attention output and apply post-attention blocks for each expert
+                end = start + seq_len
+                mixed_slice = mixed[:, start:end, :]
+                cached_expert = cached[name]
+                block = cached_expert["block"]
+                context_payload = context_all.get(name)
+
+                updated_tokens = self._apply_post_with_optional_checkpoint(
+                    block=block,
+                    residual_x=cached_expert["residual_x"],
+                    gate_msa=cached_expert["gate_msa"],
+                    shift_mlp=cached_expert["shift_mlp"],
+                    scale_mlp=cached_expert["scale_mlp"],
+                    gate_mlp=cached_expert["gate_mlp"],
+                    use_gradient_checkpointing=cached_expert["use_gradient_checkpointing"],
+                    mixed_slice=mixed_slice,
+                    context_payload=context_payload,
+                )
+
+                tokens_all[name] = updated_tokens
+                if name == "video" and video_feature_callback is not None:
+                    video_feature_callback(layer_idx, updated_tokens)
+                start = end
+
+        return tokens_all
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/schedulers/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/schedulers/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/schedulers/scheduler_continuous.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/schedulers/scheduler_continuous.py
new file mode 100644
index 0000000000000000000000000000000000000000..7c9f886fa97fb5868c4570de06f9fed284978c86
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/schedulers/scheduler_continuous.py
@@ -0,0 +1,130 @@
+import torch
+
+
+class WanContinuousFlowMatchScheduler:
+    """Continuous-time Flow-Matching scheduler with configurable training sampling."""
+
+    def __init__(
+        self,
+        num_train_timesteps: int = 1000,
+        shift: float = 5.0,
+        eps: float = 1e-10,
+        train_sampling_strategy: str = "shifted_uniform",
+        beta_alpha: float = 7.0,
+        beta_beta: float = 1.0,
+    ):
+        if num_train_timesteps <= 0:
+            raise ValueError(f"`num_train_timesteps` must be positive, got {num_train_timesteps}")
+        if shift <= 0:
+            raise ValueError(f"`shift` must be positive, got {shift}")
+        if beta_alpha <= 0 or beta_beta <= 0:
+            raise ValueError(
+                f"`beta_alpha` and `beta_beta` must be positive, got {beta_alpha}, {beta_beta}"
+            )
+        self.num_train_timesteps = int(num_train_timesteps)
+        self.shift = float(shift)
+        self.eps = float(eps)
+        self.train_sampling_strategy = str(train_sampling_strategy)
+        self.beta_alpha = float(beta_alpha)
+        self.beta_beta = float(beta_beta)
+        if self.train_sampling_strategy not in {"shifted_uniform", "beta", "pure_noise"}:
+            raise ValueError(
+                f"Unsupported train_sampling_strategy: {self.train_sampling_strategy}. "
+                "Expected one of: shifted_uniform, beta, pure_noise."
+            )
+        self._y_min, self._weight_norm_const = self._precompute_training_weight_stats()
+
+    @staticmethod
+    def _phi(u: torch.Tensor, shift: float) -> torch.Tensor:
+        return shift * u / (1.0 + (shift - 1.0) * u)
+
+    def _precompute_training_weight_stats(self) -> tuple[float, float]:
+        steps = self.num_train_timesteps
+        u_grid = torch.linspace(1.0, 0.0, steps + 1, dtype=torch.float64)[:-1]
+        t_grid = self._phi(u_grid, self.shift) * float(steps)
+        y_grid = torch.exp(-2.0 * ((t_grid - (steps / 2.0)) / steps) ** 2)
+        y_min = float(y_grid.min().item())
+        y_shifted_grid = y_grid - y_min
+        norm_const = float(y_shifted_grid.mean().item())
+        return y_min, norm_const
+
+    def sample_training_t(
+        self, batch_size: int, device: torch.device, dtype: torch.dtype
+    ) -> torch.Tensor:
+        if batch_size <= 0:
+            raise ValueError(f"`batch_size` must be positive, got {batch_size}")
+        if self.train_sampling_strategy == "shifted_uniform":
+            u = torch.rand((batch_size,), device=device, dtype=torch.float32)
+            sigma = self._phi(u, self.shift)
+        elif self.train_sampling_strategy == "beta":
+            alpha = torch.full((batch_size,), self.beta_alpha, device=device, dtype=torch.float32)
+            beta = torch.full((batch_size,), self.beta_beta, device=device, dtype=torch.float32)
+            sigma = torch._standard_gamma(alpha)
+            sigma = sigma / (sigma + torch._standard_gamma(beta)).clamp(min=self.eps)
+            sigma = sigma.clamp(min=0.0, max=1.0)
+        elif self.train_sampling_strategy == "pure_noise":
+            # sigma=1 makes add_noise return the noise exactly, with no clean sample mixed in.
+            sigma = torch.ones((batch_size,), device=device, dtype=torch.float32)
+        else:
+            raise RuntimeError(f"Unhandled train_sampling_strategy: {self.train_sampling_strategy}")
+        timestep = sigma * float(self.num_train_timesteps)
+        return timestep.to(dtype=dtype)
+
+    def training_weight(self, timestep: torch.Tensor) -> torch.Tensor:
+        if self.train_sampling_strategy == "pure_noise":
+            return torch.ones_like(timestep, dtype=torch.float32)
+        t = timestep.to(dtype=torch.float32)
+        steps = float(self.num_train_timesteps)
+        y = torch.exp(-2.0 * ((t - (steps / 2.0)) / steps) ** 2)
+        y_shifted = y - self._y_min
+        weight = y_shifted / (self._weight_norm_const + self.eps)
+        if weight.numel() == 1:
+            return weight.reshape(())
+        return weight
+
+    def add_noise(
+        self, original_samples: torch.Tensor, noise: torch.Tensor, timestep: torch.Tensor
+    ) -> torch.Tensor:
+        sigma = (timestep / float(self.num_train_timesteps)).to(
+            original_samples.device, dtype=original_samples.dtype
+        )
+        if sigma.ndim == 0:
+            return (1 - sigma) * original_samples + sigma * noise
+        sigma = sigma.view(-1, *([1] * (original_samples.ndim - 1)))
+        return (1 - sigma) * original_samples + sigma * noise
+
+    @staticmethod
+    def training_target(
+        sample: torch.Tensor, noise: torch.Tensor, timestep: torch.Tensor
+    ) -> torch.Tensor:
+        del timestep
+        return noise - sample
+
+    def build_inference_schedule(
+        self,
+        num_inference_steps: int,
+        device: torch.device,
+        dtype: torch.dtype,
+        shift_override: float | None = None,
+    ) -> tuple[torch.Tensor, torch.Tensor]:
+        if num_inference_steps <= 0:
+            raise ValueError(f"`num_inference_steps` must be positive, got {num_inference_steps}")
+        shift = self.shift if shift_override is None else float(shift_override)
+        if shift <= 0:
+            raise ValueError(f"`shift` must be positive, got {shift}")
+
+        u_steps = torch.linspace(
+            1.0, 0.0, num_inference_steps + 1, device=device, dtype=torch.float32
+        )
+        sigma_steps = self._phi(u_steps, shift)
+        timesteps = sigma_steps[:-1] * float(self.num_train_timesteps)
+        deltas = sigma_steps[1:] - sigma_steps[:-1]
+        return timesteps.to(dtype=dtype), deltas.to(dtype=dtype)
+
+    @staticmethod
+    def step(model_output: torch.Tensor, delta: torch.Tensor, sample: torch.Tensor) -> torch.Tensor:
+        delta = delta.to(sample.device, dtype=sample.dtype)
+        if delta.ndim == 0:
+            return sample + model_output * delta
+        delta = delta.view(-1, *([1] * (sample.ndim - 1)))
+        return sample + model_output * delta
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/vpp2.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/vpp2.py
new file mode 100644
index 0000000000000000000000000000000000000000..090fbe292929da406f7a48b38f52300a6fa881f6
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/vpp2.py
@@ -0,0 +1,1850 @@
+# See LICENSE and THIRD_PARTY_NOTICES.md for retained source notices.
+from contextlib import contextmanager, nullcontext
+import os
+from pathlib import Path
+from typing import Any, Optional, Sequence, Union
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from omegaconf import DictConfig, OmegaConf
+from PIL import Image
+from bagel_cp1000.utils.logging_config import get_logger
+from .action_dit import ActionDiT
+from .helpers.loader import load_wan21_i2v_14b_components
+from .mot import MoT
+from .schedulers.scheduler_continuous import WanContinuousFlowMatchScheduler
+
+logger = get_logger(__name__)
+JOINT_DENOISING_MODES = ("policy", "forward_dynamics", "inverse_dynamics")
+JOINT_DENOISING_SAMPLERS = ("euler",)
+UNIPC_SIGMA_START = 0.999
+
+
+def _normalize_joint_denoising_mode_probs(probs):
+    if probs:
+        raise ValueError("Joint denoising modes are unsupported in this release")
+    return {}
+
+
+def _normalize_joint_denoising_sampler(sampler):
+    if sampler != "euler":
+        raise ValueError("Only cached-video Euler action inference is supported")
+    return "euler"
+
+
+WAN_OFFICIAL_PEFT_VERSION = "0.14.0"
+WAN_OFFICIAL_VIDEO_LORA_TARGET_MODULES = ("q", "k", "v", "o", "ffn.0", "ffn.2")
+
+
+def _normalize_video_lora_config(config):
+    if config and config.get("enabled", False):
+        raise ValueError("This release trains the complete Video DiT; LoRA is unsupported")
+    return {"enabled": False}
+
+
+def _apply_wan_official_video_lora(video_expert, config):
+    _normalize_video_lora_config(config)
+    return video_expert
+
+
+class VPP2(torch.nn.Module):
+    """MoT world model with video/action experts."""
+
+    supports_action_only_fast_path = False
+
+    def __init__(
+        self,
+        video_expert,
+        action_expert: ActionDiT,
+        mot: MoT,
+        vae,
+        text_encoder=None,
+        tokenizer=None,
+        clip_encoder=None,
+        text_dim: Optional[int] = None,
+        proprio_dim: Optional[int] = None,
+        device: str = "cpu",
+        torch_dtype: torch.dtype = torch.float32,
+        video_train_shift: float = 5.0,
+        video_infer_shift: float = 5.0,
+        video_num_train_timesteps: int = 1000,
+        video_train_sampling_strategy: str = "shifted_uniform",
+        video_train_beta_alpha: float = 7.0,
+        video_train_beta_beta: float = 1.0,
+        action_train_shift: float = 5.0,
+        action_infer_shift: float = 5.0,
+        action_num_train_timesteps: int = 1000,
+        loss_lambda_video: float = 1.0,
+        loss_lambda_action: float = 1.0,
+        video_lora_joint_video_loss: bool = False,
+        train_wan: bool = True,
+        action_chunk_size: int = 1,
+        action_visible_video_frames: int = 1,
+        action_only_first_frame_fast_path: bool = False,
+        action_clean_first_frame_start_layer: Optional[int] = None,
+        action_clean_first_frame_timestep_cutoff: Optional[float] = None,
+        action_first_frame_kv_source: str = "clean_prefill",
+        action_video_timestep_cutoff: Optional[float] = None,
+        video_lora_config: Optional[dict[str, Any]] = None,
+        use_fixed_video_layers: Optional[int] = None,
+        use_video_tokens: bool = False,
+        video_token_latent_dim: int = 2560,
+        joint_denoising: bool = False,
+        joint_denoising_mode_probs: Optional[dict[str, float]] = None,
+        joint_denoising_sampler: str = "euler",
+    ):
+        if (
+            joint_denoising
+            or use_video_tokens
+            or use_fixed_video_layers is not None
+            or (action_only_first_frame_fast_path and not self.supports_action_only_fast_path)
+        ):
+            raise ValueError("Unsupported experimental model option")
+        super().__init__()
+        self.video_expert = video_expert
+        self.action_expert = action_expert
+        self.mot = mot
+        self.dit = self.mot
+        self.vae = vae
+        self.text_encoder = text_encoder
+        self.tokenizer = tokenizer
+        self.clip_encoder = clip_encoder
+        if text_dim is None:
+            if self.text_encoder is None:
+                raise ValueError("`text_dim` is required when `text_encoder` is not loaded.")
+            text_dim = int(self.text_encoder.dim)
+        self.text_dim = int(text_dim)
+        self.proprio_dim = None if proprio_dim is None else int(proprio_dim)
+        if self.proprio_dim is not None:
+            self.proprio_encoder = nn.Linear(self.proprio_dim, self.text_dim).to(torch_dtype)
+        else:
+            self.proprio_encoder = None
+        self.train_video_scheduler = WanContinuousFlowMatchScheduler(
+            num_train_timesteps=video_num_train_timesteps,
+            shift=video_train_shift,
+            train_sampling_strategy=video_train_sampling_strategy,
+            beta_alpha=video_train_beta_alpha,
+            beta_beta=video_train_beta_beta,
+        )
+        self.infer_video_scheduler = WanContinuousFlowMatchScheduler(
+            num_train_timesteps=video_num_train_timesteps, shift=video_infer_shift
+        )
+        self.train_action_scheduler = WanContinuousFlowMatchScheduler(
+            num_train_timesteps=action_num_train_timesteps, shift=action_train_shift
+        )
+        self.infer_action_scheduler = WanContinuousFlowMatchScheduler(
+            num_train_timesteps=action_num_train_timesteps, shift=action_infer_shift
+        )
+        self.train_scheduler = self.train_video_scheduler
+        self.infer_scheduler = self.infer_video_scheduler
+        self.device = torch.device(device)
+        self.torch_dtype = torch_dtype
+        self.loss_lambda_video = float(loss_lambda_video)
+        self.loss_lambda_action = float(loss_lambda_action)
+        self.video_lora_joint_video_loss = bool(video_lora_joint_video_loss)
+        self.train_mode = "joint"
+        self.train_wan = True
+        self.action_chunk_size = int(action_chunk_size)
+        self.action_visible_video_frames = int(action_visible_video_frames)
+        self.action_only_first_frame_fast_path = bool(action_only_first_frame_fast_path)
+        self.action_clean_first_frame_start_layer = (
+            None
+            if action_clean_first_frame_start_layer is None
+            else int(action_clean_first_frame_start_layer)
+        )
+        self.action_clean_first_frame_timestep_cutoff = (
+            None
+            if action_clean_first_frame_timestep_cutoff is None
+            else float(action_clean_first_frame_timestep_cutoff)
+        )
+        self.action_first_frame_kv_source = str(action_first_frame_kv_source).strip().lower()
+        if self.action_first_frame_kv_source not in {"clean_prefill", "noisy_cache_prefix"}:
+            raise ValueError(
+                f"`action_first_frame_kv_source` must be 'clean_prefill' or 'noisy_cache_prefix', got {action_first_frame_kv_source!r}."
+            )
+        self.action_video_timestep_cutoff = (
+            None if action_video_timestep_cutoff is None else float(action_video_timestep_cutoff)
+        )
+        self.video_lora_config = _normalize_video_lora_config(video_lora_config)
+        self.video_lora_enabled = bool(self.video_lora_config["enabled"])
+        if self.video_lora_joint_video_loss and (not False):
+            raise ValueError(
+                "`video_lora_joint_video_loss=true` requires `video_lora.enabled=true`."
+            )
+        if (
+            self.video_lora_joint_video_loss
+            and self.train_video_scheduler.train_sampling_strategy != "pure_noise"
+        ):
+            raise ValueError(
+                "Video-LoRA joint Video loss is defined at the pure-noise endpoint and requires video train_sampling_strategy='pure_noise'."
+            )
+        if self.video_lora_joint_video_loss and False:
+            raise ValueError(
+                "Video-LoRA joint Video loss requires the complete target video; set action_only_first_frame_fast_path=false."
+            )
+        self.use_fixed_video_layers = (
+            None if use_fixed_video_layers is None else int(use_fixed_video_layers)
+        )
+        self.use_video_tokens = bool(False)
+        self.video_token_latent_dim = int(video_token_latent_dim)
+        if self.action_visible_video_frames <= 0:
+            raise ValueError(
+                f"`action_visible_video_frames` must be positive, got {self.action_visible_video_frames}"
+            )
+        if self.action_chunk_size <= 0:
+            raise ValueError(f"`action_chunk_size` must be positive, got {self.action_chunk_size}")
+        if self.action_expert.action_dim % self.action_chunk_size != 0:
+            raise ValueError(
+                f"`action_expert.action_dim` must be divisible by action_chunk_size, got {self.action_expert.action_dim} and {self.action_chunk_size}"
+            )
+        if self.action_clean_first_frame_start_layer is not None:
+            if not 0 < self.action_clean_first_frame_start_layer < int(self.mot.num_layers):
+                raise ValueError(
+                    f"`action_clean_first_frame_start_layer` must be inside the ActionDiT stack [1, {int(self.mot.num_layers) - 1}], got {self.action_clean_first_frame_start_layer}."
+                )
+            if self.use_fixed_video_layers is not None or False:
+                raise ValueError(
+                    "Clean-first-frame late ActionDiT layers currently require direct per-layer video K/V (`use_fixed_video_layers=null`, `use_video_tokens=false`)."
+                )
+        if self.action_video_timestep_cutoff is not None:
+            if not 0.0 < self.action_video_timestep_cutoff <= float(action_num_train_timesteps):
+                raise ValueError(
+                    f"`action_video_timestep_cutoff` must be in (0, {int(action_num_train_timesteps)}], got {self.action_video_timestep_cutoff}."
+                )
+            if self.action_clean_first_frame_start_layer is not None:
+                raise ValueError(
+                    "Timestep-gated video conditioning and the clean-first-frame layer split are separate controlled experiments and cannot be enabled together."
+                )
+        if self.action_clean_first_frame_timestep_cutoff is not None:
+            if (
+                not 0.0
+                < self.action_clean_first_frame_timestep_cutoff
+                <= float(action_num_train_timesteps)
+            ):
+                raise ValueError(
+                    f"`action_clean_first_frame_timestep_cutoff` must be in (0, {int(action_num_train_timesteps)}], got {self.action_clean_first_frame_timestep_cutoff}."
+                )
+            if self.action_clean_first_frame_start_layer is not None:
+                raise ValueError(
+                    "Clean-first-frame layer and timestep routes are separate controlled experiments and cannot be enabled together."
+                )
+            if self.action_video_timestep_cutoff is not None:
+                raise ValueError(
+                    "Clean-first-frame timestep routing and no-video timestep gating cannot be enabled together."
+                )
+            if self.use_fixed_video_layers is not None or False:
+                raise ValueError(
+                    "Clean-first-frame timestep routing currently requires direct per-layer video K/V (`use_fixed_video_layers=null`, `use_video_tokens=false`)."
+                )
+        self.joint_denoising = bool(False)
+        self.joint_denoising_mode_probs = _normalize_joint_denoising_mode_probs(
+            joint_denoising_mode_probs
+        )
+        self.joint_denoising_sampler = _normalize_joint_denoising_sampler(joint_denoising_sampler)
+        self.raw_action_dim = self.action_expert.action_dim // self.action_chunk_size
+        self.set_train_wan(train_wan)
+        self.to(self.device)
+
+    def set_train_mode(self, mode: str):
+        if mode != "joint":
+            raise ValueError("Only joint training is supported")
+        self.train_mode = "joint"
+        self.train_wan = True
+
+    def set_train_wan(self, train_wan: bool):
+        return self.set_train_mode("joint" if train_wan else "action_only")
+
+    def _video_base_model(self):
+        return self.video_expert
+
+    def _load_video_base_state_dict(self, state: dict[str, torch.Tensor]):
+        base_model = self._video_base_model()
+        return base_model.load_state_dict(state, strict=True)
+        current_keys = set(base_model.state_dict().keys())
+        remapped = {}
+        unmatched = []
+        for key, value in state.items():
+            if key in current_keys:
+                remapped[key] = value
+                continue
+            (prefix, separator, suffix) = key.rpartition(".")
+            candidate = f"{prefix}.base_layer.{suffix}" if separator else ""
+            if candidate in current_keys:
+                remapped[candidate] = value
+            else:
+                unmatched.append(key)
+        if unmatched:
+            raise ValueError(
+                f"Video base checkpoint has keys that cannot be mapped into the PEFT-wrapped backbone: {unmatched[:8]}"
+            )
+        load_result = base_model.load_state_dict(remapped, strict=False)
+        unexpected = list(getattr(load_result, "unexpected_keys", []))
+        missing = [
+            key
+            for key in getattr(load_result, "missing_keys", [])
+            if ".lora_A." not in key and ".lora_B." not in key
+        ]
+        if unexpected or missing:
+            raise ValueError(
+                f"Video base checkpoint did not strictly cover the frozen PEFT backbone: missing={missing[:8]}, unexpected={unexpected[:8]}"
+            )
+        return load_result
+
+    def _pack_action_tokens(self, action: torch.Tensor) -> torch.Tensor:
+        if self.action_chunk_size == 1:
+            return action
+        if action.ndim != 3:
+            raise ValueError(f"`action` must be 3D [B,T,D], got {tuple(action.shape)}")
+        (batch_size, horizon, dim) = action.shape
+        if dim != self.raw_action_dim:
+            raise ValueError(f"Expected raw action dim {self.raw_action_dim}, got {dim}")
+        if horizon % self.action_chunk_size != 0:
+            raise ValueError(
+                f"Action horizon {horizon} must be divisible by action_chunk_size={self.action_chunk_size}"
+            )
+        return action.reshape(
+            batch_size, horizon // self.action_chunk_size, dim * self.action_chunk_size
+        )
+
+    def _unpack_action_tokens(self, action: torch.Tensor) -> torch.Tensor:
+        if self.action_chunk_size == 1:
+            return action
+        if action.ndim not in (2, 3):
+            raise ValueError(f"`action` must be 2D or 3D, got {tuple(action.shape)}")
+        if action.shape[-1] != self.action_expert.action_dim:
+            raise ValueError(
+                f"Expected packed action dim {self.action_expert.action_dim}, got {action.shape[-1]}"
+            )
+        leading = action.shape[:-1]
+        return action.reshape(
+            *leading[:-1], leading[-1] * self.action_chunk_size, self.raw_action_dim
+        )
+
+    def _pack_action_is_pad(self, action_is_pad: Optional[torch.Tensor]) -> Optional[torch.Tensor]:
+        if action_is_pad is None or self.action_chunk_size == 1:
+            return action_is_pad
+        if action_is_pad.ndim != 2:
+            raise ValueError(f"`action_is_pad` must be 2D [B,T], got {tuple(action_is_pad.shape)}")
+        (batch_size, horizon) = action_is_pad.shape
+        if horizon % self.action_chunk_size != 0:
+            raise ValueError(
+                f"Action pad horizon {horizon} must be divisible by action_chunk_size={self.action_chunk_size}"
+            )
+        return action_is_pad.reshape(
+            batch_size, horizon // self.action_chunk_size, self.action_chunk_size
+        ).any(dim=2)
+
+    @classmethod
+    def from_wan21_14b_pretrained(
+        cls,
+        device: str = "cuda",
+        torch_dtype: torch.dtype = torch.bfloat16,
+        model_id: str = "weights/Wan2.1-I2V-14B-480P",
+        dit_checkpoint_path: str | None = None,
+        tokenizer_model_id: str | None = None,
+        tokenizer_max_len: int = 512,
+        load_text_encoder: bool = True,
+        load_clip_encoder: bool = True,
+        proprio_dim: Optional[int] = None,
+        redirect_common_files: bool = True,
+        video_dit_config: dict[str, Any] | None = None,
+        action_dit_config: dict[str, Any] | None = None,
+        action_dit_pretrained_path: str | None = None,
+        skip_dit_load_from_pretrain: bool = False,
+        mot_checkpoint_mixed_attn: bool = True,
+        action_chunk_size: int = 1,
+        video_train_shift: float = 5.0,
+        video_infer_shift: float = 5.0,
+        video_num_train_timesteps: int = 1000,
+        video_train_sampling_strategy: str = "shifted_uniform",
+        video_train_beta_alpha: float = 7.0,
+        video_train_beta_beta: float = 1.0,
+        action_train_shift: float = 5.0,
+        action_infer_shift: float = 5.0,
+        action_num_train_timesteps: int = 1000,
+        loss_lambda_video: float = 1.0,
+        loss_lambda_action: float = 1.0,
+        video_lora_joint_video_loss: bool = False,
+        action_visible_video_frames: int = 1,
+        action_only_first_frame_fast_path: bool = False,
+        action_clean_first_frame_start_layer: Optional[int] = None,
+        action_clean_first_frame_timestep_cutoff: Optional[float] = None,
+        action_first_frame_kv_source: str = "clean_prefill",
+        action_video_timestep_cutoff: Optional[float] = None,
+        video_lora_config: Optional[dict[str, Any]] = None,
+        use_fixed_video_layers: Optional[int] = None,
+        use_video_tokens: bool = False,
+        video_token_latent_dim: int = 2560,
+        joint_denoising: bool = False,
+        joint_denoising_mode_probs: Optional[dict[str, float]] = None,
+        joint_denoising_sampler: str = "euler",
+    ):
+        if video_dit_config is None:
+            video_dit_config = {}
+        if "text_dim" not in video_dit_config:
+            video_dit_config = dict(video_dit_config)
+            video_dit_config.setdefault("text_dim", 4096)
+        components = load_wan21_i2v_14b_components(
+            device=device,
+            torch_dtype=torch_dtype,
+            model_id=model_id,
+            dit_checkpoint_path=dit_checkpoint_path,
+            tokenizer_model_id=tokenizer_model_id,
+            tokenizer_max_len=tokenizer_max_len,
+            dit_config=video_dit_config,
+            skip_dit_load_from_pretrain=skip_dit_load_from_pretrain,
+            load_text_encoder=load_text_encoder,
+            load_clip_encoder=load_clip_encoder,
+        )
+        video_expert = components.dit
+        normalized_video_lora_config = _normalize_video_lora_config(video_lora_config)
+        action_dit_config = dict(action_dit_config or {})
+        action_dit_config["use_video_tokens"] = bool(False)
+        action_expert = ActionDiT.from_pretrained(
+            action_dit_config=action_dit_config,
+            action_dit_pretrained_path=action_dit_pretrained_path,
+            skip_dit_load_from_pretrain=skip_dit_load_from_pretrain,
+            device=device,
+            torch_dtype=torch_dtype,
+        )
+        if not False and int(action_expert.num_heads) != int(video_expert.num_heads):
+            raise ValueError(
+                "ActionDiT `num_heads` must match video expert for MoT mixed attention."
+            )
+        if int(action_expert.attn_head_dim) != int(video_expert.attn_head_dim):
+            raise ValueError(
+                "ActionDiT `attn_head_dim` must match video expert for MoT mixed attention."
+            )
+        if int(len(action_expert.blocks)) != int(len(video_expert.blocks)):
+            raise ValueError("ActionDiT `num_layers` must match video expert.")
+        video_expert = _apply_wan_official_video_lora(video_expert, normalized_video_lora_config)
+        mot = MoT(
+            mixtures={"video": video_expert, "action": action_expert},
+            mot_checkpoint_mixed_attn=mot_checkpoint_mixed_attn,
+            use_fixed_video_layers=use_fixed_video_layers,
+            use_video_tokens=False,
+        )
+        model = cls(
+            video_expert=video_expert,
+            action_expert=action_expert,
+            mot=mot,
+            vae=components.vae,
+            text_encoder=components.text_encoder,
+            tokenizer=components.tokenizer,
+            clip_encoder=components.clip_encoder,
+            text_dim=int(video_dit_config["text_dim"]),
+            proprio_dim=proprio_dim,
+            device=device,
+            torch_dtype=torch_dtype,
+            video_train_shift=video_train_shift,
+            video_infer_shift=video_infer_shift,
+            video_num_train_timesteps=video_num_train_timesteps,
+            video_train_sampling_strategy=video_train_sampling_strategy,
+            video_train_beta_alpha=video_train_beta_alpha,
+            video_train_beta_beta=video_train_beta_beta,
+            action_train_shift=action_train_shift,
+            action_infer_shift=action_infer_shift,
+            action_num_train_timesteps=action_num_train_timesteps,
+            loss_lambda_video=loss_lambda_video,
+            loss_lambda_action=loss_lambda_action,
+            video_lora_joint_video_loss=video_lora_joint_video_loss,
+            action_chunk_size=action_chunk_size,
+            action_visible_video_frames=action_visible_video_frames,
+            action_only_first_frame_fast_path=action_only_first_frame_fast_path,
+            action_clean_first_frame_start_layer=action_clean_first_frame_start_layer,
+            action_clean_first_frame_timestep_cutoff=action_clean_first_frame_timestep_cutoff,
+            action_first_frame_kv_source=action_first_frame_kv_source,
+            action_video_timestep_cutoff=action_video_timestep_cutoff,
+            video_lora_config=normalized_video_lora_config,
+            use_fixed_video_layers=use_fixed_video_layers,
+            use_video_tokens=False,
+            video_token_latent_dim=video_token_latent_dim,
+            joint_denoising=False,
+            joint_denoising_mode_probs=joint_denoising_mode_probs,
+            joint_denoising_sampler=joint_denoising_sampler,
+        )
+        model.model_paths = {
+            "video_dit": components.dit_path,
+            "vae": components.vae_path,
+            "text_encoder": components.text_encoder_path,
+            "tokenizer": components.tokenizer_path,
+            "clip_encoder": components.clip_encoder_path,
+            "action_dit_backbone": "SKIPPED_PRETRAIN"
+            if skip_dit_load_from_pretrain
+            else action_dit_pretrained_path,
+        }
+        return model
+
+    def _restore_auxiliary_component_dtypes(self):
+        """Keep Wan I2V helper modules aligned with the reference generation path."""
+        if self.clip_encoder is not None:
+            self.clip_encoder.eval().requires_grad_(False).to(
+                device=self.device, dtype=torch.float16
+            )
+        self.vae.eval().requires_grad_(False).to(device=self.device, dtype=torch.float32)
+
+    def to(self, *args, **kwargs):
+        super().to(*args, **kwargs)
+        self.mot.to(*args, **kwargs)
+        if self.text_encoder is not None:
+            self.text_encoder.to(*args, **kwargs)
+        self._restore_auxiliary_component_dtypes()
+        return self
+
+    @staticmethod
+    def _check_resize_height_width(height, width, num_frames):
+        if height % 16 != 0:
+            height = (height + 15) // 16 * 16
+        if width % 16 != 0:
+            width = (width + 15) // 16 * 16
+        if num_frames % 4 != 1:
+            num_frames = (num_frames + 3) // 4 * 4 + 1
+        return (height, width, num_frames)
+
+    @torch.no_grad()
+    def encode_prompt(self, prompt: Union[str, Sequence[str]]):
+        if self.text_encoder is None or self.tokenizer is None:
+            raise ValueError(
+                "Prompt encoding requires loaded text encoder/tokenizer. Set `load_text_encoder=true` or provide precomputed `context/context_mask`."
+            )
+        (ids, mask) = self.tokenizer(prompt, return_mask=True, add_special_tokens=True)
+        ids = ids.to(self.device)
+        mask = mask.to(self.device, dtype=torch.bool)
+        prompt_emb = self.text_encoder(ids, mask)
+        seq_lens = mask.gt(0).sum(dim=1).long()
+        for i, v in enumerate(seq_lens):
+            prompt_emb[i, v:] = 0
+        mask = torch.ones_like(mask)
+        return (prompt_emb.to(device=self.device), mask)
+
+    def _append_proprio_to_context(
+        self, context: torch.Tensor, context_mask: torch.Tensor, proprio: Optional[torch.Tensor]
+    ) -> tuple[torch.Tensor, torch.Tensor]:
+        if self.proprio_encoder is None or proprio is None:
+            return (context, context_mask)
+        if proprio.ndim != 2:
+            raise ValueError(f"`proprio` must be 2D [B, D], got shape {tuple(proprio.shape)}")
+        if self.proprio_dim is None or proprio.shape[1] != self.proprio_dim:
+            raise ValueError(
+                f"`proprio` last dim must be {self.proprio_dim}, got {proprio.shape[1]}"
+            )
+        proprio_token = self.proprio_encoder(
+            proprio.to(device=self.device, dtype=context.dtype).unsqueeze(1)
+        ).to(dtype=context.dtype)
+        proprio_mask = torch.ones(
+            (context_mask.shape[0], 1), dtype=torch.bool, device=context_mask.device
+        )
+        return (
+            torch.cat([context, proprio_token], dim=1),
+            torch.cat([context_mask, proprio_mask], dim=1),
+        )
+
+    @torch.no_grad()
+    def _encode_video_latents(
+        self, video_tensor, tiled=False, tile_size=(30, 52), tile_stride=(15, 26)
+    ):
+        self._restore_auxiliary_component_dtypes()
+        videos = [
+            video_tensor[i].to(device=self.device, dtype=torch.float32)
+            for i in range(video_tensor.shape[0])
+        ]
+        z = self.vae.encode(
+            videos, device=self.device, tiled=tiled, tile_size=tile_size, tile_stride=tile_stride
+        )
+        if isinstance(z, list):
+            z = torch.stack(z)
+        return z.to(device=self.device, dtype=self.torch_dtype)
+
+    @torch.no_grad()
+    def _encode_input_image_latents_tensor(
+        self, input_image: torch.Tensor, tiled=False, tile_size=(30, 52), tile_stride=(15, 26)
+    ):
+        self._restore_auxiliary_component_dtypes()
+        if input_image.ndim == 3:
+            input_image = input_image.unsqueeze(0)
+        if input_image.ndim != 4 or input_image.shape[0] != 1 or input_image.shape[1] != 3:
+            raise ValueError(
+                f"`input_image` must have shape [1,3,H,W] or [3,H,W], got {tuple(input_image.shape)}"
+            )
+        image = input_image.to(device=self.device, dtype=torch.float32)[0].unsqueeze(1)
+        z = self.vae.encode(
+            [image], device=self.device, tiled=tiled, tile_size=tile_size, tile_stride=tile_stride
+        )
+        if isinstance(z, list):
+            z = z[0].unsqueeze(0)
+        return z.to(device=self.device, dtype=self.torch_dtype)
+
+    @torch.no_grad()
+    def _encode_separate_condition_latents(
+        self,
+        condition_video: torch.Tensor,
+        condition_observation_video: Optional[torch.Tensor] = None,
+        tiled: bool = False,
+    ) -> torch.Tensor:
+        """Encode history and the current observation as independent conditions."""
+        history_latents = self._encode_video_latents(condition_video, tiled=tiled)
+        if condition_observation_video is None:
+            return history_latents
+        if condition_observation_video.ndim == 4:
+            condition_observation_video = condition_observation_video.unsqueeze(0)
+        if condition_observation_video.ndim != 5 or condition_observation_video.shape[1] != 3:
+            raise ValueError(
+                f"`condition_observation_video` must be [B,3,1,H,W] or [3,1,H,W], got {tuple(condition_observation_video.shape)}"
+            )
+        if condition_observation_video.shape[2] != 1:
+            raise ValueError(
+                f"`condition_observation_video` must contain exactly one RGB frame, got T={condition_observation_video.shape[2]}"
+            )
+        observation_latents = self._encode_video_latents(condition_observation_video, tiled=tiled)
+        if observation_latents.shape[2] != 1:
+            raise ValueError(
+                f"One-frame observation VAE encoding must produce one latent frame, got T={observation_latents.shape[2]}"
+            )
+        if history_latents.shape[0] != observation_latents.shape[0]:
+            raise ValueError("History and observation condition batches must match.")
+        if history_latents.shape[3:] != observation_latents.shape[3:]:
+            raise ValueError(
+                f"History and observation condition latent spatial grids must match, got {tuple(history_latents.shape[3:])} vs {tuple(observation_latents.shape[3:])}."
+            )
+        return torch.cat([history_latents, observation_latents], dim=2)
+
+    @torch.no_grad()
+    def _encode_clip_fea_from_video(self, video_tensor: torch.Tensor) -> torch.Tensor:
+        self._restore_auxiliary_component_dtypes()
+        if self.clip_encoder is None:
+            raise ValueError(
+                "Wan2.1-I2V-14B requires CLIP image features. Load `clip_encoder` or provide `sample['clip_fea']` / `clip_fea` explicitly."
+            )
+        if video_tensor.ndim != 5:
+            raise ValueError(f"`video_tensor` must be [B,C,T,H,W], got {tuple(video_tensor.shape)}")
+        videos = [
+            video_tensor[i].to(device=self.device, dtype=torch.float16)
+            for i in range(video_tensor.shape[0])
+        ]
+        return self.clip_encoder(videos).to(device=self.device, dtype=self.torch_dtype)
+
+    @torch.no_grad()
+    def _encode_clip_fea_from_image(self, input_image: torch.Tensor) -> torch.Tensor:
+        if input_image.ndim == 3:
+            input_image = input_image.unsqueeze(0)
+        if input_image.ndim != 4 or input_image.shape[1] != 3:
+            raise ValueError(
+                f"`input_image` must be [B,3,H,W] or [3,H,W], got {tuple(input_image.shape)}"
+            )
+        video = input_image.to(device=self.device, dtype=torch.float16).unsqueeze(2)
+        return self._encode_clip_fea_from_video(video)
+
+    def _decode_latents(self, latents, tiled=False, tile_size=(30, 52), tile_stride=(15, 26)):
+        self._restore_auxiliary_component_dtypes()
+        video_tensor = self.vae.decode(
+            latents.to(device=self.device, dtype=torch.float32),
+            device=self.device,
+            tiled=tiled,
+            tile_size=tile_size,
+            tile_stride=tile_stride,
+        )
+        video_tensor = video_tensor.squeeze(0).detach().float().clamp(-1, 1)
+        video_tensor = ((video_tensor + 1.0) * 127.5).to(torch.uint8).cpu()
+        frames = []
+        for t in range(video_tensor.shape[1]):
+            frame = video_tensor[:, t].permute(1, 2, 0).numpy()
+            frames.append(Image.fromarray(frame))
+        return frames
+
+    def build_inputs(self, sample, tiled: bool = False, training_stage: Optional[str] = None, *, encode_video_target: bool = True):
+        video = sample["video"]
+        if "context" not in sample or "context_mask" not in sample:
+            raise ValueError(
+                "VPP2 training requires `sample['context']` and `sample['context_mask']`."
+            )
+        context = sample["context"]
+        context_mask = sample["context_mask"]
+        proprio = sample.get("proprio", None)
+        if video.ndim != 5:
+            raise ValueError(
+                f"`sample['video']` must be 5D [B, 3, T, H, W], got shape {tuple(video.shape)}"
+            )
+        if video.shape[1] != 3:
+            raise ValueError(
+                f"`sample['video']` channel dimension must be 3, got shape {tuple(video.shape)}"
+            )
+        (batch_size, _, observed_num_frames, height, width) = video.shape
+        if height % 16 != 0 or width % 16 != 0:
+            raise ValueError(
+                f"Video spatial dims must be multiples of 16, got H={height}, W={width}"
+            )
+        normalized_stage = None if training_stage is None else str(training_stage).strip().lower()
+        use_action_only_fast_path = False and (
+            not self.train_wan or ("joint" == "alternating" and normalized_stage == "action")
+        )
+        declared_num_frames = sample.get("video_num_frames", None)
+        if declared_num_frames is None:
+            num_frames = observed_num_frames
+        elif isinstance(declared_num_frames, torch.Tensor):
+            declared_values = declared_num_frames.detach().reshape(-1).cpu().tolist()
+            if not declared_values or any(
+                (int(value) != int(declared_values[0]) for value in declared_values)
+            ):
+                raise ValueError(
+                    f"`sample['video_num_frames']` must be identical across the batch, got {declared_values}."
+                )
+            num_frames = int(declared_values[0])
+        else:
+            num_frames = int(declared_num_frames)
+        if num_frames % 4 != 1:
+            raise ValueError(f"Logical video T must satisfy T % 4 == 1, got T={num_frames}")
+        if num_frames <= 1:
+            raise ValueError(
+                f"Logical video T must be > 1 for action-conditioned training, got T={num_frames}"
+            )
+        if observed_num_frames <= 0:
+            raise ValueError("`sample['video']` must contain at least one decoded frame.")
+        if not use_action_only_fast_path and observed_num_frames != num_frames:
+            raise ValueError(
+                f"A shortened decoded video is only valid for the action-only first-frame fast path: observed T={observed_num_frames}, logical T={num_frames}."
+            )
+        if "action" not in sample:
+            raise ValueError("`sample['action']` is required for VPP2 training.")
+        action = sample["action"]
+        if action.ndim != 3:
+            raise ValueError(
+                f"`sample['action']` must be 3D [B, T, a_dim], got shape {tuple(action.shape)}"
+            )
+        action_horizon = int(action.shape[1])
+        if action_horizon <= 0:
+            raise ValueError(
+                f"`sample['action']` temporal dimension must be positive, got {action_horizon}"
+            )
+        action_is_pad = sample.get("action_is_pad", None)
+        if action_is_pad is not None:
+            if action_is_pad.ndim != 2:
+                raise ValueError(
+                    f"`sample['action_is_pad']` must be 2D [B, T], got shape {tuple(action_is_pad.shape)}"
+                )
+            if action_is_pad.shape[0] != batch_size or action_is_pad.shape[1] != action_horizon:
+                raise ValueError(
+                    f"`sample['action_is_pad']` shape mismatch: got {tuple(action_is_pad.shape)} vs expected ({batch_size}, {action_horizon})"
+                )
+        action_loss_is_pad = action_is_pad
+        action_terminal_mask = sample.get("action_terminal_mask")
+        if action_terminal_mask is not None:
+            if action_is_pad is None or action_terminal_mask.shape != action_is_pad.shape:
+                raise ValueError("action_terminal_mask must match action_is_pad [B,T].")
+            terminal = action_terminal_mask.to(device=action_is_pad.device, dtype=torch.bool)
+            if bool((terminal & ~action_is_pad.bool()).any()):
+                raise ValueError("action_terminal_mask must be a subset of action_is_pad.")
+            action_loss_is_pad = action_is_pad.bool() & ~terminal
+        image_is_pad = sample.get("image_is_pad", None)
+        if image_is_pad is not None:
+            if image_is_pad.ndim != 2:
+                raise ValueError(
+                    f"`sample['image_is_pad']` must be 2D [B, T], got shape {tuple(image_is_pad.shape)}"
+                )
+            if image_is_pad.shape[0] != batch_size or image_is_pad.shape[1] != observed_num_frames:
+                raise ValueError(
+                    f"`sample['image_is_pad']` shape mismatch: got {tuple(image_is_pad.shape)} vs expected ({batch_size}, {observed_num_frames})"
+                )
+        if not encode_video_target and sample.get("condition_video") is None:
+            raise ValueError("Skipping unused target encoding requires an independently supplied observation/history condition.")
+        input_latents = None
+        input_video = None
+        if not use_action_only_fast_path and encode_video_target:
+            input_video = video.to(device=self.device, dtype=self.torch_dtype, non_blocking=True)
+            input_latents = self._encode_video_latents(input_video, tiled=tiled)
+        condition_video = sample.get("condition_video")
+        condition_observation_video = sample.get("condition_observation_video")
+        if condition_video is None and condition_observation_video is not None:
+            raise ValueError("`condition_observation_video` requires `condition_video` history.")
+        if condition_video is None:
+            if input_video is None:
+                input_video = video[:, :, :1].to(
+                    device=self.device, dtype=self.torch_dtype, non_blocking=True
+                )
+                condition_latents = self._encode_video_latents(input_video, tiled=tiled)
+            else:
+                condition_latents = input_latents[:, :, :1]
+            clip_condition_video = input_video
+        else:
+            if condition_video.ndim != 5 or condition_video.shape[:2] != (batch_size, 3):
+                raise ValueError(
+                    f"`sample['condition_video']` must be [B,3,T,H,W], got {tuple(condition_video.shape)}."
+                )
+            if condition_video.shape[-2:] != (height, width):
+                raise ValueError(
+                    f"Condition and target video spatial sizes must match, got {tuple(condition_video.shape[-2:])} vs {(height, width)}."
+                )
+            condition_num_frames = int(condition_video.shape[2])
+            if condition_num_frames <= 0 or condition_num_frames % 4 != 1:
+                raise ValueError(
+                    f"Condition temporal length must be positive and satisfy T % 4 == 1, got T={condition_num_frames}."
+                )
+            condition_input_video = condition_video.to(
+                device=self.device, dtype=self.torch_dtype, non_blocking=True
+            )
+            if condition_observation_video is not None:
+                if condition_observation_video.ndim != 5 or condition_observation_video.shape[
+                    :2
+                ] != (batch_size, 3):
+                    raise ValueError(
+                        f"`sample['condition_observation_video']` must be [B,3,1,H,W], got {tuple(condition_observation_video.shape)}"
+                    )
+                if condition_observation_video.shape[2] != 1:
+                    raise ValueError(
+                        f"`sample['condition_observation_video']` must contain one frame, got T={condition_observation_video.shape[2]}"
+                    )
+                condition_observation_video = condition_observation_video.to(
+                    device=self.device, dtype=self.torch_dtype, non_blocking=True
+                )
+            condition_latents = self._encode_separate_condition_latents(
+                condition_input_video,
+                condition_observation_video=condition_observation_video,
+                tiled=tiled,
+            )
+            logical_latent_frames = (num_frames - 1) // int(self.vae.temporal_downsample_factor) + 1
+            if condition_latents.shape[2] > logical_latent_frames:
+                raise ValueError(
+                    f"Condition cannot be longer than the logical video latent sequence, got {condition_latents.shape[2]} > {logical_latent_frames}."
+                )
+            if input_latents is not None and condition_latents.shape[3:] != input_latents.shape[3:]:
+                raise ValueError(
+                    f"Condition and target latent spatial grids must match, got {tuple(condition_latents.shape[3:])} vs {tuple(input_latents.shape[3:])}."
+                )
+            if condition_observation_video is not None and input_latents is not None:
+                if condition_latents.shape[2] > input_latents.shape[2]:
+                    raise ValueError(
+                        f"Separate condition cannot be longer than the target latent sequence, got {condition_latents.shape[2]} > {input_latents.shape[2]}."
+                    )
+                input_latents = torch.cat(
+                    [condition_latents, input_latents[:, :, condition_latents.shape[2] :]], dim=2
+                )
+            clip_condition_video = condition_input_video[:, :, :1]
+        clip_fea = sample.get("clip_fea", None)
+        if clip_fea is None:
+            clip_fea = self._encode_clip_fea_from_video(clip_condition_video)
+        else:
+            clip_fea = clip_fea.to(device=self.device, dtype=self.torch_dtype, non_blocking=True)
+        if context.ndim != 3 or context_mask.ndim != 2:
+            raise ValueError(
+                f"`context/context_mask` must be [B,L,D]/[B,L], got {tuple(context.shape)} and {tuple(context_mask.shape)}"
+            )
+        context = context.to(device=self.device, dtype=self.torch_dtype, non_blocking=True)
+        context_mask = context_mask.to(device=self.device, dtype=torch.bool, non_blocking=True)
+        use_proprio = normalized_stage != "video"
+        if self.proprio_encoder is not None and use_proprio:
+            if proprio is None:
+                raise ValueError("`sample['proprio']` is required when `proprio_dim` is enabled.")
+            if proprio.ndim != 3:
+                raise ValueError(
+                    f"`sample['proprio']` must be 3D [B, T, d], got shape {tuple(proprio.shape)}"
+                )
+            if proprio.shape[2] != self.proprio_dim:
+                raise ValueError(
+                    f"`sample['proprio']` last dim must be {self.proprio_dim}, got {proprio.shape[2]}"
+                )
+            proprio = proprio[:, 0, :]
+            (context, context_mask) = self._append_proprio_to_context(
+                context=context,
+                context_mask=context_mask,
+                proprio=proprio.to(device=self.device, dtype=self.torch_dtype),
+            )
+        action = action.to(device=self.device, dtype=self.torch_dtype, non_blocking=True)
+        if action_is_pad is not None:
+            action_is_pad = action_is_pad.to(
+                device=self.device, dtype=torch.bool, non_blocking=True
+            )
+        if action_loss_is_pad is not None:
+            action_loss_is_pad = action_loss_is_pad.to(
+                device=self.device, dtype=torch.bool, non_blocking=True
+            )
+        action = self._pack_action_tokens(action)
+        action_is_pad = self._pack_action_is_pad(action_is_pad)
+        action_loss_is_pad = self._pack_action_is_pad(action_loss_is_pad)
+        if image_is_pad is not None:
+            image_is_pad = image_is_pad.to(device=self.device, dtype=torch.bool, non_blocking=True)
+        return {
+            "context": context,
+            "context_mask": context_mask,
+            "input_latents": input_latents,
+            "condition_latents": condition_latents,
+            "clip_fea": clip_fea,
+            "first_frame_latents": None,
+            "fuse_vae_embedding_in_latents": False,
+            "action": action,
+            "action_is_pad": action_is_pad,
+            "action_loss_is_pad": action_loss_is_pad,
+            "image_is_pad": image_is_pad,
+            "num_video_frames": num_frames,
+            "action_only_first_frame_fast_path": use_action_only_fast_path,
+        }
+
+    @torch.no_grad()
+    def _build_mot_attention_mask(
+        self,
+        video_seq_len: int,
+        action_seq_len: int,
+        video_tokens_per_frame: int,
+        device: torch.device,
+        action_visible_video_frames: Optional[int] = None,
+    ) -> torch.Tensor:
+        total_seq_len = video_seq_len + action_seq_len
+        mask = torch.zeros((total_seq_len, total_seq_len), dtype=torch.bool, device=device)
+        mask[:video_seq_len, :video_seq_len] = self.video_expert.build_video_to_video_mask(
+            video_seq_len=video_seq_len,
+            video_tokens_per_frame=video_tokens_per_frame,
+            device=device,
+        )
+        mask[video_seq_len:, video_seq_len:] = True
+        visible_frames = (
+            self.action_visible_video_frames
+            if action_visible_video_frames is None
+            else int(action_visible_video_frames)
+        )
+        if visible_frames <= 0:
+            raise ValueError(
+                f"`action_visible_video_frames` must be positive, got {visible_frames}."
+            )
+        visible_video_tokens = min(video_tokens_per_frame * visible_frames, video_seq_len)
+        mask[video_seq_len:, :visible_video_tokens] = True
+        return mask
+
+    def _build_action_video_conditioning_mask(
+        self, timestep_action: torch.Tensor
+    ) -> Optional[torch.Tensor]:
+        """Select samples whose action queries may read cached video K/V."""
+        cutoff = getattr(self, "action_video_timestep_cutoff", None)
+        if cutoff is None:
+            return None
+        if timestep_action.ndim == 0:
+            timestep_action = timestep_action.reshape(1)
+        if timestep_action.ndim != 1:
+            raise ValueError(
+                f"`timestep_action` must be scalar or 1D [B] for timestep-gated video conditioning, got shape {tuple(timestep_action.shape)}"
+            )
+        return timestep_action.to(dtype=torch.float32) >= float(cutoff)
+
+    def _build_action_clean_first_frame_conditioning_mask(
+        self, timestep_action: torch.Tensor
+    ) -> Optional[torch.Tensor]:
+        """Select samples that replace noisy-video K/V with clean frame-0 K/V."""
+        cutoff = getattr(self, "action_clean_first_frame_timestep_cutoff", None)
+        if cutoff is None:
+            return None
+        if timestep_action.ndim == 0:
+            timestep_action = timestep_action.reshape(1)
+        if timestep_action.ndim != 1:
+            raise ValueError(
+                f"`timestep_action` must be scalar or 1D [B] for timestep-gated clean-first-frame conditioning, got shape {tuple(timestep_action.shape)}"
+            )
+        return timestep_action.to(dtype=torch.float32) < float(cutoff)
+
+    @staticmethod
+    def _masked_batch_mean(values: torch.Tensor, mask: Optional[torch.Tensor]) -> torch.Tensor:
+        """Batch mean with inactive samples contributing zero.
+
+        Normalizing by the batch size (not the local active count) keeps every
+        active sample equally weighted after DDP averages the per-rank losses;
+        with no mask / an all-true mask this is exactly ``values.mean()``.
+        """
+        if mask is None:
+            return values.mean()
+        weights = mask.to(device=values.device, dtype=values.dtype)
+        return (values * weights).sum() / values.numel()
+
+    def _compute_video_loss_per_sample(
+        self,
+        pred_video: torch.Tensor,
+        target_video: torch.Tensor,
+        image_is_pad: Optional[torch.Tensor],
+        include_initial_video_step: bool,
+    ) -> torch.Tensor:
+        video_loss_token = F.mse_loss(
+            pred_video.float(), target_video.float(), reduction="none"
+        ).mean(dim=(1, 3, 4))
+        if image_is_pad is None:
+            return video_loss_token.mean(dim=1)
+        temporal_factor = int(self.vae.temporal_downsample_factor)
+        if temporal_factor <= 0:
+            raise ValueError(
+                f"`vae.temporal_downsample_factor` must be positive, got {temporal_factor}."
+            )
+        if image_is_pad.shape[1] < 1:
+            raise ValueError("`image_is_pad` must contain at least one frame.")
+        if (image_is_pad.shape[1] - 1) % temporal_factor != 0:
+            raise ValueError(
+                f"Cannot align `image_is_pad` with video latent steps: num_frames={image_is_pad.shape[1]}, temporal_downsample_factor={temporal_factor}."
+            )
+        tail_is_pad = image_is_pad[:, 1:]
+        latent_tail_is_pad = tail_is_pad.view(image_is_pad.shape[0], -1, temporal_factor).all(dim=2)
+        if include_initial_video_step:
+            video_is_pad = torch.cat([image_is_pad[:, :1], latent_tail_is_pad], dim=1)
+        else:
+            video_is_pad = latent_tail_is_pad
+        if video_is_pad.shape[1] != video_loss_token.shape[1]:
+            raise ValueError(
+                f"Video-loss mask shape mismatch: mask steps={video_is_pad.shape[1]}, loss steps={video_loss_token.shape[1]}."
+            )
+        valid = (~video_is_pad).to(device=video_loss_token.device, dtype=video_loss_token.dtype)
+        valid_sum = valid.sum(dim=1).clamp(min=1.0)
+        return (video_loss_token * valid).sum(dim=1) / valid_sum
+
+    @contextmanager
+    def _disable_video_cache_checkpointing(self):
+        old_mot_checkpoint = bool(getattr(self.mot, "mot_checkpoint_mixed_attn", False))
+        old_video_checkpoint = bool(getattr(self.video_expert, "use_gradient_checkpointing", False))
+        self.mot.mot_checkpoint_mixed_attn = False
+        if hasattr(self.video_expert, "use_gradient_checkpointing"):
+            self.video_expert.use_gradient_checkpointing = False
+        try:
+            yield
+        finally:
+            self.mot.mot_checkpoint_mixed_attn = old_mot_checkpoint
+            if hasattr(self.video_expert, "use_gradient_checkpointing"):
+                self.video_expert.use_gradient_checkpointing = old_video_checkpoint
+
+    def _build_frozen_video_cache(
+        self,
+        latents: torch.Tensor,
+        timestep_video: torch.Tensor,
+        context: torch.Tensor,
+        context_mask: torch.Tensor,
+        condition_latents: torch.Tensor,
+        clip_fea: torch.Tensor,
+        action_seq_len: int,
+        fuse_vae_embedding_in_latents: bool,
+        return_video_prediction: bool = False,
+        allow_trainable_video: bool = False,
+        action_visible_video_frames: Optional[int] = None,
+        max_video_layers: Optional[int] = None,
+    ):
+        if latents.ndim != 5:
+            raise ValueError(
+                f"Action-only video context must be [B,C,T,H,W], got shape {tuple(latents.shape)}"
+            )
+        visible_frames = (
+            self.action_visible_video_frames
+            if action_visible_video_frames is None
+            else int(action_visible_video_frames)
+        )
+        if visible_frames <= 0:
+            raise ValueError(
+                f"`action_visible_video_frames` must be positive, got {visible_frames}."
+            )
+        if latents.shape[2] < visible_frames:
+            raise ValueError(
+                f"Action-only video context requires at least {visible_frames} video latent frames so all action-visible frames exist, got shape {tuple(latents.shape)}"
+            )
+        if action_seq_len <= 0:
+            raise ValueError(f"`action_seq_len` must be positive, got {action_seq_len}")
+        lora_graph_requested = allow_trainable_video and bool(False)
+        if self.training or lora_graph_requested:
+            valid_alternating_cache = "joint" == "alternating" and allow_trainable_video
+            valid_lora_cache = "joint" == "action_only" and lora_graph_requested
+            if "joint" != "action_only" and (not valid_alternating_cache):
+                raise RuntimeError(
+                    f"Frozen video K/V cache is only valid for action-only training or an explicit alternating action block; got train_mode={'joint'!r}."
+                )
+            if valid_lora_cache:
+                allowed = {id(parameter) for (_, parameter) in self.video_lora_named_parameters()}
+                invalid_trainable = next(
+                    (
+                        name
+                        for (name, parameter) in self.video_expert.named_parameters()
+                        if parameter.requires_grad and id(parameter) not in allowed
+                    ),
+                    None,
+                )
+                if invalid_trainable is not None:
+                    raise RuntimeError(
+                        f"Video-LoRA cache permits only PEFT A/B tensors to train; first unexpected trainable parameter is {invalid_trainable!r}."
+                    )
+                if not any(
+                    (
+                        parameter.requires_grad
+                        for (_, parameter) in self.video_lora_named_parameters()
+                    )
+                ):
+                    raise RuntimeError(
+                        "Video LoRA is enabled but no adapter A/B tensor is trainable."
+                    )
+            elif not valid_alternating_cache:
+                trainable_video_parameter = next(
+                    (
+                        name
+                        for (name, parameter) in self.video_expert.named_parameters()
+                        if parameter.requires_grad
+                    ),
+                    None,
+                )
+                if trainable_video_parameter is not None:
+                    raise RuntimeError(
+                        f"Frozen video K/V cache requires every video-expert parameter to have requires_grad=False during training; first trainable parameter is {trainable_video_parameter!r}."
+                    )
+        retain_video_graph = "joint" == "action_only" and lora_graph_requested
+        graph_context = nullcontext() if retain_video_graph else torch.no_grad()
+        checkpoint_context = (
+            nullcontext() if retain_video_graph else self._disable_video_cache_checkpointing()
+        )
+        with graph_context:
+            video_pre = self.video_expert.pre_dit(
+                x=latents,
+                timestep=timestep_video,
+                context=context,
+                context_mask=context_mask,
+                action=None,
+                fuse_vae_embedding_in_latents=fuse_vae_embedding_in_latents,
+                clip_fea=clip_fea,
+                condition_latents=condition_latents,
+            )
+            video_seq_len = int(video_pre["tokens"].shape[1])
+            attention_mask = self._build_mot_attention_mask(
+                video_seq_len=video_seq_len,
+                action_seq_len=action_seq_len,
+                video_tokens_per_frame=int(video_pre["meta"]["tokens_per_frame"]),
+                device=video_pre["tokens"].device,
+                action_visible_video_frames=visible_frames,
+            )
+            with checkpoint_context:
+                prefill_kwargs = {
+                    "video_tokens": video_pre["tokens"],
+                    "video_freqs": video_pre["freqs"],
+                    "video_t_mod": video_pre["t_mod"],
+                    "video_context_payload": {
+                        "context": video_pre["context"],
+                        "mask": video_pre["context_mask"],
+                    },
+                    "video_attention_mask": attention_mask[:video_seq_len, :video_seq_len],
+                }
+                if return_video_prediction:
+                    if max_video_layers is not None:
+                        raise ValueError(
+                            "`return_video_prediction=true` requires all video layers; `max_video_layers` must be null."
+                        )
+                    (video_kv_cache, final_video_tokens) = self.mot.prefill_video_cache(
+                        **prefill_kwargs, return_final_tokens=True
+                    )
+                    video_prediction = self.video_expert.post_dit(final_video_tokens, video_pre)
+                else:
+                    video_kv_cache = self.mot.prefill_video_cache(
+                        **prefill_kwargs, max_layers=max_video_layers
+                    )
+                    video_prediction = None
+        if not self.training and bool(False):
+            with torch.no_grad():
+                video_kv_cache = self.mot.materialize_action_video_kv_cache(video_kv_cache)
+        if return_video_prediction:
+            return (video_kv_cache, attention_mask, video_seq_len, video_prediction)
+        return (video_kv_cache, attention_mask, video_seq_len)
+
+    @torch.no_grad()
+    def _build_noisy_first_frame_cache_view(
+        self,
+        *,
+        video_kv_cache: list[dict[str, torch.Tensor]],
+        video_seq_len: int,
+        video_latent_frames: int,
+        action_seq_len: int,
+    ) -> tuple[list[dict[str, torch.Tensor]], torch.Tensor, int]:
+        """View frame-0 K/V from a complete noisy-video per-layer cache.
+
+        This deliberately does not claim clean-frame equivalence: frame-0 was
+        embedded from noisy latent channels concatenated with the I2V condition,
+        and deeper frame-0 states were mixed with the remaining noisy video by
+        bidirectional self-attention. The returned tensors are prefix views, so
+        no second Video Expert prefill or duplicate K/V storage is introduced.
+        """
+        if getattr(self, "use_fixed_video_layers", None) is not None or bool(False):
+            raise ValueError(
+                "Noisy first-frame cache reuse requires direct per-layer video K/V (`use_fixed_video_layers=null`, `use_video_tokens=false`)."
+            )
+        expected_layers = int(self.mot.num_layers)
+        if len(video_kv_cache) != expected_layers:
+            raise ValueError(
+                f"Noisy first-frame cache reuse requires a complete video cache; expected {expected_layers} layers, got {len(video_kv_cache)}."
+            )
+        if video_latent_frames <= 0:
+            raise ValueError(f"`video_latent_frames` must be positive, got {video_latent_frames}.")
+        if video_seq_len <= 0 or video_seq_len % video_latent_frames != 0:
+            raise ValueError(
+                f"Noisy video token count must divide evenly across latent frames; tokens={video_seq_len}, latent_frames={video_latent_frames}."
+            )
+        first_frame_seq_len = video_seq_len // video_latent_frames
+        first_frame_cache: list[dict[str, torch.Tensor]] = []
+        cache_device = None
+        for layer_idx, layer_cache in enumerate(video_kv_cache):
+            if "k" not in layer_cache or "v" not in layer_cache:
+                raise ValueError(
+                    f"Noisy first-frame cache reuse requires `k` and `v` in video_kv_cache[{layer_idx}]."
+                )
+            k = layer_cache["k"]
+            v = layer_cache["v"]
+            if int(k.shape[1]) != video_seq_len or int(v.shape[1]) != video_seq_len:
+                raise ValueError(
+                    f"Noisy video K/V sequence mismatch at layer {layer_idx}: k={int(k.shape[1])}, v={int(v.shape[1])}, expected={video_seq_len}."
+                )
+            first_frame_cache.append(
+                {"k": k[:, :first_frame_seq_len], "v": v[:, :first_frame_seq_len]}
+            )
+            if cache_device is None:
+                cache_device = k.device
+        if cache_device is None:
+            raise ValueError("Noisy video cache is empty.")
+        first_frame_attention_mask = self._build_mot_attention_mask(
+            video_seq_len=first_frame_seq_len,
+            action_seq_len=action_seq_len,
+            video_tokens_per_frame=first_frame_seq_len,
+            device=cache_device,
+            action_visible_video_frames=1,
+        )
+        return (first_frame_cache, first_frame_attention_mask, first_frame_seq_len)
+
+    def _sample_action_video_latents(
+        self,
+        *,
+        first_frame_latents: torch.Tensor,
+        num_video_frames: int,
+        generator: Optional[torch.Generator],
+        rand_device: str,
+    ) -> torch.Tensor:
+        if first_frame_latents.ndim != 5 or first_frame_latents.shape[2] < 1:
+            raise ValueError(
+                f"`first_frame_latents` must be [B,C,T>=1,H,W], got {tuple(first_frame_latents.shape)}."
+            )
+        temporal_factor = int(self.vae.temporal_downsample_factor)
+        if temporal_factor <= 0:
+            raise ValueError(
+                f"`vae.temporal_downsample_factor` must be positive, got {temporal_factor}."
+            )
+        if num_video_frames <= 1:
+            raise ValueError(
+                f"Action-only inference requires more than one RGB frame, got {num_video_frames}."
+            )
+        if (num_video_frames - 1) % temporal_factor != 0:
+            raise ValueError(
+                f"Action-only inference RGB transitions must be divisible by `vae.temporal_downsample_factor`: num_video_frames={num_video_frames}, temporal_downsample_factor={temporal_factor}."
+            )
+        latent_t = (num_video_frames - 1) // temporal_factor + 1
+        if latent_t < self.action_visible_video_frames:
+            raise ValueError(
+                f"Action-only inference requires at least {self.action_visible_video_frames} video latent frames, got {latent_t} from num_video_frames={num_video_frames} and temporal_downsample_factor={temporal_factor}."
+            )
+        (batch_size, channels, _, latent_h, latent_w) = first_frame_latents.shape
+        return torch.randn(
+            (batch_size, channels, latent_t, latent_h, latent_w),
+            generator=generator,
+            device=rand_device,
+            dtype=torch.float32,
+        ).to(device=self.device, dtype=self.torch_dtype)
+
+    def training_loss(
+        self,
+        sample,
+        tiled: bool = False,
+        training_stage: Optional[str] = None,
+        video_feature_callback=None,
+    ):
+        resolved_stage = "joint"
+        if training_stage is not None and str(training_stage).strip().lower() != "joint":
+            raise ValueError("VPP2 joint training only accepts training_stage='joint'.")
+        if video_feature_callback is not None and "joint" != "joint":
+            raise ValueError("Video feature collection requires joint training")
+        inputs = self.build_inputs(sample, tiled=tiled, training_stage="joint")
+        input_latents = inputs["input_latents"]
+        context = inputs["context"]
+        context_mask = inputs["context_mask"]
+        action = inputs["action"]
+        action_is_pad = inputs.get("action_loss_is_pad", inputs["action_is_pad"])
+        image_is_pad = inputs["image_is_pad"]
+        condition_latents = inputs["condition_latents"]
+        clip_fea = inputs["clip_fea"]
+        batch_size = int(condition_latents.shape[0])
+        if inputs["action_only_first_frame_fast_path"]:
+            temporal_factor = int(self.vae.temporal_downsample_factor)
+            if temporal_factor <= 0:
+                raise ValueError(
+                    f"`vae.temporal_downsample_factor` must be positive, got {temporal_factor}."
+                )
+            num_video_frames = int(inputs["num_video_frames"])
+            if (num_video_frames - 1) % temporal_factor != 0:
+                raise ValueError(
+                    f"Action-only first-frame fast path RGB transitions must be divisible by `vae.temporal_downsample_factor`: num_video_frames={num_video_frames}, temporal_downsample_factor={temporal_factor}."
+                )
+            latent_t = (num_video_frames - 1) // temporal_factor + 1
+            if latent_t < self.action_visible_video_frames:
+                raise ValueError(
+                    f"Action-only first-frame fast path requires at least {self.action_visible_video_frames} video latent frames, got {latent_t}."
+                )
+            noise_video = torch.randn(
+                (
+                    batch_size,
+                    int(condition_latents.shape[1]),
+                    latent_t,
+                    int(condition_latents.shape[3]),
+                    int(condition_latents.shape[4]),
+                ),
+                device=condition_latents.device,
+                dtype=condition_latents.dtype,
+            )
+        else:
+            if input_latents is None:
+                raise RuntimeError("Full video training path requires encoded input latents.")
+            noise_video = torch.randn_like(input_latents)
+        joint_denoising = bool(False)
+        video_loss_mask = None
+        action_loss_mask = None
+        shared_action_timestep = None
+        timestep_video = self.train_video_scheduler.sample_training_t(
+            batch_size=batch_size, device=self.device, dtype=condition_latents.dtype
+        )
+        joint_video_loss = bool(
+            "joint" == "action" and getattr(self, "video_lora_joint_video_loss", False)
+        )
+        if "joint" == "action" and (not joint_video_loss):
+            latents = noise_video
+            target_video = None
+        elif inputs["action_only_first_frame_fast_path"]:
+            latents = noise_video
+            target_video = None
+        else:
+            latents = self.train_video_scheduler.add_noise(
+                input_latents, noise_video, timestep_video
+            )
+            target_video = self.train_video_scheduler.training_target(
+                input_latents, noise_video, timestep_video
+            )
+        noise_action = torch.randn_like(action)
+        if shared_action_timestep is not None:
+            timestep_action = shared_action_timestep.to(dtype=action.dtype)
+        else:
+            timestep_action = self.train_action_scheduler.sample_training_t(
+                batch_size=batch_size, device=self.device, dtype=action.dtype
+            )
+        noisy_action = self.train_action_scheduler.add_noise(action, noise_action, timestep_action)
+        target_action = self.train_action_scheduler.training_target(
+            action, noise_action, timestep_action
+        )
+        if target_video is None:
+            raise RuntimeError("Joint training requires a video flow-matching target.")
+        video_pre = self.video_expert.pre_dit(
+            x=latents,
+            timestep=timestep_video,
+            context=context,
+            context_mask=context_mask,
+            action=None,
+            fuse_vae_embedding_in_latents=inputs["fuse_vae_embedding_in_latents"],
+            clip_fea=clip_fea,
+            condition_latents=condition_latents,
+        )
+        action_pre = self.action_expert.pre_dit(
+            action_tokens=noisy_action,
+            timestep=timestep_action,
+            context=context,
+            context_mask=context_mask,
+        )
+        video_tokens = video_pre["tokens"]
+        action_tokens = action_pre["tokens"]
+        attention_mask = self._build_mot_attention_mask(
+            video_seq_len=video_tokens.shape[1],
+            action_seq_len=action_tokens.shape[1],
+            video_tokens_per_frame=int(video_pre["meta"]["tokens_per_frame"]),
+            device=video_tokens.device,
+        )
+        tokens_out = self.mot(
+            embeds_all={"video": video_tokens, "action": action_tokens},
+            attention_mask=attention_mask,
+            freqs_all={"video": video_pre["freqs"], "action": action_pre["freqs"]},
+            context_all={
+                "video": {"context": video_pre["context"], "mask": video_pre["context_mask"]},
+                "action": {"context": action_pre["context"], "mask": action_pre["context_mask"]},
+            },
+            t_mod_all={"video": video_pre["t_mod"], "action": action_pre["t_mod"]},
+            **{"video_feature_callback": video_feature_callback}
+            if video_feature_callback is not None
+            else {},
+        )
+        pred_video = self.video_expert.post_dit(tokens_out["video"], video_pre)
+        pred_action = self.action_expert.post_dit(tokens_out["action"], action_pre)
+        include_initial_video_step = inputs["first_frame_latents"] is None
+        if inputs["first_frame_latents"] is not None:
+            pred_video = pred_video[:, :, 1:]
+            target_video = target_video[:, :, 1:]
+        loss_video_per_sample = self._compute_video_loss_per_sample(
+            pred_video=pred_video,
+            target_video=target_video,
+            image_is_pad=image_is_pad,
+            include_initial_video_step=include_initial_video_step,
+        )
+        video_weight = self.train_video_scheduler.training_weight(timestep_video).to(
+            loss_video_per_sample.device, dtype=loss_video_per_sample.dtype
+        )
+        loss_video = self._masked_batch_mean(loss_video_per_sample * video_weight, video_loss_mask)
+        action_loss_token = F.mse_loss(
+            pred_action.float(), target_action.float(), reduction="none"
+        ).mean(dim=2)
+        if action_is_pad is not None:
+            valid = (~action_is_pad).to(
+                device=action_loss_token.device, dtype=action_loss_token.dtype
+            )
+            valid_sum = valid.sum(dim=1).clamp(min=1.0)
+            action_loss_per_sample = (action_loss_token * valid).sum(dim=1) / valid_sum
+        else:
+            action_loss_per_sample = action_loss_token.mean(dim=1)
+        action_weight = self.train_action_scheduler.training_weight(timestep_action).to(
+            action_loss_per_sample.device, dtype=action_loss_per_sample.dtype
+        )
+        loss_action = self._masked_batch_mean(
+            action_loss_per_sample * action_weight, action_loss_mask
+        )
+        loss_total = self.loss_lambda_video * loss_video + self.loss_lambda_action * loss_action
+        loss_dict = {
+            "loss_video": self.loss_lambda_video * float(loss_video.detach().item()),
+            "loss_action": self.loss_lambda_action * float(loss_action.detach().item()),
+        }
+        return (loss_total, loss_dict)
+
+    @torch.no_grad()
+    def _predict_action_noise_with_cache(
+        self,
+        latents_action: torch.Tensor,
+        timestep_action: torch.Tensor,
+        context: torch.Tensor,
+        context_mask: torch.Tensor,
+        video_kv_cache: list[dict[str, torch.Tensor]],
+        attention_mask: torch.Tensor,
+        video_seq_len: int,
+        clean_first_frame_kv_cache: Optional[list[dict[str, torch.Tensor]]] = None,
+        clean_first_frame_attention_mask: Optional[torch.Tensor] = None,
+        clean_first_frame_video_seq_len: Optional[int] = None,
+        clean_first_frame_start_layer: Optional[int] = None,
+    ) -> torch.Tensor:
+        action_pre = self.action_expert.pre_dit(
+            action_tokens=latents_action,
+            timestep=timestep_action,
+            context=context,
+            context_mask=context_mask,
+        )
+        action_tokens = self.mot.forward_action_with_video_cache(
+            action_tokens=action_pre["tokens"],
+            action_freqs=action_pre["freqs"],
+            action_t_mod=action_pre["t_mod"],
+            action_context_payload={
+                "context": action_pre["context"],
+                "mask": action_pre["context_mask"],
+            },
+            video_kv_cache=video_kv_cache,
+            attention_mask=attention_mask,
+            video_seq_len=video_seq_len,
+            clean_first_frame_kv_cache=clean_first_frame_kv_cache,
+            clean_first_frame_attention_mask=clean_first_frame_attention_mask,
+            clean_first_frame_video_seq_len=clean_first_frame_video_seq_len,
+            clean_first_frame_start_layer=clean_first_frame_start_layer,
+            action_video_conditioning_mask=self._build_action_video_conditioning_mask(
+                timestep_action
+            ),
+            action_clean_first_frame_conditioning_mask=self._build_action_clean_first_frame_conditioning_mask(
+                timestep_action
+            ),
+        )
+        return self.action_expert.post_dit(action_tokens, action_pre)
+
+    @torch.no_grad()
+    def infer_action(
+        self,
+        prompt: Optional[str],
+        input_image: torch.Tensor,
+        action_horizon: int,
+        condition_video: Optional[torch.Tensor] = None,
+        proprio: Optional[torch.Tensor] = None,
+        context: Optional[torch.Tensor] = None,
+        context_mask: Optional[torch.Tensor] = None,
+        negative_prompt: Optional[str] = None,
+        text_cfg_scale: float = 1.0,
+        num_inference_steps: int = 20,
+        sigma_shift: Optional[float] = None,
+        seed: Optional[int] = None,
+        video_seed: Optional[int] = None,
+        rand_device: str = "cpu",
+        tiled: bool = False,
+        num_video_frames: int = 17,
+        return_one_step_video: bool = False,
+        return_full_video: bool = False,
+        full_video_num_inference_steps: Optional[int] = None,
+        condition_observation_video: Optional[torch.Tensor] = None,
+    ) -> dict[str, Any]:
+        self.eval()
+        allowed_action_modes = {"bidirectional", "first_frame_causal", "three_frame"}
+        if (
+            str(getattr(self.video_expert, "video_attention_mask_mode", ""))
+            not in allowed_action_modes
+        ):
+            raise ValueError(
+                f"`infer_action` requires `video_attention_mask_mode` in {sorted(allowed_action_modes)}."
+            )
+        if input_image.ndim == 3:
+            input_image = input_image.unsqueeze(0)
+        if input_image.ndim != 4 or input_image.shape[0] != 1 or input_image.shape[1] != 3:
+            raise ValueError(
+                f"`input_image` must have shape [1,3,H,W] or [3,H,W], got {tuple(input_image.shape)}"
+            )
+        (_, _, height, width) = input_image.shape
+        if height % 16 != 0 or width % 16 != 0:
+            raise ValueError(
+                f"`input_image` must be resized before infer, expected multiples of 16 but got HxW=({height},{width})"
+            )
+        if proprio is not None:
+            if self.proprio_dim is None:
+                raise ValueError(
+                    "`proprio` was provided but `proprio_dim=None` so `proprio_encoder` is disabled."
+                )
+            if proprio.ndim == 1:
+                proprio = proprio.unsqueeze(0)
+            elif proprio.ndim == 2 and proprio.shape[0] == 1:
+                pass
+            else:
+                raise ValueError(
+                    f"`proprio` must be [D] or [1,D], got shape {tuple(proprio.shape)}"
+                )
+            if proprio.shape[1] != self.proprio_dim:
+                raise ValueError(
+                    f"`proprio` last dim must be {self.proprio_dim}, got {proprio.shape[1]}"
+                )
+            proprio = proprio.to(device=self.device, dtype=self.torch_dtype)
+        action_generator = (
+            None if seed is None else torch.Generator(device=rand_device).manual_seed(seed)
+        )
+        video_generator = (
+            None
+            if video_seed is None and seed is None
+            else torch.Generator(device=rand_device).manual_seed(
+                seed if video_seed is None else video_seed
+            )
+        )
+        if action_horizon % self.action_chunk_size != 0:
+            raise ValueError(
+                f"`action_horizon` must be divisible by action_chunk_size={self.action_chunk_size}, got {action_horizon}"
+            )
+        action_token_horizon = action_horizon // self.action_chunk_size
+        latents_action = torch.randn(
+            (1, action_token_horizon, self.action_expert.action_dim),
+            generator=action_generator,
+            device=rand_device,
+            dtype=torch.float32,
+        ).to(device=self.device, dtype=self.torch_dtype)
+        input_image = input_image.to(device=self.device, dtype=self.torch_dtype)
+        if condition_video is None:
+            if condition_observation_video is not None:
+                raise ValueError(
+                    "`condition_observation_video` requires `condition_video` history."
+                )
+            condition_latents = self._encode_input_image_latents_tensor(
+                input_image=input_image, tiled=tiled
+            )
+            clip_fea = self._encode_clip_fea_from_image(input_image)
+        else:
+            if condition_video.ndim == 4:
+                condition_video = condition_video.unsqueeze(0)
+            if condition_video.ndim != 5 or condition_video.shape[:2] != (1, 3):
+                raise ValueError(
+                    f"`condition_video` must be [1,3,T,H,W] or [3,T,H,W], got {tuple(condition_video.shape)}."
+                )
+            if condition_video.shape[-2:] != (height, width):
+                raise ValueError(
+                    f"Condition and input image spatial sizes must match, got {tuple(condition_video.shape[-2:])} vs {(height, width)}."
+                )
+            condition_num_frames = int(condition_video.shape[2])
+            if condition_num_frames <= 0 or condition_num_frames % 4 != 1:
+                raise ValueError(
+                    f"Condition temporal length must be positive and satisfy T % 4 == 1, got T={condition_num_frames}."
+                )
+            condition_video = condition_video.to(device=self.device, dtype=self.torch_dtype)
+            if condition_observation_video is not None:
+                if condition_observation_video.ndim == 4:
+                    condition_observation_video = condition_observation_video.unsqueeze(0)
+                if condition_observation_video.shape != (1, 3, 1, height, width):
+                    raise ValueError(
+                        f"`condition_observation_video` must be [1,3,1,H,W] with the same spatial size as the input image, got {tuple(condition_observation_video.shape)}"
+                    )
+                condition_observation_video = condition_observation_video.to(
+                    device=self.device, dtype=self.torch_dtype
+                )
+            condition_latents = self._encode_separate_condition_latents(
+                condition_video,
+                condition_observation_video=condition_observation_video,
+                tiled=tiled,
+            )
+            clip_fea = self._encode_clip_fea_from_video(condition_video[:, :, :1])
+        first_frame_latents = condition_latents[:, :, :1]
+        fuse_flag = bool(getattr(self.video_expert, "fuse_vae_embedding_in_latents", False))
+        use_prompt = prompt is not None
+        use_context = context is not None or context_mask is not None
+        if use_prompt and use_context:
+            raise ValueError("`prompt` and `context/context_mask` are mutually exclusive.")
+        if not use_prompt and (not use_context):
+            raise ValueError("Either `prompt` or both `context/context_mask` must be provided.")
+        if use_prompt:
+            (context, context_mask) = self.encode_prompt(prompt)
+        else:
+            if context is None or context_mask is None:
+                raise ValueError("`context` and `context_mask` must be both provided together.")
+            if context.ndim == 2:
+                context = context.unsqueeze(0)
+            if context_mask.ndim == 1:
+                context_mask = context_mask.unsqueeze(0)
+            if context.ndim != 3 or context_mask.ndim != 2:
+                raise ValueError(
+                    f"`context/context_mask` must be [B,L,D]/[B,L], got {tuple(context.shape)} and {tuple(context_mask.shape)}"
+                )
+            context = context.to(device=self.device, dtype=self.torch_dtype, non_blocking=True)
+            context_mask = context_mask.to(device=self.device, dtype=torch.bool, non_blocking=True)
+        if proprio is not None:
+            (context, context_mask) = self._append_proprio_to_context(
+                context=context, context_mask=context_mask, proprio=proprio
+            )
+        latents_video = self._sample_action_video_latents(
+            first_frame_latents=condition_latents,
+            num_video_frames=num_video_frames,
+            generator=video_generator,
+            rand_device=rand_device,
+        )
+        timestep_video = torch.full(
+            (latents_video.shape[0],),
+            float(self.train_video_scheduler.num_train_timesteps),
+            dtype=condition_latents.dtype,
+            device=self.device,
+        )
+        capture_video_prediction = return_one_step_video or return_full_video
+        clean_start_layer = getattr(self, "action_clean_first_frame_start_layer", None)
+        clean_timestep_cutoff = getattr(self, "action_clean_first_frame_timestep_cutoff", None)
+        first_frame_kv_source = getattr(self, "action_first_frame_kv_source", "clean_prefill")
+        first_frame_route_enabled = (
+            clean_start_layer is not None or clean_timestep_cutoff is not None
+        )
+        video_cache_result = self._build_frozen_video_cache(
+            latents=latents_video,
+            timestep_video=timestep_video,
+            context=context,
+            context_mask=context_mask,
+            condition_latents=condition_latents,
+            clip_fea=clip_fea,
+            action_seq_len=latents_action.shape[1],
+            fuse_vae_embedding_in_latents=fuse_flag,
+            return_video_prediction=capture_video_prediction,
+            max_video_layers=None
+            if capture_video_prediction
+            or (first_frame_route_enabled and first_frame_kv_source == "noisy_cache_prefix")
+            else clean_start_layer,
+        )
+        video_prediction = None
+        if capture_video_prediction:
+            (video_kv_cache, attention_mask, video_seq_len, video_prediction) = video_cache_result
+        else:
+            (video_kv_cache, attention_mask, video_seq_len) = video_cache_result
+        del video_cache_result
+        clean_cache_kwargs = {}
+        if first_frame_route_enabled:
+            if first_frame_kv_source == "noisy_cache_prefix":
+                (
+                    clean_first_frame_kv_cache,
+                    clean_first_frame_attention_mask,
+                    clean_first_frame_video_seq_len,
+                ) = self._build_noisy_first_frame_cache_view(
+                    video_kv_cache=video_kv_cache,
+                    video_seq_len=video_seq_len,
+                    video_latent_frames=int(latents_video.shape[2]),
+                    action_seq_len=int(latents_action.shape[1]),
+                )
+            else:
+                (
+                    clean_first_frame_kv_cache,
+                    clean_first_frame_attention_mask,
+                    clean_first_frame_video_seq_len,
+                ) = self._build_frozen_video_cache(
+                    latents=first_frame_latents,
+                    timestep_video=torch.zeros_like(timestep_video),
+                    context=context,
+                    context_mask=context_mask,
+                    condition_latents=first_frame_latents,
+                    clip_fea=clip_fea,
+                    action_seq_len=latents_action.shape[1],
+                    fuse_vae_embedding_in_latents=fuse_flag,
+                    action_visible_video_frames=1,
+                )
+            clean_cache_kwargs = {
+                "clean_first_frame_kv_cache": clean_first_frame_kv_cache,
+                "clean_first_frame_attention_mask": clean_first_frame_attention_mask,
+                "clean_first_frame_video_seq_len": clean_first_frame_video_seq_len,
+            }
+            if clean_start_layer is not None:
+                clean_cache_kwargs["clean_first_frame_start_layer"] = clean_start_layer
+        (infer_timesteps_action, infer_deltas_action) = (
+            self.infer_action_scheduler.build_inference_schedule(
+                num_inference_steps=num_inference_steps,
+                device=self.device,
+                dtype=latents_action.dtype,
+                shift_override=sigma_shift,
+            )
+        )
+        for step_t_action, step_delta_action in zip(infer_timesteps_action, infer_deltas_action):
+            timestep_action = step_t_action.unsqueeze(0).to(
+                dtype=latents_action.dtype, device=self.device
+            )
+            pred_action_posi = self._predict_action_noise_with_cache(
+                latents_action=latents_action,
+                timestep_action=timestep_action,
+                context=context,
+                context_mask=context_mask,
+                video_kv_cache=video_kv_cache,
+                attention_mask=attention_mask,
+                video_seq_len=video_seq_len,
+                **clean_cache_kwargs,
+            )
+            pred_action = pred_action_posi
+            latents_action = self.infer_action_scheduler.step(
+                pred_action, step_delta_action, latents_action
+            )
+        result = {
+            "action": self._unpack_action_tokens(latents_action)[0]
+            .detach()
+            .to(device="cpu", dtype=torch.float32)
+        }
+        del video_kv_cache
+        if first_frame_route_enabled:
+            del clean_first_frame_kv_cache
+        if capture_video_prediction:
+            if video_prediction is None:
+                raise RuntimeError("One-step video capture did not return a video prediction.")
+            (one_step_timesteps, one_step_deltas) = (
+                self.infer_video_scheduler.build_inference_schedule(
+                    num_inference_steps=1,
+                    device=self.device,
+                    dtype=latents_video.dtype,
+                    shift_override=sigma_shift,
+                )
+            )
+            if len(one_step_timesteps) != 1 or len(one_step_deltas) != 1:
+                raise RuntimeError(
+                    f"One-step video capture returned an unexpected scheduler length: timesteps={len(one_step_timesteps)}, deltas={len(one_step_deltas)}."
+                )
+            one_step_timestep = one_step_timesteps[0].to(
+                device=timestep_video.device, dtype=timestep_video.dtype
+            )
+            if not torch.allclose(timestep_video, one_step_timestep.expand_as(timestep_video)):
+                raise RuntimeError(
+                    f"Action video-cache timestep does not match standalone one-step video scheduler: cache={timestep_video.detach().cpu().tolist()} schedule={float(one_step_timestep.detach().cpu())}."
+                )
+            one_step_video_latents = self.infer_video_scheduler.step(
+                video_prediction, one_step_deltas[0], latents_video
+            )
+            full_video_inference_steps = int(full_video_num_inference_steps or num_inference_steps)
+            if return_full_video and full_video_inference_steps <= 0:
+                raise ValueError(
+                    f"`full_video_num_inference_steps` must be positive, got {full_video_inference_steps}."
+                )
+            if return_full_video:
+                (video_timesteps, video_deltas) = (
+                    self.infer_video_scheduler.build_inference_schedule(
+                        num_inference_steps=full_video_inference_steps,
+                        device=self.device,
+                        dtype=latents_video.dtype,
+                        shift_override=sigma_shift,
+                    )
+                )
+                if (
+                    len(video_timesteps) != full_video_inference_steps
+                    or len(video_deltas) != full_video_inference_steps
+                ):
+                    raise RuntimeError(
+                        f"Full video trajectory capture returned an unexpected scheduler length: expected={full_video_inference_steps}, timesteps={len(video_timesteps)}, deltas={len(video_deltas)}."
+                    )
+                scheduled_timestep = video_timesteps[0].to(
+                    device=timestep_video.device, dtype=timestep_video.dtype
+                )
+                if not torch.allclose(timestep_video, scheduled_timestep.expand_as(timestep_video)):
+                    raise RuntimeError(
+                        f"Action video-cache timestep does not match full-video scheduler: cache={timestep_video.detach().cpu().tolist()} schedule={float(scheduled_timestep.detach().cpu())}."
+                    )
+                trajectory_latents = self.infer_video_scheduler.step(
+                    video_prediction, video_deltas[0], latents_video
+                )
+                for step_t_video, step_delta_video in zip(video_timesteps[1:], video_deltas[1:]):
+                    timestep_video_next = step_t_video.unsqueeze(0).to(
+                        dtype=trajectory_latents.dtype, device=self.device
+                    )
+                    next_cache_result = self._build_frozen_video_cache(
+                        latents=trajectory_latents,
+                        timestep_video=timestep_video_next,
+                        context=context,
+                        context_mask=context_mask,
+                        condition_latents=condition_latents,
+                        clip_fea=clip_fea,
+                        action_seq_len=latents_action.shape[1],
+                        fuse_vae_embedding_in_latents=fuse_flag,
+                        return_video_prediction=True,
+                    )
+                    (_, _, _, next_video_prediction) = next_cache_result
+                    del next_cache_result
+                    trajectory_latents = self.infer_video_scheduler.step(
+                        next_video_prediction, step_delta_video, trajectory_latents
+                    )
+            result["video_one_step"] = self._decode_latents(one_step_video_latents, tiled=tiled)
+            result["video"] = result["video_one_step"]
+            result["video_inference_steps"] = 1
+            result["video_one_step_source"] = "action_context_cache_forward"
+            if return_full_video:
+                result["video_full"] = self._decode_latents(trajectory_latents, tiled=tiled)
+                result["video_full_inference_steps"] = full_video_inference_steps
+                result["video_trajectory_num_inference_steps"] = full_video_inference_steps
+                result["video_full_shares_action_context_initial_noise"] = True
+                result["video_full_shares_action_context_initial_prediction"] = True
+        return result
+
+    def save_checkpoint(self, path, optimizer=None, step=None):
+        payload = {"mot": self.mot.state_dict(), "step": step, "torch_dtype": str(self.torch_dtype)}
+        if self.proprio_encoder is not None:
+            payload["proprio_encoder"] = self.proprio_encoder.state_dict()
+        if optimizer is not None:
+            payload["optimizer"] = optimizer.state_dict()
+        torch.save(payload, path)
+
+    @staticmethod
+    def _load_video_checkpoint_state(path):
+        payload = torch.load(path, map_location="cpu", mmap=True, weights_only=True)
+        if not isinstance(payload, dict):
+            raise ValueError(
+                f"Video checkpoint must contain a state dictionary, got {type(payload)}"
+            )
+        mot = payload.get("mot")
+        if isinstance(mot, dict):
+            payload = {
+                key[len("mixtures.video.") :]: value
+                for (key, value) in mot.items()
+                if key.startswith("mixtures.video.")
+            }
+        for key in ("dit", "state_dict", "module", "model_state"):
+            nested = payload.get(key)
+            if isinstance(nested, dict):
+                payload = nested
+                break
+        state = {key: value for (key, value) in payload.items() if isinstance(value, torch.Tensor)}
+        if not state:
+            raise ValueError(f"No video tensors found in checkpoint: {path}")
+        return state
+
+    def load_checkpoint(
+        self,
+        path,
+        optimizer=None,
+        video_checkpoint_override=None,
+        load_video_from_action_checkpoint=True,
+    ):
+        payload = torch.load(path, map_location="cpu", weights_only=True, mmap=True)
+        if "mot" not in payload:
+            raise ValueError("Resume requires a full training checkpoint containing mot")
+        self.mot.load_state_dict(payload["mot"], strict=True)
+        if self.proprio_encoder is not None:
+            self.proprio_encoder.load_state_dict(payload["proprio_encoder"], strict=True)
+        if optimizer is not None and "optimizer" in payload:
+            optimizer.load_state_dict(payload["optimizer"])
+        return payload.get("step")
+
+    def forward(self, *args, **kwargs):
+        return self.training_loss(*args, **kwargs)
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan21.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan21.py
new file mode 100644
index 0000000000000000000000000000000000000000..2333ecb3611bbacd42eed41c1b04b35e0ba5a249
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan21.py
@@ -0,0 +1,775 @@
+from pathlib import Path
+from typing import Any, Optional, Sequence, Union
+
+import torch
+import torch.nn.functional as F
+from PIL import Image
+
+from bagel_cp1000.utils.logging_config import get_logger
+
+from .helpers.loader import load_wan21_i2v_14b_components
+from .schedulers.scheduler_continuous import WanContinuousFlowMatchScheduler
+
+logger = get_logger(__name__)
+
+
+class Wan21VideoModel(torch.nn.Module):
+    """Standalone Wan2.1-I2V model for video-only flow-matching training."""
+
+    def __init__(
+        self,
+        dit,
+        vae,
+        text_encoder=None,
+        tokenizer=None,
+        clip_encoder=None,
+        text_dim: int = 4096,
+        device: str = "cpu",
+        torch_dtype: torch.dtype = torch.float32,
+        video_train_shift: float = 5.0,
+        video_infer_shift: float = 3.0,
+        video_num_train_timesteps: int = 1000,
+        video_train_sampling_strategy: str = "beta",
+        video_train_beta_alpha: float = 7.0,
+        video_train_beta_beta: float = 1.0,
+        condition_clip_frame: str = "first",
+        condition_latent_loss_weight: float = 1.0,
+    ):
+        super().__init__()
+        self.dit = dit
+        self.vae = vae
+        self.text_encoder = text_encoder
+        self.tokenizer = tokenizer
+        self.clip_encoder = clip_encoder
+        self.text_dim = int(text_dim)
+        self.device = torch.device(device)
+        self.torch_dtype = torch_dtype
+        self.train_mode = "video_only"
+        self.condition_clip_frame = str(condition_clip_frame).strip().lower()
+        if self.condition_clip_frame not in {"first", "latest"}:
+            raise ValueError(
+                "`condition_clip_frame` must be one of: first, latest; got "
+                f"{condition_clip_frame!r}."
+            )
+        self.condition_latent_loss_weight = float(condition_latent_loss_weight)
+        if self.condition_latent_loss_weight < 0.0:
+            raise ValueError(
+                "`condition_latent_loss_weight` must be non-negative, got "
+                f"{condition_latent_loss_weight}."
+            )
+
+        self.train_video_scheduler = WanContinuousFlowMatchScheduler(
+            num_train_timesteps=video_num_train_timesteps,
+            shift=video_train_shift,
+            train_sampling_strategy=video_train_sampling_strategy,
+            beta_alpha=video_train_beta_alpha,
+            beta_beta=video_train_beta_beta,
+        )
+        self.infer_video_scheduler = WanContinuousFlowMatchScheduler(
+            num_train_timesteps=video_num_train_timesteps,
+            shift=video_infer_shift,
+        )
+        self.train_scheduler = self.train_video_scheduler
+        self.infer_scheduler = self.infer_video_scheduler
+
+        self.dit.to(device=self.device, dtype=self.torch_dtype)
+        if self.text_encoder is not None:
+            self.text_encoder.to(device=self.device, dtype=self.torch_dtype)
+        self._restore_auxiliary_component_dtypes()
+
+    def set_train_mode(self, train_mode: str):
+        normalized = str(train_mode).strip().lower()
+        if normalized != "video_only":
+            raise ValueError(
+                f"Wan21VideoModel only supports train_mode=video_only, got {normalized}"
+            )
+        self.train_mode = normalized
+        return self
+
+    def _restore_auxiliary_component_dtypes(self):
+        if self.clip_encoder is not None:
+            self.clip_encoder.eval().requires_grad_(False).to(
+                device=self.device,
+                dtype=torch.float16,
+            )
+        self.vae.eval().requires_grad_(False).to(
+            device=self.device,
+            dtype=torch.float32,
+        )
+
+    @staticmethod
+    def _check_resize_height_width(height: int, width: int, num_frames: int):
+        if height % 16 != 0:
+            height = (height + 15) // 16 * 16
+        if width % 16 != 0:
+            width = (width + 15) // 16 * 16
+        if num_frames % 4 != 1:
+            num_frames = (num_frames + 3) // 4 * 4 + 1
+        return height, width, num_frames
+
+    @torch.no_grad()
+    def encode_prompt(self, prompt: Union[str, Sequence[str]]):
+        if self.text_encoder is None or self.tokenizer is None:
+            raise ValueError(
+                "Prompt encoding requires a loaded text encoder/tokenizer; "
+                "otherwise provide cached context/context_mask."
+            )
+        ids, mask = self.tokenizer(
+            prompt,
+            return_mask=True,
+            add_special_tokens=True,
+        )
+        ids = ids.to(self.device)
+        mask = mask.to(self.device, dtype=torch.bool)
+        prompt_emb = self.text_encoder(ids, mask)
+        sequence_lengths = mask.gt(0).sum(dim=1).long()
+        for index, length in enumerate(sequence_lengths):
+            prompt_emb[index, length:] = 0
+        return prompt_emb.to(device=self.device), torch.ones_like(mask)
+
+    @torch.no_grad()
+    def _encode_video_latents(
+        self,
+        video_tensor,
+        tiled=False,
+        tile_size=(30, 52),
+        tile_stride=(15, 26),
+    ):
+        self._restore_auxiliary_component_dtypes()
+        videos = [
+            video_tensor[index].to(device=self.device, dtype=torch.float32)
+            for index in range(video_tensor.shape[0])
+        ]
+        latents = self.vae.encode(
+            videos,
+            device=self.device,
+            tiled=tiled,
+            tile_size=tile_size,
+            tile_stride=tile_stride,
+        )
+        if isinstance(latents, list):
+            latents = torch.stack(latents)
+        return latents.to(device=self.device, dtype=self.torch_dtype)
+
+    @torch.no_grad()
+    def _encode_input_image_latents_tensor(
+        self,
+        input_image: torch.Tensor,
+        tiled=False,
+        tile_size=(30, 52),
+        tile_stride=(15, 26),
+    ):
+        self._restore_auxiliary_component_dtypes()
+        if input_image.ndim == 3:
+            input_image = input_image.unsqueeze(0)
+        if input_image.ndim != 4 or input_image.shape[0] != 1 or input_image.shape[1] != 3:
+            raise ValueError(
+                "`input_image` must have shape [1,3,H,W] or [3,H,W], "
+                f"got {tuple(input_image.shape)}"
+            )
+        image_video = input_image.to(device=self.device, dtype=torch.float32)[0].unsqueeze(1)
+        latents = self.vae.encode(
+            [image_video],
+            device=self.device,
+            tiled=tiled,
+            tile_size=tile_size,
+            tile_stride=tile_stride,
+        )
+        if isinstance(latents, list):
+            latents = torch.stack(latents)
+        return latents.to(device=self.device, dtype=self.torch_dtype)
+
+    @torch.no_grad()
+    def _encode_separate_condition_latents(
+        self,
+        condition_video: torch.Tensor,
+        condition_observation_video: Optional[torch.Tensor] = None,
+        tiled: bool = False,
+    ) -> torch.Tensor:
+        """Encode history and the observation frame as separate VAE inputs.
+
+        The history prefix is encoded as a normal ``4k+1`` video (two latent
+        frames for the current five-RGB-frame contract).  The current
+        observation is encoded independently as a one-frame image, yielding
+        one additional clean latent.  Keeping this operation separate from
+        the full target VAE encode makes train and inference use the same
+        three clean condition latents.
+        """
+        history_latents = self._encode_video_latents(condition_video, tiled=tiled)
+        if condition_observation_video is None:
+            return history_latents
+        if condition_observation_video.ndim == 4:
+            condition_observation_video = condition_observation_video.unsqueeze(0)
+        if condition_observation_video.ndim != 5 or condition_observation_video.shape[1] != 3:
+            raise ValueError(
+                "`condition_observation_video` must be [B,3,1,H,W] or [3,1,H,W], got "
+                f"{tuple(condition_observation_video.shape)}"
+            )
+        if condition_observation_video.shape[2] != 1:
+            raise ValueError(
+                "`condition_observation_video` must contain exactly one RGB frame, got "
+                f"T={condition_observation_video.shape[2]}"
+            )
+        observation_latents = self._encode_video_latents(
+            condition_observation_video,
+            tiled=tiled,
+        )
+        if observation_latents.shape[2] != 1:
+            raise ValueError(
+                "One-frame observation VAE encoding must produce one latent frame, got "
+                f"T={observation_latents.shape[2]}"
+            )
+        if history_latents.shape[0] != observation_latents.shape[0]:
+            raise ValueError("History and observation condition batches must match.")
+        if history_latents.shape[3:] != observation_latents.shape[3:]:
+            raise ValueError(
+                "History and observation condition latent spatial grids must match, got "
+                f"{tuple(history_latents.shape[3:])} vs {tuple(observation_latents.shape[3:])}."
+            )
+        return torch.cat([history_latents, observation_latents], dim=2)
+
+    @torch.no_grad()
+    def _encode_clip_fea_from_video(self, video_tensor: torch.Tensor):
+        self._restore_auxiliary_component_dtypes()
+        if self.clip_encoder is None:
+            raise ValueError(
+                "Wan2.1-I2V-14B requires CLIP image features; load clip_encoder "
+                "or provide sample['clip_fea']."
+            )
+        if video_tensor.ndim != 5:
+            raise ValueError(f"`video_tensor` must be [B,C,T,H,W], got {tuple(video_tensor.shape)}")
+        videos = [
+            video_tensor[index].to(device=self.device, dtype=torch.float16)
+            for index in range(video_tensor.shape[0])
+        ]
+        return self.clip_encoder(videos).to(
+            device=self.device,
+            dtype=self.torch_dtype,
+        )
+
+    @torch.no_grad()
+    def _encode_clip_fea_from_image(self, input_image: torch.Tensor):
+        if input_image.ndim == 3:
+            input_image = input_image.unsqueeze(0)
+        return self._encode_clip_fea_from_video(input_image.unsqueeze(2))
+
+    def _select_clip_condition_video(
+        self,
+        condition_video: torch.Tensor,
+    ) -> torch.Tensor:
+        if condition_video.ndim != 5:
+            raise ValueError(
+                f"`condition_video` must be [B,3,T,H,W], got {tuple(condition_video.shape)}"
+            )
+        if self.condition_clip_frame == "first":
+            return condition_video[:, :, :1]
+        return condition_video[:, :, -1:]
+
+    @torch.no_grad()
+    def _decode_latents(
+        self,
+        latents,
+        tiled=False,
+        tile_size=(30, 52),
+        tile_stride=(15, 26),
+    ):
+        self._restore_auxiliary_component_dtypes()
+        video_tensor = self.vae.decode(
+            latents.to(device=self.device, dtype=torch.float32),
+            device=self.device,
+            tiled=tiled,
+            tile_size=tile_size,
+            tile_stride=tile_stride,
+        )
+        if isinstance(video_tensor, list):
+            video_tensor = torch.stack(video_tensor)
+        if video_tensor.ndim != 5 or video_tensor.shape[0] != 1:
+            raise ValueError(
+                f"Decoded video must have shape [1,C,T,H,W], got {tuple(video_tensor.shape)}"
+            )
+        video_tensor = video_tensor[0].detach().float().clamp(-1, 1)
+        video_tensor = ((video_tensor + 1.0) * 127.5).to(torch.uint8).cpu()
+        return [
+            Image.fromarray(video_tensor[:, frame].permute(1, 2, 0).numpy())
+            for frame in range(video_tensor.shape[1])
+        ]
+
+    def build_inputs(self, sample, tiled: bool = False):
+        if "video" not in sample:
+            raise ValueError("Wan21VideoModel training requires sample['video'].")
+        if "context" not in sample or "context_mask" not in sample:
+            raise ValueError("Wan21VideoModel training requires cached context/context_mask.")
+
+        video = sample["video"]
+        context = sample["context"]
+        context_mask = sample["context_mask"]
+        if video.ndim != 5 or video.shape[1] != 3:
+            raise ValueError(f"`sample['video']` must be [B,3,T,H,W], got {tuple(video.shape)}")
+        batch_size, _, num_frames, height, width = video.shape
+        if height % 16 != 0 or width % 16 != 0:
+            raise ValueError(
+                f"Video spatial dims must be multiples of 16, got H={height}, W={width}"
+            )
+        if num_frames <= 1 or num_frames % 4 != 1:
+            raise ValueError(
+                f"Video temporal length must be greater than 1 and satisfy T % 4 == 1, got {num_frames}"
+            )
+        if context.ndim != 3 or context_mask.ndim != 2:
+            raise ValueError(
+                "`context/context_mask` must be [B,L,D]/[B,L], "
+                f"got {tuple(context.shape)} and {tuple(context_mask.shape)}"
+            )
+        if context.shape[:2] != context_mask.shape or context.shape[0] != batch_size:
+            raise ValueError("Cached text context batch/sequence dimensions do not match.")
+
+        input_video = video.to(
+            device=self.device,
+            dtype=self.torch_dtype,
+            non_blocking=True,
+        )
+        input_latents = self._encode_video_latents(input_video, tiled=tiled)
+        condition_video = sample.get("condition_video")
+        if condition_video is None:
+            condition_latents = input_latents[:, :, :1]
+            clip_condition_video = input_video[:, :, :1]
+        else:
+            if condition_video.ndim != 5 or condition_video.shape[:2] != (
+                batch_size,
+                3,
+            ):
+                raise ValueError(
+                    "`sample['condition_video']` must be [B,3,T,H,W], got "
+                    f"{tuple(condition_video.shape)}"
+                )
+            if condition_video.shape[-2:] != (height, width):
+                raise ValueError(
+                    "Condition and target video spatial sizes must match, got "
+                    f"{tuple(condition_video.shape[-2:])} vs {(height, width)}."
+                )
+            condition_num_frames = int(condition_video.shape[2])
+            if condition_num_frames <= 1 or condition_num_frames % 4 != 1:
+                raise ValueError(
+                    "Condition temporal length must be > 1 and satisfy T % 4 == 1, "
+                    f"got T={condition_num_frames}."
+                )
+            condition_video = condition_video.to(
+                device=self.device,
+                dtype=self.torch_dtype,
+                non_blocking=True,
+            )
+            condition_observation_video = sample.get("condition_observation_video")
+            if condition_observation_video is not None:
+                if condition_observation_video.ndim != 5 or condition_observation_video.shape[
+                    :2
+                ] != (
+                    batch_size,
+                    3,
+                ):
+                    raise ValueError(
+                        "`sample['condition_observation_video']` must be [B,3,1,H,W], got "
+                        f"{tuple(condition_observation_video.shape)}"
+                    )
+                if condition_observation_video.shape[2] != 1:
+                    raise ValueError(
+                        "`sample['condition_observation_video']` must contain one frame, got "
+                        f"T={condition_observation_video.shape[2]}"
+                    )
+                condition_observation_video = condition_observation_video.to(
+                    device=self.device,
+                    dtype=self.torch_dtype,
+                    non_blocking=True,
+                )
+            condition_latents = self._encode_separate_condition_latents(
+                condition_video,
+                condition_observation_video=condition_observation_video,
+                tiled=tiled,
+            )
+            if condition_latents.shape[3:] != input_latents.shape[3:]:
+                raise ValueError(
+                    "History condition and target must produce the same latent spatial "
+                    "grid for direct channel conditioning, got "
+                    f"{tuple(condition_latents.shape[2:])} vs "
+                    f"{tuple(input_latents.shape[2:])}."
+                )
+            if condition_latents.shape[2] > input_latents.shape[2]:
+                raise ValueError(
+                    "History condition cannot be longer than the target latent sequence, "
+                    f"got {condition_latents.shape[2]} > {input_latents.shape[2]}."
+                )
+            clip_condition_video = self._select_clip_condition_video(condition_video)
+            if condition_observation_video is not None:
+                input_latents = torch.cat(
+                    [
+                        condition_latents,
+                        input_latents[:, :, condition_latents.shape[2] :],
+                    ],
+                    dim=2,
+                )
+        clip_fea = sample.get("clip_fea")
+        if clip_fea is None:
+            clip_fea = self._encode_clip_fea_from_video(clip_condition_video)
+        else:
+            clip_fea = clip_fea.to(
+                device=self.device,
+                dtype=self.torch_dtype,
+                non_blocking=True,
+            )
+        image_is_pad = sample.get("image_is_pad")
+        if image_is_pad is not None:
+            if image_is_pad.shape != (batch_size, num_frames):
+                raise ValueError(
+                    "`image_is_pad` must match [B,T], "
+                    f"got {tuple(image_is_pad.shape)} vs {(batch_size, num_frames)}"
+                )
+            image_is_pad = image_is_pad.to(
+                device=self.device,
+                dtype=torch.bool,
+                non_blocking=True,
+            )
+
+        return {
+            "input_latents": input_latents,
+            "condition_latents": condition_latents,
+            "clip_fea": clip_fea,
+            "context": context.to(
+                device=self.device,
+                dtype=self.torch_dtype,
+                non_blocking=True,
+            ),
+            "context_mask": context_mask.to(
+                device=self.device,
+                dtype=torch.bool,
+                non_blocking=True,
+            ),
+            "image_is_pad": image_is_pad,
+        }
+
+    def _compute_video_loss_per_sample(
+        self,
+        pred_video: torch.Tensor,
+        target_video: torch.Tensor,
+        image_is_pad: Optional[torch.Tensor],
+        condition_latent_frames: int = 0,
+    ):
+        video_loss_token = F.mse_loss(
+            pred_video.float(),
+            target_video.float(),
+            reduction="none",
+        ).mean(dim=(1, 3, 4))
+        condition_latent_frames = int(condition_latent_frames)
+        if condition_latent_frames < 0 or condition_latent_frames > video_loss_token.shape[1]:
+            raise ValueError(
+                "`condition_latent_frames` must be in [0, latent_frames], got "
+                f"{condition_latent_frames} for latent_frames={video_loss_token.shape[1]}"
+            )
+        token_weight = torch.ones(
+            (video_loss_token.shape[1],),
+            device=video_loss_token.device,
+            dtype=video_loss_token.dtype,
+        )
+        if condition_latent_frames:
+            token_weight[:condition_latent_frames] = self.condition_latent_loss_weight
+
+        if image_is_pad is None:
+            weighted = video_loss_token * token_weight.unsqueeze(0)
+            return weighted.sum(dim=1) / token_weight.sum().clamp(min=1.0)
+
+        temporal_factor = int(self.vae.temporal_downsample_factor)
+        if temporal_factor <= 0:
+            raise ValueError(
+                f"`vae.temporal_downsample_factor` must be positive, got {temporal_factor}"
+            )
+        if (image_is_pad.shape[1] - 1) % temporal_factor != 0:
+            raise ValueError(
+                "Cannot align image padding with video latents: "
+                f"frames={image_is_pad.shape[1]}, temporal_factor={temporal_factor}"
+            )
+        latent_tail_is_pad = (
+            image_is_pad[:, 1:]
+            .reshape(
+                image_is_pad.shape[0],
+                -1,
+                temporal_factor,
+            )
+            .all(dim=2)
+        )
+        video_is_pad = torch.cat([image_is_pad[:, :1], latent_tail_is_pad], dim=1)
+        if video_is_pad.shape != video_loss_token.shape:
+            raise ValueError(
+                "Video loss padding mask mismatch: "
+                f"mask={tuple(video_is_pad.shape)}, loss={tuple(video_loss_token.shape)}"
+            )
+        valid = (~video_is_pad).to(
+            device=video_loss_token.device,
+            dtype=video_loss_token.dtype,
+        )
+        weighted_valid = valid * token_weight.unsqueeze(0)
+        return (video_loss_token * weighted_valid).sum(dim=1) / weighted_valid.sum(dim=1).clamp(
+            min=1.0
+        )
+
+    def training_loss(self, sample, tiled: bool = False):
+        inputs = self.build_inputs(sample, tiled=tiled)
+        input_latents = inputs["input_latents"]
+        batch_size = input_latents.shape[0]
+        noise = torch.randn_like(input_latents)
+        timestep = self.train_video_scheduler.sample_training_t(
+            batch_size=batch_size,
+            device=self.device,
+            dtype=input_latents.dtype,
+        )
+        noisy_latents = self.train_video_scheduler.add_noise(
+            input_latents,
+            noise,
+            timestep,
+        )
+        target = self.train_video_scheduler.training_target(
+            input_latents,
+            noise,
+            timestep,
+        )
+        pred = self.dit(
+            x=noisy_latents,
+            timestep=timestep,
+            context=inputs["context"],
+            context_mask=inputs["context_mask"],
+            action=None,
+            clip_fea=inputs["clip_fea"],
+            condition_latents=inputs["condition_latents"],
+        )
+        loss_per_sample = self._compute_video_loss_per_sample(
+            pred_video=pred,
+            target_video=target,
+            image_is_pad=inputs["image_is_pad"],
+            condition_latent_frames=inputs["condition_latents"].shape[2],
+        )
+        weight = self.train_video_scheduler.training_weight(timestep).to(
+            device=loss_per_sample.device,
+            dtype=loss_per_sample.dtype,
+        )
+        loss_video = (loss_per_sample * weight).mean()
+        return loss_video, {"loss_video": float(loss_video.detach().item())}
+
+    @torch.no_grad()
+    def infer(
+        self,
+        prompt: Optional[str],
+        input_image: torch.Tensor,
+        num_frames: int,
+        action=None,
+        action_horizon=None,
+        proprio=None,
+        context: Optional[torch.Tensor] = None,
+        context_mask: Optional[torch.Tensor] = None,
+        negative_prompt: Optional[str] = None,
+        text_cfg_scale: float = 1.0,
+        action_cfg_scale: float = 1.0,
+        num_inference_steps: int = 20,
+        sigma_shift: Optional[float] = None,
+        seed: Optional[int] = None,
+        rand_device: str = "cpu",
+        generator: Optional[torch.Generator] = None,
+        tiled: bool = False,
+        condition_video: Optional[torch.Tensor] = None,
+        condition_observation_video: Optional[torch.Tensor] = None,
+    ) -> dict[str, Any]:
+        del action, action_horizon, proprio, negative_prompt
+        del text_cfg_scale, action_cfg_scale
+        self.eval()
+        if input_image.ndim == 3:
+            input_image = input_image.unsqueeze(0)
+        if input_image.ndim != 4 or input_image.shape[:2] != (1, 3):
+            raise ValueError(
+                f"`input_image` must be [1,3,H,W] or [3,H,W], got {tuple(input_image.shape)}"
+            )
+        _, _, height, width = input_image.shape
+        checked = self._check_resize_height_width(height, width, int(num_frames))
+        if checked != (height, width, int(num_frames)):
+            raise ValueError(
+                "Input H/W must be multiples of 16 and num_frames must satisfy T % 4 == 1, "
+                f"got H={height}, W={width}, T={num_frames}"
+            )
+
+        use_prompt = prompt is not None
+        use_context = context is not None or context_mask is not None
+        if use_prompt == use_context:
+            raise ValueError("Provide exactly one of prompt or cached context/context_mask.")
+        if use_prompt:
+            context, context_mask = self.encode_prompt(prompt)
+        else:
+            if context is None or context_mask is None:
+                raise ValueError("context and context_mask must be provided together.")
+            if context.ndim == 2:
+                context = context.unsqueeze(0)
+            if context_mask.ndim == 1:
+                context_mask = context_mask.unsqueeze(0)
+            if context.ndim != 3 or context_mask.ndim != 2:
+                raise ValueError(
+                    "`context/context_mask` must be [B,L,D]/[B,L], "
+                    f"got {tuple(context.shape)} and {tuple(context_mask.shape)}"
+                )
+            context = context.to(device=self.device, dtype=self.torch_dtype)
+            context_mask = context_mask.to(device=self.device, dtype=torch.bool)
+
+        latent_t = (int(num_frames) - 1) // int(self.vae.temporal_downsample_factor) + 1
+        latent_h = height // int(self.vae.upsampling_factor)
+        latent_w = width // int(self.vae.upsampling_factor)
+        if generator is None and seed is not None:
+            generator = torch.Generator(device=rand_device).manual_seed(seed)
+        noise_device = generator.device if generator is not None else torch.device(rand_device)
+        latents = torch.randn(
+            (1, int(self.vae.model.z_dim), latent_t, latent_h, latent_w),
+            generator=generator,
+            device=noise_device,
+            dtype=torch.float32,
+        ).to(device=self.device, dtype=self.torch_dtype)
+
+        input_image = input_image.to(device=self.device, dtype=self.torch_dtype)
+        if condition_video is None:
+            if condition_observation_video is not None:
+                raise ValueError(
+                    "`condition_observation_video` requires `condition_video` history."
+                )
+            condition_latents = self._encode_input_image_latents_tensor(
+                input_image,
+                tiled=tiled,
+            )
+            clip_fea = self._encode_clip_fea_from_image(input_image)
+        else:
+            if condition_video.ndim == 4:
+                condition_video = condition_video.unsqueeze(0)
+            if condition_video.ndim != 5 or condition_video.shape[:2] != (1, 3):
+                raise ValueError(
+                    "`condition_video` must be [1,3,T,H,W] or [3,T,H,W], got "
+                    f"{tuple(condition_video.shape)}"
+                )
+            if condition_video.shape[-2:] != (height, width):
+                raise ValueError(
+                    "Condition and output spatial sizes must match, got "
+                    f"{tuple(condition_video.shape[-2:])} vs {(height, width)}."
+                )
+            condition_video = condition_video.to(
+                device=self.device,
+                dtype=self.torch_dtype,
+            )
+            if condition_observation_video is not None:
+                if condition_observation_video.ndim == 4:
+                    condition_observation_video = condition_observation_video.unsqueeze(0)
+                if condition_observation_video.shape != (1, 3, 1, height, width):
+                    raise ValueError(
+                        "`condition_observation_video` must be [1,3,1,H,W] with the same "
+                        f"spatial size as the output, got {tuple(condition_observation_video.shape)}"
+                    )
+                condition_observation_video = condition_observation_video.to(
+                    device=self.device,
+                    dtype=self.torch_dtype,
+                )
+            condition_latents = self._encode_separate_condition_latents(
+                condition_video,
+                condition_observation_video=condition_observation_video,
+                tiled=tiled,
+            )
+            if int(condition_latents.shape[2]) > latent_t:
+                raise ValueError(
+                    "Condition cannot be longer than the output latent sequence for "
+                    f"direct conditioning, got {condition_latents.shape[2]} > {latent_t}."
+                )
+            clip_fea = self._encode_clip_fea_from_video(
+                self._select_clip_condition_video(condition_video)
+            )
+        timesteps, deltas = self.infer_video_scheduler.build_inference_schedule(
+            num_inference_steps=num_inference_steps,
+            device=self.device,
+            dtype=latents.dtype,
+            shift_override=sigma_shift,
+        )
+        for step_t, step_delta in zip(timesteps, deltas):
+            timestep = step_t.reshape(1).to(
+                device=self.device,
+                dtype=latents.dtype,
+            )
+            pred = self.dit(
+                x=latents,
+                timestep=timestep,
+                context=context,
+                context_mask=context_mask,
+                action=None,
+                clip_fea=clip_fea,
+                condition_latents=condition_latents,
+            )
+            latents = self.infer_video_scheduler.step(pred, step_delta, latents)
+        return {"video": self._decode_latents(latents, tiled=tiled)}
+
+    @classmethod
+    def from_wan21_14b_pretrained(
+        cls,
+        model_id,
+        dit_checkpoint_path,
+        video_dit_config,
+        video_scheduler,
+        tokenizer_model_id,
+        tokenizer_max_len=512,
+        load_text_encoder=False,
+        load_clip_encoder=True,
+        condition_clip_frame="first",
+        condition_latent_loss_weight=1.0,
+        redirect_common_files=False,
+        torch_dtype=torch.bfloat16,
+        device="cpu",
+    ):
+        del redirect_common_files
+        components = load_wan21_i2v_14b_components(
+            device=device,
+            torch_dtype=torch_dtype,
+            model_id=model_id,
+            dit_checkpoint_path=dit_checkpoint_path,
+            tokenizer_model_id=tokenizer_model_id,
+            tokenizer_max_len=tokenizer_max_len,
+            dit_config=video_dit_config,
+            load_text_encoder=load_text_encoder,
+            load_clip_encoder=load_clip_encoder,
+        )
+        model = cls(
+            dit=components.dit,
+            vae=components.vae,
+            text_encoder=components.text_encoder,
+            tokenizer=components.tokenizer,
+            clip_encoder=components.clip_encoder,
+            text_dim=int(video_dit_config.get("text_dim", 4096)),
+            video_train_shift=float(video_scheduler["train_shift"]),
+            video_infer_shift=float(video_scheduler["infer_shift"]),
+            video_num_train_timesteps=int(video_scheduler["num_train_timesteps"]),
+            video_train_sampling_strategy=str(video_scheduler["train_sampling_strategy"]),
+            video_train_beta_alpha=float(video_scheduler["beta_alpha"]),
+            video_train_beta_beta=float(video_scheduler["beta_beta"]),
+            condition_clip_frame=condition_clip_frame,
+            condition_latent_loss_weight=condition_latent_loss_weight,
+            torch_dtype=torch_dtype,
+            device=device,
+        )
+        model.model_paths = {
+            "video_dit": components.dit_path,
+            "vae": components.vae_path,
+            "text_encoder": components.text_encoder_path,
+            "tokenizer": components.tokenizer_path,
+            "clip_encoder": components.clip_encoder_path,
+        }
+        return model
+
+    def save_checkpoint(self, path, optimizer=None, step=None):
+        payload = {
+            "format": "wan21_video_v1",
+            "dit": self.dit.state_dict(),
+            "step": step,
+        }
+        if optimizer is not None:
+            payload["optimizer"] = optimizer.state_dict()
+        torch.save(payload, Path(path))
+
+    def load_checkpoint(self, path, optimizer=None):
+        payload = torch.load(path, map_location="cpu")
+        state = payload["dit"] if "dit" in payload else payload
+        self.dit.load_state_dict(state, strict=True)
+        if optimizer is not None and "optimizer" in payload:
+            optimizer.load_state_dict(payload["optimizer"])
+        return payload
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_clip.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_clip.py
new file mode 100644
index 0000000000000000000000000000000000000000..ac7e4b6518349c648b77af9b98052f229514b31b
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_clip.py
@@ -0,0 +1,194 @@
+import math
+from typing import Sequence
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torchvision.transforms as T
+
+
+def pos_interpolate(pos, seq_len):
+    if pos.size(1) == seq_len:
+        return pos
+    src_grid = int(math.sqrt(pos.size(1)))
+    tar_grid = int(math.sqrt(seq_len))
+    n = pos.size(1) - src_grid * src_grid
+    return torch.cat(
+        [
+            pos[:, :n],
+            F.interpolate(
+                pos[:, n:].float().reshape(1, src_grid, src_grid, -1).permute(0, 3, 1, 2),
+                size=(tar_grid, tar_grid),
+                mode="bicubic",
+                align_corners=False,
+            )
+            .flatten(2)
+            .transpose(1, 2),
+        ],
+        dim=1,
+    )
+
+
+class QuickGELU(nn.Module):
+    def forward(self, x):
+        return x * torch.sigmoid(1.702 * x)
+
+
+class LayerNorm(nn.LayerNorm):
+    def forward(self, x):
+        weight = self.weight.float() if self.weight is not None else None
+        bias = self.bias.float() if self.bias is not None else None
+        return F.layer_norm(x.float(), self.normalized_shape, weight, bias, self.eps).type_as(x)
+
+
+class SelfAttention(nn.Module):
+    def __init__(self, dim, num_heads, causal=False, attn_dropout=0.0, proj_dropout=0.0):
+        super().__init__()
+        self.dim = dim
+        self.num_heads = num_heads
+        self.head_dim = dim // num_heads
+        self.causal = causal
+        self.attn_dropout = attn_dropout
+        self.proj_dropout = proj_dropout
+        self.to_qkv = nn.Linear(dim, dim * 3)
+        self.proj = nn.Linear(dim, dim)
+
+    def forward(self, x):
+        b, s, c = x.shape
+        q, k, v = self.to_qkv(x).view(b, s, 3, self.num_heads, self.head_dim).unbind(2)
+        q = q.transpose(1, 2)
+        k = k.transpose(1, 2)
+        v = v.transpose(1, 2)
+        p = self.attn_dropout if self.training else 0.0
+        x = F.scaled_dot_product_attention(q, k, v, is_causal=self.causal, dropout_p=p)
+        x = x.transpose(1, 2).reshape(b, s, c)
+        return F.dropout(self.proj(x), self.proj_dropout, self.training)
+
+
+class AttentionBlock(nn.Module):
+    def __init__(
+        self,
+        dim,
+        mlp_ratio,
+        num_heads,
+        post_norm=False,
+        causal=False,
+        activation="gelu",
+        attn_dropout=0.0,
+        proj_dropout=0.0,
+        norm_eps=1e-5,
+    ):
+        super().__init__()
+        self.post_norm = post_norm
+        self.norm1 = LayerNorm(dim, eps=norm_eps)
+        self.attn = SelfAttention(dim, num_heads, causal, attn_dropout, proj_dropout)
+        self.norm2 = LayerNorm(dim, eps=norm_eps)
+        act = QuickGELU() if activation == "quick_gelu" else nn.GELU()
+        self.mlp = nn.Sequential(
+            nn.Linear(dim, int(dim * mlp_ratio)),
+            act,
+            nn.Linear(int(dim * mlp_ratio), dim),
+            nn.Dropout(proj_dropout),
+        )
+
+    def forward(self, x):
+        if self.post_norm:
+            x = x + self.norm1(self.attn(x))
+            return x + self.norm2(self.mlp(x))
+        x = x + self.attn(self.norm1(x))
+        return x + self.mlp(self.norm2(x))
+
+
+class VisionTransformer(nn.Module):
+    def __init__(
+        self,
+        image_size=224,
+        patch_size=14,
+        dim=1280,
+        mlp_ratio=4,
+        out_dim=1024,
+        num_heads=16,
+        num_layers=32,
+        pool_type="token",
+        pre_norm=True,
+        post_norm=False,
+        activation="gelu",
+        attn_dropout=0.0,
+        proj_dropout=0.0,
+        embedding_dropout=0.0,
+        norm_eps=1e-5,
+    ):
+        super().__init__()
+        self.image_size = image_size
+        self.patch_size = patch_size
+        self.num_patches = (image_size // patch_size) ** 2
+        self.pool_type = pool_type
+        gain = 1.0 / math.sqrt(dim)
+        self.patch_embedding = nn.Conv2d(
+            3, dim, kernel_size=patch_size, stride=patch_size, bias=not pre_norm
+        )
+        self.cls_embedding = nn.Parameter(gain * torch.randn(1, 1, dim))
+        self.pos_embedding = nn.Parameter(gain * torch.randn(1, self.num_patches + 1, dim))
+        self.dropout = nn.Dropout(embedding_dropout)
+        self.pre_norm = LayerNorm(dim, eps=norm_eps) if pre_norm else None
+        self.transformer = nn.Sequential(
+            *[
+                AttentionBlock(
+                    dim,
+                    mlp_ratio,
+                    num_heads,
+                    post_norm,
+                    False,
+                    activation,
+                    attn_dropout,
+                    proj_dropout,
+                    norm_eps,
+                )
+                for _ in range(num_layers)
+            ]
+        )
+        self.post_norm = LayerNorm(dim, eps=norm_eps)
+        self.head = nn.Parameter(gain * torch.randn(dim, out_dim))
+
+    def forward(self, x, interpolation=False, use_31_block=False):
+        b = x.size(0)
+        x = self.patch_embedding(x).flatten(2).permute(0, 2, 1)
+        x = torch.cat([self.cls_embedding.expand(b, -1, -1), x], dim=1)
+        x = self.dropout(
+            x
+            + (
+                pos_interpolate(self.pos_embedding, x.size(1))
+                if interpolation
+                else self.pos_embedding
+            )
+        )
+        if self.pre_norm is not None:
+            x = self.pre_norm(x)
+        if use_31_block:
+            return self.transformer[:-1](x)
+        return self.transformer(x)
+
+
+class WanImageEncoderCLIP(nn.Module):
+    def __init__(self):
+        super().__init__()
+        self.visual = VisionTransformer()
+        self.log_scale = nn.Parameter(math.log(1 / 0.07) * torch.ones([]))
+        self.normalize = T.Normalize(
+            mean=[0.48145466, 0.4578275, 0.40821073],
+            std=[0.26862954, 0.26130258, 0.27577711],
+        )
+
+    def forward(self, videos: Sequence[torch.Tensor]):
+        size = (self.visual.image_size,) * 2
+        frames = torch.cat(
+            [
+                F.interpolate(
+                    u[:, :1].transpose(0, 1), size=size, mode="bicubic", align_corners=False
+                )
+                for u in videos
+            ],
+            dim=0,
+        )
+        frames = self.normalize(frames.mul(0.5).add(0.5))
+        return self.visual(frames, use_31_block=True)
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_dit.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_dit.py
new file mode 100644
index 0000000000000000000000000000000000000000..872666ae828415088a0a76517f3efe1860356d14
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_dit.py
@@ -0,0 +1,616 @@
+import math
+from typing import Any, Dict, Optional, Tuple
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from einops import rearrange
+
+from .helpers.gradient import gradient_checkpoint_forward
+from bagel_cp1000.utils.logging_config import get_logger
+
+logger = get_logger(__name__)
+
+
+def flash_attention(
+    q: torch.Tensor,
+    k: torch.Tensor,
+    v: torch.Tensor,
+    num_heads: int,
+    ctx_mask: Optional[torch.Tensor] = None,
+):
+    q = rearrange(q, "b s (n d) -> b n s d", n=num_heads)
+    k = rearrange(k, "b s (n d) -> b n s d", n=num_heads)
+    v = rearrange(v, "b s (n d) -> b n s d", n=num_heads)
+    if ctx_mask is not None:
+        if ctx_mask.dim() == 2:
+            ctx_mask = ctx_mask.unsqueeze(0).unsqueeze(0)
+        elif ctx_mask.dim() == 3:
+            ctx_mask = ctx_mask.unsqueeze(1)
+        ctx_mask = ctx_mask.to(device=q.device, dtype=torch.bool)
+    x = F.scaled_dot_product_attention(q, k, v, attn_mask=ctx_mask)
+    return rearrange(x, "b n s d -> b s (n d)", n=num_heads)
+
+
+def modulate(x: torch.Tensor, shift: torch.Tensor, scale: torch.Tensor):
+    return x * (1 + scale) + shift
+
+
+def sinusoidal_embedding_1d(dim: int, position: torch.Tensor):
+    assert dim % 2 == 0
+    half = dim // 2
+    dtype = position.dtype if position.is_floating_point() else torch.float32
+    position = position.type(torch.float64)
+    sinusoid = torch.outer(
+        position,
+        torch.pow(
+            10000, -torch.arange(half, dtype=torch.float64, device=position.device).div(half)
+        ),
+    )
+    return torch.cat([torch.cos(sinusoid), torch.sin(sinusoid)], dim=1).to(dtype)
+
+
+def precompute_freqs_cis(dim: int, end: int = 1024, theta: float = 10000.0):
+    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2).double()[: (dim // 2)] / dim))
+    freqs = torch.outer(torch.arange(end, device=freqs.device), freqs)
+    return torch.polar(torch.ones_like(freqs), freqs)
+
+
+def precompute_freqs_cis_3d(dim: int, end: int = 1024, theta: float = 10000.0):
+    return (
+        precompute_freqs_cis(dim - 4 * (dim // 6), end, theta),
+        precompute_freqs_cis(2 * (dim // 6), end, theta),
+        precompute_freqs_cis(2 * (dim // 6), end, theta),
+    )
+
+
+def rope_apply(x: torch.Tensor, freqs: torch.Tensor, num_heads: int):
+    x = rearrange(x, "b s (n d) -> b s n d", n=num_heads)
+    x_out = torch.view_as_complex(x.to(torch.float64).reshape(*x.shape[:-1], -1, 2))
+    freqs = freqs.to(device=x.device)
+    if freqs.device.type == "npu":
+        freqs = freqs.to(torch.complex64)
+    x_out = torch.view_as_real(x_out * freqs).flatten(2)
+    return x_out.to(x.dtype)
+
+
+class WanRMSNorm(nn.Module):
+    def __init__(self, dim: int, eps: float = 1e-5):
+        super().__init__()
+        self.eps = eps
+        self.weight = nn.Parameter(torch.ones(dim))
+
+    def forward(self, x: torch.Tensor):
+        dtype = x.dtype
+        x = x.float()
+        x = x * torch.rsqrt(x.pow(2).mean(dim=-1, keepdim=True) + self.eps)
+        return x.to(dtype) * self.weight
+
+
+class WanLayerNorm(nn.LayerNorm):
+    def __init__(self, dim: int, eps: float = 1e-6, elementwise_affine: bool = False):
+        super().__init__(dim, eps=eps, elementwise_affine=elementwise_affine)
+
+    def forward(self, x: torch.Tensor):
+        weight = self.weight.float() if self.weight is not None else None
+        bias = self.bias.float() if self.bias is not None else None
+        return F.layer_norm(x.float(), self.normalized_shape, weight, bias, self.eps).type_as(x)
+
+
+class SelfAttention(nn.Module):
+    def __init__(
+        self,
+        hidden_dim: int,
+        attn_head_dim: Optional[int] = None,
+        num_heads: int = 16,
+        eps: float = 1e-6,
+        qk_norm: bool = True,
+    ):
+        super().__init__()
+        if attn_head_dim is None:
+            if hidden_dim % num_heads != 0:
+                raise ValueError(
+                    "`hidden_dim` must be divisible by `num_heads` when attn_head_dim is omitted."
+                )
+            attn_head_dim = hidden_dim // num_heads
+        self.hidden_dim = hidden_dim
+        self.num_heads = num_heads
+        self.attn_head_dim = attn_head_dim
+        self.attn_hidden_dim = num_heads * attn_head_dim
+
+        self.q = nn.Linear(hidden_dim, self.attn_hidden_dim)
+        self.k = nn.Linear(hidden_dim, self.attn_hidden_dim)
+        self.v = nn.Linear(hidden_dim, self.attn_hidden_dim)
+        self.o = nn.Linear(self.attn_hidden_dim, hidden_dim)
+        self.norm_q = WanRMSNorm(self.attn_hidden_dim, eps=eps) if qk_norm else nn.Identity()
+        self.norm_k = WanRMSNorm(self.attn_hidden_dim, eps=eps) if qk_norm else nn.Identity()
+
+    def forward(self, x, freqs, self_attn_mask: Optional[torch.Tensor] = None):
+        q = self.norm_q(self.q(x))
+        k = self.norm_k(self.k(x))
+        v = self.v(x)
+        q = rope_apply(q, freqs, self.num_heads)
+        k = rope_apply(k, freqs, self.num_heads)
+        return self.o(flash_attention(q, k, v, self.num_heads, ctx_mask=self_attn_mask))
+
+
+class CrossAttention(nn.Module):
+    def __init__(
+        self,
+        hidden_dim: int,
+        attn_head_dim: Optional[int] = None,
+        num_heads: int = 16,
+        eps: float = 1e-6,
+        qk_norm: bool = True,
+    ):
+        super().__init__()
+        if attn_head_dim is None:
+            if hidden_dim % num_heads != 0:
+                raise ValueError(
+                    "`hidden_dim` must be divisible by `num_heads` when attn_head_dim is omitted."
+                )
+            attn_head_dim = hidden_dim // num_heads
+        self.hidden_dim = hidden_dim
+        self.num_heads = num_heads
+        self.attn_head_dim = attn_head_dim
+        self.attn_hidden_dim = num_heads * attn_head_dim
+
+        self.q = nn.Linear(hidden_dim, self.attn_hidden_dim)
+        self.k = nn.Linear(hidden_dim, self.attn_hidden_dim)
+        self.v = nn.Linear(hidden_dim, self.attn_hidden_dim)
+        self.o = nn.Linear(self.attn_hidden_dim, hidden_dim)
+        self.norm_q = WanRMSNorm(self.attn_hidden_dim, eps=eps) if qk_norm else nn.Identity()
+        self.norm_k = WanRMSNorm(self.attn_hidden_dim, eps=eps) if qk_norm else nn.Identity()
+
+    def forward(self, x: torch.Tensor, ctx: torch.Tensor, ctx_mask: Optional[torch.Tensor] = None):
+        q = self.norm_q(self.q(x))
+        k = self.norm_k(self.k(ctx))
+        v = self.v(ctx)
+        x = flash_attention(q, k, v, self.num_heads, ctx_mask=ctx_mask)
+        return self.o(x)
+
+
+class I2VCrossAttention(CrossAttention):
+    def __init__(
+        self,
+        hidden_dim: int,
+        attn_head_dim: Optional[int],
+        num_heads: int,
+        eps: float,
+        qk_norm: bool = True,
+    ):
+        super().__init__(hidden_dim, attn_head_dim, num_heads, eps, qk_norm)
+        self.k_img = nn.Linear(hidden_dim, self.attn_hidden_dim)
+        self.v_img = nn.Linear(hidden_dim, self.attn_hidden_dim)
+        self.norm_k_img = WanRMSNorm(self.attn_hidden_dim, eps=eps) if qk_norm else nn.Identity()
+
+    def forward(self, x: torch.Tensor, ctx: torch.Tensor, ctx_mask: Optional[torch.Tensor] = None):
+        if ctx.shape[1] < 257:
+            raise ValueError(
+                f"Wan2.1 i2v context must include 257 image tokens, got {ctx.shape[1]}."
+            )
+        ctx_img = ctx[:, :257]
+        ctx_txt = ctx[:, 257:]
+        text_mask = None if ctx_mask is None else ctx_mask[..., 257:]
+
+        q = self.norm_q(self.q(x))
+        k_img = self.norm_k_img(self.k_img(ctx_img))
+        v_img = self.v_img(ctx_img)
+        img_x = flash_attention(q, k_img, v_img, self.num_heads)
+
+        k = self.norm_k(self.k(ctx_txt))
+        v = self.v(ctx_txt)
+        txt_x = flash_attention(q, k, v, self.num_heads, ctx_mask=text_mask)
+        return self.o(txt_x + img_x)
+
+
+class GateModule(nn.Module):
+    def forward(self, x, gate, residual):
+        return x + gate * residual
+
+
+class DiTBlock(nn.Module):
+    def __init__(
+        self,
+        hidden_dim: int,
+        attn_head_dim: Optional[int],
+        num_heads: int,
+        ffn_dim: int,
+        eps: float = 1e-6,
+        cross_attn_type: str = "t2v",
+        qk_norm: bool = True,
+        cross_attn_norm: bool = False,
+    ):
+        super().__init__()
+        self.hidden_dim = hidden_dim
+        self.attn_head_dim = attn_head_dim if attn_head_dim is not None else hidden_dim // num_heads
+        self.num_heads = num_heads
+        self.ffn_dim = ffn_dim
+
+        self.self_attn = SelfAttention(hidden_dim, self.attn_head_dim, num_heads, eps, qk_norm)
+        cross_cls = I2VCrossAttention if cross_attn_type == "i2v" else CrossAttention
+        self.cross_attn = cross_cls(hidden_dim, self.attn_head_dim, num_heads, eps, qk_norm)
+        self.norm1 = WanLayerNorm(hidden_dim, eps=eps, elementwise_affine=False)
+        self.norm2 = WanLayerNorm(hidden_dim, eps=eps, elementwise_affine=False)
+        self.norm3 = WanLayerNorm(hidden_dim, eps=eps, elementwise_affine=bool(cross_attn_norm))
+        self.ffn = nn.Sequential(
+            nn.Linear(hidden_dim, ffn_dim),
+            nn.GELU(approximate="tanh"),
+            nn.Linear(ffn_dim, hidden_dim),
+        )
+        self.modulation = nn.Parameter(torch.randn(1, 6, hidden_dim) / hidden_dim**0.5)
+        self.gate = GateModule()
+
+    def forward(
+        self,
+        x,
+        context,
+        t_mod,
+        freqs,
+        context_mask=None,
+        self_attn_mask: Optional[torch.Tensor] = None,
+    ):
+        if context_mask is not None and context_mask.dim() == 3:
+            context_mask = context_mask.unsqueeze(1)
+        has_seq = len(t_mod.shape) == 4
+        chunk_dim = 2 if has_seq else 1
+        shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = (
+            self.modulation.to(dtype=t_mod.dtype, device=t_mod.device) + t_mod
+        ).chunk(6, dim=chunk_dim)
+        if has_seq:
+            shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = (
+                shift_msa.squeeze(2),
+                scale_msa.squeeze(2),
+                gate_msa.squeeze(2),
+                shift_mlp.squeeze(2),
+                scale_mlp.squeeze(2),
+                gate_mlp.squeeze(2),
+            )
+        y = self.self_attn(
+            modulate(self.norm1(x), shift_msa, scale_msa), freqs, self_attn_mask=self_attn_mask
+        )
+        x = self.gate(x, gate_msa, y)
+        x = x + self.cross_attn(self.norm3(x), context, ctx_mask=context_mask)
+        y = self.ffn(modulate(self.norm2(x), shift_mlp, scale_mlp))
+        x = self.gate(x, gate_mlp, y)
+        return x
+
+
+class Head(nn.Module):
+    def __init__(self, dim: int, out_dim: int, patch_size: Tuple[int, int, int], eps: float = 1e-6):
+        super().__init__()
+        self.dim = dim
+        self.out_dim = out_dim
+        self.patch_size = tuple(patch_size)
+        self.norm = WanLayerNorm(dim, eps=eps, elementwise_affine=False)
+        self.head = nn.Linear(dim, out_dim * math.prod(self.patch_size))
+        self.modulation = nn.Parameter(torch.randn(1, 2, dim) / dim**0.5)
+
+    def forward(self, x, t):
+        if len(t.shape) == 3:
+            shift, scale = (
+                self.modulation.unsqueeze(0).to(dtype=t.dtype, device=t.device) + t.unsqueeze(2)
+            ).chunk(2, dim=2)
+            return self.head(self.norm(x) * (1 + scale.squeeze(2)) + shift.squeeze(2))
+        shift, scale = (self.modulation.to(dtype=t.dtype, device=t.device) + t.unsqueeze(1)).chunk(
+            2, dim=1
+        )
+        return self.head(self.norm(x) * (1 + scale) + shift)
+
+
+class MLPProj(nn.Module):
+    def __init__(self, in_dim: int, out_dim: int):
+        super().__init__()
+        self.proj = nn.Sequential(
+            nn.LayerNorm(in_dim),
+            nn.Linear(in_dim, in_dim),
+            nn.GELU(),
+            nn.Linear(in_dim, out_dim),
+            nn.LayerNorm(out_dim),
+        )
+
+    def forward(self, image_embeds):
+        return self.proj(image_embeds)
+
+
+class WanVideoDiT(nn.Module):
+    def __init__(
+        self,
+        model_type: str = "i2v",
+        patch_size: Tuple[int, int, int] = (1, 2, 2),
+        text_len: int = 512,
+        in_dim: int = 36,
+        dim: Optional[int] = None,
+        hidden_dim: Optional[int] = None,
+        ffn_dim: int = 13824,
+        freq_dim: int = 256,
+        text_dim: int = 4096,
+        out_dim: int = 16,
+        num_heads: int = 40,
+        attn_head_dim: Optional[int] = None,
+        num_layers: int = 40,
+        qk_norm: bool = True,
+        cross_attn_norm: bool = True,
+        eps: float = 1e-6,
+        use_gradient_checkpointing: bool = False,
+        video_attention_mask_mode: str = "first_frame_causal",
+        condition_channels: int = 20,
+        **_: Any,
+    ):
+        super().__init__()
+        if model_type != "i2v":
+            raise ValueError(
+                f"Wan2.1-14B VPP2 supports only model_type='i2v', got {model_type!r}."
+            )
+        hidden_dim = int(
+            hidden_dim if hidden_dim is not None else (dim if dim is not None else 5120)
+        )
+        if attn_head_dim is None:
+            if hidden_dim % num_heads != 0:
+                raise ValueError("`hidden_dim` must be divisible by `num_heads`.")
+            attn_head_dim = hidden_dim // num_heads
+
+        self.model_type = model_type
+        self.patch_size = tuple(patch_size)
+        self.text_len = int(text_len)
+        self.in_dim = int(in_dim)
+        self.dim = hidden_dim
+        self.hidden_dim = hidden_dim
+        self.ffn_dim = int(ffn_dim)
+        self.freq_dim = int(freq_dim)
+        self.text_dim = int(text_dim)
+        self.out_dim = int(out_dim)
+        self.num_heads = int(num_heads)
+        self.attn_head_dim = int(attn_head_dim)
+        self.num_layers = int(num_layers)
+        self.video_attention_mask_mode = str(video_attention_mask_mode)
+        self.condition_channels = int(condition_channels)
+        self.fuse_vae_embedding_in_latents = False
+
+        self.patch_embedding = nn.Conv3d(
+            self.in_dim, hidden_dim, kernel_size=self.patch_size, stride=self.patch_size
+        )
+        self.text_embedding = nn.Sequential(
+            nn.Linear(text_dim, hidden_dim),
+            nn.GELU(approximate="tanh"),
+            nn.Linear(hidden_dim, hidden_dim),
+        )
+        self.time_embedding = nn.Sequential(
+            nn.Linear(freq_dim, hidden_dim), nn.SiLU(), nn.Linear(hidden_dim, hidden_dim)
+        )
+        self.time_projection = nn.Sequential(nn.SiLU(), nn.Linear(hidden_dim, hidden_dim * 6))
+        self.blocks = nn.ModuleList(
+            [
+                DiTBlock(
+                    hidden_dim,
+                    self.attn_head_dim,
+                    self.num_heads,
+                    self.ffn_dim,
+                    eps,
+                    cross_attn_type="i2v",
+                    qk_norm=qk_norm,
+                    cross_attn_norm=cross_attn_norm,
+                )
+                for _ in range(self.num_layers)
+            ]
+        )
+        self.head = Head(hidden_dim, self.out_dim, self.patch_size, eps)
+        self.img_emb = MLPProj(1280, hidden_dim)
+        self.freqs = precompute_freqs_cis_3d(self.attn_head_dim)
+        self.use_gradient_checkpointing = bool(use_gradient_checkpointing)
+
+    def patchify(self, x: torch.Tensor):
+        return self.patch_embedding(x)
+
+    def unpatchify(self, x: torch.Tensor, grid_size):
+        f, h, w = grid_size
+        return rearrange(
+            x,
+            "b (f h w) (x y z c) -> b c (f x) (h y) (w z)",
+            f=f,
+            h=h,
+            w=w,
+            x=self.patch_size[0],
+            y=self.patch_size[1],
+            z=self.patch_size[2],
+        )
+
+    def _build_condition(self, x: torch.Tensor, condition_latents: Optional[torch.Tensor]):
+        if condition_latents is None:
+            condition_latents = x[:, :, :1]
+        if condition_latents.ndim != 5:
+            raise ValueError(
+                f"`condition_latents` must be 5D [B,C,T,H,W], got {tuple(condition_latents.shape)}"
+            )
+        b, _, lat_t, lat_h, lat_w = x.shape
+        expected = (b, self.out_dim, lat_h, lat_w)
+        actual = (
+            condition_latents.shape[0],
+            condition_latents.shape[1],
+            condition_latents.shape[3],
+            condition_latents.shape[4],
+        )
+        if actual != expected:
+            raise ValueError(
+                "`condition_latents` batch/channel/spatial shape mismatch: "
+                f"got {tuple(condition_latents.shape)}, expected "
+                f"[B={b},C={self.out_dim},T<={lat_t},H={lat_h},W={lat_w}]."
+            )
+        if condition_latents.shape[2] > lat_t:
+            raise ValueError(
+                "`condition_latents` cannot be longer than the noisy video latent "
+                f"sequence, got {condition_latents.shape[2]} > {lat_t}."
+            )
+        mask = torch.zeros((b, 4, lat_t, lat_h, lat_w), dtype=x.dtype, device=x.device)
+        mask[:, :, : condition_latents.shape[2]] = 1
+        y = torch.zeros((b, self.out_dim, lat_t, lat_h, lat_w), dtype=x.dtype, device=x.device)
+        y[:, :, : condition_latents.shape[2]] = condition_latents.to(device=x.device, dtype=x.dtype)
+        return torch.cat([mask, y], dim=1)
+
+    def build_video_to_video_mask(
+        self, video_seq_len: int, video_tokens_per_frame: int, device: torch.device
+    ):
+        if self.video_attention_mask_mode in {"bidirectional", "three_frame"}:
+            return torch.ones((video_seq_len, video_seq_len), dtype=torch.bool, device=device)
+        if self.video_attention_mask_mode == "first_frame_causal":
+            mask = torch.ones((video_seq_len, video_seq_len), dtype=torch.bool, device=device)
+            first_frame_tokens = min(video_tokens_per_frame, video_seq_len)
+            mask[:first_frame_tokens, first_frame_tokens:] = False
+            return mask
+        if self.video_attention_mask_mode == "per_frame_causal":
+            if video_seq_len % video_tokens_per_frame != 0:
+                raise ValueError("`video_seq_len` must be divisible by `video_tokens_per_frame`.")
+            frames = video_seq_len // video_tokens_per_frame
+            mask = torch.tril(torch.ones((frames, frames), dtype=torch.bool, device=device))
+            return mask.repeat_interleave(video_tokens_per_frame, 0).repeat_interleave(
+                video_tokens_per_frame, 1
+            )
+        raise ValueError(f"Unsupported video attention mask mode: {self.video_attention_mask_mode}")
+
+    def pre_dit(
+        self,
+        x: torch.Tensor,
+        timestep: torch.Tensor,
+        context: torch.Tensor,
+        context_mask: Optional[torch.Tensor] = None,
+        clip_fea: Optional[torch.Tensor] = None,
+        condition_latents: Optional[torch.Tensor] = None,
+        **_: Any,
+    ) -> Dict[str, Any]:
+        if x.ndim != 5:
+            raise ValueError(f"`x` must be 5D [B,C,T,H,W], got {tuple(x.shape)}")
+        if x.shape[1] == self.out_dim:
+            y = self._build_condition(x, condition_latents)
+            x = torch.cat([x, y], dim=1)
+        if x.shape[1] != self.in_dim:
+            raise ValueError(
+                f"Wan2.1-14B DiT expected {self.in_dim} channels after conditioning, got {x.shape[1]}."
+            )
+        if context.ndim != 3:
+            raise ValueError(f"`context` must be 3D [B,L,D], got {tuple(context.shape)}")
+        batch_size = x.shape[0]
+        if timestep.ndim != 1 or timestep.shape[0] not in (1, batch_size):
+            raise ValueError(f"`timestep` must have shape [B] or [1], got {tuple(timestep.shape)}")
+        if timestep.shape[0] == 1 and batch_size > 1:
+            if self.training:
+                raise ValueError("During training, timestep length must match batch size.")
+            timestep = timestep.expand(batch_size)
+        if context_mask is None:
+            context_mask = torch.ones(
+                (batch_size, context.shape[1]), dtype=torch.bool, device=context.device
+            )
+        if clip_fea is None:
+            clip_fea = torch.zeros(
+                (batch_size, 257, 1280), dtype=context.dtype, device=context.device
+            )
+        if clip_fea.ndim != 3 or clip_fea.shape[1:] != (257, 1280):
+            raise ValueError(f"`clip_fea` must be [B,257,1280], got {tuple(clip_fea.shape)}")
+
+        x = self.patchify(x)
+        f, h, w = x.shape[2:]
+        tokens_per_frame = h * w
+        seq_len = f * h * w
+
+        time_weight = self.time_embedding[0].weight
+        time_input = sinusoidal_embedding_1d(self.freq_dim, timestep).to(
+            device=time_weight.device,
+            dtype=time_weight.dtype,
+        )
+        t = self.time_embedding(time_input)
+        t_mod = self.time_projection(t).unflatten(1, (6, self.hidden_dim))
+
+        text_context = self.text_embedding(context)
+        image_context = self.img_emb(
+            clip_fea.to(device=text_context.device, dtype=text_context.dtype)
+        )
+        context_emb = torch.cat([image_context, text_context], dim=1)
+        image_mask = torch.ones(
+            (batch_size, image_context.shape[1]), dtype=torch.bool, device=context_mask.device
+        )
+        combined_mask = torch.cat(
+            [image_mask, context_mask.to(dtype=torch.bool, device=context_mask.device)], dim=1
+        )
+        context_attn_mask = combined_mask.unsqueeze(1).expand(-1, seq_len, -1)
+
+        x_tokens = rearrange(x, "b c f h w -> b (f h w) c").contiguous()
+        freqs = (
+            torch.cat(
+                [
+                    self.freqs[0][:f].view(f, 1, 1, -1).expand(f, h, w, -1),
+                    self.freqs[1][:h].view(1, h, 1, -1).expand(f, h, w, -1),
+                    self.freqs[2][:w].view(1, 1, w, -1).expand(f, h, w, -1),
+                ],
+                dim=-1,
+            )
+            .reshape(seq_len, 1, -1)
+            .to(x_tokens.device)
+        )
+
+        return {
+            "tokens": x_tokens,
+            "freqs": freqs,
+            "t": t,
+            "t_mod": t_mod,
+            "context": context_emb,
+            "context_mask": context_attn_mask,
+            "meta": {
+                "grid_size": (f, h, w),
+                "tokens_per_frame": tokens_per_frame,
+                "batch_size": batch_size,
+            },
+        }
+
+    def post_dit(self, x_tokens: torch.Tensor, pre_state: Dict[str, Any]):
+        f, h, w = pre_state["meta"]["grid_size"]
+        x = self.head(x_tokens, pre_state["t"])
+        return self.unpatchify(x, (f, h, w))
+
+    def forward(
+        self,
+        x,
+        timestep,
+        context,
+        context_mask=None,
+        clip_fea=None,
+        condition_latents=None,
+        **kwargs,
+    ):
+        pre = self.pre_dit(
+            x=x,
+            timestep=timestep,
+            context=context,
+            context_mask=context_mask,
+            clip_fea=clip_fea,
+            condition_latents=condition_latents,
+            **kwargs,
+        )
+        tokens = pre["tokens"]
+        self_attn_mask = None
+        if self.video_attention_mask_mode not in {"bidirectional", "three_frame"}:
+            self_attn_mask = self.build_video_to_video_mask(
+                tokens.shape[1],
+                int(pre["meta"]["tokens_per_frame"]),
+                tokens.device,
+            )
+        for block in self.blocks:
+            if self.use_gradient_checkpointing:
+                tokens = gradient_checkpoint_forward(
+                    block,
+                    True,
+                    tokens,
+                    pre["context"],
+                    pre["t_mod"],
+                    pre["freqs"],
+                    context_mask=pre["context_mask"],
+                    self_attn_mask=self_attn_mask,
+                )
+            else:
+                tokens = block(
+                    tokens,
+                    pre["context"],
+                    pre["t_mod"],
+                    pre["freqs"],
+                    context_mask=pre["context_mask"],
+                    self_attn_mask=self_attn_mask,
+                )
+        return self.post_dit(tokens, pre)
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_text_encoder.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_text_encoder.py
new file mode 100644
index 0000000000000000000000000000000000000000..6b4de7a919bdcaac13e90787551c5e530c9edf87
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_text_encoder.py
@@ -0,0 +1,341 @@
+import math
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from transformers import AutoTokenizer
+import html
+import string
+
+try:
+    import ftfy
+except ModuleNotFoundError:
+
+    class _FTFY:
+        @staticmethod
+        def fix_text(text):
+            return text
+
+    ftfy = _FTFY()
+try:
+    import regex as re
+except ModuleNotFoundError:
+    import re
+
+
+def fp16_clamp(x):
+    if x.dtype == torch.float16 and torch.isinf(x).any():
+        clamp = torch.finfo(x.dtype).max - 1000
+        x = torch.clamp(x, min=-clamp, max=clamp)
+    return x
+
+
+class GELU(nn.Module):
+    def forward(self, x):
+        return (
+            0.5
+            * x
+            * (1.0 + torch.tanh(math.sqrt(2.0 / math.pi) * (x + 0.044715 * torch.pow(x, 3.0))))
+        )
+
+
+class T5LayerNorm(nn.Module):
+    def __init__(self, dim, eps=1e-6):
+        super(T5LayerNorm, self).__init__()
+        self.dim = dim
+        self.eps = eps
+        self.weight = nn.Parameter(torch.ones(dim))
+
+    def forward(self, x):
+        x = x * torch.rsqrt(x.float().pow(2).mean(dim=-1, keepdim=True) + self.eps)
+        if self.weight.dtype in [torch.float16, torch.bfloat16]:
+            x = x.type_as(self.weight)
+        return self.weight * x
+
+
+class T5Attention(nn.Module):
+    def __init__(self, dim, dim_attn, num_heads, dropout=0.1):
+        assert dim_attn % num_heads == 0
+        super(T5Attention, self).__init__()
+        self.dim = dim
+        self.dim_attn = dim_attn
+        self.num_heads = num_heads
+        self.head_dim = dim_attn // num_heads
+
+        # layers
+        self.q = nn.Linear(dim, dim_attn, bias=False)
+        self.k = nn.Linear(dim, dim_attn, bias=False)
+        self.v = nn.Linear(dim, dim_attn, bias=False)
+        self.o = nn.Linear(dim_attn, dim, bias=False)
+        self.dropout = nn.Dropout(dropout)
+
+    def forward(self, x, context=None, mask=None, pos_bias=None):
+        """
+        x:          [B, L1, C].
+        context:    [B, L2, C] or None.
+        mask:       [B, L2] or [B, L1, L2] or None.
+        """
+        # check inputs
+        context = x if context is None else context
+        b, n, c = x.size(0), self.num_heads, self.head_dim
+
+        # compute query, key, value
+        q = self.q(x).view(b, -1, n, c)
+        k = self.k(context).view(b, -1, n, c)
+        v = self.v(context).view(b, -1, n, c)
+
+        # attention bias
+        attn_bias = x.new_zeros(b, n, q.size(1), k.size(1))
+        if pos_bias is not None:
+            attn_bias += pos_bias
+        if mask is not None:
+            assert mask.ndim in [2, 3]
+            mask = mask.view(b, 1, 1, -1) if mask.ndim == 2 else mask.unsqueeze(1)
+            attn_bias.masked_fill_(mask == 0, torch.finfo(x.dtype).min)
+
+        # compute attention (T5 does not use scaling)
+        attn = torch.einsum("binc,bjnc->bnij", q, k) + attn_bias
+        attn = F.softmax(attn.float(), dim=-1).type_as(attn)
+        x = torch.einsum("bnij,bjnc->binc", attn, v)
+
+        # output
+        x = x.reshape(b, -1, n * c)
+        x = self.o(x)
+        x = self.dropout(x)
+        return x
+
+
+class T5FeedForward(nn.Module):
+    def __init__(self, dim, dim_ffn, dropout=0.1):
+        super(T5FeedForward, self).__init__()
+        self.dim = dim
+        self.dim_ffn = dim_ffn
+
+        # layers
+        self.gate = nn.Sequential(nn.Linear(dim, dim_ffn, bias=False), GELU())
+        self.fc1 = nn.Linear(dim, dim_ffn, bias=False)
+        self.fc2 = nn.Linear(dim_ffn, dim, bias=False)
+        self.dropout = nn.Dropout(dropout)
+
+    def forward(self, x):
+        x = self.fc1(x) * self.gate(x)
+        x = self.dropout(x)
+        x = self.fc2(x)
+        x = self.dropout(x)
+        return x
+
+
+class T5SelfAttention(nn.Module):
+    def __init__(
+        self, dim, dim_attn, dim_ffn, num_heads, num_buckets, shared_pos=True, dropout=0.1
+    ):
+        super(T5SelfAttention, self).__init__()
+        self.dim = dim
+        self.dim_attn = dim_attn
+        self.dim_ffn = dim_ffn
+        self.num_heads = num_heads
+        self.num_buckets = num_buckets
+        self.shared_pos = shared_pos
+
+        # layers
+        self.norm1 = T5LayerNorm(dim)
+        self.attn = T5Attention(dim, dim_attn, num_heads, dropout)
+        self.norm2 = T5LayerNorm(dim)
+        self.ffn = T5FeedForward(dim, dim_ffn, dropout)
+        self.pos_embedding = (
+            None if shared_pos else T5RelativeEmbedding(num_buckets, num_heads, bidirectional=True)
+        )
+
+    def forward(self, x, mask=None, pos_bias=None):
+        e = pos_bias if self.shared_pos else self.pos_embedding(x.size(1), x.size(1))
+        x = fp16_clamp(x + self.attn(self.norm1(x), mask=mask, pos_bias=e))
+        x = fp16_clamp(x + self.ffn(self.norm2(x)))
+        return x
+
+
+class T5RelativeEmbedding(nn.Module):
+    def __init__(self, num_buckets, num_heads, bidirectional, max_dist=128):
+        super(T5RelativeEmbedding, self).__init__()
+        self.num_buckets = num_buckets
+        self.num_heads = num_heads
+        self.bidirectional = bidirectional
+        self.max_dist = max_dist
+
+        # layers
+        self.embedding = nn.Embedding(num_buckets, num_heads)
+
+    def forward(self, lq, lk):
+        device = self.embedding.weight.device
+        # rel_pos = torch.arange(lk).unsqueeze(0).to(device) - \
+        #     torch.arange(lq).unsqueeze(1).to(device)
+        rel_pos = torch.arange(lk, device=device).unsqueeze(0) - torch.arange(
+            lq, device=device
+        ).unsqueeze(1)
+        rel_pos = self._relative_position_bucket(rel_pos)
+        rel_pos_embeds = self.embedding(rel_pos)
+        rel_pos_embeds = rel_pos_embeds.permute(2, 0, 1).unsqueeze(0)  # [1, N, Lq, Lk]
+        return rel_pos_embeds.contiguous()
+
+    def _relative_position_bucket(self, rel_pos):
+        # preprocess
+        if self.bidirectional:
+            num_buckets = self.num_buckets // 2
+            rel_buckets = (rel_pos > 0).long() * num_buckets
+            rel_pos = torch.abs(rel_pos)
+        else:
+            num_buckets = self.num_buckets
+            rel_buckets = 0
+            rel_pos = -torch.min(rel_pos, torch.zeros_like(rel_pos))
+
+        # embeddings for small and large positions
+        max_exact = num_buckets // 2
+        rel_pos_large = (
+            max_exact
+            + (
+                torch.log(rel_pos.float() / max_exact)
+                / math.log(self.max_dist / max_exact)
+                * (num_buckets - max_exact)
+            ).long()
+        )
+        rel_pos_large = torch.min(rel_pos_large, torch.full_like(rel_pos_large, num_buckets - 1))
+        rel_buckets += torch.where(rel_pos < max_exact, rel_pos, rel_pos_large)
+        return rel_buckets
+
+
+def init_weights(m):
+    if isinstance(m, T5LayerNorm):
+        nn.init.ones_(m.weight)
+    elif isinstance(m, T5FeedForward):
+        nn.init.normal_(m.gate[0].weight, std=m.dim**-0.5)
+        nn.init.normal_(m.fc1.weight, std=m.dim**-0.5)
+        nn.init.normal_(m.fc2.weight, std=m.dim_ffn**-0.5)
+    elif isinstance(m, T5Attention):
+        nn.init.normal_(m.q.weight, std=(m.dim * m.dim_attn) ** -0.5)
+        nn.init.normal_(m.k.weight, std=m.dim**-0.5)
+        nn.init.normal_(m.v.weight, std=m.dim**-0.5)
+        nn.init.normal_(m.o.weight, std=(m.num_heads * m.dim_attn) ** -0.5)
+    elif isinstance(m, T5RelativeEmbedding):
+        nn.init.normal_(m.embedding.weight, std=(2 * m.num_buckets * m.num_heads) ** -0.5)
+
+
+class WanTextEncoder(torch.nn.Module):
+    def __init__(
+        self,
+        vocab=256384,
+        dim=4096,
+        dim_attn=4096,
+        dim_ffn=10240,
+        num_heads=64,
+        num_layers=24,
+        num_buckets=32,
+        shared_pos=False,
+        dropout=0.1,
+    ):
+        super(WanTextEncoder, self).__init__()
+        self.dim = dim
+        self.dim_attn = dim_attn
+        self.dim_ffn = dim_ffn
+        self.num_heads = num_heads
+        self.num_layers = num_layers
+        self.num_buckets = num_buckets
+        self.shared_pos = shared_pos
+
+        # layers
+        self.token_embedding = (
+            vocab if isinstance(vocab, nn.Embedding) else nn.Embedding(vocab, dim)
+        )
+        self.pos_embedding = (
+            T5RelativeEmbedding(num_buckets, num_heads, bidirectional=True) if shared_pos else None
+        )
+        self.dropout = nn.Dropout(dropout)
+        self.blocks = nn.ModuleList(
+            [
+                T5SelfAttention(dim, dim_attn, dim_ffn, num_heads, num_buckets, shared_pos, dropout)
+                for _ in range(num_layers)
+            ]
+        )
+        self.norm = T5LayerNorm(dim)
+
+        # Skip costly random init when loading pretrained checkpoints.
+        # Verified: checkpoint fully covers this module (missing/unexpected keys are both 0).
+        # self.apply(init_weights)
+
+    def forward(self, ids, mask=None):
+        x = self.token_embedding(ids)
+        x = self.dropout(x)
+        e = self.pos_embedding(x.size(1), x.size(1)) if self.shared_pos else None
+        for block in self.blocks:
+            x = block(x, mask, pos_bias=e)
+        x = self.norm(x)
+        x = self.dropout(x)
+        return x
+
+
+def basic_clean(text):
+    text = ftfy.fix_text(text)
+    text = html.unescape(html.unescape(text))
+    return text.strip()
+
+
+def whitespace_clean(text):
+    text = re.sub(r"\s+", " ", text)
+    text = text.strip()
+    return text
+
+
+def canonicalize(text, keep_punctuation_exact_string=None):
+    text = text.replace("_", " ")
+    if keep_punctuation_exact_string:
+        text = keep_punctuation_exact_string.join(
+            part.translate(str.maketrans("", "", string.punctuation))
+            for part in text.split(keep_punctuation_exact_string)
+        )
+    else:
+        text = text.translate(str.maketrans("", "", string.punctuation))
+    text = text.lower()
+    text = re.sub(r"\s+", " ", text)
+    return text.strip()
+
+
+class HuggingfaceTokenizer:
+    def __init__(self, name, seq_len=None, clean=None, **kwargs):
+        assert clean in (None, "whitespace", "lower", "canonicalize")
+        self.name = name
+        self.seq_len = seq_len
+        self.clean = clean
+
+        # init tokenizer
+        self.tokenizer = AutoTokenizer.from_pretrained(name, **kwargs)
+        self.vocab_size = self.tokenizer.vocab_size
+
+    def __call__(self, sequence, **kwargs):
+        return_mask = kwargs.pop("return_mask", False)
+
+        # arguments
+        _kwargs = {"return_tensors": "pt"}
+        if self.seq_len is not None:
+            _kwargs.update(
+                {"padding": "max_length", "truncation": True, "max_length": self.seq_len}
+            )
+        _kwargs.update(**kwargs)
+
+        # tokenization
+        if isinstance(sequence, str):
+            sequence = [sequence]
+        if self.clean:
+            sequence = [self._clean(u) for u in sequence]
+        ids = self.tokenizer(sequence, **_kwargs)
+
+        # output
+        if return_mask:
+            return ids.input_ids, ids.attention_mask
+        else:
+            return ids.input_ids
+
+    def _clean(self, text):
+        if self.clean == "whitespace":
+            text = whitespace_clean(basic_clean(text))
+        elif self.clean == "lower":
+            text = whitespace_clean(basic_clean(text)).lower()
+        elif self.clean == "canonicalize":
+            text = canonicalize(basic_clean(text))
+        return text
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_vae.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_vae.py
new file mode 100644
index 0000000000000000000000000000000000000000..cae08a02ff7d8cb925c5c864cb8e4822d9860b02
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/models/wan21_14b/wan_video_vae.py
@@ -0,0 +1,863 @@
+from einops import rearrange, repeat
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from tqdm import tqdm
+
+CACHE_T = 2
+
+
+def check_is_instance(model, module_class):
+    if isinstance(model, module_class):
+        return True
+    if hasattr(model, "module") and isinstance(model.module, module_class):
+        return True
+    return False
+
+
+class CausalConv3d(nn.Conv3d):
+    """
+    Causal 3d convolusion.
+    """
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._padding = (
+            self.padding[2],
+            self.padding[2],
+            self.padding[1],
+            self.padding[1],
+            2 * self.padding[0],
+            0,
+        )
+        self.padding = (0, 0, 0)
+
+    def forward(self, x, cache_x=None):
+        padding = list(self._padding)
+        if cache_x is not None and self._padding[4] > 0:
+            cache_x = cache_x.to(x.device)
+            x = torch.cat([cache_x, x], dim=2)
+            padding[4] -= cache_x.shape[2]
+        x = F.pad(x, padding)
+
+        return super().forward(x)
+
+
+class RMS_norm(nn.Module):
+    def __init__(self, dim, channel_first=True, images=True, bias=False):
+        super().__init__()
+        broadcastable_dims = (1, 1, 1) if not images else (1, 1)
+        shape = (dim, *broadcastable_dims) if channel_first else (dim,)
+
+        self.channel_first = channel_first
+        self.scale = dim**0.5
+        self.gamma = nn.Parameter(torch.ones(shape))
+        self.bias = nn.Parameter(torch.zeros(shape)) if bias else 0.0
+
+    def forward(self, x):
+        return (
+            F.normalize(x, dim=(1 if self.channel_first else -1)) * self.scale * self.gamma
+            + self.bias
+        )
+
+
+class Upsample(nn.Upsample):
+    def forward(self, x):
+        """
+        Fix bfloat16 support for nearest neighbor interpolation.
+        """
+        return super().forward(x.float()).type_as(x)
+
+
+class Resample(nn.Module):
+    def __init__(self, dim, mode):
+        assert mode in ("none", "upsample2d", "upsample3d", "downsample2d", "downsample3d")
+        super().__init__()
+        self.dim = dim
+        self.mode = mode
+
+        # layers
+        if mode == "upsample2d":
+            self.resample = nn.Sequential(
+                Upsample(scale_factor=(2.0, 2.0), mode="nearest-exact"),
+                nn.Conv2d(dim, dim // 2, 3, padding=1),
+            )
+        elif mode == "upsample3d":
+            self.resample = nn.Sequential(
+                Upsample(scale_factor=(2.0, 2.0), mode="nearest-exact"),
+                nn.Conv2d(dim, dim // 2, 3, padding=1),
+            )
+            self.time_conv = CausalConv3d(dim, dim * 2, (3, 1, 1), padding=(1, 0, 0))
+
+        elif mode == "downsample2d":
+            self.resample = nn.Sequential(
+                nn.ZeroPad2d((0, 1, 0, 1)), nn.Conv2d(dim, dim, 3, stride=(2, 2))
+            )
+        elif mode == "downsample3d":
+            self.resample = nn.Sequential(
+                nn.ZeroPad2d((0, 1, 0, 1)), nn.Conv2d(dim, dim, 3, stride=(2, 2))
+            )
+            self.time_conv = CausalConv3d(dim, dim, (3, 1, 1), stride=(2, 1, 1), padding=(0, 0, 0))
+
+        else:
+            self.resample = nn.Identity()
+
+    def forward(self, x, feat_cache=None, feat_idx=[0]):
+        b, c, t, h, w = x.size()
+        if self.mode == "upsample3d":
+            if feat_cache is not None:
+                idx = feat_idx[0]
+                if feat_cache[idx] is None:
+                    feat_cache[idx] = "Rep"
+                    feat_idx[0] += 1
+                else:
+                    cache_x = x[:, :, -CACHE_T:, :, :].clone()
+                    if (
+                        cache_x.shape[2] < 2
+                        and feat_cache[idx] is not None
+                        and feat_cache[idx] != "Rep"
+                    ):
+                        # cache last frame of last two chunk
+                        cache_x = torch.cat(
+                            [
+                                feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device),
+                                cache_x,
+                            ],
+                            dim=2,
+                        )
+                    if (
+                        cache_x.shape[2] < 2
+                        and feat_cache[idx] is not None
+                        and feat_cache[idx] == "Rep"
+                    ):
+                        cache_x = torch.cat(
+                            [torch.zeros_like(cache_x).to(cache_x.device), cache_x], dim=2
+                        )
+                    if feat_cache[idx] == "Rep":
+                        x = self.time_conv(x)
+                    else:
+                        x = self.time_conv(x, feat_cache[idx])
+                    feat_cache[idx] = cache_x
+                    feat_idx[0] += 1
+
+                    x = x.reshape(b, 2, c, t, h, w)
+                    x = torch.stack((x[:, 0, :, :, :, :], x[:, 1, :, :, :, :]), 3)
+                    x = x.reshape(b, c, t * 2, h, w)
+        t = x.shape[2]
+        x = rearrange(x, "b c t h w -> (b t) c h w")
+        x = self.resample(x)
+        x = rearrange(x, "(b t) c h w -> b c t h w", t=t)
+
+        if self.mode == "downsample3d":
+            if feat_cache is not None:
+                idx = feat_idx[0]
+                if feat_cache[idx] is None:
+                    feat_cache[idx] = x.clone()
+                    feat_idx[0] += 1
+                else:
+                    cache_x = x[:, :, -1:, :, :].clone()
+                    x = self.time_conv(torch.cat([feat_cache[idx][:, :, -1:, :, :], x], 2))
+                    feat_cache[idx] = cache_x
+                    feat_idx[0] += 1
+        return x, feat_cache, feat_idx
+
+    def init_weight(self, conv):
+        conv_weight = conv.weight
+        nn.init.zeros_(conv_weight)
+        c1, c2, t, h, w = conv_weight.size()
+        one_matrix = torch.eye(c1, c2)
+        init_matrix = one_matrix
+        nn.init.zeros_(conv_weight)
+        conv_weight.data[:, :, 1, 0, 0] = init_matrix
+        conv.weight.data.copy_(conv_weight)
+        nn.init.zeros_(conv.bias.data)
+
+    def init_weight2(self, conv):
+        conv_weight = conv.weight.data
+        nn.init.zeros_(conv_weight)
+        c1, c2, t, h, w = conv_weight.size()
+        init_matrix = torch.eye(c1 // 2, c2)
+        conv_weight[: c1 // 2, :, -1, 0, 0] = init_matrix
+        conv_weight[c1 // 2 :, :, -1, 0, 0] = init_matrix
+        conv.weight.data.copy_(conv_weight)
+        nn.init.zeros_(conv.bias.data)
+
+
+class ResidualBlock(nn.Module):
+    def __init__(self, in_dim, out_dim, dropout=0.0):
+        super().__init__()
+        self.in_dim = in_dim
+        self.out_dim = out_dim
+
+        # layers
+        self.residual = nn.Sequential(
+            RMS_norm(in_dim, images=False),
+            nn.SiLU(),
+            CausalConv3d(in_dim, out_dim, 3, padding=1),
+            RMS_norm(out_dim, images=False),
+            nn.SiLU(),
+            nn.Dropout(dropout),
+            CausalConv3d(out_dim, out_dim, 3, padding=1),
+        )
+        self.shortcut = CausalConv3d(in_dim, out_dim, 1) if in_dim != out_dim else nn.Identity()
+
+    def forward(self, x, feat_cache=None, feat_idx=[0]):
+        h = self.shortcut(x)
+        for layer in self.residual:
+            if check_is_instance(layer, CausalConv3d) and feat_cache is not None:
+                idx = feat_idx[0]
+                cache_x = x[:, :, -CACHE_T:, :, :].clone()
+                if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                    # cache last frame of last two chunk
+                    cache_x = torch.cat(
+                        [feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x],
+                        dim=2,
+                    )
+                x = layer(x, feat_cache[idx])
+                feat_cache[idx] = cache_x
+                feat_idx[0] += 1
+            else:
+                x = layer(x)
+        return x + h, feat_cache, feat_idx
+
+
+class AttentionBlock(nn.Module):
+    """
+    Causal self-attention with a single head.
+    """
+
+    def __init__(self, dim):
+        super().__init__()
+        self.dim = dim
+
+        # layers
+        self.norm = RMS_norm(dim)
+        self.to_qkv = nn.Conv2d(dim, dim * 3, 1)
+        self.proj = nn.Conv2d(dim, dim, 1)
+
+        # zero out the last layer params
+        nn.init.zeros_(self.proj.weight)
+
+    def forward(self, x):
+        identity = x
+        b, c, t, h, w = x.size()
+        x = rearrange(x, "b c t h w -> (b t) c h w")
+        x = self.norm(x)
+        # compute query, key, value
+        q, k, v = (
+            self.to_qkv(x)
+            .reshape(b * t, 1, c * 3, -1)
+            .permute(0, 1, 3, 2)
+            .contiguous()
+            .chunk(3, dim=-1)
+        )
+
+        # apply attention
+        x = F.scaled_dot_product_attention(
+            q,
+            k,
+            v,
+            # attn_mask=block_causal_mask(q, block_size=h * w)
+        )
+        x = x.squeeze(1).permute(0, 2, 1).reshape(b * t, c, h, w)
+
+        # output
+        x = self.proj(x)
+        x = rearrange(x, "(b t) c h w-> b c t h w", t=t)
+        return x + identity
+
+
+class Encoder3d(nn.Module):
+    def __init__(
+        self,
+        dim=128,
+        z_dim=4,
+        dim_mult=[1, 2, 4, 4],
+        num_res_blocks=2,
+        attn_scales=[],
+        temperal_downsample=[True, True, False],
+        dropout=0.0,
+    ):
+        super().__init__()
+        self.dim = dim
+        self.z_dim = z_dim
+        self.dim_mult = dim_mult
+        self.num_res_blocks = num_res_blocks
+        self.attn_scales = attn_scales
+        self.temperal_downsample = temperal_downsample
+
+        # dimensions
+        dims = [dim * u for u in [1] + dim_mult]
+        scale = 1.0
+
+        # init block
+        self.conv1 = CausalConv3d(3, dims[0], 3, padding=1)
+
+        # downsample blocks
+        downsamples = []
+        for i, (in_dim, out_dim) in enumerate(zip(dims[:-1], dims[1:])):
+            # residual (+attention) blocks
+            for _ in range(num_res_blocks):
+                downsamples.append(ResidualBlock(in_dim, out_dim, dropout))
+                if scale in attn_scales:
+                    downsamples.append(AttentionBlock(out_dim))
+                in_dim = out_dim
+
+            # downsample block
+            if i != len(dim_mult) - 1:
+                mode = "downsample3d" if temperal_downsample[i] else "downsample2d"
+                downsamples.append(Resample(out_dim, mode=mode))
+                scale /= 2.0
+        self.downsamples = nn.Sequential(*downsamples)
+
+        # middle blocks
+        self.middle = nn.Sequential(
+            ResidualBlock(out_dim, out_dim, dropout),
+            AttentionBlock(out_dim),
+            ResidualBlock(out_dim, out_dim, dropout),
+        )
+
+        # output blocks
+        self.head = nn.Sequential(
+            RMS_norm(out_dim, images=False), nn.SiLU(), CausalConv3d(out_dim, z_dim, 3, padding=1)
+        )
+
+    def forward(self, x, feat_cache=None, feat_idx=[0]):
+        if feat_cache is not None:
+            idx = feat_idx[0]
+            cache_x = x[:, :, -CACHE_T:, :, :].clone()
+            if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                # cache last frame of last two chunk
+                cache_x = torch.cat(
+                    [feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x],
+                    dim=2,
+                )
+            x = self.conv1(x, feat_cache[idx])
+            feat_cache[idx] = cache_x
+            feat_idx[0] += 1
+        else:
+            x = self.conv1(x)
+
+        ## downsamples
+        for layer in self.downsamples:
+            if feat_cache is not None:
+                x, feat_cache, feat_idx = layer(x, feat_cache, feat_idx)
+            else:
+                x = layer(x)
+
+        ## middle
+        for layer in self.middle:
+            if check_is_instance(layer, ResidualBlock) and feat_cache is not None:
+                x, feat_cache, feat_idx = layer(x, feat_cache, feat_idx)
+            else:
+                x = layer(x)
+
+        ## head
+        for layer in self.head:
+            if check_is_instance(layer, CausalConv3d) and feat_cache is not None:
+                idx = feat_idx[0]
+                cache_x = x[:, :, -CACHE_T:, :, :].clone()
+                if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                    # cache last frame of last two chunk
+                    cache_x = torch.cat(
+                        [feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x],
+                        dim=2,
+                    )
+                x = layer(x, feat_cache[idx])
+                feat_cache[idx] = cache_x
+                feat_idx[0] += 1
+            else:
+                x = layer(x)
+        return x, feat_cache, feat_idx
+
+
+class Decoder3d(nn.Module):
+    def __init__(
+        self,
+        dim=128,
+        z_dim=4,
+        dim_mult=[1, 2, 4, 4],
+        num_res_blocks=2,
+        attn_scales=[],
+        temperal_upsample=[False, True, True],
+        dropout=0.0,
+    ):
+        super().__init__()
+        self.dim = dim
+        self.z_dim = z_dim
+        self.dim_mult = dim_mult
+        self.num_res_blocks = num_res_blocks
+        self.attn_scales = attn_scales
+        self.temperal_upsample = temperal_upsample
+
+        # dimensions
+        dims = [dim * u for u in [dim_mult[-1]] + dim_mult[::-1]]
+        scale = 1.0 / 2 ** (len(dim_mult) - 2)
+
+        # init block
+        self.conv1 = CausalConv3d(z_dim, dims[0], 3, padding=1)
+
+        # middle blocks
+        self.middle = nn.Sequential(
+            ResidualBlock(dims[0], dims[0], dropout),
+            AttentionBlock(dims[0]),
+            ResidualBlock(dims[0], dims[0], dropout),
+        )
+
+        # upsample blocks
+        upsamples = []
+        for i, (in_dim, out_dim) in enumerate(zip(dims[:-1], dims[1:])):
+            # residual (+attention) blocks
+            if i == 1 or i == 2 or i == 3:
+                in_dim = in_dim // 2
+            for _ in range(num_res_blocks + 1):
+                upsamples.append(ResidualBlock(in_dim, out_dim, dropout))
+                if scale in attn_scales:
+                    upsamples.append(AttentionBlock(out_dim))
+                in_dim = out_dim
+
+            # upsample block
+            if i != len(dim_mult) - 1:
+                mode = "upsample3d" if temperal_upsample[i] else "upsample2d"
+                upsamples.append(Resample(out_dim, mode=mode))
+                scale *= 2.0
+        self.upsamples = nn.Sequential(*upsamples)
+
+        # output blocks
+        self.head = nn.Sequential(
+            RMS_norm(out_dim, images=False), nn.SiLU(), CausalConv3d(out_dim, 3, 3, padding=1)
+        )
+
+    def forward(self, x, feat_cache=None, feat_idx=[0]):
+        ## conv1
+        if feat_cache is not None:
+            idx = feat_idx[0]
+            cache_x = x[:, :, -CACHE_T:, :, :].clone()
+            if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                # cache last frame of last two chunk
+                cache_x = torch.cat(
+                    [feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x],
+                    dim=2,
+                )
+            x = self.conv1(x, feat_cache[idx])
+            feat_cache[idx] = cache_x
+            feat_idx[0] += 1
+        else:
+            x = self.conv1(x)
+
+        ## middle
+        for layer in self.middle:
+            if check_is_instance(layer, ResidualBlock) and feat_cache is not None:
+                x, feat_cache, feat_idx = layer(x, feat_cache, feat_idx)
+            else:
+                x = layer(x)
+
+        ## upsamples
+        for layer in self.upsamples:
+            if feat_cache is not None:
+                x, feat_cache, feat_idx = layer(x, feat_cache, feat_idx)
+            else:
+                x = layer(x)
+
+        ## head
+        for layer in self.head:
+            if check_is_instance(layer, CausalConv3d) and feat_cache is not None:
+                idx = feat_idx[0]
+                cache_x = x[:, :, -CACHE_T:, :, :].clone()
+                if cache_x.shape[2] < 2 and feat_cache[idx] is not None:
+                    # cache last frame of last two chunk
+                    cache_x = torch.cat(
+                        [feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x],
+                        dim=2,
+                    )
+                x = layer(x, feat_cache[idx])
+                feat_cache[idx] = cache_x
+                feat_idx[0] += 1
+            else:
+                x = layer(x)
+        return x, feat_cache, feat_idx
+
+
+def count_conv3d(model):
+    count = 0
+    for m in model.modules():
+        if isinstance(m, CausalConv3d):
+            count += 1
+    return count
+
+
+class VideoVAE_(nn.Module):
+    def __init__(
+        self,
+        dim=96,
+        z_dim=16,
+        dim_mult=[1, 2, 4, 4],
+        num_res_blocks=2,
+        attn_scales=[],
+        temperal_downsample=[False, True, True],
+        dropout=0.0,
+    ):
+        super().__init__()
+        self.dim = dim
+        self.z_dim = z_dim
+        self.dim_mult = dim_mult
+        self.num_res_blocks = num_res_blocks
+        self.attn_scales = attn_scales
+        self.temperal_downsample = temperal_downsample
+        self.temperal_upsample = temperal_downsample[::-1]
+
+        # modules
+        self.encoder = Encoder3d(
+            dim, z_dim * 2, dim_mult, num_res_blocks, attn_scales, self.temperal_downsample, dropout
+        )
+        self.conv1 = CausalConv3d(z_dim * 2, z_dim * 2, 1)
+        self.conv2 = CausalConv3d(z_dim, z_dim, 1)
+        self.decoder = Decoder3d(
+            dim, z_dim, dim_mult, num_res_blocks, attn_scales, self.temperal_upsample, dropout
+        )
+
+    def forward(self, x):
+        mu, log_var = self.encode(x)
+        z = self.reparameterize(mu, log_var)
+        x_recon = self.decode(z)
+        return x_recon, mu, log_var
+
+    def encode(self, x, scale):
+        self.clear_cache()
+        ## cache
+        t = x.shape[2]
+        iter_ = 1 + (t - 1) // 4
+
+        for i in range(iter_):
+            self._enc_conv_idx = [0]
+            if i == 0:
+                out, self._enc_feat_map, self._enc_conv_idx = self.encoder(
+                    x[:, :, :1, :, :], feat_cache=self._enc_feat_map, feat_idx=self._enc_conv_idx
+                )
+            else:
+                out_, self._enc_feat_map, self._enc_conv_idx = self.encoder(
+                    x[:, :, 1 + 4 * (i - 1) : 1 + 4 * i, :, :],
+                    feat_cache=self._enc_feat_map,
+                    feat_idx=self._enc_conv_idx,
+                )
+                out = torch.cat([out, out_], 2)
+        mu, log_var = self.conv1(out).chunk(2, dim=1)
+        if isinstance(scale[0], torch.Tensor):
+            scale = [s.to(dtype=mu.dtype, device=mu.device) for s in scale]
+            mu = (mu - scale[0].view(1, self.z_dim, 1, 1, 1)) * scale[1].view(
+                1, self.z_dim, 1, 1, 1
+            )
+        else:
+            scale = scale.to(dtype=mu.dtype, device=mu.device)
+            mu = (mu - scale[0]) * scale[1]
+        return mu
+
+    def decode(self, z, scale):
+        self.clear_cache()
+        # z: [b,c,t,h,w]
+        if isinstance(scale[0], torch.Tensor):
+            scale = [s.to(dtype=z.dtype, device=z.device) for s in scale]
+            z = z / scale[1].view(1, self.z_dim, 1, 1, 1) + scale[0].view(1, self.z_dim, 1, 1, 1)
+        else:
+            scale = scale.to(dtype=z.dtype, device=z.device)
+            z = z / scale[1] + scale[0]
+        iter_ = z.shape[2]
+        x = self.conv2(z)
+        for i in range(iter_):
+            self._conv_idx = [0]
+            if i == 0:
+                out, self._feat_map, self._conv_idx = self.decoder(
+                    x[:, :, i : i + 1, :, :], feat_cache=self._feat_map, feat_idx=self._conv_idx
+                )
+            else:
+                out_, self._feat_map, self._conv_idx = self.decoder(
+                    x[:, :, i : i + 1, :, :], feat_cache=self._feat_map, feat_idx=self._conv_idx
+                )
+                out = torch.cat([out, out_], 2)  # may add tensor offload
+        return out
+
+    def reparameterize(self, mu, log_var):
+        std = torch.exp(0.5 * log_var)
+        eps = torch.randn_like(std)
+        return eps * std + mu
+
+    def sample(self, imgs, deterministic=False):
+        mu, log_var = self.encode(imgs)
+        if deterministic:
+            return mu
+        std = torch.exp(0.5 * log_var.clamp(-30.0, 20.0))
+        return mu + std * torch.randn_like(std)
+
+    def clear_cache(self):
+        self._conv_num = count_conv3d(self.decoder)
+        self._conv_idx = [0]
+        self._feat_map = [None] * self._conv_num
+        # cache encode
+        self._enc_conv_num = count_conv3d(self.encoder)
+        self._enc_conv_idx = [0]
+        self._enc_feat_map = [None] * self._enc_conv_num
+
+
+class WanVideoVAE(nn.Module):
+    def __init__(self, z_dim=16):
+        super().__init__()
+
+        mean = [
+            -0.7571,
+            -0.7089,
+            -0.9113,
+            0.1075,
+            -0.1745,
+            0.9653,
+            -0.1517,
+            1.5508,
+            0.4134,
+            -0.0715,
+            0.5517,
+            -0.3632,
+            -0.1922,
+            -0.9497,
+            0.2503,
+            -0.2921,
+        ]
+        std = [
+            2.8184,
+            1.4541,
+            2.3275,
+            2.6558,
+            1.2196,
+            1.7708,
+            2.6052,
+            2.0743,
+            3.2687,
+            2.1526,
+            2.8652,
+            1.5579,
+            1.6382,
+            1.1253,
+            2.8251,
+            1.9160,
+        ]
+        self.mean = torch.tensor(mean)
+        self.std = torch.tensor(std)
+        self.scale = [self.mean, 1.0 / self.std]
+
+        # init model
+        self.model = VideoVAE_(z_dim=z_dim).eval().requires_grad_(False)
+        self.upsampling_factor = 8
+        self.temporal_downsample_factor = 4
+        self.z_dim = z_dim
+
+    def build_1d_mask(self, length, left_bound, right_bound, border_width):
+        x = torch.ones((length,))
+        if not left_bound:
+            x[:border_width] = (torch.arange(border_width) + 1) / border_width
+        if not right_bound:
+            x[-border_width:] = torch.flip(
+                (torch.arange(border_width) + 1) / border_width, dims=(0,)
+            )
+        return x
+
+    def build_mask(self, data, is_bound, border_width):
+        _, _, _, H, W = data.shape
+        h = self.build_1d_mask(H, is_bound[0], is_bound[1], border_width[0])
+        w = self.build_1d_mask(W, is_bound[2], is_bound[3], border_width[1])
+
+        h = repeat(h, "H -> H W", H=H, W=W)
+        w = repeat(w, "W -> H W", H=H, W=W)
+
+        mask = torch.stack([h, w]).min(dim=0).values
+        mask = rearrange(mask, "H W -> 1 1 1 H W")
+        return mask
+
+    def tiled_decode(self, hidden_states, device, tile_size, tile_stride):
+        _, _, T, H, W = hidden_states.shape
+        size_h, size_w = tile_size
+        stride_h, stride_w = tile_stride
+
+        # Split tasks
+        tasks = []
+        for h in range(0, H, stride_h):
+            if h - stride_h >= 0 and h - stride_h + size_h >= H:
+                continue
+            for w in range(0, W, stride_w):
+                if w - stride_w >= 0 and w - stride_w + size_w >= W:
+                    continue
+                h_, w_ = h + size_h, w + size_w
+                tasks.append((h, h_, w, w_))
+
+        data_device = "cpu"
+        computation_device = device
+
+        out_T = T * 4 - 3
+        weight = torch.zeros(
+            (1, 1, out_T, H * self.upsampling_factor, W * self.upsampling_factor),
+            dtype=hidden_states.dtype,
+            device=data_device,
+        )
+        values = torch.zeros(
+            (1, 3, out_T, H * self.upsampling_factor, W * self.upsampling_factor),
+            dtype=hidden_states.dtype,
+            device=data_device,
+        )
+
+        for h, h_, w, w_ in tqdm(tasks, desc="VAE decoding"):
+            hidden_states_batch = hidden_states[:, :, :, h:h_, w:w_].to(computation_device)
+            hidden_states_batch = self.model.decode(hidden_states_batch, self.scale).to(data_device)
+
+            mask = self.build_mask(
+                hidden_states_batch,
+                is_bound=(h == 0, h_ >= H, w == 0, w_ >= W),
+                border_width=(
+                    (size_h - stride_h) * self.upsampling_factor,
+                    (size_w - stride_w) * self.upsampling_factor,
+                ),
+            ).to(dtype=hidden_states.dtype, device=data_device)
+
+            target_h = h * self.upsampling_factor
+            target_w = w * self.upsampling_factor
+            values[
+                :,
+                :,
+                :,
+                target_h : target_h + hidden_states_batch.shape[3],
+                target_w : target_w + hidden_states_batch.shape[4],
+            ] += hidden_states_batch * mask
+            weight[
+                :,
+                :,
+                :,
+                target_h : target_h + hidden_states_batch.shape[3],
+                target_w : target_w + hidden_states_batch.shape[4],
+            ] += mask
+        values = values / weight
+        values = values.clamp_(-1, 1)
+        return values
+
+    def tiled_encode(self, video, device, tile_size, tile_stride):
+        _, _, T, H, W = video.shape
+        size_h, size_w = tile_size
+        stride_h, stride_w = tile_stride
+
+        # Split tasks
+        tasks = []
+        for h in range(0, H, stride_h):
+            if h - stride_h >= 0 and h - stride_h + size_h >= H:
+                continue
+            for w in range(0, W, stride_w):
+                if w - stride_w >= 0 and w - stride_w + size_w >= W:
+                    continue
+                h_, w_ = h + size_h, w + size_w
+                tasks.append((h, h_, w, w_))
+
+        data_device = "cpu"
+        computation_device = device
+
+        out_T = (T + 3) // 4
+        weight = torch.zeros(
+            (1, 1, out_T, H // self.upsampling_factor, W // self.upsampling_factor),
+            dtype=video.dtype,
+            device=data_device,
+        )
+        values = torch.zeros(
+            (1, self.z_dim, out_T, H // self.upsampling_factor, W // self.upsampling_factor),
+            dtype=video.dtype,
+            device=data_device,
+        )
+
+        for h, h_, w, w_ in tqdm(tasks, desc="VAE encoding"):
+            hidden_states_batch = video[:, :, :, h:h_, w:w_].to(computation_device)
+            hidden_states_batch = self.model.encode(hidden_states_batch, self.scale).to(data_device)
+
+            mask = self.build_mask(
+                hidden_states_batch,
+                is_bound=(h == 0, h_ >= H, w == 0, w_ >= W),
+                border_width=(
+                    (size_h - stride_h) // self.upsampling_factor,
+                    (size_w - stride_w) // self.upsampling_factor,
+                ),
+            ).to(dtype=video.dtype, device=data_device)
+
+            target_h = h // self.upsampling_factor
+            target_w = w // self.upsampling_factor
+            values[
+                :,
+                :,
+                :,
+                target_h : target_h + hidden_states_batch.shape[3],
+                target_w : target_w + hidden_states_batch.shape[4],
+            ] += hidden_states_batch * mask
+            weight[
+                :,
+                :,
+                :,
+                target_h : target_h + hidden_states_batch.shape[3],
+                target_w : target_w + hidden_states_batch.shape[4],
+            ] += mask
+        values = values / weight
+        return values
+
+    def single_encode(self, video, device):
+        video = video.to(device)
+        x = self.model.encode(video, self.scale)
+        return x
+
+    def single_decode(self, hidden_state, device):
+        hidden_state = hidden_state.to(device)
+        video = self.model.decode(hidden_state, self.scale)
+        return video.clamp_(-1, 1)
+
+    def encode(self, videos, device, tiled=False, tile_size=(34, 34), tile_stride=(18, 16)):
+        # videos = [video.to("cpu") for video in videos]
+        hidden_states = []
+        for video in videos:
+            video = video.unsqueeze(0)
+            if tiled:
+                raise NotImplementedError("Tiled encoding is not allowed yet.")
+                tile_size = (
+                    tile_size[0] * self.upsampling_factor,
+                    tile_size[1] * self.upsampling_factor,
+                )
+                tile_stride = (
+                    tile_stride[0] * self.upsampling_factor,
+                    tile_stride[1] * self.upsampling_factor,
+                )
+                hidden_state = self.tiled_encode(video, device, tile_size, tile_stride)
+            else:
+                hidden_state = self.single_encode(video, device)
+            hidden_state = hidden_state.squeeze(0)
+            hidden_states.append(hidden_state)
+        hidden_states = torch.stack(hidden_states)
+        return hidden_states
+
+    def decode(self, hidden_states, device, tiled=False, tile_size=(34, 34), tile_stride=(18, 16)):
+        hidden_states = [hidden_state.to("cpu") for hidden_state in hidden_states]
+        videos = []
+        for hidden_state in hidden_states:
+            hidden_state = hidden_state.unsqueeze(0)
+            if tiled:
+                video = self.tiled_decode(hidden_state, device, tile_size, tile_stride)
+            else:
+                video = self.single_decode(hidden_state, device)
+            video = video.squeeze(0)
+            videos.append(video)
+        videos = torch.stack(videos)
+        return videos
+
+    @staticmethod
+    def state_dict_converter():
+        return WanVideoVAEStateDictConverter()
+
+
+class WanVideoVAEStateDictConverter:
+    def __init__(self):
+        pass
+
+    def from_civitai(self, state_dict):
+        state_dict_ = {}
+        if "model_state" in state_dict:
+            state_dict = state_dict["model_state"]
+        for name in state_dict:
+            state_dict_["model." + name] = state_dict[name]
+        return state_dict_
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/policy.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/policy.py
new file mode 100644
index 0000000000000000000000000000000000000000..8c971fde0f74d94e3e6db135d6efdeca4575990d
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/policy.py
@@ -0,0 +1,457 @@
+"""EE16 RoboDojo policy: T-shaped RGB, history8, cached-video Euler action inference."""
+
+from collections import deque
+from pathlib import Path
+from typing import Any, Dict, Optional
+import numpy as np
+import torch
+import torchvision.transforms.functional as transforms_F
+from hydra.utils import instantiate
+from omegaconf import OmegaConf
+from PIL import Image
+from .data import DEFAULT_PROMPT
+from .checkpoint_compat import action_config
+from .datasets.lerobot.utils.normalizer import load_dataset_stats_from_json
+
+ROBODOJO_TSHAPE_VIDEO_SIZE = (240, 416)
+ROBODOJO_HORIZONTAL_VIDEO_SIZE = (240, 960)
+ROBODOJO_HORIZONTAL_PER_CAMERA_SIZE = (240, 320)
+DELTA_EEF14_REPRESENTATION = "unsupported_delta_eef"
+DELTA_JOINT14_REPRESENTATION = "unsupported_delta_joint"
+
+
+def _normalized_to_uint8(image: torch.Tensor) -> torch.Tensor:
+    """Keep one composed RGB frame compactly in the CPU history buffer."""
+    frame = image.detach().float().cpu()
+    if frame.ndim == 4:
+        if frame.shape[0] != 1:
+            raise ValueError(f"Expected one image batch, got {tuple(image.shape)}")
+        frame = frame[0]
+    if frame.ndim != 3 or frame.shape[0] != 3:
+        raise ValueError(f"Expected normalized [3,H,W], got {tuple(image.shape)}")
+    return frame.clamp(-1, 1).add(1.0).mul(127.5).round().to(torch.uint8).contiguous()
+
+
+def _uint8_to_normalized(frames: list[torch.Tensor]) -> torch.Tensor:
+    """Stack compact CPU RGB frames into normalized ``[3,T,H,W]`` form."""
+    if not frames:
+        raise ValueError("Cannot build a history condition from no frames.")
+    stacked = torch.stack(frames, dim=1).to(torch.float32)
+    return stacked.div(255.0).mul(2.0).sub(1.0)
+
+
+def _model_dtype(value: Any) -> torch.dtype:
+    key = str(value or "bf16").strip().lower()
+    if key == "bf16":
+        return torch.bfloat16
+    if key == "fp16":
+        return torch.float16
+    if key in {"no", "fp32", "float32"}:
+        return torch.float32
+    raise ValueError(f"Unsupported mixed precision: {value!r}")
+
+
+def _as_uint8_hwc(image: Any) -> np.ndarray:
+    array = np.asarray(image)
+    if array.ndim != 3:
+        raise ValueError(f"RoboDojo RGB view must be HWC, got {array.shape}")
+    if array.shape[0] in {1, 3} and array.shape[-1] not in {1, 3, 4}:
+        array = np.transpose(array, (1, 2, 0))
+    if array.shape[-1] < 3:
+        raise ValueError(f"RoboDojo RGB view needs three channels, got {array.shape}")
+    array = array[..., :3]
+    if np.issubdtype(array.dtype, np.floating):
+        scale = 255.0 if float(np.nanmax(array)) <= 1.5 else 1.0
+        array = np.clip(array * scale, 0, 255).astype(np.uint8)
+    else:
+        array = np.clip(array, 0, 255).astype(np.uint8)
+    return np.ascontiguousarray(array)
+
+
+def compose_training_condition(
+    head: Any,
+    left_wrist: Any,
+    right_wrist: Any,
+    *,
+    layout: str = "robodojo_tshape",
+    video_size: tuple[int, int] = ROBODOJO_TSHAPE_VIDEO_SIZE,
+    per_camera_video_size: Optional[tuple[int, int]] = None,
+) -> torch.Tensor:
+    """Reproduce the checkpoint's camera composition and normalization."""
+    views = []
+    for image in (head, left_wrist, right_wrist):
+        tensor = torch.from_numpy(np.array(_as_uint8_hwc(image), copy=True))
+        views.append(tensor.permute(2, 0, 1).to(torch.float32).div_(255.0))
+    layout = str(layout).strip().lower()
+    target_size = tuple((int(value) for value in video_size))
+    if layout == "horizontal":
+        camera_size = tuple(
+            (int(value) for value in per_camera_video_size or ROBODOJO_HORIZONTAL_PER_CAMERA_SIZE)
+        )
+        resized_views = [
+            transforms_F.resize(
+                view,
+                size=list(camera_size),
+                interpolation=transforms_F.InterpolationMode.BILINEAR,
+                antialias=True,
+            )
+            for view in views
+        ]
+        condition = torch.cat(resized_views, dim=2)
+        if tuple(condition.shape[-2:]) != target_size:
+            condition = transforms_F.resize(
+                condition,
+                size=list(target_size),
+                interpolation=transforms_F.InterpolationMode.BILINEAR,
+                antialias=True,
+            )
+    elif layout == "robodojo_tshape":
+        if target_size != ROBODOJO_TSHAPE_VIDEO_SIZE:
+            raise ValueError(
+                f"RoboDojo T-shape deployment requires video_size={ROBODOJO_TSHAPE_VIDEO_SIZE}, got {target_size}."
+            )
+        head_tensor = transforms_F.resize(
+            views[0],
+            size=[480, 640],
+            interpolation=transforms_F.InterpolationMode.BILINEAR,
+            antialias=True,
+        )
+        left_tensor = transforms_F.resize(
+            views[1],
+            size=[240, 240],
+            interpolation=transforms_F.InterpolationMode.BILINEAR,
+            antialias=True,
+        )
+        right_tensor = transforms_F.resize(
+            views[2],
+            size=[240, 240],
+            interpolation=transforms_F.InterpolationMode.BILINEAR,
+            antialias=True,
+        )
+        raw_layout = torch.cat([head_tensor, torch.cat([left_tensor, right_tensor], dim=1)], dim=2)
+        condition = transforms_F.resize(
+            raw_layout,
+            size=list(target_size),
+            interpolation=transforms_F.InterpolationMode.BILINEAR,
+            antialias=True,
+        )
+    else:
+        raise ValueError(f"Unsupported RoboDojo camera layout: {layout!r}.")
+    return condition.mul(2.0).sub(1.0).unsqueeze(0).contiguous()
+
+
+class RoboDojoVPP2Policy:
+    def __init__(
+        self,
+        checkpoint_path,
+        wan_model_dir,
+        dataset_stats_path=None,
+        video_checkpoint_path=None,
+        device="cuda",
+        mixed_precision="bf16",
+        num_inference_steps=10,
+        sigma_shift=1.0,
+        seed=1,
+        replan_steps=16,
+    ):
+        checkpoint_path = Path(checkpoint_path).resolve()
+        payload = torch.load(checkpoint_path, map_location="cpu", mmap=True, weights_only=True)
+        cfg = OmegaConf.create(action_config(payload))
+        data = cfg.data.train
+        if (
+            int(cfg.model.action_dit_config.hidden_dim),
+            int(data.processor.action_output_dim),
+            int(data.processor.proprio_output_dim),
+        ) != (1024, 16, 16):
+            raise ValueError("Expected Action2B / EE16 architecture")
+        if list(data.video_size) != [240, 416] or data.video_resize_mode != "direct":
+            raise ValueError("Expected direct 240x416 T-shape preprocessing")
+        if (data.condition_history_frames, data.condition_history_stride, data.action_horizon) != (
+            8,
+            25,
+            32,
+        ):
+            raise ValueError("Expected history8/stride25/horizon32")
+        if not data.condition_include_episode_first or not data.condition_history_target_prefix:
+            raise ValueError("Episode anchor and history prefix are required")
+        video_path = Path(video_checkpoint_path or payload["video_checkpoint"]["path"])
+        if not video_path.is_absolute():
+            video_path = checkpoint_path.parent / video_path
+        if not video_path.is_file():
+            raise FileNotFoundError(video_path)
+        video_meta = torch.load(video_path, map_location="cpu", mmap=True, weights_only=True)
+        if video_meta.get("step") != payload.get("step") or "dit" not in video_meta:
+            raise ValueError("Video/Action step or format mismatch")
+        del video_meta
+        model_cfg = cfg.model
+        model_cfg._target_ = "bagel_cp1000.runtime.create_vpp2_wan21_14b"
+        model_cfg.model_id = str(Path(wan_model_dir).resolve())
+        model_cfg.tokenizer_model_id = str(Path(model_cfg.model_id) / "google/umt5-xxl")
+        model_cfg.dit_checkpoint_path = str(video_path.resolve())
+        model_cfg.action_dit_pretrained_path = None
+        model_cfg.load_text_encoder = True
+        model_cfg.video_dit_config.use_gradient_checkpointing = False
+        model_cfg.action_dit_config.use_gradient_checkpointing = False
+        model = instantiate(model_cfg, model_dtype=_model_dtype(mixed_precision), device=device)
+        model.action_expert.load_state_dict(payload["action_expert"], strict=True)
+        model.proprio_encoder.load_state_dict(payload["proprio_encoder"], strict=True)
+        self.model = model.eval()
+        self.processor = instantiate(data.processor).eval()
+        stats = (
+            Path(dataset_stats_path)
+            if dataset_stats_path
+            else checkpoint_path.parent / "dataset_stats.json"
+        )
+        self.processor.set_normalizer_from_stats(load_dataset_stats_from_json(str(stats)))
+        self.action_dim = self.proprio_dim = self.output_action_dim = 16
+        self.action_representation = "absolute_dual_arm_ee_pose_qwxyz_plus_gripper"
+        self.action_horizon = 32
+        self.replan_steps = int(replan_steps)
+        self.num_inference_steps = int(num_inference_steps)
+        self.sigma_shift = float(sigma_shift)
+        if (
+            not 1 <= self.replan_steps <= 32
+            or self.num_inference_steps < 1
+            or self.sigma_shift <= 0
+        ):
+            raise ValueError("Invalid inference horizon/steps/shift")
+        self.base_seed = int(seed)
+        self.video_seed_offset = 1000003
+        self.text_cfg_scale = 1.0
+        self.negative_prompt = ""
+        self.rand_device = "cpu"
+        self.tiled = False
+        self.video_size = (240, 416)
+        self.visual_layout = "robodojo_tshape"
+        self.per_camera_video_size = None
+        self.num_video_frames = 17
+        self.history_condition_frames = 8
+        self.history_stride = 25
+        self.history_include_episode_first = True
+        self.history_buffer_size = 177
+        self.visual_planning_output_root = None
+        self.visual_planning_full_steps = 10
+        self._history_by_env = {}
+        self._history_anchor_by_env = {}
+        self._history_step_by_env = {}
+        self._latest_image_by_env = {}
+        self._inference_index_by_env = {}
+        self._inference_index = 0
+        self._request_context = {}
+
+    _PER_ENV_STATE = (
+        "_history_by_env",
+        "_history_anchor_by_env",
+        "_history_step_by_env",
+        "_latest_image_by_env",
+        "_inference_index_by_env",
+    )
+
+    def set_request_context(self, context: Dict[str, Any]) -> None:
+        self._request_context = dict(context or {})
+
+    def _compose_observation_image(self, observation: Dict[str, Any]) -> torch.Tensor:
+        images = observation["observation"]
+        return compose_training_condition(
+            images["head_camera"]["rgb"],
+            images["left_camera"]["rgb"],
+            images["right_camera"]["rgb"],
+            layout=self.visual_layout,
+            video_size=self.video_size,
+            per_camera_video_size=self.per_camera_video_size,
+        ).to(device=self.model.device, dtype=self.model.torch_dtype)
+
+    def observe(self, observation: Dict[str, Any]) -> None:
+        """Consume one native simulator observation for history conditioning.
+
+        The official RoboDojo client calls ``update_obs`` for every action
+        inside a returned chunk.  Recording here, rather than only when a new
+        chunk is requested, preserves the native 25-Hz history timeline.
+        """
+        env_idx = int(observation.get("env_idx", 0))
+        image = self._compose_observation_image(observation)
+        self._latest_image_by_env[env_idx] = image
+        if self.history_condition_frames > 0:
+            self._record_history_frame(image, env_idx)
+
+    def _normalize_state(self, state: np.ndarray) -> torch.Tensor:
+        state = np.asarray(state, dtype=np.float32)
+        if state.shape != (self.output_action_dim,):
+            raise ValueError(
+                f"Expected RoboDojo state [{self.output_action_dim}], got {state.shape}."
+            )
+        state_meta = self.processor.shape_meta["state"]
+        if len(state_meta) != 1:
+            raise ValueError("Expected one merged state key.")
+        key = state_meta[0]["key"]
+        batch = {"state": {key: torch.as_tensor(state, dtype=torch.float32).unsqueeze(0)}}
+        batch = self.processor.action_state_transform(batch)
+        batch = self.processor.normalizer.forward(batch)
+        return batch["state"][key]
+
+    def _denormalize_action(self, action: torch.Tensor) -> np.ndarray:
+        if action.ndim == 2:
+            action = action.unsqueeze(0)
+        key = self.processor.shape_meta["action"][0]["key"]
+        normalizer = self.processor.normalizer.normalizers["action"][key]
+        return normalizer.backward(action.float().cpu()).numpy()
+
+    def _record_history_frame(self, image: torch.Tensor, env_idx: int) -> None:
+        """Append one simulator observation to the per-environment history."""
+        env_idx = int(env_idx)
+        history = self._history_by_env.get(env_idx)
+        if history is None:
+            history = deque(maxlen=self.history_buffer_size)
+            self._history_by_env[env_idx] = history
+            step = 0
+            self._history_anchor_by_env[env_idx] = _normalized_to_uint8(image)
+        else:
+            step = int(self._history_step_by_env[env_idx]) + 1
+        frame = _normalized_to_uint8(image)
+        history.append((step, frame))
+        self._history_step_by_env[env_idx] = step
+
+    def _build_history_condition(
+        self, image: torch.Tensor, env_idx: int
+    ) -> tuple[Optional[torch.Tensor], Optional[dict[str, Any]]]:
+        """Build the exact anchor+history RGB condition expected by aligned history."""
+        if self.history_condition_frames <= 0:
+            return (None, None)
+        env_idx = int(env_idx)
+        history = self._history_by_env.get(env_idx)
+        if not history:
+            self._record_history_frame(image, env_idx)
+            history = self._history_by_env[env_idx]
+        current_step = int(self._history_step_by_env[env_idx])
+        anchor = self._history_anchor_by_env.get(env_idx)
+        if anchor is None:
+            anchor = history[0][1]
+            self._history_anchor_by_env[env_idx] = anchor
+        requested_offsets = [
+            -(self.history_condition_frames - 1 - index) * self.history_stride
+            for index in range(self.history_condition_frames)
+        ]
+        step_to_frame = {int(step): frame for (step, frame) in history}
+        selected_frames: list[torch.Tensor] = []
+        selected_steps: list[int] = []
+        for offset in requested_offsets:
+            requested_step = current_step + int(offset)
+            if requested_step <= 0:
+                selected_step = 0
+                selected_frame = anchor
+            elif requested_step in step_to_frame:
+                selected_step = requested_step
+                selected_frame = step_to_frame[requested_step]
+            else:
+                prior = [
+                    (int(step), frame) for (step, frame) in history if int(step) <= requested_step
+                ]
+                if prior:
+                    (selected_step, selected_frame) = prior[-1]
+                else:
+                    (selected_step, selected_frame) = (0, anchor)
+            selected_steps.append(int(selected_step))
+            selected_frames.append(selected_frame)
+        condition_frames = [anchor, *selected_frames]
+        condition = _uint8_to_normalized(condition_frames).unsqueeze(0)
+        condition = condition.to(device=image.device, dtype=image.dtype)
+        metadata = {
+            "enabled": True,
+            "env_idx": env_idx,
+            "current_step": current_step,
+            "requested_offsets": requested_offsets,
+            "selected_steps": [0, *selected_steps],
+            "rgb_frames": len(condition_frames),
+            "latent_frames": (len(condition_frames) - 1) // 4 + 1,
+            "stride_native_steps": self.history_stride,
+        }
+        return (condition.contiguous(), metadata)
+
+    def _seeds(self, env_idx: int = 0) -> tuple[Optional[int], Optional[int]]:
+        env_idx = int(env_idx)
+        replan_index = int(self._inference_index_by_env.get(env_idx, 0))
+        self._inference_index_by_env[env_idx] = replan_index + 1
+        self._inference_index += 1
+        if self.base_seed is None:
+            return (None, None)
+        action_seed = int(self.base_seed + replan_index)
+        video_seed = int(action_seed + self.video_seed_offset)
+        return (action_seed, video_seed)
+
+    def _infer_action_chunk(self, observation: Dict[str, Any], instruction: str) -> np.ndarray:
+        state = np.asarray(observation["joint_action"]["vector"], dtype=np.float32)
+        env_idx = int(observation.get("env_idx", 0))
+        image = self._latest_image_by_env.get(env_idx)
+        if image is None:
+            image = self._compose_observation_image(observation)
+            if self.history_condition_frames > 0:
+                self._record_history_frame(image, env_idx)
+        history_condition = None
+        history_metadata = None
+        if self.history_condition_frames > 0:
+            (history_condition, history_metadata) = self._build_history_condition(
+                image=image, env_idx=env_idx
+            )
+        delta_eef_reference = state
+        proprio = self._normalize_state(delta_eef_reference)
+        (action_seed, video_seed) = self._seeds(env_idx)
+        capture_visual_plan = self.visual_planning_output_root is not None
+        infer_kwargs = {
+            "prompt": DEFAULT_PROMPT.format(task=instruction),
+            "input_image": image,
+            "action_horizon": self.action_horizon,
+            "proprio": proprio,
+            "negative_prompt": self.negative_prompt,
+            "text_cfg_scale": self.text_cfg_scale,
+            "num_inference_steps": self.num_inference_steps,
+            "sigma_shift": self.sigma_shift,
+            "seed": action_seed,
+            "video_seed": video_seed,
+            "rand_device": self.rand_device,
+            "tiled": self.tiled,
+            "num_video_frames": self.num_video_frames,
+            "return_one_step_video": capture_visual_plan,
+            "return_full_video": capture_visual_plan,
+            "full_video_num_inference_steps": self.visual_planning_full_steps
+            if capture_visual_plan
+            else None,
+        }
+        if history_condition is not None:
+            infer_kwargs["condition_video"] = history_condition
+        with torch.no_grad():
+            prediction = self.model.infer_action(**infer_kwargs)
+        action = self._denormalize_action(prediction["action"])[0]
+        if action.ndim != 2 or action.shape[1] != self.output_action_dim:
+            raise ValueError(
+                f"Model returned an invalid RoboDojo action chunk: expected [T,{self.output_action_dim}], got {action.shape}."
+            )
+        return action
+
+    def reset(self) -> None:
+        self._inference_index = 0
+        for name in self._PER_ENV_STATE:
+            state = getattr(self, name, None)
+            if state is not None:
+                state.clear()
+
+    def reset_env(self, env_idx: int) -> None:
+        """Clear the episode state of one environment only."""
+        for name in self._PER_ENV_STATE:
+            state = getattr(self, name, None)
+            if state is not None:
+                state.pop(int(env_idx), None)
+
+
+def get_model(args):
+    return RoboDojoVPP2Policy(
+        checkpoint_path=args.get("checkpoint_path") or args.get("ckpt_setting"),
+        wan_model_dir=args["wan_model_dir"],
+        dataset_stats_path=args.get("dataset_stats_path"),
+        video_checkpoint_path=args.get("video_checkpoint_path"),
+        device=args.get("device") or "cuda",
+        mixed_precision=args.get("mixed_precision") or "bf16",
+        num_inference_steps=args.get("num_inference_steps") or 10,
+        sigma_shift=args.get("sigma_shift") or 1,
+        seed=args.get("seed", 1),
+        replan_steps=args.get("replan_steps") or 16,
+    )
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/runtime.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/runtime.py
new file mode 100644
index 0000000000000000000000000000000000000000..385a60b46e1d14bac3be1c639bac9a87b2969645
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/runtime.py
@@ -0,0 +1,170 @@
+"""Inference factory extracted from the evaluated runtime."""
+from pathlib import Path
+import torch
+from omegaconf import DictConfig, OmegaConf
+
+def create_vpp2_wan21_14b(
+    model_id: str = "weights/Wan2.1-I2V-14B-480P",
+    dit_checkpoint_path: str | None = None,
+    tokenizer_model_id: str | None = None,
+    video_dit_config=None,
+    tokenizer_max_len: int = 512,
+    load_text_encoder: bool = True,
+    load_clip_encoder: bool = True,
+    proprio_dim: int | None = None,
+    action_dit_config=None,
+    action_dit_pretrained_path: str | None = None,
+    skip_dit_load_from_pretrain: bool = False,
+    video_scheduler=None,
+    action_scheduler=None,
+    loss=None,
+    mot_checkpoint_mixed_attn: bool = True,
+    action_chunk_size: int = 1,
+    action_visible_video_frames: int = 1,
+    action_only_first_frame_fast_path: bool = False,
+    action_clean_first_frame_start_layer: int | None = None,
+    action_clean_first_frame_timestep_cutoff: float | None = None,
+    action_first_frame_kv_source: str = "clean_prefill",
+    action_video_timestep_cutoff: float | None = None,
+    video_lora=None,
+    use_fixed_video_layers: int | None = None,
+    use_video_tokens: bool = False,
+    video_token_latent_dim: int = 2560,
+    joint_denoising: bool = False,
+    joint_denoising_mode_probs=None,
+    joint_denoising_sampler: str = "euler",
+    redirect_common_files: bool = True,
+    model_dtype: torch.dtype = torch.bfloat16,
+    device: str = "cuda",
+):
+    from .models.wan21_14b.vpp2 import VPP2
+
+    if joint_denoising or video_lora and bool(video_lora.get("enabled", False)):
+        raise ValueError(
+            "This repository supports full joint training and cached-video action inference only."
+        )
+    if action_chunk_size != 1 or use_fixed_video_layers is not None or use_video_tokens:
+        raise ValueError("Unsupported experimental architecture option")
+
+    if isinstance(video_dit_config, DictConfig):
+        video_dit_config = OmegaConf.to_container(video_dit_config, resolve=True)
+    if video_dit_config is None:
+        video_dit_config = {}
+    if not isinstance(video_dit_config, dict):
+        raise ValueError(f"`video_dit_config` must resolve to a dict, got {type(video_dit_config)}")
+
+    if isinstance(action_dit_config, DictConfig):
+        action_dit_config = OmegaConf.to_container(action_dit_config, resolve=True)
+    if action_dit_config is None:
+        action_dit_config = {}
+    if not isinstance(action_dit_config, dict):
+        raise ValueError(
+            f"`action_dit_config` must resolve to a dict, got {type(action_dit_config)}"
+        )
+    action_chunk_size = int(action_chunk_size)
+    if action_chunk_size <= 0:
+        raise ValueError(f"`action_chunk_size` must be positive, got {action_chunk_size}")
+    if action_chunk_size > 1:
+        action_dit_config = dict(action_dit_config)
+        action_dit_config["action_dim"] = int(action_dit_config["action_dim"]) * action_chunk_size
+
+    if isinstance(video_scheduler, DictConfig):
+        video_scheduler = OmegaConf.to_container(video_scheduler, resolve=True)
+    if video_scheduler is None:
+        video_scheduler = {}
+    if not isinstance(video_scheduler, dict):
+        raise ValueError(f"`video_scheduler` must be dict-like, got {type(video_scheduler)}")
+
+    if isinstance(action_scheduler, DictConfig):
+        action_scheduler = OmegaConf.to_container(action_scheduler, resolve=True)
+    if action_scheduler is None:
+        raise ValueError("`action_scheduler` is required for VPP2 Wan2.1-14B.")
+    if not isinstance(action_scheduler, dict):
+        raise ValueError(f"`action_scheduler` must be dict-like, got {type(action_scheduler)}")
+    required_action_scheduler_keys = {"train_shift", "infer_shift", "num_train_timesteps"}
+    missing_keys = required_action_scheduler_keys - set(action_scheduler.keys())
+    if missing_keys:
+        raise ValueError(
+            f"`action_scheduler` missing required keys: {sorted(missing_keys)}. "
+            "Expected keys: train_shift, infer_shift, num_train_timesteps."
+        )
+
+    if isinstance(loss, DictConfig):
+        loss = OmegaConf.to_container(loss, resolve=True)
+    if loss is None:
+        loss = {}
+    if not isinstance(loss, dict):
+        raise ValueError(f"`loss` must be dict-like, got {type(loss)}")
+
+    if isinstance(video_lora, DictConfig):
+        video_lora = OmegaConf.to_container(video_lora, resolve=True)
+    if video_lora is None:
+        video_lora = {}
+    if not isinstance(video_lora, dict):
+        raise ValueError(f"`video_lora` must be dict-like, got {type(video_lora)}")
+
+    # Explicit checkpoint paths in the release YAML are relative to the checkout.
+    if dit_checkpoint_path is not None and Path(dit_checkpoint_path).is_file():
+        dit_checkpoint_path = str(Path(dit_checkpoint_path).resolve())
+
+    return VPP2.from_wan21_14b_pretrained(
+        device=device,
+        torch_dtype=model_dtype,
+        model_id=model_id,
+        dit_checkpoint_path=dit_checkpoint_path,
+        tokenizer_model_id=tokenizer_model_id,
+        tokenizer_max_len=int(tokenizer_max_len),
+        load_text_encoder=bool(load_text_encoder),
+        load_clip_encoder=bool(load_clip_encoder),
+        proprio_dim=(None if proprio_dim is None else int(proprio_dim)),
+        redirect_common_files=bool(redirect_common_files),
+        video_dit_config=video_dit_config,
+        action_dit_config=action_dit_config,
+        action_dit_pretrained_path=action_dit_pretrained_path,
+        skip_dit_load_from_pretrain=bool(skip_dit_load_from_pretrain),
+        mot_checkpoint_mixed_attn=bool(mot_checkpoint_mixed_attn),
+        action_chunk_size=action_chunk_size,
+        action_visible_video_frames=int(action_visible_video_frames),
+        action_only_first_frame_fast_path=bool(action_only_first_frame_fast_path),
+        action_clean_first_frame_start_layer=(
+            None
+            if action_clean_first_frame_start_layer is None
+            else int(action_clean_first_frame_start_layer)
+        ),
+        action_clean_first_frame_timestep_cutoff=(
+            None
+            if action_clean_first_frame_timestep_cutoff is None
+            else float(action_clean_first_frame_timestep_cutoff)
+        ),
+        action_first_frame_kv_source=str(action_first_frame_kv_source),
+        action_video_timestep_cutoff=(
+            None if action_video_timestep_cutoff is None else float(action_video_timestep_cutoff)
+        ),
+        video_lora_config=video_lora,
+        use_fixed_video_layers=(
+            None if use_fixed_video_layers is None else int(use_fixed_video_layers)
+        ),
+        use_video_tokens=bool(use_video_tokens),
+        video_token_latent_dim=int(video_token_latent_dim),
+        video_train_shift=float(video_scheduler.get("train_shift", 5.0)),
+        video_infer_shift=float(video_scheduler.get("infer_shift", 5.0)),
+        video_num_train_timesteps=int(video_scheduler.get("num_train_timesteps", 1000)),
+        video_train_sampling_strategy=str(
+            video_scheduler.get("train_sampling_strategy", "shifted_uniform")
+        ),
+        video_train_beta_alpha=float(video_scheduler.get("beta_alpha", 7.0)),
+        video_train_beta_beta=float(video_scheduler.get("beta_beta", 1.0)),
+        action_train_shift=float(action_scheduler["train_shift"]),
+        action_infer_shift=float(action_scheduler["infer_shift"]),
+        action_num_train_timesteps=int(action_scheduler["num_train_timesteps"]),
+        loss_lambda_video=float(loss.get("lambda_video", 1.0)),
+        loss_lambda_action=float(loss.get("lambda_action", 1.0)),
+        video_lora_joint_video_loss=bool(loss.get("video_lora_joint_video_loss", False)),
+        joint_denoising=bool(joint_denoising),
+        joint_denoising_mode_probs=(
+            OmegaConf.to_container(joint_denoising_mode_probs, resolve=True)
+            if isinstance(joint_denoising_mode_probs, DictConfig)
+            else joint_denoising_mode_probs
+        ),
+        joint_denoising_sampler=str(joint_denoising_sampler),
+    )
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/__init__.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/logging_config.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/logging_config.py
new file mode 100644
index 0000000000000000000000000000000000000000..9d059792f23e05916293b2b6c3edb89acac13c9e
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/logging_config.py
@@ -0,0 +1,155 @@
+import logging
+from typing import Optional
+import os
+from pathlib import Path
+
+import torch.distributed as dist
+
+from rich.logging import RichHandler
+
+
+def setup_logging(
+    log_level: int = logging.INFO,
+    is_main_process: Optional[bool] = None,
+    rich_handler_kwargs: Optional[dict] = None,
+    formatter_kwargs: Optional[dict] = None,
+    preserve_hydra_handlers: bool = True,
+    log_file: Optional[str | os.PathLike[str]] = None,
+) -> None:
+    """
+    Configure the logging system for the entire codebase with Rich formatting.
+
+    In distributed training, only the main process outputs logs while other processes are silenced.
+    This function configures the root logger so all child loggers inherit the same configuration.
+
+    Args:
+        log_level: Logging level (default INFO), only applies to main process
+        is_main_process: Whether this is the main process. If None, infer automatically.
+        rich_handler_kwargs: Additional kwargs to pass to RichHandler
+        formatter_kwargs: Additional kwargs to pass to Formatter (fmt and datefmt)
+        preserve_hydra_handlers: Keep existing FileHandlers from Hydra (default True)
+        log_file: Optional extra file path to write logs to, e.g. output_dir/train.log
+
+    Example:
+        ```python
+        # In a single-machine script
+        from bagel_cp1000.utils.logging_config import setup_logging
+        setup_logging()
+
+        # In a distributed training script
+        from accelerate import PartialState
+        from bagel_cp1000.utils.logging_config import setup_logging
+
+        distributed_state = PartialState()
+        setup_logging(
+            log_level=logging.INFO,
+            is_main_process=distributed_state.is_main_process
+        )
+        ```
+    """
+    if is_main_process is None:
+        is_main_process = _is_main_process()
+
+    root_logger = logging.getLogger()
+
+    if is_main_process:
+        # Save existing FileHandlers (e.g., from Hydra) if requested
+        existing_file_handlers = []
+        if preserve_hydra_handlers:
+            existing_file_handlers = [
+                h for h in root_logger.handlers if isinstance(h, logging.FileHandler)
+            ]
+
+        # Clear all default handlers on the root logger
+        root_logger.handlers.clear()
+
+        # Configure RichHandler parameters
+        default_rich_kwargs = {
+            "markup": True,
+            "rich_tracebacks": True,
+            "show_level": True,
+            "show_path": True,
+            "show_time": True,
+        }
+        if rich_handler_kwargs:
+            default_rich_kwargs.update(rich_handler_kwargs)
+
+        # Create RichHandler
+        rich_handler = RichHandler(**default_rich_kwargs)
+
+        # Configure Formatter
+        default_formatter_kwargs = {
+            "fmt": "| >> %(message)s",
+            "datefmt": "%m/%d [%H:%M:%S]",
+        }
+        if formatter_kwargs:
+            default_formatter_kwargs.update(formatter_kwargs)
+
+        formatter = logging.Formatter(**default_formatter_kwargs)
+        rich_handler.setFormatter(formatter)
+
+        # Add handler and set logging level
+        root_logger.addHandler(rich_handler)
+
+        # Restore existing FileHandlers from Hydra
+        for handler in existing_file_handlers:
+            root_logger.addHandler(handler)
+
+        if log_file is not None:
+            log_path = Path(log_file).expanduser()
+            log_path.parent.mkdir(parents=True, exist_ok=True)
+            file_handler = logging.FileHandler(log_path, mode="a", encoding="utf-8")
+            file_handler.setFormatter(
+                logging.Formatter(
+                    fmt="[%(asctime)s][%(name)s][%(levelname)s] - %(message)s",
+                    datefmt="%Y-%m-%d %H:%M:%S",
+                )
+            )
+            root_logger.addHandler(file_handler)
+
+        root_logger.setLevel(log_level)
+
+    else:
+        # In non-main processes, set root logger level to ERROR to silence all logs
+        root_logger.setLevel(logging.ERROR)
+
+
+def _is_main_process() -> bool:
+    """
+    Best-effort check for main process without any synchronization.
+    """
+    # Prefer torch.distributed state if initialized.
+    if dist is not None and dist.is_available() and dist.is_initialized():
+        return dist.get_rank() == 0
+
+    # Fallback to environment variables commonly set by launchers.
+    for key in ("RANK", "SLURM_PROCID", "LOCAL_RANK"):
+        if key in os.environ:
+            return os.environ.get(key, "0") in ("0", "0\n", "")
+
+    return True
+
+
+def get_logger(name: str = __name__, level: int = logging.INFO) -> logging.Logger:
+    """
+    Drop-in replacement for accelerate.logging.get_logger:
+    - No implicit barriers.
+    - Only the main process emits log records by default.
+    """
+    logger = logging.getLogger(name)
+    logger.setLevel(level)
+
+    # if not logger.handlers and _is_main_process():
+    #     handler = logging.StreamHandler()
+    #     formatter = logging.Formatter(
+    #         fmt="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+    #         datefmt="%Y-%m-%d %H:%M:%S",
+    #     )
+    #     handler.setFormatter(formatter)
+    #     logger.addHandler(handler)
+
+    if not _is_main_process():
+        logger.propagate = False
+        logger.disabled = True
+
+    return logger
diff --git a/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/pytorch_utils.py b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/pytorch_utils.py
new file mode 100644
index 0000000000000000000000000000000000000000..279423a6de82fe1c4e015c4dbb6b0ec56aee4f76
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/runtime/src/bagel_cp1000/utils/pytorch_utils.py
@@ -0,0 +1,141 @@
+from typing import Dict, List, Callable, Optional
+import os
+import random
+import collections
+
+import torch
+import torch.distributed as dist
+import numpy as np
+
+
+def _resolve_global_rank() -> int:
+    if torch.distributed.is_available() and torch.distributed.is_initialized():
+        return int(torch.distributed.get_rank())
+    return int(
+        os.environ.get("RANK", os.environ.get("SLURM_PROCID", os.environ.get("LOCAL_RANK", "0")))
+    )
+
+
+def set_global_seed(seed: int, get_worker_init_fn: bool = False) -> Optional[Callable[[int], None]]:
+    """Sets seed for all randomness libraries (mostly random, numpy, torch) and produces a `worker_init_fn`"""
+    assert np.iinfo(np.uint32).min < seed < np.iinfo(np.uint32).max, (
+        "Seed outside the np.uint32 bounds!"
+    )
+
+    # Set Seed as an Environment Variable
+    os.environ["EXPERIMENT_GLOBAL_SEED"] = str(seed)
+
+    # Process-specific seeding: offset by global rank so each process gets a different seed
+    global_rank = _resolve_global_rank()
+    process_seed = seed + global_rank
+
+    random.seed(process_seed)
+    np.random.seed(process_seed)
+    torch.manual_seed(process_seed)
+    if torch.cuda.is_available():
+        torch.cuda.manual_seed_all(process_seed)
+
+    return worker_init_function if get_worker_init_fn else None
+
+
+def worker_init_function(worker_id: int) -> None:
+    """
+    Borrowed directly from PyTorch-Lightning; inspired by this issue comment in the PyTorch repo:
+        > Ref: https://github.com/pytorch/pytorch/issues/5059#issuecomment-817392562
+
+    Intuition: You can think of the seed sequence spawn function as a "janky" torch.Generator() or jax.PRNGKey that
+    you can run iterative splitting on to get new (predictable) randomness.
+
+    :param worker_id: Identifier for the given worker [0, num_workers) for the Dataloader in question.
+    """
+    # Get current global `rank` (if running distributed) and `process_seed`
+    process_seed = torch.initial_seed()
+    global_rank = _resolve_global_rank()
+
+    # Back out the "base" (original) seed - the per-worker seed is set in PyTorch:
+    #   > https://pytorch.org/docs/stable/data.html#data-loading-randomness
+    base_seed = process_seed - worker_id
+
+    # "Magic" code --> basically creates a seed sequence that mixes different "sources" and seeds every library...
+    seed_seq = np.random.SeedSequence([base_seed, worker_id, global_rank])
+
+    # Use 128 bits (4 x 32-bit words) to represent seed --> generate_state(k) produces a `k` element array!
+    np.random.seed(seed_seq.generate_state(4))
+
+    # Spawn distinct child sequences for PyTorch (reseed) and stdlib random
+    torch_seed_seq, random_seed_seq, tf_seed_seq = seed_seq.spawn(3)
+
+    # Torch Manual seed takes 64 bits (so just specify a dtype of uint64
+    torch.manual_seed(torch_seed_seq.generate_state(1, dtype=np.uint64)[0])
+
+    # Use 128 Bits for `random`, but express as integer instead of as an array
+    random_seed = (
+        random_seed_seq.generate_state(2, dtype=np.uint64).astype(list) * [1 << 64, 1]
+    ).sum()
+    random.seed(random_seed)
+
+
+def dict_apply(
+    x: Dict[str, torch.Tensor], func: Callable[[torch.Tensor], torch.Tensor]
+) -> Dict[str, torch.Tensor]:
+    result = dict()
+    for key, value in x.items():
+        if isinstance(value, dict):
+            result[key] = dict_apply(value, func)
+        else:
+            result[key] = func(value)
+    return result
+
+
+def dict_to_array(x):
+    data = np.concatenate([item for _, item in x.items()], axis=-1)
+    return data
+
+
+def pad_remaining_dims(x, target):
+    assert x.shape == target.shape[: len(x.shape)]
+    return x.reshape(x.shape + (1,) * (len(target.shape) - len(x.shape)))
+
+
+def dict_apply_split(
+    x: Dict[str, torch.Tensor], split_func: Callable[[torch.Tensor], Dict[str, torch.Tensor]]
+) -> Dict[str, torch.Tensor]:
+    results = collections.defaultdict(dict)
+    for key, value in x.items():
+        result = split_func(value)
+        for k, v in result.items():
+            results[k][key] = v
+    return results
+
+
+def dict_apply_reduce(
+    x: List[Dict[str, torch.Tensor]], reduce_func: Callable[[List[torch.Tensor]], torch.Tensor]
+) -> Dict[str, torch.Tensor]:
+    result = dict()
+    for key in x[0].keys():
+        result[key] = reduce_func([x_[key] for x_ in x])
+    return result
+
+
+def optimizer_to(optimizer, device):
+    for state in optimizer.state.values():
+        for k, v in state.items():
+            if isinstance(v, torch.Tensor):
+                state[k] = v.to(device=device)
+    return optimizer
+
+
+def is_rank0() -> bool:
+    """
+    Best-effort check for main process without any synchronization.
+    """
+    # Prefer torch.distributed state if initialized.
+    if dist is not None and dist.is_available() and dist.is_initialized():
+        return dist.get_rank() == 0
+
+    # Fallback to environment variables commonly set by launchers.
+    for key in ("RANK", "SLURM_PROCID", "LOCAL_RANK"):
+        if key in os.environ:
+            return os.environ.get(key, "0") in ("0", "0\n", "")
+
+    return True
diff --git a/inference/policy/bagel_robotics_1/setup_eval_env_client.sh b/inference/policy/bagel_robotics_1/setup_eval_env_client.sh
new file mode 100644
index 0000000000000000000000000000000000000000..99681874d0a20dd1883821e3919381f082f8bf96
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/setup_eval_env_client.sh
@@ -0,0 +1,48 @@
+#!/bin/bash
+set -euo pipefail
+
+bench_name=${1}
+task_name=${2}
+ckpt_name=${3}
+env_cfg_type=${4}
+action_type=${5}
+seed=${6}
+env_gpu_id=${7}
+eval_env_conda_env=${8}
+additional_info=${9}
+policy_server_port=${10}
+policy_server_ip=${11:-"localhost"}
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+XPL_ROOT="$(cd "${SCRIPT_DIR}/../.." && pwd)"
+BENCH_ROOT="$(cd "${XPL_ROOT}/.." && pwd)"
+UTILS_DIR="${XPL_ROOT}/utils"
+
+policy_name="$(basename "${SCRIPT_DIR}")"
+yaml_file="${XPL_ROOT}/policy/${policy_name}/deploy.yml"
+
+echo "[CLIENT] policy=${policy_name}, task=${task_name}, server=${policy_server_ip}:${policy_server_port}"
+
+if [[ "${EVAL_ENV_TYPE:-sim}" == "debug" && -x "${eval_env_conda_env}/bin/python" ]]; then
+    export PYTHONPATH="${BENCH_ROOT}:${XPL_ROOT}${PYTHONPATH:+:${PYTHONPATH}}"
+    exec "${eval_env_conda_env}/bin/python" "${UTILS_DIR}/debug_env_client.py" \
+        --bench_name "${bench_name}" --task_name "${task_name}" \
+        --env_cfg_type "${env_cfg_type}" --policy_name "${policy_name}" \
+        --protocol ws --host "${policy_server_ip}" --port "${policy_server_port}" \
+        --eval_batch false --eval_episode_num "${DEBUG_EVAL_EPISODES:-1}"
+fi
+
+bash "${UTILS_DIR}/setup_env_client.sh" \
+    "${UTILS_DIR}" \
+    "${yaml_file}" \
+    "${eval_env_conda_env}" \
+    "${policy_server_port}" \
+    "${bench_name}" \
+    "${task_name}" \
+    "${env_cfg_type}" \
+    "${policy_name}" \
+    "${additional_info}" \
+    "${BENCH_ROOT}" \
+    "${seed}" \
+    "${env_gpu_id}" \
+    "${policy_server_ip}"
diff --git a/inference/policy/bagel_robotics_1/setup_eval_policy_server.sh b/inference/policy/bagel_robotics_1/setup_eval_policy_server.sh
new file mode 100644
index 0000000000000000000000000000000000000000..22c818c68e1233002bc4597b921a47c02c486fbf
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/setup_eval_policy_server.sh
@@ -0,0 +1,55 @@
+#!/bin/bash
+set -euo pipefail
+
+bench_name=${1}
+task_name=${2}
+ckpt_name=${3}
+env_cfg_type=${4}
+action_type=${5}
+seed=${6}
+policy_gpu_id=${7}
+policy_conda_env=${8}
+policy_server_port=${9}
+policy_server_host=${10:-"localhost"}
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+XPL_ROOT="$(cd "${SCRIPT_DIR}/../.." && pwd)"
+UTILS_DIR="${XPL_ROOT}/utils"
+
+policy_name="$(basename "${SCRIPT_DIR}")"
+yaml_file="${XPL_ROOT}/policy/${policy_name}/deploy.yml"
+
+echo "[SERVER] policy=${policy_name}, task=${task_name}, policy_server_port=${policy_server_port}"
+
+if [[ -x "${policy_conda_env}/bin/python" ]]; then
+    POLICY_PYTHON="${policy_conda_env}/bin/python"
+else
+    source "$(conda info --base)/etc/profile.d/conda.sh"
+    conda activate "${policy_conda_env}"
+    POLICY_PYTHON="$(command -v python)"
+fi
+BENCH_ROOT="$(cd "${XPL_ROOT}/.." && pwd)"
+export PYTHONPATH="${BENCH_ROOT}:${XPL_ROOT}${PYTHONPATH:+:${PYTHONPATH}}"
+
+SERVER_ENTRYPOINT="${XPL_ROOT}/setup_policy_server.py"
+if [[ -n "${BAGEL_VALIDATION_TRACE:-}" ]]; then
+    SERVER_ENTRYPOINT="${SCRIPT_DIR}/validation_server.py"
+fi
+
+exec env \
+    PYTHONHASHSEED=1 \
+    CUBLAS_WORKSPACE_CONFIG=:4096:8 \
+    PYTHONWARNINGS=ignore::UserWarning \
+    CUDA_VISIBLE_DEVICES="${policy_gpu_id}" \
+    "${POLICY_PYTHON}" "${SERVER_ENTRYPOINT}" \
+        --config_path "${yaml_file}" \
+        --overrides \
+            port="${policy_server_port}" \
+            host="${policy_server_host}" \
+            bench_name="${bench_name}" \
+            task_name="${task_name}" \
+            ckpt_name="${ckpt_name}" \
+            env_cfg_type="${env_cfg_type}" \
+            seed="${seed}" \
+            policy_name="${policy_name}" \
+            action_type="${action_type}"
diff --git a/inference/policy/bagel_robotics_1/source-provenance.json b/inference/policy/bagel_robotics_1/source-provenance.json
new file mode 100644
index 0000000000000000000000000000000000000000..7094072700656681a64c401ab4b4db57fcffdccd
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/source-provenance.json
@@ -0,0 +1,213 @@
+{
+  "reference_protocol": "83627b59eb149de1a7550edf5997418fa89b89905de1bd90ec9dbad66726af30",
+  "xpolicylab_base": "6baccd57ff958570d0db1588370289ee36c8bb70",
+  "files": [
+    {
+      "source": "__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "checkpoint_compat.py",
+      "source_sha256": "462c5f93812483b5871633b200bc2ffecb9c201a2caf6b14f8f36dbe95fba50a",
+      "packaged_sha256": "70e67d196cb46263677ebe873c5837657d6b8f90ca40cb6845f4782f5377ba2d"
+    },
+    {
+      "source": "data.py",
+      "source_sha256": "8574e1ae3c6ca6e4b84633ee1815f2b3f2ea17c1e55f12fda0b00c5c5a6a7b53",
+      "packaged_sha256": "8729d8e1386c74a0e3bd1c800a7c883725c9d00441f19bcbe7e40c2f8ff7af84"
+    },
+    {
+      "source": "datasets/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "datasets/lerobot/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "datasets/lerobot/processors/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "datasets/lerobot/processors/base_processor.py",
+      "source_sha256": "d3327cbe398d77761df960014d41e70d349bb0fa42bca874333b9ee970222447",
+      "packaged_sha256": "df79b044f93bcc9f4e93a2575ed5a7bc191595e3916753fcc26d3dd00db892c3"
+    },
+    {
+      "source": "datasets/lerobot/processors/vpp2_processor.py",
+      "source_sha256": "43c5604cbfc315b7377316cb97c9a6f3322124404397abfd01c6c3d0090cd414",
+      "packaged_sha256": "b00acaa43c3d18323f56ec9899b2de3ee5a7df8c2289ceea283725ca62aae773"
+    },
+    {
+      "source": "datasets/lerobot/transforms/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "datasets/lerobot/transforms/action_state_merger.py",
+      "source_sha256": "39fd44f254bd7fb71b9269d75d8f9b5a5bbbe167ea438dda92d40eda1d256734",
+      "packaged_sha256": "39fd44f254bd7fb71b9269d75d8f9b5a5bbbe167ea438dda92d40eda1d256734"
+    },
+    {
+      "source": "datasets/lerobot/transforms/image.py",
+      "source_sha256": "f494d262ce29770e959bdff60b997f7a07e5f00d10ef515eee7be813d1db038f",
+      "packaged_sha256": "f494d262ce29770e959bdff60b997f7a07e5f00d10ef515eee7be813d1db038f"
+    },
+    {
+      "source": "datasets/lerobot/utils/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "datasets/lerobot/utils/normalizer.py",
+      "source_sha256": "8207d813b698a91dfdbe6910a4aa1fe0e775a2a8edf5c05b2d7a42591e9f8adb",
+      "packaged_sha256": "28cedb2f4d69257a5f4da10e406dddba78ddd2a95408a9a55d6f74be5899d8dd"
+    },
+    {
+      "source": "models/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "models/wan21_14b/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "models/wan21_14b/action_dit.py",
+      "source_sha256": "9a654edd860ce77c1308923bd338c7f14d09f68686642531d7143c16b9f6f331",
+      "packaged_sha256": "780171144ff43f9282875af6c42a16db96b1f9ff56d01742a1024f7888fbb0fb"
+    },
+    {
+      "source": "models/wan21_14b/helpers/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "models/wan21_14b/helpers/compact_channels.py",
+      "source_sha256": "6f7ad787624844970985117002ab809bd777f7474969daca121e7101381c6d0c",
+      "packaged_sha256": "6f7ad787624844970985117002ab809bd777f7474969daca121e7101381c6d0c"
+    },
+    {
+      "source": "models/wan21_14b/helpers/gradient.py",
+      "source_sha256": "594249b6eb931048e352fd3ffef9b1ac1db9e4fc74659e04216d6b823dd174f3",
+      "packaged_sha256": "594249b6eb931048e352fd3ffef9b1ac1db9e4fc74659e04216d6b823dd174f3"
+    },
+    {
+      "source": "models/wan21_14b/helpers/io.py",
+      "source_sha256": "1b68cf2f2fd674c7a59140f8632d4464044bffc2e9277ef5f9a975e7c699966a",
+      "packaged_sha256": "39cb43aafb25c9046d19921cdddc6df2bae2b3eb267f5f1c271254dd216908e5"
+    },
+    {
+      "source": "models/wan21_14b/helpers/loader.py",
+      "source_sha256": "b9180b9c9177af192e5cb92387b4a09bd0dbc08fb75a94e843b67d315c94e339",
+      "packaged_sha256": "ce632db536b479c96a7160968eb314deb4e2dd1ab1b986e68f5dd8dea0ed6468"
+    },
+    {
+      "source": "models/wan21_14b/mot.py",
+      "source_sha256": "cc14a909bfb181a441bc9d8d6de12f3c31d89f3dc3415031e34a5f8c21410019",
+      "packaged_sha256": "5f42547a8571a2758ea6d4285834d5eaafc67285c6db42d44ec3c317e8402cc2"
+    },
+    {
+      "source": "models/wan21_14b/schedulers/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "models/wan21_14b/schedulers/scheduler_continuous.py",
+      "source_sha256": "15de04574c1d8ccb794ca6560b1e501e5c83cc86104bb80f1eb3cc3b0324ffe0",
+      "packaged_sha256": "15de04574c1d8ccb794ca6560b1e501e5c83cc86104bb80f1eb3cc3b0324ffe0"
+    },
+    {
+      "source": "models/wan21_14b/vpp2.py",
+      "source_sha256": "8c419af905a8ffa93d77c827a283b7aeff744f3a461c390783eef8a333c31756",
+      "packaged_sha256": "41e8632ad7898bf3a63cf7c874179adae609583ceb1b4dadf2b4c22669865323"
+    },
+    {
+      "source": "models/wan21_14b/wan21.py",
+      "source_sha256": "01b45cc71f69a6914f7b286559317c8e8485d0b0fec93bb28ebc37c183c295f4",
+      "packaged_sha256": "d203d5ec0f6c913b052e22202eff0fe238e087040ae0de9843b7a22a5533b5f3"
+    },
+    {
+      "source": "models/wan21_14b/wan_video_clip.py",
+      "source_sha256": "f65e44c9d1b7811d2996b3fa22e6fe4c2ac5e6edf1fc2c14251cf582bf8d1f05",
+      "packaged_sha256": "f65e44c9d1b7811d2996b3fa22e6fe4c2ac5e6edf1fc2c14251cf582bf8d1f05"
+    },
+    {
+      "source": "models/wan21_14b/wan_video_dit.py",
+      "source_sha256": "0d1b5837ea923508c05394d56aaa465813e7b4f5aea3b31ceceae39b2de9b0f9",
+      "packaged_sha256": "516b8dd09d576898c8760968ada4b617f5fe6a7c3fd945787898ae326df8e9a1"
+    },
+    {
+      "source": "models/wan21_14b/wan_video_text_encoder.py",
+      "source_sha256": "902a1e46b21b5c48e166fe03defabb94b3773c6bbe836fc09b9a0d0894eb1f6e",
+      "packaged_sha256": "902a1e46b21b5c48e166fe03defabb94b3773c6bbe836fc09b9a0d0894eb1f6e"
+    },
+    {
+      "source": "models/wan21_14b/wan_video_vae.py",
+      "source_sha256": "33f7d804677b44a6f3ab334684017dcdb1e9dfb748279de76ad94a5bb71a7e51",
+      "packaged_sha256": "33f7d804677b44a6f3ab334684017dcdb1e9dfb748279de76ad94a5bb71a7e51"
+    },
+    {
+      "source": "policy.py",
+      "source_sha256": "2a7696cd73ccb99575bf901185cf578c3775945f61f6e67aeb2b4ec6bfc4e30b",
+      "packaged_sha256": "35ca975c19a1290531c0027e5cf849ceabee724ff771ed4b49dd653c125c8deb"
+    },
+    {
+      "source": "runtime.py",
+      "source_sha256": "e26885a3732f5fdac3d8381da870a1673d0c03354f02f9deee7fe779d645f8a7",
+      "packaged_sha256": "dcf2942f3b9a02590902212b3c4f24b297c485ed0be65702ee540529dd4f55c0"
+    },
+    {
+      "source": "utils/__init__.py",
+      "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+      "packaged_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+    },
+    {
+      "source": "utils/logging_config.py",
+      "source_sha256": "1de4abf66d9d201b49e7bff48291643b9dc5ec9b6e7918ca73e1bd18887b1c43",
+      "packaged_sha256": "dd3afea396e0ba3e619a1a66121f3c41dd0ca4174e1cf1f9f459b200f23e2690"
+    },
+    {
+      "source": "utils/pytorch_utils.py",
+      "source_sha256": "9b0ba19de9a64458b12a0fb24f0811f49d9e9fca5d8b815f5f5ecd3b7886fd63",
+      "packaged_sha256": "9b0ba19de9a64458b12a0fb24f0811f49d9e9fca5d8b815f5f5ecd3b7886fd63"
+    }
+  ],
+  "packaging_changes": [
+    "Namespace isolation: vpp2 -> bagel_cp1000",
+    "Inference factory extraction; no trainer/data-processing imports",
+    "Compact-only strict loader; no dense/legacy fallback",
+    "Pinned artifact/config validation and shared robot-dimension lookup",
+    "Full reset for one-client deployment; stop actions at episode termination",
+    "Default replan corrected to 16; sampling/history/ensemble math retained",
+    "Recorded deterministic backend controls applied at startup",
+    "Remove unreachable adapter branches for unsupported action spaces; preserve EE16 math",
+    "Restore evaluated processor, image transform and action/state merger dependencies under the isolated namespace",
+    "Recursively map serialized Hydra targets to bagel_cp1000; preserve tensor keys and all non-target configuration values"
+  ],
+  "adapter_files": [
+    {
+      "source": "simulator/XPolicyLab/policy/VPP2/model.py",
+      "source_sha256": "699acc0c459c5997db4e3155e2b1b7eb64bf750552436074a369c4eb6e9ffef9",
+      "packaged_path": "model.py",
+      "packaged_sha256": "56418464b2a88e7aea31d436ee42851dfababad97db9178abdce318b5a7e621b"
+    },
+    {
+      "source": "simulator/XPolicyLab/policy/VPP2/deploy.py",
+      "source_sha256": "ab94a8569e6168ab7cde6660890035e11b34961ff2648854a7b61d188c555b6e",
+      "packaged_path": "deploy.py",
+      "packaged_sha256": "68f337b21b1e517e1df27550e64fbc542b644528adc8f3f47fb8e44849185af6"
+    },
+    {
+      "source": "simulator/XPolicyLab/policy/VPP2/temporal_ensemble.py",
+      "source_sha256": "42be4b09c56377f4686b2e722de760fa3bafd1c6b92605ada269d94e19efc95e",
+      "packaged_path": "temporal_ensemble.py",
+      "packaged_sha256": "42be4b09c56377f4686b2e722de760fa3bafd1c6b92605ada269d94e19efc95e"
+    }
+  ]
+}
diff --git a/inference/policy/bagel_robotics_1/temporal_ensemble.py b/inference/policy/bagel_robotics_1/temporal_ensemble.py
new file mode 100644
index 0000000000000000000000000000000000000000..8272b786c349436ba8ad29a1fbd7638a876edccc
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/temporal_ensemble.py
@@ -0,0 +1,45 @@
+"""Blend aligned absolute EE16 predictions, keeping the new chunk dominant."""
+import numpy as np
+
+
+class TemporalEnsemble:
+    def __init__(self, horizon=32, replan_steps=16, new_weight=0.9, old_weight=0.1):
+        if (horizon, replan_steps) != (32, 16):
+            raise ValueError('This two-chunk ensemble requires horizon 32 and replan 16')
+        if (new_weight, old_weight) != (0.9, 0.1):
+            raise ValueError('This evaluation requires new=0.9, old=0.1')
+        self.new_weight, self.old_weight = new_weight, old_weight
+        self.previous = None
+        self.step = 0
+        self.last_info = None
+
+    def apply(self, chunk):
+        chunk = np.asarray(chunk, dtype=np.float32)
+        if chunk.shape != (32, 16) or not np.isfinite(chunk).all():
+            raise ValueError('Expected a finite [32,16] action prediction')
+        new = chunk[:16]
+        old = None if self.previous is None else self.previous[16:32]
+        if old is None:
+            output = new.copy()
+        else:
+            output = self.new_weight * new + self.old_weight * old
+            for sl in (slice(3, 7), slice(11, 15)):
+                qnew, qold = new[:, sl].copy(), old[:, sl].copy()
+                nnew = np.linalg.norm(qnew, axis=-1, keepdims=True)
+                nold = np.linalg.norm(qold, axis=-1, keepdims=True)
+                if np.any(nnew < 1e-8) or np.any(nold < 1e-8):
+                    raise ValueError('Cannot ensemble a zero-norm quaternion')
+                qnew /= nnew
+                qold /= nold
+                qold *= np.where(np.sum(qnew * qold, axis=-1, keepdims=True) < 0, -1, 1)
+                blended = self.new_weight * qnew + self.old_weight * qold
+                output[:, sl] = blended / np.linalg.norm(blended, axis=-1, keepdims=True)
+        self.last_info = dict(start_step=self.step, horizon=32, executed=16,
+                              new_weight=self.new_weight, old_weight=self.old_weight,
+                              overlap=old is not None, new_chunk=chunk.copy(),
+                              old_aligned=None if old is None else old.copy(),
+                              output=output.copy())
+        # Keep raw predictions: blending must not recursively change the old target.
+        self.previous = chunk.copy()
+        self.step += 16
+        return output
diff --git a/inference/policy/bagel_robotics_1/tests/fixtures/checkpoint_processor.json b/inference/policy/bagel_robotics_1/tests/fixtures/checkpoint_processor.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bcaf73176881ad84bf4803873e14eead843e23e
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/tests/fixtures/checkpoint_processor.json
@@ -0,0 +1,59 @@
+{
+  "format": "vpp2_wan21_action_v1",
+  "processor": {
+    "_target_": "vpp2.datasets.lerobot.processors.vpp2_processor.VPP2Processor",
+    "shape_meta": {
+      "images": [
+        {
+          "key": "image",
+          "raw_shape": [
+            3,
+            480,
+            880
+          ],
+          "shape": [
+            3,
+            480,
+            880
+          ]
+        }
+      ],
+      "action": [
+        {
+          "key": "default",
+          "raw_shape": 16,
+          "shape": 16
+        }
+      ],
+      "state": [
+        {
+          "key": "default",
+          "raw_shape": 16,
+          "shape": 16
+        }
+      ]
+    },
+    "num_obs_steps": 65,
+    "num_output_cameras": 1,
+    "action_output_dim": 16,
+    "proprio_output_dim": 16,
+    "action_state_transforms": null,
+    "use_stepwise_action_norm": false,
+    "norm_default_mode": "z-score",
+    "norm_exception_mode": null,
+    "action_state_merger": {
+      "_target_": "vpp2.datasets.lerobot.transforms.action_state_merger.ConcatLeftAlign"
+    },
+    "train_transforms": [
+      {
+        "_target_": "vpp2.datasets.lerobot.transforms.image.ToTensor"
+      }
+    ],
+    "val_transforms": [
+      {
+        "_target_": "vpp2.datasets.lerobot.transforms.image.ToTensor"
+      }
+    ],
+    "num_image_obs_steps": 17
+  }
+}
diff --git a/inference/policy/bagel_robotics_1/tests/test_backend.py b/inference/policy/bagel_robotics_1/tests/test_backend.py
new file mode 100644
index 0000000000000000000000000000000000000000..4cd23a304bab640cb7159bf9a7972af5036a71ca
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/tests/test_backend.py
@@ -0,0 +1,47 @@
+"""Check required launch controls without touching a GPU."""
+from pathlib import Path
+import json
+import os
+import subprocess
+import sys
+
+import pytest
+
+
+def test_backend_refuses_missing_startup_controls():
+    policy = Path(__file__).resolve().parents[1]
+    source = "from backend import configure_backend; configure_backend(1)"
+    env = dict(os.environ, PYTHONPATH=str(policy), CUDA_VISIBLE_DEVICES="")
+    env.pop("PYTHONHASHSEED", None)
+    result = subprocess.run([sys.executable, "-c", source], env=env, text=True, capture_output=True)
+    assert result.returncode != 0 and "PYTHONHASHSEED=1" in result.stderr
+    env["PYTHONHASHSEED"] = "1"
+    env.pop("CUBLAS_WORKSPACE_CONFIG", None)
+    result = subprocess.run([sys.executable, "-c", source], env=env, text=True, capture_output=True)
+    assert result.returncode != 0 and "CUBLAS_WORKSPACE_CONFIG" in result.stderr
+
+
+def test_backend_cpu_configuration_and_seed_reproducibility():
+    policy = Path(__file__).resolve().parents[1]
+    source = """
+import json, random
+import numpy as np
+import torch
+from backend import configure_backend
+report = configure_backend(42)
+first = (random.random(), float(np.random.random()), float(torch.rand(1)))
+configure_backend(42)
+second = (random.random(), float(np.random.random()), float(torch.rand(1)))
+assert first == second
+assert torch.backends.cudnn.deterministic and not torch.backends.cudnn.benchmark
+assert torch.backends.cudnn.allow_tf32 and not torch.backends.cuda.matmul.allow_tf32
+assert not torch.is_deterministic_algorithms_warn_only_enabled()
+print(json.dumps(report))
+"""
+    env = dict(os.environ, PYTHONPATH=str(policy), CUDA_VISIBLE_DEVICES="",
+               PYTHONHASHSEED="1", CUBLAS_WORKSPACE_CONFIG=":4096:8")
+    result = subprocess.run([sys.executable, "-c", source], env=env, text=True, capture_output=True)
+    assert result.returncode == 0, result.stderr
+    report = json.loads(result.stdout)
+    assert report["deterministic_algorithms"] is True
+    assert report["float32_matmul_precision"] == "highest"
diff --git a/inference/policy/bagel_robotics_1/tests/test_checkpoint_processor.py b/inference/policy/bagel_robotics_1/tests/test_checkpoint_processor.py
new file mode 100644
index 0000000000000000000000000000000000000000..cb1fcfe066ae082c44870a83016ce5480d0e5d93
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/tests/test_checkpoint_processor.py
@@ -0,0 +1,67 @@
+"""Load real checkpoint processor metadata without importing the research package."""
+import copy
+import importlib.abc
+import json
+from pathlib import Path
+import sys
+
+from hydra.utils import instantiate
+import pytest
+import torch
+
+from bagel_cp1000.checkpoint_compat import action_config
+
+FIXTURE = json.loads((Path(__file__).parent / 'fixtures/checkpoint_processor.json').read_text())
+
+
+class RejectResearchImports(importlib.abc.MetaPathFinder):
+    def find_spec(self, fullname, path=None, target=None):
+        if fullname.split('.')[0] in {'vpp2', 'fastwam', 'robodojo_joint2b'}:
+            raise AssertionError(f'Standalone processor requested research import: {fullname}')
+
+
+def replace_targets(value, prefix):
+    if isinstance(value, list):
+        return [replace_targets(x, prefix) for x in value]
+    if not isinstance(value, dict):
+        return value
+    return {key: (item.replace('vpp2.', prefix, 1) if key == '_target_' else replace_targets(item, prefix))
+            for key, item in value.items()}
+
+
+@pytest.mark.parametrize('prefix', ['vpp2.', 'fastwam.', 'robodojo_joint2b.'])
+def test_real_checkpoint_processor_instantiates_without_research_imports(prefix):
+    config = replace_targets(copy.deepcopy(FIXTURE['processor']), prefix)
+    original = copy.deepcopy(config)
+    payload = {'format': FIXTURE['format'], 'resolved_config': {'processor': config}}
+    guard = RejectResearchImports()
+    sys.meta_path.insert(0, guard)
+    try:
+        processor = instantiate(action_config(payload)['processor']).eval()
+        assert type(processor).__module__.startswith('bagel_cp1000.')
+        assert type(processor.action_state_merger).__module__.startswith('bagel_cp1000.')
+        assert processor.action_output_dim == processor.proprio_output_dim == 16
+        # EE16 inference must preserve values/order before normalization.
+        state = torch.arange(16, dtype=torch.float32).reshape(1, 16)
+        action = torch.arange(32 * 16, dtype=torch.float32).reshape(32, 16)
+        batch = {'state': {'default': state.clone()}, 'action': {'default': action.clone()}}
+        result = processor.action_state_merger.forward(processor.action_state_transform(batch))
+        torch.testing.assert_close(result['state'], state, rtol=0, atol=0)
+        torch.testing.assert_close(result['action'], action, rtol=0, atol=0)
+        assert not result['action_dim_is_pad'].any()
+        pixels = torch.tensor([0, 127, 255], dtype=torch.uint8)
+        torch.testing.assert_close(processor.val_transforms[0](pixels), pixels.float() / 255, rtol=0, atol=0)
+        assert config == original  # serialized metadata must not be mutated
+    finally:
+        sys.meta_path.remove(guard)
+
+
+def test_target_isolation_preserves_non_target_strings_and_input():
+    source = {'_target_': 'fastwam.datasets.lerobot.processors.fastwam_processor.FastWAMProcessor',
+              'path': 'vpp2/checkpoint.pt', 'prompt': 'vpp2.example',
+              'nested': [{'_target_': 'torch.nn.Identity'}]}
+    expected = copy.deepcopy(source)
+    result = action_config({'format': 'fastwam_wan21_action_v1', 'resolved_config': source})
+    assert result['_target_'] == 'bagel_cp1000.datasets.lerobot.processors.vpp2_processor.VPP2Processor'
+    assert result['path'] == expected['path'] and result['prompt'] == expected['prompt']
+    assert result['nested'] == expected['nested'] and source == expected
diff --git a/inference/policy/bagel_robotics_1/tests/test_contract.py b/inference/policy/bagel_robotics_1/tests/test_contract.py
new file mode 100644
index 0000000000000000000000000000000000000000..155de51d6481a963fdd4d0366b33fe896d75d302
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/tests/test_contract.py
@@ -0,0 +1,162 @@
+"""CPU checks of scientifically important packaging contracts."""
+from pathlib import Path
+import copy
+import importlib
+import json
+import sys
+from types import SimpleNamespace
+
+import numpy as np
+import pytest
+import torch
+
+POLICY = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(POLICY.parents[2]))
+from XPolicyLab.policy.bagel_robotics_1 import assets, deploy
+from XPolicyLab.policy.bagel_robotics_1.temporal_ensemble import TemporalEnsemble
+from bagel_cp1000.models.wan21_14b.helpers.compact_channels import build
+
+
+def observation(step=0):
+    return {"env_idx": 0, "instruction": "stack blocks",
+            "vision": {name: {"color": np.full((24, 32, 3), step, dtype=np.uint8)}
+                       for name in ["cam_head", "cam_left_wrist", "cam_right_wrist"]},
+            "state": {f"{side}_{key}": np.array(value, np.float32)
+                      for side in ["left", "right"]
+                      for key, value in [("ee_pose", [step, 0, 0, 1, 0, 0, 0]),
+                                         ("ee_joint_state", [0.2])]}}
+
+
+def test_frozen_settings_and_seed_override():
+    cfg = assets.validate_settings({"seed": 42})
+    assert cfg["seed"] == 42 and cfg["replan_steps"] == 16
+    for key, value in [("replan_steps", 24), ("num_inference_steps", 8),
+                       ("mixed_precision", "fp16"), ("eval_batch", True),
+                       ("env_cfg_type", "piper"), ("seed", -1)]:
+        with pytest.raises(ValueError):
+            assets.validate_settings({key: value})
+
+
+def test_blending_uses_aligned_raw_predictions_and_quaternion_hemisphere():
+    def chunk(value):
+        x = np.full((32, 16), value, np.float32)
+        x[:, 3:7] = [1, 0, 0, 0]
+        x[:, 11:15] = [1, 0, 0, 0]
+        return x
+    ensemble = TemporalEnsemble()
+    first = chunk(1)
+    first[16:, 0] = 4
+    assert np.array_equal(ensemble.apply(first), first[:16])
+    second = chunk(2)
+    second[:, 3:7] *= -1
+    result = ensemble.apply(second)
+    assert np.allclose(result[:, 0], 2.2)
+    assert np.allclose(np.linalg.norm(result[:, 3:7], axis=-1), 1)
+    third = ensemble.apply(chunk(3))
+    assert np.allclose(third[:, 0], 2.9)  # previous raw chunk, not previous blend
+
+
+def test_actual_architecture_on_meta_device():
+    # Optional real manifest from metadata download; requires no model tensors/GPU.
+    path = POLICY / "checkpoints/cp1000/compact_architecture.json"
+    if not path.exists():
+        pytest.skip("Run download_checkpoint.py --metadata-only to check the real architecture")
+    spec = json.loads(path.read_text())
+    model = build(spec, device="meta")
+    assert len(model.blocks) == 40
+    assert sum(p.numel() for p in model.parameters()) == 12468409664
+    assert [b.ffn[0].out_features for b in model.blocks] == [len(x) for x in spec["selection"]["ffn_indices"]]
+    assert [b.cross_attn.num_heads for b in model.blocks] == [len(x) for x in spec["selection"]["cross_head_indices"]]
+    assert not any("mask" in name for name, _ in model.named_parameters())
+    malformed = copy.deepcopy(spec)
+    malformed["selection"]["ffn_indices"][0][1] = malformed["selection"]["ffn_indices"][0][0]
+    with pytest.raises(ValueError, match="Invalid per-layer"):
+        build(malformed, device="meta")
+
+
+def test_tampered_checkpoint_rejected_before_tensor_loading(tmp_path, monkeypatch):
+    content = b"approved tensor bytes"
+    (tmp_path / "video.pt").write_bytes(content)
+    manifest = {"files": {"video.pt": {"size": len(content), "sha256": "0" * 64}}}
+    monkeypatch.setattr(assets, "MANIFEST", manifest)
+    with pytest.raises(ValueError, match="SHA-256"):
+        assets.verify_bundle(tmp_path)
+    (tmp_path / "video.pt").write_bytes(content[:3])
+    with pytest.raises(ValueError, match="size mismatch"):
+        assets.verify_bundle(tmp_path)
+
+
+class FakePolicy:
+    def __init__(self):
+        self.action_horizon, self.replan_steps = 32, 16
+        self.history_condition_frames = 8
+        self._history_step_by_env = {}
+        self.model = SimpleNamespace(video_expert=SimpleNamespace(
+            parameters=lambda: [SimpleNamespace(numel=lambda: 12468409664)],
+            named_parameters=lambda: []))
+
+    def observe(self, obs):
+        key = obs["env_idx"]
+        self._history_step_by_env[key] = self._history_step_by_env.get(key, -1) + 1
+
+    def _infer_action_chunk(self, obs, instruction):
+        return np.repeat(obs["joint_action"]["vector"][None], 32, axis=0)
+
+    def reset(self):
+        self._history_step_by_env.clear()
+
+    def reset_env(self, key):
+        self._history_step_by_env.pop(key, None)
+
+
+def make_adapter(monkeypatch):
+    module = importlib.import_module("XPolicyLab.policy.bagel_robotics_1.model")
+    runtime = importlib.import_module("bagel_cp1000.policy")
+    cfg = assets.validate_settings({"checkpoint_path": "/unused/action.pt", "dataset_stats_path": "/unused/stats.json"})
+    monkeypatch.setattr(module, "prepare_model_config", lambda c, p: cfg)
+    monkeypatch.setattr(module, "configure_backend", lambda seed: {})
+    monkeypatch.setattr(module, "get_robot_action_dim_info", lambda _: {"arm_dim": [6, 6], "ee_dim": [1, 1]})
+    monkeypatch.setattr(runtime, "get_model", lambda _: FakePolicy())
+    return module.Model(cfg)
+
+
+def test_reset_restores_first_episode_behavior_with_client_ids(monkeypatch):
+    model = make_adapter(monkeypatch)
+    obs = observation()
+    obs[model.CLIENT_ID_FIELD] = "client-a"
+    model.update_obs(obs)
+    initial = model.get_action({model.CLIENT_ID_FIELD: "client-a"})
+    assert len(initial) == 16
+    assert set(initial[0]) == {"left_ee_pose", "right_ee_pose", "left_ee_joint_state", "right_ee_joint_state"}
+    model.reset()
+    assert not model._ensemble_by_virtual and not model.model._history_step_by_env
+    model.update_obs(obs)
+    after = model.get_action({model.CLIENT_ID_FIELD: "client-a"})
+    for a, b in zip(initial, after):
+        for key in a:
+            np.testing.assert_array_equal(a[key], b[key])
+    assert model.history_guard_mismatches == 0
+
+
+def test_deployment_records_each_step_and_stops_at_episode_end(monkeypatch):
+    model = make_adapter(monkeypatch)
+    class Environment:
+        step = 0
+        observations = []
+        def get_obs(self):
+            self.observations.append(self.step)
+            return observation(self.step)
+        def is_episode_end(self):
+            return self.step >= 20
+        def take_action(self, action):
+            assert self.step < 20
+            self.step += 1
+    class Client:
+        def call(self, func_name, **kwargs):
+            return getattr(model, func_name)(**kwargs)
+    env = Environment()
+    deploy.eval_one_episode(env, Client())
+    assert env.step == 20
+    assert env.observations == list(range(20))
+    assert model.history_guard_mismatches == 0
+    assert model._last_ensemble["overlap"] is True
diff --git a/inference/policy/bagel_robotics_1/tests/test_validation.py b/inference/policy/bagel_robotics_1/tests/test_validation.py
new file mode 100644
index 0000000000000000000000000000000000000000..3926c18fc8ca257decdb21415aad962b0b259126
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/tests/test_validation.py
@@ -0,0 +1,52 @@
+"""Protect evidence integrity and detect numerical differences, without a GPU."""
+import json
+from pathlib import Path
+import sys
+
+import numpy as np
+import pytest
+
+POLICY = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(POLICY))
+from validation import decode, encode
+from validation_compare import compare
+
+
+def test_lossless_camera_capture_rejects_corruption_and_missing_arrays(tmp_path):
+    frame = np.arange(180, dtype=np.uint8).reshape(6, 10, 3)
+    encoded = encode({"camera": frame, "state": np.array([0.125], np.float32), "scalar": np.array(3, np.int64)}, tmp_path)
+    restored = decode(encoded, tmp_path)
+    np.testing.assert_array_equal(restored["camera"], frame)
+    assert restored["scalar"].shape == () and restored["scalar"].item() == 3
+    image_path = tmp_path / (encoded["camera"]["array"] + ".npy")
+    np.save(image_path, frame + 1, allow_pickle=False)
+    with pytest.raises(ValueError, match="integrity"):
+        decode(encoded, tmp_path)
+    image_path.unlink()
+    with pytest.raises(FileNotFoundError):
+        decode(encoded, tmp_path)
+
+
+def test_comparison_detects_single_action_change_and_missing_events(tmp_path):
+    left, right = tmp_path / "left", tmp_path / "right"
+    for root in [left, right]:
+        root.mkdir()
+        row = {"kind": "commands", "data": encode(np.zeros((16, 16), np.float32), root / "arrays")}
+        (root / "trace.jsonl").write_text(json.dumps(row) + "\n")
+    assert compare(left, right)["passed"]
+    changed = np.zeros((16, 16), np.float32)
+    changed[7, 2] = 0.25
+    row["data"] = encode(changed, right / "arrays")
+    (right / "trace.jsonl").write_text(json.dumps(row) + "\n")
+    report = compare(left, right)
+    assert not report["passed"] and report["max_abs_stored_array_difference"] == 0.25
+    (right / "trace.jsonl").write_text("")
+    assert not compare(left, right)["passed"]
+
+
+def test_empty_evidence_is_not_accepted(tmp_path):
+    for name in ["a", "b"]:
+        root = tmp_path / name
+        root.mkdir()
+        (root / "trace.jsonl").write_text("")
+    assert not compare(tmp_path / "a", tmp_path / "b")["passed"]
diff --git a/inference/policy/bagel_robotics_1/validation.py b/inference/policy/bagel_robotics_1/validation.py
new file mode 100644
index 0000000000000000000000000000000000000000..8681dc90a8db4c004da1c02b59234c0e448455e0
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/validation.py
@@ -0,0 +1,176 @@
+"""Opt-in lossless capture; default inference does not import this module."""
+from collections import deque
+from functools import wraps
+import hashlib
+import json
+from pathlib import Path
+import sys
+
+import numpy as np
+
+
+def encode(value, arrays=None):
+    torch = sys.modules.get("torch")
+    if torch is not None and isinstance(value, torch.Tensor):
+        value = value.detach().cpu().contiguous()
+        if value.dtype == torch.bfloat16:
+            return {"bfloat16_bits": encode(value.view(torch.uint16).numpy(), arrays)}
+        value = value.numpy()
+    if isinstance(value, np.ndarray):
+        value = np.array(value, copy=True, order="C")
+        if value.dtype.hasobject:
+            raise TypeError("Object arrays are not valid replay evidence")
+        header = json.dumps([value.dtype.str, list(value.shape)]).encode()
+        digest = hashlib.sha256(header + value.tobytes()).hexdigest()
+        if arrays is not None:
+            arrays = Path(arrays)
+            arrays.mkdir(parents=True, exist_ok=True)
+            target = arrays / (digest + ".npy")
+            if not target.exists():
+                with target.open("xb") as stream:
+                    np.save(stream, value, allow_pickle=False)
+        return {"array": digest, "dtype": value.dtype.str, "shape": list(value.shape)}
+    if isinstance(value, np.generic):
+        return encode(value.item(), arrays)
+    if isinstance(value, dict):
+        return {str(k): encode(v, arrays) for k, v in value.items()}
+    if isinstance(value, (list, tuple, deque)):
+        return [encode(v, arrays) for v in value]
+    if value is None or isinstance(value, (str, bool, int, float)):
+        return value
+    raise TypeError(f"Unsupported evidence type: {type(value).__name__}")
+
+
+def decode(value, arrays):
+    if isinstance(value, dict) and "array" in value:
+        digest = value["array"]
+        if len(digest) != 64 or any(c not in "0123456789abcdef" for c in digest):
+            raise ValueError("Invalid array identifier")
+        result = np.load(Path(arrays) / (digest + ".npy"), allow_pickle=False)
+        if encode(result) != value:
+            raise ValueError("Replay array integrity check failed")
+        return result
+    if isinstance(value, dict):
+        return {k: decode(v, arrays) for k, v in value.items()}
+    if isinstance(value, list):
+        return [decode(v, arrays) for v in value]
+    return value
+
+
+def records(directory):
+    with (Path(directory) / "trace.jsonl").open() as stream:
+        return [json.loads(line) for line in stream]
+
+
+def history_state(model):
+    # Normalize virtual IDs: routing bookkeeping is not a numerical input.
+    native = model.model
+    return {name: [state[k] for k in sorted(state)]
+            for name in native._PER_ENV_STATE
+            for state in [getattr(native, name)]}
+
+
+class Recorder:
+    def __init__(self, model, root, limit=209):
+        self.model, self.root, self.limit = model, Path(root), limit
+        self.root.mkdir(parents=True, exist_ok=False)
+        self.episode = -1
+        self.stream = None
+        self.observations = self.replans = 0
+        self.directory = None
+
+    @property
+    def active(self):
+        return self.stream is not None and self.observations <= self.limit
+
+    def event(self, kind, data, *, save=False):
+        if self.stream is not None:
+            row = {"kind": kind, "data": encode(data, self.directory / "arrays" if save else None)}
+            self.stream.write(json.dumps(row, sort_keys=True, allow_nan=False) + "\n")
+            self.stream.flush()
+
+    def start(self):
+        if self.stream is not None:
+            self.stream.close()
+        self.episode += 1
+        self.directory = self.root / f"episode-{self.episode:03d}"
+        self.directory.mkdir()
+        self.stream = (self.directory / "trace.jsonl").open("x")
+        self.observations = self.replans = 0
+
+    def install(self):
+        model = self.model
+        original_reset = model.reset
+        original_client_reset = model.reset_client
+        original_update = model.update_obs
+        original_action = model.get_action
+        original_infer = model.model.model.infer_action
+
+        @wraps(original_reset)
+        def reset():
+            result = original_reset()
+            self.start()
+            return result
+
+        @wraps(original_client_reset)
+        def reset_client(obs=None):
+            result = original_client_reset(obs=obs)
+            if self.stream is None:
+                self.start()
+            sizes = {name: len(getattr(model.model, name)) for name in model.model._PER_ENV_STATE}
+            adapter_sizes = {name: len(getattr(model, name)) for name in (
+                "_last_infer_step", "_infer_count_by_virtual", "_ensemble_by_virtual",
+                "_last_obs_by_client", "_last_instruction_by_client", "_last_virtual_by_client")}
+            if any(sizes.values()) or any(adapter_sizes.values()) or model.last_obs is not None:
+                raise RuntimeError("Episode reset leaked numerical state")
+            self.event("reset", {"native": sizes, "adapter": adapter_sizes})
+            return result
+
+        @wraps(original_update)
+        def update_obs(obs):
+            if self.stream is None:
+                self.start()
+            result = original_update(obs)
+            if self.observations < self.limit:
+                numerical = {k: v for k, v in obs.items() if k != model.CLIENT_ID_FIELD}
+                self.event("observation", {"step": self.observations, "observation": numerical}, save=True)
+            self.observations += 1
+            return result
+
+        @wraps(original_infer)
+        def infer_action(**kwargs):
+            if self.active:
+                self.event("inputs", kwargs)
+                self.event("history", {"state": history_state(model),
+                           "mature": any(s >= 175 for s in model.model._history_step_by_env.values())})
+            result = original_infer(**kwargs)
+            if self.active:
+                self.event("prediction", result["action"], save=True)
+            return result
+
+        @wraps(original_action)
+        def get_action(obs=None):
+            result = original_action(obs=obs)
+            if model.history_guard_mismatches:
+                raise RuntimeError("Observation history cadence mismatch")
+            if self.active:
+                self.event("ensemble", model._last_ensemble, save=True)
+                self.event("commands", {"replan": self.replans,
+                           "observations": self.observations, "actions": result}, save=True)
+                (self.directory / "coverage.json").write_text(json.dumps({
+                    "observations": self.observations, "action_chunks": self.replans + 1,
+                    "history_mature": any(s >= 175 for s in model.model._history_step_by_env.values()),
+                    "capture_limit": self.limit}, indent=2) + "\n")
+            self.replans += 1
+            return result
+
+        model.reset = reset
+        model.reset_client = reset_client
+        model.update_obs = update_obs
+        model.get_action = get_action
+        model.model.model.infer_action = infer_action
+        return self
+
+
+def attach(model, root):
+    return Recorder(model, root).install()
diff --git a/inference/policy/bagel_robotics_1/validation_compare.py b/inference/policy/bagel_robotics_1/validation_compare.py
new file mode 100644
index 0000000000000000000000000000000000000000..c1e949e05a86145be9019bb8ac551236a365c89b
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/validation_compare.py
@@ -0,0 +1,88 @@
+"""Exact comparison of captured or replayed model evidence."""
+import argparse
+import json
+from pathlib import Path
+
+import numpy as np
+
+from validation import decode, records
+
+KINDS = {"reset", "observation", "inputs", "history", "prediction", "ensemble", "commands"}
+
+
+def compare(left, right):
+    left, right = Path(left), Path(right)
+    a = [r for r in records(left) if r["kind"] in KINDS]
+    b = [r for r in records(right) if r["kind"] in KINDS]
+    differences = []
+    max_abs = 0.0
+
+    def visit(x, y, path):
+        nonlocal max_abs
+        if x == y:
+            return
+        if isinstance(x, dict) and isinstance(y, dict):
+            if "array" in x and "array" in y:
+                item = {"path": path, "left": x, "right": y}
+                # Inputs/history store fingerprints only; actions are lossless.
+                if (left / "arrays" / (x["array"] + ".npy")).exists() and (right / "arrays" / (y["array"] + ".npy")).exists():
+                    av, bv = decode(x, left / "arrays"), decode(y, right / "arrays")
+                    if av.shape == bv.shape and av.dtype == bv.dtype:
+                        delta = float(np.max(np.abs(av.astype(np.float64) - bv.astype(np.float64)), initial=0))
+                        item["max_abs_difference"] = delta
+                        max_abs = max(max_abs, delta)
+                differences.append(item)
+                return
+            if set(x) != set(y):
+                differences.append({"path": path, "reason": "different keys"})
+                return
+            for key in x:
+                visit(x[key], y[key], path + "/" + key)
+        elif isinstance(x, list) and isinstance(y, list) and len(x) == len(y):
+            for i, (av, bv) in enumerate(zip(x, y)):
+                visit(av, bv, path + "/" + str(i))
+        else:
+            differences.append({"path": path, "left": x, "right": y})
+
+    if not a or not b:
+        differences.append({"reason": "empty evidence is not a passing comparison"})
+    if len(a) != len(b):
+        differences.append({"reason": "different event counts", "left": len(a), "right": len(b)})
+    for i, (av, bv) in enumerate(zip(a, b)):
+        visit(av, bv, f"event-{i}")
+    return {"passed": not differences, "events": [len(a), len(b)],
+            "max_abs_stored_array_difference": max_abs,
+            "difference_count": len(differences), "differences": differences[:20]}
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--fixtures", type=Path, required=True)
+    parser.add_argument("--reference", type=Path, required=True)
+    parser.add_argument("--packaged", type=Path, required=True)
+    parser.add_argument("--output", type=Path, required=True)
+    args = parser.parse_args()
+    fixtures = json.loads(args.fixtures.read_text())["fixtures"]
+    checks = {}
+    for i in range(3):
+        checks[f"reference_vs_packaged_{i}"] = compare(
+            args.reference / f"episode-{i:03d}", args.packaged / f"episode-{i:03d}")
+    for i in range(2):
+        checks[f"captured_vs_packaged_{i}"] = compare(fixtures[i], args.packaged / f"episode-{i:03d}")
+    for name, root in [("reference", args.reference), ("packaged", args.packaged)]:
+        checks[f"{name}_reset_A_to_B_to_A"] = compare(root / "episode-000", root / "episode-002")
+    a = json.loads((args.reference / "environment.json").read_text())
+    b = json.loads((args.packaged / "environment.json").read_text())
+    keys = ["torch", "cuda", "cudnn", "gpu", "deterministic", "warn_only", "cudnn_benchmark",
+            "cudnn_deterministic", "cudnn_tf32", "matmul_tf32", "matmul_precision"]
+    checks["matching_backend"] = {"passed": all(a[k] == b[k] for k in keys),
+                                 "reference": {k: a[k] for k in keys}, "packaged": {k: b[k] for k in keys}}
+    report = {"passed": all(c["passed"] for c in checks.values()), "checks": checks,
+              "scope": "Same observations, seed 1, hardware and backend; not a benchmark score."}
+    args.output.write_text(json.dumps(report, indent=2) + "\n")
+    print(json.dumps({"passed": report["passed"], "checks": {k: v["passed"] for k, v in checks.items()}}))
+    raise SystemExit(0 if report["passed"] else 1)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/inference/policy/bagel_robotics_1/validation_replay.py b/inference/policy/bagel_robotics_1/validation_replay.py
new file mode 100644
index 0000000000000000000000000000000000000000..28e41c307ef185e0fa6cd4135f6009d74253f67d
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/validation_replay.py
@@ -0,0 +1,78 @@
+"""Replay two lossless real episodes in A/B/A order in one loaded model."""
+import argparse
+import importlib
+import json
+import os
+from pathlib import Path
+import sys
+
+from validation import attach, decode, records
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--implementation", choices=["packaged", "reference"], required=True)
+    parser.add_argument("--fixtures", type=Path, required=True)
+    parser.add_argument("--output", type=Path, required=True)
+    parser.add_argument("--xpl-root", type=Path)
+    parser.add_argument("--reference-root", type=Path)
+    args = parser.parse_args()
+    os.environ.pop("BAGEL_VALIDATION_TRACE", None)
+    from backend import configure_backend
+    configure_backend(1)
+    if args.implementation == "reference":
+        if args.reference_root is None:
+            parser.error("--reference-root is required")
+        root = args.reference_root.resolve()
+        sys.path[:0] = [str(root / "source/vpp2"), str(root / "source/vpp2/src"),
+                       str(root / "source/simulator"), str(root / "source/simulator/XPolicyLab")]
+        module = importlib.import_module("XPolicyLab.policy.VPP2.model")
+        config = json.loads((root / "inputs/vpp2-policy.json").read_text())
+    else:
+        if args.xpl_root is None:
+            parser.error("--xpl-root is required")
+        root = args.xpl_root.resolve()
+        sys.path[:0] = [str(root.parent), str(root)]
+        module = importlib.import_module("XPolicyLab.policy.bagel_robotics_1.model")
+        import yaml
+        config = yaml.safe_load((root / "policy/bagel_robotics_1/deploy.yml").read_text())
+        config.update(seed=1, bench_name="RoboDojo", env_cfg_type="arx_x5", action_type="ee",
+                      ckpt_name="cp1000", task_name=None)
+    model = module.Model(config)
+    recorder = attach(model, args.output)
+    import torch
+    environment = {
+        "implementation": args.implementation, "python": sys.version,
+        "torch": torch.__version__, "cuda": torch.version.cuda,
+        "cudnn": torch.backends.cudnn.version(), "gpu": torch.cuda.get_device_name(0),
+        "deterministic": torch.are_deterministic_algorithms_enabled(),
+        "warn_only": torch.is_deterministic_algorithms_warn_only_enabled(),
+        "cudnn_benchmark": torch.backends.cudnn.benchmark,
+        "cudnn_deterministic": torch.backends.cudnn.deterministic,
+        "cudnn_tf32": torch.backends.cudnn.allow_tf32,
+        "matmul_tf32": torch.backends.cuda.matmul.allow_tf32,
+        "matmul_precision": torch.get_float32_matmul_precision(),
+        "model_source": module.__file__,
+    }
+    (args.output / "environment.json").write_text(json.dumps(environment, indent=2) + "\n")
+    fixtures = json.loads(args.fixtures.read_text())["fixtures"]
+    if len(fixtures) != 2:
+        raise ValueError("Two independent fixtures are required for A/B/A reset validation")
+    for source in [fixtures[0], fixtures[1], fixtures[0]]:
+        model.reset()
+        client = {model.CLIENT_ID_FIELD: "validation-replay"}
+        model.reset_client(obs=client)
+        for row in records(source):
+            if row["kind"] == "observation":
+                obs = decode(row["data"]["observation"], Path(source) / "arrays")
+                obs.update(client)
+                model.update_obs(obs)
+            elif row["kind"] == "commands":
+                model.get_action(obs=client)
+        recorder.stream.flush()
+    recorder.stream.close()
+    print("Completed A/B/A replay:", args.implementation, flush=True)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/inference/policy/bagel_robotics_1/validation_server.py b/inference/policy/bagel_robotics_1/validation_server.py
new file mode 100644
index 0000000000000000000000000000000000000000..f09b71ff0d1e94f8437a0370d957fa336fc1be89
--- /dev/null
+++ b/inference/policy/bagel_robotics_1/validation_server.py
@@ -0,0 +1,32 @@
+"""Official server with an optional recorder; no inference or scoring changes."""
+import os
+from pathlib import Path
+import sys
+
+POLICY = Path(__file__).resolve().parent
+XPL = POLICY.parents[1]
+sys.path[:0] = [str(XPL.parent), str(XPL)]
+
+import setup_policy_server as official
+from validation import attach
+
+
+def main():
+    original = official.eval_function_decorator
+
+    def load(name, attribute):
+        factory = original(name, attribute)
+
+        def construct(config):
+            model = factory(config)
+            root = Path(os.environ["BAGEL_VALIDATION_TRACE"])
+            attach(model, root / str(config["task_name"]) / f"server-{os.getpid()}")
+            return model
+        return construct
+
+    official.eval_function_decorator = load
+    official.main(official.parse_args_and_config())
+
+
+if __name__ == "__main__":
+    main()
diff --git a/provenance/config.json b/provenance/config.json
deleted file mode 100644
index 3b751e327433aeebfc8963c278175b23688caabe..0000000000000000000000000000000000000000
--- a/provenance/config.json
+++ /dev/null
@@ -1,267 +0,0 @@
-{
-  "artifacts": {
-    "audit_manifest": {
-      "path": "/data3/praveen/experiments/vpp2-adaptive-latekv-mask-search-20261030/audit/manifest.json",
-      "sha256": "160d6bfd909032fdc907dc58f30ff7504b35ebe91cd2674ef0ec913e8fe3930e"
-    },
-    "data_cpu_receipt": {
-      "path": "/data3/praveen/experiments/vpp2-full30k-mask-search-20261030/checks/cpu_checks.json",
-      "sha256": "45b5a6ce431e1c5ac15d601180be16a337970d52ef96f141a7505434512cf876"
-    },
-    "normalization": {
-      "path": "/data1/praveen/experiments/vpp2-online-distill-20261015/demonstrations_native_text/dataset_stats.json",
-      "sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16"
-    },
-    "source_manifest": {
-      "path": "/data1/praveen/experiments/vpp2-online-distill-20261015/demonstrations_native_text/manifest.json",
-      "sha256": "fbbeae4b8ccf12ac9bd843f51147f78045a1789ad765b9f0e3c24a37b6b2e01f"
-    },
-    "visual_review": {
-      "path": "/data3/praveen/experiments/vpp2-critical-calibration-20261029/revisions/3ca392bfb574898f/visual_review_receipt.json",
-      "sha256": "9cbb622356fd0538f8e071d3feba6df7ea7efd2a0ccc18c1414cc634dc2d1e2f"
-    }
-  },
-  "curation": {
-    "audit_episodes_per_task": 8,
-    "audit_locked": true,
-    "auto_expand": false,
-    "base_windows_per_episode": 4,
-    "candidate_stride": 8,
-    "max_windows_per_episode": 16,
-    "min_selection_episodes_per_task": 83,
-    "monitor_episodes_per_task": 8,
-    "monitor_windows_per_task": 32,
-    "output": "/data3/praveen/experiments/vpp2-critical-calibration-20261029/revisions/3ca392bfb574898f",
-    "previous_panel": "/data3/praveen/experiments/vpp2-full-robodojo-mask-calibration-20261028/data/curated-v2/panel.json",
-    "quality": {
-      "action_latent": 32,
-      "check_every": 200,
-      "learning_rate": 0.001,
-      "max_updates": 2400,
-      "method": "deminf_inspired_episode_tiebreaker",
-      "minimum_rank_correlation": 0.8,
-      "neighbors": [
-        5,
-        6,
-        7
-      ],
-      "obs_latent": 64,
-      "patience": 4,
-      "repeats": 4
-    },
-    "random_fraction": 0.2,
-    "require_model_probe": true,
-    "require_visual_review": true,
-    "schema": "robodojo-curated-calibration-v3",
-    "seed": 100008,
-    "target_total": 30000,
-    "target_windows_per_task": null,
-    "vision": {
-      "batch_size": 32,
-      "sha256": "983f1562536e84ff750a1576fb08e54de751dbf2e17c0d8a4a13704341fdcd3d",
-      "weights": "/home/praveen/.cache/torch/hub/checkpoints/convnext_tiny-983f1562.pth"
-    },
-    "visual_backend": "frozen_convnext_spatial_history_v1",
-    "visual_descriptor_scope": "frozen pretrained spatial image features with nine native history slots; not verified contact or memory-cue labels"
-  },
-  "curation_artifacts": "/data3/praveen/experiments/vpp2-full-robodojo-mask-calibration-20261028/data/curated-v3",
-  "demonstrations": "/data1/praveen/experiments/vpp2-online-distill-20261015/demonstrations_native_text",
-  "execution": {
-    "cached_norm_flags": true,
-    "calibration_micro_batch": 2,
-    "deduplicate_conditioning_reads": true,
-    "fill_uninitialized_memory": false,
-    "mixed_attention_checkpointing": false,
-    "skip_nograd_checkpointing": true
-  },
-  "geometry": {
-    "ffn": 13824,
-    "ffn_floor": 4096,
-    "head_dim": 128,
-    "head_floor": 8,
-    "heads": 40,
-    "hidden": 5120,
-    "hidden_floor": 4864,
-    "layers": 40,
-    "pack": 128
-  },
-  "initial_masks": "/data3/praveen/experiments/vpp2-adaptive-hidden-latekv-mask-search-20261030/runs/masks_attempt_000870.pt",
-  "initial_masks_sha256": "680ba1058447b05d83d0295efbf1bdbcbe33163d8c4bba13e15ab7893db58c80",
-  "initial_preferred": {
-    "cross_scores": 32,
-    "ffn_scores": 9728,
-    "hidden_scores": 4864
-  },
-  "launch_policy": {
-    "automatic_submission": false,
-    "compact_recovery_automatic": false,
-    "minimum_disk_free_gib": 800,
-    "simulation_automatic": false
-  },
-  "native_inference": {
-    "history_stride": 25,
-    "horizon": 32,
-    "rand_device": "cpu",
-    "replan": 24,
-    "seed": 1,
-    "shift": 1.0,
-    "steps": 10
-  },
-  "normalization_sha256": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
-  "original_video_parameters": 16394878784,
-  "panel_file": "/data3/praveen/experiments/vpp2-critical-calibration-20261029/revisions/3ca392bfb574898f/panel.json",
-  "panel_sha256": "bc6df13b64b5f96bbe17c64244240cc44a40d704babb70893f79c0f30d389680",
-  "python_deps": "/data3/praveen/experiments/vpp2-adaptive-latekv-mask-search-20261030/python-deps",
-  "recovery": {
-    "final_lr_fraction": 0.1,
-    "learning_rate": 1e-06,
-    "updates": 2344,
-    "warmup_updates": 0
-  },
-  "recovery_conditioning_cache_entries": 128,
-  "recovery_extension": {
-    "condition": "gt_better_than_historical250",
-    "gradient_updates": [
-      500,
-      1000
-    ],
-    "parent_identity": "491ec49869e4d6a098c87ff99e6285ece09d841db4be9b3a638f9d6ae07aff9d",
-    "parent_pins": {
-      "checks/cpu.json": "4aca2e892c511978bb119d550c07931b92267dcbd6ac7ce3e66d64eac42991e0",
-      "checks/gradient_noninterference_gpu.json": "8efbe5a29cf58a4ba4a8038e3e41cf22afb49f9b1acccb4a5e3642ef2a9e4892",
-      "config.json": "3b3a91b77d0dd8e599e17941c5c8080674431bce101f743e00c7820d56a64f28",
-      "inputs.json": "56b484b54833aa547167083529a21c23d923739bd3c98a1a340206f3e18f7dd8",
-      "source_manifest.json": "3108de14e4b73e822511acc16fbbf5213a3916c79b6ba6da8046783793551d48"
-    },
-    "parent_root": "/data3/praveen/experiments/vpp2-recovery55-no-kv-20261102",
-    "parent_tag": "update_000250",
-    "parent_update": 250,
-    "schema": "no-kv-recovery250-to1000-v1",
-    "selection_metric": "task_macro_ground_truth_native_action_mse",
-    "target_update": 1000
-  },
-  "recovery_handoff": {
-    "cross_environment_checks_requested": false,
-    "early_recovery_authorized": true,
-    "mask_convergence_claimed": false,
-    "protocol_identity": "3adc9422e373a3cd7b386b2aded18c150fb57d38762de3f86393156f9bff69c3",
-    "schema": "milestone-selection-recovery-v1",
-    "selection_root": "/data3/praveen/experiments/vpp2-mask-checkpoint-selection-20261101/node3",
-    "selection_rule_sha256": "ca74561d69f065c73f6c083839a28a2b9a1d03fd45624030109b27b98884c560"
-  },
-  "recovery_lr_branch": {
-    "artifact_storage": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103/artifacts",
-    "baseline55_validation": "/data3/praveen/experiments/vpp2-selected-mask-recovery-curated30k-20261101/recovery/diagnostics/native_000055.json",
-    "baseline55_validation_sha256": "5c13ef2b51d76be5b108cc15b12577083c7dc2f461ebc92e1869084aac043eb4",
-    "checkpoint_every": 250,
-    "checkpoint_storage": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103/artifacts/checkpoints",
-    "comparison_identity": "0c94de36c72c55171c825364757390237142fbb18eafd27253be93c0fa5ca0ef",
-    "comparison_pins": {
-      "config.json": "c086c3ae472e9e0087ad6d70930e094952d62a80a233d67b43cd31f1365e6cff",
-      "inputs.json": "88b34bd5def0064551093a34290cb82c22545f26328be3f71b990bf214e3ff3a",
-      "recovery/diagnostics/native_000250.json": "71306f667cba71f9d9a5838ae3d9f015fc93a62096e4207d95a50ed26ebcd721",
-      "source_manifest.json": "becac462d4a2b25d36e9f915a586413c5f4b2ddbe7134365e3ef9d538abc3a80"
-    },
-    "comparison_root": "/data3/praveen/experiments/vpp2-selected-mask-recovery-curated30k-20261101",
-    "diagnostic_noise_draws": 2,
-    "diagnostic_windows_per_task": 2,
-    "final_lr_fraction": 0.1,
-    "gradient_updates": [
-      55,
-      150,
-      250
-    ],
-    "hold_through": 250,
-    "learning_rate": 1e-06,
-    "load_scratch": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103/artifacts/load-scratch",
-    "parent_identity": "d85b815fd1e304baea64a9f1fd246a2b84ef3b86893a7bba72710e34bea26127",
-    "parent_root": "/data3/praveen/experiments/vpp2-selected-mask-recovery-20261101",
-    "parent_tag": "update_000055",
-    "parent_update": 55,
-    "paused_replay_root": "/data3/praveen/experiments/vpp2-recovery250-faithful-replay-20261102",
-    "review_update": 250,
-    "schedule_endpoint": 2399,
-    "schema": "recovery55-low-lr-no-kv-v1",
-    "validation_every": 50
-  },
-  "recovery_subset": {
-    "export_manifest_sha256": "2917adfeadbc2c6d4da60b7739dbee861d7f318f8fc66877f1ca6bd691938172",
-    "export_path": "/data3/praveen/experiments/vpp2-selected-mask-recovery-20261101/exports/update_001500",
-    "parent_checkpoint_identity": "333fa7a74168926559bf6be86291644f369c11e5680c76539580b951524fa8a7",
-    "parent_config_sha256": "d1fcf70b6a625698866a175bd10c24238067a3c6a0e4f945c48e360e7b0cff77",
-    "parent_export_receipt_sha256": "a8fec6af95b6243fdae2e17dc82f089b316b1a77b19fc8f816e545b12c5fcf2f",
-    "parent_identity": "d85b815fd1e304baea64a9f1fd246a2b84ef3b86893a7bba72710e34bea26127",
-    "parent_inputs_sha256": "296441c7ac166873cc40e1b4056f814844f7fa9d738d46064cea9f5bb8100030",
-    "parent_receipt_sha256": "038a56811ed0cc6aee63fbea71727f985b64db06bbd89a3f53192c44c98e6594",
-    "parent_root": "/data3/praveen/experiments/vpp2-selected-mask-recovery-20261101",
-    "parent_selected_sha256": "aaf1290c299704293b1dd936cbc92dcf0b4aa0043ef8947be86836957c499e9e",
-    "parent_sources_sha256": "ed356b55e1b676522cb7d99de77780306a02b6e8887c91421a9c0b6b83dd3421",
-    "parent_tag": "update_000055",
-    "parent_update": 55,
-    "sampler": "original_curated_task_balanced",
-    "schema": "curated30k-recovery-continuation-v1",
-    "updates": 2344,
-    "windows": 30000
-  },
-  "reference_split": "/data1/praveen/experiments/vpp2-online-distill-20261015/prepared_split",
-  "reliability": {
-    "floor": 1e-08,
-    "margin_fraction": 0.01,
-    "method": "relative_student_advantage_with_baseline_component_veto"
-  },
-  "root": "/data3/praveen/experiments/vpp2-recovery55-no-kv-to1000-20261103",
-  "runtime_conditioning_cache": null,
-  "schema": "vpp2-flexible-joint-mask-v1",
-  "seed": 100008,
-  "size_coefficient": {
-    "first_update": 193,
-    "invalid_measurement": "checkpoint_and_stop",
-    "last_regular_update": 200,
-    "last_replacement_update": 232,
-    "ramp_updates": 100,
-    "samples_per_family": 8,
-    "target_gradient_fraction": 0.1,
-    "upper_cap": 1.0
-  },
-  "teacher_bundle": "/data1/praveen/experiments/vpp2-joint-feature-cache-20261013/assets/model/checkpoints/joint2b_s100000",
-  "train": {
-    "action_gt_weight": 1.0,
-    "betas": [
-      0.9,
-      0.999
-    ],
-    "checkpoint_every": 250,
-    "count_budget_weight": 0.0,
-    "final_lr_fraction": 0.1,
-    "flow_monitor_every": 500,
-    "frozen_action": true,
-    "frozen_video": false,
-    "global_batch": 256,
-    "gradient_clip": 1.0,
-    "kv_layers": [],
-    "kv_weight": 0.0,
-    "mask_lr": 0.01,
-    "mask_weight_decay": 0.0,
-    "max_updates": 3000,
-    "micro_batch": 4,
-    "minimum_updates": 2000,
-    "native_windows_per_task": 4,
-    "teacher_action_weight": 0.0,
-    "teacher_video_weight": 0.05,
-    "video_lr": 1e-06,
-    "video_weight_decay": 0.01,
-    "warmup_updates": 50,
-    "world_size": 8
-  },
-  "validation": {
-    "bf16_action_video_relative_tolerance": 0.01,
-    "fp32_relative_tolerance": 2e-05
-  },
-  "wan_root": "/data1/praveen/experiments/vpp2-joint-feature-cache-20261013/assets/model/checkpoints/Wan2.1-I2V-14B-480P",
-  "wandb": {
-    "entity": "bakery_ismail-bagel-network",
-    "group": "recovery55-no-kv-20261102",
-    "name": "recovery55-no-kv-continue250-to1000",
-    "project": "vpp2-thaw-robodojo"
-  }
-}
diff --git a/provenance/source_manifest.json b/provenance/source_manifest.json
deleted file mode 100644
index 89bd3862dd425007b21f1bde121b3f4d27ce05fb..0000000000000000000000000000000000000000
--- a/provenance/source_manifest.json
+++ /dev/null
@@ -1,317 +0,0 @@
-{
-  "files": {
-    "configs/adaptive_mask/full30k.json": "75ce014340a61990680f0bb7fb1bfdd46f4f1f6b807709e5823906a84a4f043e",
-    "configs/adaptive_mask/hidden_0to5.json": "9d7c3cd234aac10abfd40c67048aa1f1717b8390c476e7d58136fb377eada5b8",
-    "configs/compact_kv/precision.json": "60af967dfb7c8e45438f0131d76a3cbe1740e4cb5f0cbb1deea20bb00b36f597",
-    "configs/flexible_mask/full30k.json": "f084dc3e5b0a8e0b003a138494d9fef04923af89f4764bf176135fa677523c0f",
-    "configs/flexible_mask/resident_teacher.json": "321b8e8b57f1e49e854d904f4e821824d7cd7fef04e214bfb0636ec9b7098387",
-    "configs/learned_width_kv/precision.json": "bf0d2b572ef63ab2c56aac8c620f91d6c2a2603dc518a27768d6d97a7e3c9a01",
-    "configs/libero/accelerate_zero2.yaml": "f0f265e577da2492d785dfd5a7031cacb833235ccefd8c07aa048ff42002e242",
-    "configs/libero/dataset_stats.json": "4717a109760fcd1d504c7d301ff95106b14446a89b019a026427ddbbcc866d1b",
-    "configs/libero/deepspeed_zero2.json": "57a47cd51dc5434250e2975ab46fe6f9f0e742f5e5b6ae50a587368959bae77a",
-    "configs/libero/eval_ood.yaml": "d1ee29118dad9d0117b7fc6c4a6a4d40025239af2cfc73e861c269b1ee69b773",
-    "configs/libero/eval_pro.yaml": "aba2f40cb089993a73901c081c059730a140f7d7ce0059f2cd5ee9d4f8541082",
-    "configs/libero/eval_standard.yaml": "45a5d165d8fb92f392a278bf8a9713d49db55f12e45c60e66b8c7462359cf4d0",
-    "configs/libero/train_action.yaml": "d2a4fb641c847f0e9f66994e425c0ef8f976228b6a744561982d7dc238f23e29",
-    "configs/libero/train_video.yaml": "3f8c1fa0c831fbffb862a50d5156097efd08b4df4b9230baf902f058f5172db0",
-    "configs/robodojo/accelerate.yaml": "0a0535e4c3e35016d0176370167fb435b1c59e15921fc8d4037ab509ff9dd467",
-    "configs/robodojo/deepspeed_zero1.json": "9ef633d5a346924567ae18a801945389ee037136a4ff8ca46bce99ab3c926b58",
-    "configs/robodojo/eval_reference54.json": "588ffeb92796bd34c4dc7a2c1570a8dcb61adc9386c60da7d7f0a3bd63022ca5",
-    "configs/robodojo/holdout.json": "7bf4892dc9271c6aea6fa20acdc5f76057b35851097c152e3cff44a3c959a0e0",
-    "configs/robodojo/normalization_ee16.json": "2649d4e3ff3d6d400c9953b7c20c79541474f9c38a8a9916b9caacecd8578f16",
-    "configs/robodojo/train_100k.yaml": "1eb185bb258d0ee6808df8a7b338d1fdd4882ada8a34532211291c03d5517930",
-    "configs/robodojo/train_video_10k.yaml": "cfe17410517dd29e11e98bbb7bb61a02ea45f18788024429f783aa43614e78d4",
-    "configs/width_kv/precision.json": "2f09d40747010590e2e30be24ceefb35432c77c544b3e9bf0ac99f16eded24f2",
-    "scripts/adaptive_mask/benchmark_checkpointing.py": "ce81ecde2782be8957f1defcd21ae506717b6a3857bf5867c205c4612cde24ab",
-    "scripts/adaptive_mask/benchmark_resume_execution.py": "b21dd24c00900e1c311f7be042d0051eef9744472743e24f3e2e0e0e1886a067",
-    "scripts/adaptive_mask/diagnose_export_rounding.py": "1b50f4cdbf021e10b901f760f6314df2582eab5bdc85368273fe3eefc71893f6",
-    "scripts/adaptive_mask/diagnose_full_precision.py": "6895dc9a0cebbc699dd4bc4364e7e8b1521b0ed460681e4b089ee57038c97a97",
-    "scripts/adaptive_mask/migrate_execution_resume.py": "c321210712b267fde9d8c6106023128c99226a939c8df6facb002c0eac7c21a7",
-    "scripts/adaptive_mask/migrate_quality_policy.py": "d24aa2588fd93e148757e42af509009518dcf0edc7b5cb415f2e8699b72472ad",
-    "scripts/adaptive_mask/probe_bf16_kernels.py": "6362a6fae3dca344d073ab49653ba0a608671689d42e587453429f22d63b94fc",
-    "scripts/adaptive_mask/publish.py": "8a2fe20a2b10d1a8dfd4439aa32db0460e730c98f77f049c559eeccc5d876168",
-    "scripts/adaptive_mask/submit.py": "7fbe10a1766c1fdbece9400534dd140ae42be09e2db6e441d1fb11152a18e27f",
-    "scripts/adaptive_mask/verify_launch.py": "9484f189947deb9de3b9ac2cda0e5f0cd8952ce48db52b4d8a6cb7dce07307e9",
-    "scripts/flexible_mask/check_zero_gradient.py": "319736845e2c16962cef8aeace9ea02ccab23f40cc0d5127d4a3f033485130ca",
-    "scripts/flexible_mask/publish.py": "a2240c2e7960a96784089f611a615b9c94353345a538ba925c672ba9fc69b9bd",
-    "scripts/flexible_mask/verify_launch.py": "dde4be6c730231a85c7bb68601c1eafce4b218f2f1469e60d58668bba98c662b",
-    "src/vpp2/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/adaptive_mask/__init__.py": "a73f5515ffb77d323f9c324776e9c166d13817c470506cfd5db4ef0546cbbdda",
-    "src/vpp2/adaptive_mask/audit.py": "c949909a2e60c3c4389c0221c79459961f2c4925baa07f7985774a52257a1d17",
-    "src/vpp2/adaptive_mask/budget.py": "74e32285903dbf673ecadb6a60d917a990b9c6619beac9f926c1b3cbdae3887c",
-    "src/vpp2/adaptive_mask/controller.py": "328f636201a845ebdc211340a9014a5076fa895e10cf9c2e1f91b96761a2514b",
-    "src/vpp2/adaptive_mask/data.py": "e3b633fedaa503b101c80610cbed0933e9e245acf56a6c0692c4d99620faf84c",
-    "src/vpp2/adaptive_mask/equivalence.py": "8658cb69d549fa0575570364fe3aa1e0b121fcf8b9965b17cd174d459dddf5ab",
-    "src/vpp2/adaptive_mask/execution.py": "efec52c9b2eabf0300c72edea18f9cd4cc16a0889496793abb6a50161f2328ef",
-    "src/vpp2/adaptive_mask/export.py": "0ce51e74787d28bbe9f4e73a3774297b424aa131fa78968d24154ff56db98b0f",
-    "src/vpp2/adaptive_mask/feedback.py": "519fa4e6416723fc000754815a2f05d3415299ae35724f37bb806ebc22393d38",
-    "src/vpp2/adaptive_mask/geometry.py": "f1ce4110c3a3a893f4f240c49e2ccc4e93215ac4ba9e447de8042f9b19afab98",
-    "src/vpp2/adaptive_mask/hidden.py": "968560bd6dae0fce59bd28084091b10d559cf165c81764f6ad24b7492101810d",
-    "src/vpp2/adaptive_mask/losses.py": "4a7dea6f87ef0fa38a7202d85d98d37460f8f4dc13ca5336f32b1da8b84a4322",
-    "src/vpp2/adaptive_mask/masking.py": "31b595916410c60f772fff1ab63e92c854c8fd0d2261e6b865fd05ce2c61b9c6",
-    "src/vpp2/adaptive_mask/monitor.py": "48b71f499dd5a0a79557eeb918f5fb1a641dc8654e5d412aa628ca9c57b5dcf8",
-    "src/vpp2/adaptive_mask/physical.py": "f54e8b4fae8026cc89f296d74bf4cbea06fa4e126905b572a22c7d8226f0238c",
-    "src/vpp2/adaptive_mask/preflight.py": "b646e7ca785afc87356e3e1184e48f535792ebafc84936450fc83bfd3f9f5ea8",
-    "src/vpp2/adaptive_mask/priority.py": "68ccff1a384081f0c9a82e8f54ad2c720293f8500e7440a469b5e8fac1ee608a",
-    "src/vpp2/adaptive_mask/quality.py": "c3b3a2264ca0b4773805f4485ebfe91bb4ab861e226ff4bd27bd667ab372b59b",
-    "src/vpp2/adaptive_mask/quality_policy.py": "f100b7d4a412592777d936a6cf3c73a9f787cc4a62d5f1b8a369fed74acf897f",
-    "src/vpp2/adaptive_mask/reliability.py": "306b41720154810bd972fe4d9c4c388e6967539d076ddbedd3a432ac62d13104",
-    "src/vpp2/adaptive_mask/report.py": "f71401e09297daed8bbb72cb3af3b887e5fdf5ec282f47637fbb01392d280171",
-    "src/vpp2/adaptive_mask/runtime.py": "ed74bfedd147389ea01f3be8c2979fa06c7152326ccc29ee347175690cb9c6cc",
-    "src/vpp2/adaptive_mask/state.py": "5e8f95021eeb35511f5481e4e91ae687fe605f9e88d17b3b98d377cc967c17e1",
-    "src/vpp2/adaptive_mask/train.py": "ad95349bfb242bcecd7ddf82fc58b31d63f2938a2101f9560bbc9e305f5f5b20",
-    "src/vpp2/benchmark/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/benchmark/audit.py": "8587be371653c38b9926aa0ace006a244f270e0ef0020147a7bf5b0fc85d4538",
-    "src/vpp2/benchmark/packed.py": "5047fd02f3fcc68e41f543fe83bc9202769a05794d898524d6d25e308f2b5125",
-    "src/vpp2/benchmark/scheduler_io.py": "d6df06259549faed2da9f2888b540eac8e1ff2895c339eea58b4ea6a7e3b3cb0",
-    "src/vpp2/checkpoint_compat.py": "462c5f93812483b5871633b200bc2ffecb9c201a2caf6b14f8f36dbe95fba50a",
-    "src/vpp2/cli.py": "54c19a0f7e25b53bbf56ec13b0f537c9d7b81282e1e80575bddfb470172164e3",
-    "src/vpp2/compact_kv/__init__.py": "f34f2f77f68399f45f6656b562a5c331f2504672dc145df368e086532191f766",
-    "src/vpp2/compact_kv/blocks.py": "662c1f35055c2be2e2a14f2d92615d046f5e454879e34af39e3c7aec3af34098",
-    "src/vpp2/compact_kv/checkpointing.py": "7fdecdea02a851ff7abb4952ad3f162b4b4b352d3fdf04e29f7fc876a6be6589",
-    "src/vpp2/compact_kv/eval_prepare.py": "bdb33444056245e3cc773b0849b4db608b454cd24a7ac5b752f1ba26ff978865",
-    "src/vpp2/compact_kv/export_check.py": "e6cba70b47027882a34d587aba0da5f9371128ec2db2282b91ad2d49e40065b9",
-    "src/vpp2/compact_kv/gpu_check.py": "fedb6a9ab69b51080161ff2a4feae3f85262cc50f932f48e8decf04c1d147a20",
-    "src/vpp2/compact_kv/losses.py": "914437ebf3ad91fd2cffa893a65ecb04ab01f2b337658a781a472feeb50c0a06",
-    "src/vpp2/compact_kv/model.py": "332ebaebf543d40a34d188ebbe81055ec3e4bcff164bfe90d15aa6cef0748b8d",
-    "src/vpp2/compact_kv/prepare.py": "7b01764cad0fafc0cc9441952b2e928048b0a43a6105dbd0eb64c349f4b9bf1f",
-    "src/vpp2/compact_kv/provenance.py": "103e0f7100be56233aa86041ba9bc427ccb0db4fe045c4e6b8d3481ab6bb42da",
-    "src/vpp2/compact_kv/reporting.py": "0b9ff602fc3425d1b9c122a77da3a426814e11c9cbdcb6b9a75a2a48130087a7",
-    "src/vpp2/compact_kv/sampling.py": "2e8d82704f07853a7b487d5aaedef1b2f1fcfc5247e9100a56ed21d69262eae2",
-    "src/vpp2/compact_kv/spec.py": "9bc920fc7d4398e424a0bde38ce97f87473aa417cf68b48e2c292a59491bc7d7",
-    "src/vpp2/compact_kv/train.py": "a9e4410c5cde5df002d9808554538c29b77ed9e2d2bedc9ac5f8465eed70cc3b",
-    "src/vpp2/compact_kv/validation.py": "73b62d02ce1d518cba7b9462e85c3c8d272ae9974e1161b1b6adea6f0040e2b5",
-    "src/vpp2/convert_robodojo.py": "7ec2d6bfffbb4f5891ee33a4bf630cf27b17628506aa3cee1c46b3050027b290",
-    "src/vpp2/critical_windows/__init__.py": "22721b77e25ec4874c4853d2f2991a14dff6aaa665da500f2830826b06545478",
-    "src/vpp2/critical_windows/contracts.py": "ded4a239f7599785d74ff57ff892c5aa74f0280d6759460c93942e91f505d47a",
-    "src/vpp2/critical_windows/coverage.py": "6c18bf995aacf413bbe4536eaeed92ab4b2f15f9eaaaa03a2730256ea7922ce4",
-    "src/vpp2/critical_windows/inference.py": "df301bd187b27d20b6623febcb5e85815e1b7d4a1e9524a8a65bdf5904bf7232",
-    "src/vpp2/critical_windows/metrics.py": "ae2bab145555d55696ef479758a5bae6780527fd3085a4f46d035b667ff32736",
-    "src/vpp2/critical_windows/packets.py": "8f7d2c411407d95b1ff4d0e5b9055589f6114b1bb1020eeb7e1b106f330b853a",
-    "src/vpp2/critical_windows/prepare_staged.py": "694cb5699de2d9f5ce1bc4667627b47d6bf0698715b4aad86907139859caffe5",
-    "src/vpp2/critical_windows/reselect.py": "8d0e31ca1c17906d74de6439276e49ff7ea78abfc66fa556d8a306070b3dbf4c",
-    "src/vpp2/critical_windows/score.py": "32daceb9a3fd957d9be75823b4d519629bc60a64d1da62f465e172cb1ec98d55",
-    "src/vpp2/critical_windows/selection.py": "a7c35f6719d3308bfb738965d17544a15985c6d684bb47eac7272d6155d96786",
-    "src/vpp2/critical_windows/staged_review.py": "d5cfb3f49ec5acc6914da98f5413d1549b2695e7c85c9aed9c0db83dee6be754",
-    "src/vpp2/data.py": "8574e1ae3c6ca6e4b84633ee1815f2b3f2ea17c1e55f12fda0b00c5c5a6a7b53",
-    "src/vpp2/datasets/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/datasets/dataset_utils.py": "e1a1471b0aa70917eaf699f85a9acf3ac70d9fb6eb9ebaf02d2bf2f218d6ab2f",
-    "src/vpp2/datasets/lerobot/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/datasets/lerobot/base_lerobot_dataset.py": "729fb3fd4ca531212eadb4a4e9c7e9f97ad34cca091d66380acd6b179f0c29c6",
-    "src/vpp2/datasets/lerobot/constants.py": "11ac9260341e1560c5eb0dd7acb79ed63c74180b8ab4113c27622ff213db417f",
-    "src/vpp2/datasets/lerobot/critical_phase_dataset.py": "bfe75bb9593ec646236777d90d9fa309fa0e1188db07e07a3be0810edbe3da56",
-    "src/vpp2/datasets/lerobot/critical_phase_sampling.py": "266b0b7b023f014f4ccd935b6015b0a05b4659f372264fdefd997ea9692b244a",
-    "src/vpp2/datasets/lerobot/episode_instruction.py": "74767839a813c7902861f4561101541c282b72e261df1fd28fe8d06c91231b69",
-    "src/vpp2/datasets/lerobot/lerobot/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "src/vpp2/datasets/lerobot/lerobot/datasets/__init__.py": "01ba4719c80b6fe911b091a7c05124b64eeece964e09c058ef8f9805daca546b",
-    "src/vpp2/datasets/lerobot/lerobot/datasets/compute_stats.py": "7d7daab219332b5abdfd5ec67ccaf3502e8eb583a17a4e482c3f3d4f7a691e89",
-    "src/vpp2/datasets/lerobot/lerobot/datasets/utils.py": "4fdb4ac0653201e6743f2161c485213429948c1e1d5b068d03b4d97803254006",
-    "src/vpp2/datasets/lerobot/lerobot/datasets/video_utils.py": "4f66c410ae6aa5a98acfc3320f3dd7b017d0285117054f14226f9e5d82395087",
-    "src/vpp2/datasets/lerobot/lerobot/lerobot_dataset.py": "5659e992010bb6674ec50ab0d33b28a8e9dab0ecc895c199662af4c40fe56629",
-    "src/vpp2/datasets/lerobot/processors/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/datasets/lerobot/processors/base_processor.py": "d3327cbe398d77761df960014d41e70d349bb0fa42bca874333b9ee970222447",
-    "src/vpp2/datasets/lerobot/processors/vpp2_processor.py": "43c5604cbfc315b7377316cb97c9a6f3322124404397abfd01c6c3d0090cd414",
-    "src/vpp2/datasets/lerobot/robot_video_dataset.py": "8eabfbd649116b0b4d6c9704c0df34febbf8875f6ab863dd75f4107dbbcdef73",
-    "src/vpp2/datasets/lerobot/transforms/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/datasets/lerobot/transforms/action_state_merger.py": "39fd44f254bd7fb71b9269d75d8f9b5a5bbbe167ea438dda92d40eda1d256734",
-    "src/vpp2/datasets/lerobot/transforms/image.py": "f494d262ce29770e959bdff60b997f7a07e5f00d10ef515eee7be813d1db038f",
-    "src/vpp2/datasets/lerobot/transforms/rotation.py": "feb59d26539be6cfdb86488f0c10c11488ed73901658bc657e67a755ddb9e67e",
-    "src/vpp2/datasets/lerobot/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/datasets/lerobot/utils/normalizer.py": "8207d813b698a91dfdbe6910a4aa1fe0e775a2a8edf5c05b2d7a42591e9f8adb",
-    "src/vpp2/datasets/lerobot/utils/rotation.py": "8ea2c72565707f5979967a1abeb5edf1eaa2c3fec55a7b50e640de869fc9fbcf",
-    "src/vpp2/datasets/lerobot/video_event_dataset.py": "c20d3f5e2bc05e1b7ed94ae38872883a976a179e09b7fa051e94f94e2836c02d",
-    "src/vpp2/enhance_prompt.py": "f31b3190c24bad073c5e3121ded4f564232729566e8c1fcc3e76fe75b28faf45",
-    "src/vpp2/evaluation.py": "9b76882149e3c3a8c417a0e23514af1bb305c01bf0b6d21fff1f70bc9843f55a",
-    "src/vpp2/flexible_mask/__init__.py": "64bed1808976d1f3c708154c36a97288a75abc6463d49d7f0065cb8360a2fb16",
-    "src/vpp2/flexible_mask/calibration_batch.py": "dc5d83fc9e2bb595b9e3aeb1923115ebb9345283290a7c3f95c79fdeef7d3e76",
-    "src/vpp2/flexible_mask/checkpoint.py": "c8531727e2388ec3d7a9770a504ceda119afcf03d64b2462befcd45a48949f9d",
-    "src/vpp2/flexible_mask/coefficient.py": "07a3cc8d8e6c8b60119b9f94ae85074da017b08158fa0a0912cc9051b6df8ef7",
-    "src/vpp2/flexible_mask/config.py": "a1076ef48b85b88c50e6d213e04107a2f1c8c8111b4e70655670591ed835c6ac",
-    "src/vpp2/flexible_mask/continuation.py": "2016463b4efe01b5d0a9ee84f03c3d95ee16f8e24f230faa465598cc978da29b",
-    "src/vpp2/flexible_mask/controller.py": "263f1c9091e0c4318aeefb61e91549ba905ff046a625e35e1b18c94da4b4668d",
-    "src/vpp2/flexible_mask/data.py": "d245bafd6f04f33ae7cef7c01d5db876a1a2a1a138c4de8a139d3ea78095521c",
-    "src/vpp2/flexible_mask/diagnose.py": "1440691074a8b352a8587c26a46070610981d3dcd191740245c18347c0f44fcd",
-    "src/vpp2/flexible_mask/export.py": "3064dfd82bb55a145c16c77bbe17244970d07d18e944b793f4b7368dbcede764",
-    "src/vpp2/flexible_mask/gradient_probe.py": "6fe3f13540cb9d8dd20c0b6429268ca74625dfad1e41806412c0f980fac76509",
-    "src/vpp2/flexible_mask/losses.py": "7a962ac5f29b88d01ef3c9fec676a240edfff7bb6147f111ab08a8fb774540cd",
-    "src/vpp2/flexible_mask/low_lr.py": "a155933ccab8a6957c06297897c613b1ae6ac01d941cf88d45f729907104a78a",
-    "src/vpp2/flexible_mask/low_lr_entry.py": "e49e39ecc89db87da56bd1b4bc54db69756520ea75a666566b96203dd66e121d",
-    "src/vpp2/flexible_mask/metrics.py": "e976dd066e4b5e7e7ae79db77b9d1be64d17c212c61d0bc113b87a048e8f2b6f",
-    "src/vpp2/flexible_mask/performance.py": "95b3977fd739cc98f7de94b14fb490a7c132021e742e7179dc5b5dddd6557c68",
-    "src/vpp2/flexible_mask/preflight.py": "f794b997a492911c4d77656310c89d3b85e54983b3b889ba9904be64fb961e43",
-    "src/vpp2/flexible_mask/prepare.py": "d6cf82e27555ecc13973727f4b71ce7e40558fef7d95805665fc7e458a42fad4",
-    "src/vpp2/flexible_mask/pressure.py": "6cf74edcf3625fce9eec689d70841c6c1481f7c1abd87e5bf70bb38eb34632a2",
-    "src/vpp2/flexible_mask/recover.py": "418b2d5ecdd08b08686b2a07bbeb0f969471eb73c04ac6952dcfbcda3b123f30",
-    "src/vpp2/flexible_mask/recovery_best.py": "9a1f08ec2b1627dd4935599cba1d502051014fdb5501b390ffd4a7d9978807ce",
-    "src/vpp2/flexible_mask/recovery_export.py": "c50433aa1b7e528fb23c514dddb5b9197f44c662829838ecda106980ab0dd54d",
-    "src/vpp2/flexible_mask/recovery_extension.py": "ab537157a19dce6fada71261960f4e6bc10d16897f7341d3864353605f87f0f5",
-    "src/vpp2/flexible_mask/recovery_extension_entry.py": "5c1fe4c591c2b979554691e88fd7523d4205f74ff7f3a4a69ff07938e17336da",
-    "src/vpp2/flexible_mask/recovery_gradients.py": "bcbd757ea2670dd96562a749c32dbaa6918d601b14e0b5f4c2d6a1bb6910e59a",
-    "src/vpp2/flexible_mask/recovery_handoff.py": "aa590f707738e4b7c7250f901e29795c7441430f961bfa764437cf34bdc514b1",
-    "src/vpp2/flexible_mask/recovery_subset.py": "acb1253714f8424996df5df3558a6b2d0ac9aeee09c6538594b404b50e56501b",
-    "src/vpp2/flexible_mask/replay_entry.py": "feeaeef4c21cb5caf938f1a25878431f9c2016cfbdf0bdfcbd05ea0a7f47116d",
-    "src/vpp2/flexible_mask/report.py": "5a90a5fd6eff25b62f2b7cab77194be62a55b4d0bb27e03a78564039cb9330f6",
-    "src/vpp2/flexible_mask/rng.py": "01680b5b7e8a29ce92e96d756e12ceb80bb0e0ce1ebb642a14fa2a3a34776daf",
-    "src/vpp2/flexible_mask/runtime.py": "81555e82b21f9b7d61fc011323237d2e86cf3dc2b71c579a24d4604b3755d200",
-    "src/vpp2/flexible_mask/select.py": "b547857bad61ae7ebd25b888fba73c2ca8f50a8eb31551ca54a909c75c8984be",
-    "src/vpp2/flexible_mask/selection.py": "b9453c8735051b54ae7dc8610d2ca332eae8ca1a3e099aaadaf5ab75bfb0f8b0",
-    "src/vpp2/flexible_mask/speed_review.py": "9d7099d1e173026274eb439169df2ad4812ec1d56a5f0b21242f972cd835ffd2",
-    "src/vpp2/flexible_mask/test_gradient_gpu.py": "0e5efe20a33242f8a78ad0af824d95d04c9134fb3dbf06b97266925873f5599e",
-    "src/vpp2/flexible_mask/test_recovery_resume_gpu.py": "50729fffb7f9ade6f919baf3560b83053c83ed395cf18405fa93689f12f3159f",
-    "src/vpp2/flexible_mask/throughput.py": "ce21bdfda51ae10d1470ac537d910d9f46f60a38910888808ffb3f5a80174d58",
-    "src/vpp2/flexible_mask/train.py": "1fd8ffeab47be9ca7fd9a7c7c46d372e4b31e8d459ce132b896574c5e1c67c50",
-    "src/vpp2/flexible_mask/verify_export.py": "fb68e62b0da98e6f96cfa694bd061162bbeec1763e5787e057a1cbac5b2ce8f6",
-    "src/vpp2/full_mask/__init__.py": "27c9ede277380e8d64586371f2a03334e68ab2d3bfe516d95e8cb96099827929",
-    "src/vpp2/full_mask/calibration.py": "5d97e0385eba82252e12cc1562d1a77da2f2bce29332d758c1b5b6b84c5bb47a",
-    "src/vpp2/full_mask/conditioning_cache.py": "07adf95165db4c3c1c469695d6b7b6fe4cf4fa9aff0a3d360f4229bffe9be17e",
-    "src/vpp2/full_mask/controller.py": "0e6a969b89a4275352cc5a1e00bdfe8e3f8754a6ebd7f9266b92179cbe6bdd10",
-    "src/vpp2/full_mask/cpu_contracts.py": "1a46a53a68d2d8a202c5f98d9a78f6d5f5041dada0c2239ed81e7916a28f0758",
-    "src/vpp2/full_mask/curation.py": "4ffd7b0212bdbca1daf73ba7c24423523437012150fc044a43bbb33cce747545",
-    "src/vpp2/full_mask/curation_v3.py": "9dba386d66869de0fba321ade79b2692d47f111ccfebc7f367194cef7a4b4e8c",
-    "src/vpp2/full_mask/curation_v3_checks.py": "688ea4e7e44c4ecc1a3cd81cf682e2cbf611d9ad2d52744ba1308324da70febc",
-    "src/vpp2/full_mask/curation_v3_features.py": "11011a089f1586045e2972cf267773aaf4602cf08c8b0e755bbf017ede1e6e43",
-    "src/vpp2/full_mask/curation_v3_finalize.py": "a2a7812eb6c670686a7a4f4adff0a2c6363dbc078c71d948537f1c5fd546254f",
-    "src/vpp2/full_mask/curation_v3_probe.py": "2755e9cabd0fe02e70cb8330c083af881a5e4696697d58cf7aa01bd00ef57f6d",
-    "src/vpp2/full_mask/curation_v3_quality.py": "b9170398762112a23bb133dde53c6a49193c671c767647239c8b0ff15159530c",
-    "src/vpp2/full_mask/data.py": "668a7c5f88a3ab17854eb0df240c974082f4aa64604cf11d3143b18337f73181",
-    "src/vpp2/full_mask/distributed_smoke.py": "f699d4f330309f29a537fbabb61c60f61eb3322d7304605351dba3d2acb48e8e",
-    "src/vpp2/full_mask/gpu_prepare.py": "92ce96dc4c1b6d6ebc22d99935472c15d86ba2a7c050e31640d2c32e08748ec5",
-    "src/vpp2/full_mask/monitor.py": "2724a934d1f96444227c854f75ef84df6025c02485aa68eca34f3b28d509fe19",
-    "src/vpp2/full_mask/native_monitor.py": "5572712a340f17692a669d2a4c41d1d691164925eb7bd06a56fff1dc2afd69fa",
-    "src/vpp2/full_mask/panel.py": "2844b663340cbfb54f7622f3447a7831e2809050739a52dc076fd6ae32ed9e46",
-    "src/vpp2/full_mask/review_decisions.py": "ed1184bcda7b5bb3fb11bb4262c4461c408e3d3ac0d23e80d2f8cafcf39c7e18",
-    "src/vpp2/full_mask/review_report.py": "752ba9e8a3caf5b8cefa2d18e6fe14860e6f998f57c03af0940af0943b09a9b1",
-    "src/vpp2/full_mask/runtime.py": "ae30b4d3fd3c888bef17672376dc3b34491cb296a1883959674ad465579efe7b",
-    "src/vpp2/full_mask/search_quality.py": "ee3238d3366799e97d3f9a90c11f579ff8f51d5bd0642f52c37d9dcdbcf02e6e",
-    "src/vpp2/full_mask/train.py": "f4127e6580522d4d9e5a50db10d201526e948d4d3479ddfbbc12e60ec61c7501",
-    "src/vpp2/full_mask/verify_data.py": "72ee60ba2016d5f73eb2ccb01b666a5dac37138620efc4a79a20607df3658fb3",
-    "src/vpp2/inference.py": "42b3057a85aaacbb87440bc740f13b09b9c82cd51d8c5af4e080716aabdcf38e",
-    "src/vpp2/learned_width_kv/__init__.py": "0c59cb8424b6cd64508a005625e9edf63c84d24618a5d2caf6e6fa5d63cc859e",
-    "src/vpp2/learned_width_kv/calibrate.py": "b89bbf103b49fd85f85f8c300a7868bb42ae7a13f29bd9878dbd5a11061ae51d",
-    "src/vpp2/learned_width_kv/checkpointing.py": "7d30e46b95d91df8aff088346ae75809958622b3a8e8d953db4f30adfbaebadd",
-    "src/vpp2/learned_width_kv/common.py": "52dff98c6317b34df2a13313a641998acf4f9e18ac31ea4e8e25227afa14620d",
-    "src/vpp2/learned_width_kv/controller.py": "7b5a8793a540e1e7464185229189f4f1f1acadcd7cf3c2bf477bda5de7d18787",
-    "src/vpp2/learned_width_kv/data_scope.py": "0ddde6805f646afbf5f5966c53530f829becd083802f0e0a8697dc1004e2003e",
-    "src/vpp2/learned_width_kv/distributed_check.py": "f89f153a20730a6bc3c70133479f34918d30c5a48775bb3f6c14bb25701801ce",
-    "src/vpp2/learned_width_kv/gradient_probe.py": "3bed964f6488e95b386cc58058e7f28aff50c9a98dd526f3365891a9049a6fab",
-    "src/vpp2/learned_width_kv/lifecycle.py": "798afa47ec5dcc8749982da2d7853eff184e450a32d51bcf70eb4578f2bb0b09",
-    "src/vpp2/learned_width_kv/masking.py": "3272ea09271130dcb3cab6935f6bb7d61bda66b9982c3baabcd61ee0efd914a0",
-    "src/vpp2/learned_width_kv/performance.py": "a2fd9597c51bf0cf5c435b48e8f57792df24d2723a6d7b6b841d6fa24f17ab2f",
-    "src/vpp2/learned_width_kv/reliability.py": "f74399fb67427093b3e4c7df9749379f9291b7ac58bd6632db226b6e78fdbfd0",
-    "src/vpp2/learned_width_kv/reporting.py": "f22c1d8806a5b7757bb6dd99b15153d0305b2eeb0790bb37763337690d642036",
-    "src/vpp2/learned_width_kv/runtime.py": "513da1e06ba9509164f017e0b8dbf6b219d7e1cff82edb2d2fd8d2d7839e666f",
-    "src/vpp2/learned_width_kv/train.py": "5813ebd4fefec6d3a8cead9e0c645d10734f5ec88c4d74a5742587f2eb4de6c3",
-    "src/vpp2/learned_width_kv/validation.py": "d5594ce875c8011bff12d3c77083dca68878178ed022b8079761267d04a82386",
-    "src/vpp2/learned_width_kv/verification.py": "1bc43d48ef0ec3056af614adf43ca0744d94b1e77fe970168fef87c7cf2e6fe8",
-    "src/vpp2/libero/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/libero/action_model.py": "178fbe4cf02b9a2e133ef31389b6bb2f9baf0504bd0818a265b6bff32ebc3914",
-    "src/vpp2/libero/cli.py": "fcf68a88e2db0a9b786c6199d9b719802fe4b84514741658b49e2b06e2473e45",
-    "src/vpp2/libero/data.py": "19b699381e7d6fa769b5090c7125ae01b7584576e969468660ee65b1585ef5f1",
-    "src/vpp2/libero/diagnostics.py": "9540e31dfd2def06bec942ffe73262278700a17a340e14cb5bd41a295706e147",
-    "src/vpp2/libero/eval_helpers.py": "ac338d4ade195ceb3a87620d162befa9679e2d7ed3d91b09927e5d59d86d97f4",
-    "src/vpp2/libero/eval_libero_single.py": "87e6b457dc412ef4fe5d01aa48b1a6d20a28dbce3152848042e2103769e4efef",
-    "src/vpp2/libero/eval_libero_task_manifest.py": "f0c721a7ddc768edddc8b0e065a4d8aeec35e1a546c88deebd9097ad93f6227b",
-    "src/vpp2/libero/evaluate.py": "24216f9814f8304e077d72ff8d8c778fd588239313b34cd7d23b960c8f924c67",
-    "src/vpp2/libero/libero_utils.py": "c7cc23976c86eabcada4846d46345d3605fc2ce03929761646486be3901697d4",
-    "src/vpp2/libero/prepare_pro.py": "0999050352d924c11d29b22f63e902162d8cfb23f94815c93c01295c11ed81ae",
-    "src/vpp2/libero/pro_alias.py": "90b5c8c8061bc727a6e413a8746fcc98cbd06e1c7c4175fd223b45af4f10b700",
-    "src/vpp2/libero/pro_runtime.py": "adf94ed64b032aaeb0d88391e7b653f88f67e76b9411419ece25faa31ccc1340",
-    "src/vpp2/libero/protocols.py": "68225c95eeeee25032966a7f90c1ba18067c53225397f59a463d764030fd0d93",
-    "src/vpp2/libero/runtime.py": "1e2e2900a604d33cbdf8774cfa33dcee1065f21dc0b59fcf30799b5098c1662a",
-    "src/vpp2/libero/trainer.py": "ea29b4e000047a6b0c8aee3cb6f32df9187987a657924085dbafd9628315107b",
-    "src/vpp2/libero/worker.py": "9df6350c4ab92dea70863c08f1bdcebd0d77574a3a2d8d7ac9899a5a825be6b0",
-    "src/vpp2/models/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/models/wan21_14b/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/models/wan21_14b/action_dit.py": "9a654edd860ce77c1308923bd338c7f14d09f68686642531d7143c16b9f6f331",
-    "src/vpp2/models/wan21_14b/helpers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/models/wan21_14b/helpers/compact_channels.py": "6f7ad787624844970985117002ab809bd777f7474969daca121e7101381c6d0c",
-    "src/vpp2/models/wan21_14b/helpers/gradient.py": "594249b6eb931048e352fd3ffef9b1ac1db9e4fc74659e04216d6b823dd174f3",
-    "src/vpp2/models/wan21_14b/helpers/io.py": "1b68cf2f2fd674c7a59140f8632d4464044bffc2e9277ef5f9a975e7c699966a",
-    "src/vpp2/models/wan21_14b/helpers/loader.py": "b9180b9c9177af192e5cb92387b4a09bd0dbc08fb75a94e843b67d315c94e339",
-    "src/vpp2/models/wan21_14b/mot.py": "cc14a909bfb181a441bc9d8d6de12f3c31d89f3dc3415031e34a5f8c21410019",
-    "src/vpp2/models/wan21_14b/schedulers/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/models/wan21_14b/schedulers/scheduler_continuous.py": "15de04574c1d8ccb794ca6560b1e501e5c83cc86104bb80f1eb3cc3b0324ffe0",
-    "src/vpp2/models/wan21_14b/vpp2.py": "8c419af905a8ffa93d77c827a283b7aeff744f3a461c390783eef8a333c31756",
-    "src/vpp2/models/wan21_14b/wan21.py": "01b45cc71f69a6914f7b286559317c8e8485d0b0fec93bb28ebc37c183c295f4",
-    "src/vpp2/models/wan21_14b/wan_video_clip.py": "f65e44c9d1b7811d2996b3fa22e6fe4c2ac5e6edf1fc2c14251cf582bf8d1f05",
-    "src/vpp2/models/wan21_14b/wan_video_dit.py": "0d1b5837ea923508c05394d56aaa465813e7b4f5aea3b31ceceae39b2de9b0f9",
-    "src/vpp2/models/wan21_14b/wan_video_text_encoder.py": "902a1e46b21b5c48e166fe03defabb94b3773c6bbe836fc09b9a0d0894eb1f6e",
-    "src/vpp2/models/wan21_14b/wan_video_vae.py": "33f7d804677b44a6f3ab334684017dcdb1e9dfb748279de76ad94a5bb71a7e51",
-    "src/vpp2/policy.py": "2a7696cd73ccb99575bf901185cf578c3775945f61f6e67aeb2b4ec6bfc4e30b",
-    "src/vpp2/preflight.py": "7a91e486675a4cdf38565e0f4b58e60a68bba7de5654cfff736b5df58f5c53a6",
-    "src/vpp2/robodojo.py": "183db869a622ef17ee92c487949db2e22a7d1365558fe703bfb8dd8fef83e9c6",
-    "src/vpp2/runtime.py": "e26885a3732f5fdac3d8381da870a1673d0c03354f02f9deee7fe779d645f8a7",
-    "src/vpp2/trainer.py": "feaa0ac4ed94a376836eb60cfe6f52460392b11f7cafafec9c6d9f151f5b61f1",
-    "src/vpp2/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-    "src/vpp2/utils/conditioning.py": "43d43f0f3ca0add64498c3d64a3791513a53e3fce0ea53f77ef09580a86b7b2d",
-    "src/vpp2/utils/fs.py": "69e6df236b43360428a613686c8f37c531c0d16a3a2c5b06a725c46b49d657c5",
-    "src/vpp2/utils/logging_config.py": "1de4abf66d9d201b49e7bff48291643b9dc5ec9b6e7918ca73e1bd18887b1c43",
-    "src/vpp2/utils/lr_continuation.py": "72ea9f2af52041db19aab681d7a99fdf6d87d5ed5486f2499383a211c6c1ab76",
-    "src/vpp2/utils/misc.py": "7e42c330c24604e118f1878473dc804aa8a272944df502ef1414c05be4d5bba2",
-    "src/vpp2/utils/pytorch_utils.py": "9b0ba19de9a64458b12a0fb24f0811f49d9e9fca5d8b815f5f5ecd3b7886fd63",
-    "src/vpp2/utils/samplers.py": "4d3bd9b2d498f4ad8f2cf9765d60c4bf5368fa6fa2a7d42d0db6210f66899652",
-    "src/vpp2/utils/training_profiler.py": "91f6d98577e63d85e474c738a9bc1901edc194bdf318b5387d527c31b6708832",
-    "src/vpp2/utils/video_caption.py": "d9572deb3690aecb0f16c052016cedc5f94da8ce8f016b4dff1a4e358e18d5e5",
-    "src/vpp2/utils/video_io.py": "9e9898b021a44e87aa1b4e6d13e3755a9869a2e27a7eb53c72a6802808d5bd80",
-    "src/vpp2/utils/video_metrics.py": "1840a37685d686336c7326a2a3d4fbe048d1e8e7e431fb14519d360a8b7aaae8",
-    "src/vpp2/video_training.py": "20f55bf8b0a7dba6422a1115be1c8adf31a0db6a9f53a8f6f53f5d970188bfb2",
-    "src/vpp2/weight_resize.py": "269edd4b9d32a1c7f363ee082dde7ce57f2176190daab6d5bf4a1c98ab43f0b4",
-    "src/vpp2/weights.py": "5dbdff347de1b57e30433efca6da28334d0263d27aaa9b2278d95614db1b8641",
-    "src/vpp2/width_kv/__init__.py": "0183a7106da7e45223550a552db2905d6b59c2a4259f8c86cb98629b03e66b43",
-    "src/vpp2/width_kv/calibrate.py": "93a06910cb23a9e1dbd961ee9ef9794d70aaaa6186b08483260c8215109598f5",
-    "src/vpp2/width_kv/common.py": "45f5f7087f5f3dc067358838e3fc738698b796defecb2ea721545d275f767766",
-    "src/vpp2/width_kv/diagnostics.py": "cf184ae58f5e1c5f4453774eee1e10c85931e618bbbf768ab42f649cf301f5f1",
-    "src/vpp2/width_kv/reproducibility.py": "89c5377e1314dd6e8de7f67c4385d30f5bd91fa4a5742449f1c1ff8dd578f451",
-    "src/vpp2/width_kv/scoring.py": "7644358d0b2d29d3b91515fcea56090dbd2a7a0dad2d5e5c88e8d585bff15533",
-    "src/vpp2/width_kv/slicing.py": "856203c14e0b06d19585a96392d00a00703779ebdb9dbdc9ab1323d50daf8ba4",
-    "tests/adaptive_mask/test_core.py": "af02f3e6d3ef682dfeca4a3b1fbfa8abe2cdd1827eab40f29fac9a260a2b6a4f",
-    "tests/adaptive_mask/test_equivalence.py": "a3a39a950bab5bf22620f13b893e36b2f69edd545e7305c0e24db5611f99cb7a",
-    "tests/adaptive_mask/test_execution.py": "d1fae7d45d4fc5885cbcb9448848d54df21936570b89b42ddf13eeb225fe2429",
-    "tests/adaptive_mask/test_execution_cache.py": "10305169b0138082966faaf2949fbff0d7b6cb472c5b67f2938d627b82495637",
-    "tests/adaptive_mask/test_feedback.py": "a78e5960437b1cb398103ef5f94fe38257bdbdeec23e668d8546e08f9ba65ed5",
-    "tests/adaptive_mask/test_hidden.py": "090b5260e5919524725915bab5acc4569641a079639b816de785a45c2db38158",
-    "tests/adaptive_mask/test_physical.py": "1064d96d8fd9ade92c7189d2bab3cb4df9e205cd7d309a7d3bded4a8487b782e",
-    "tests/adaptive_mask/test_priority.py": "db50aeacd7b17660040e377321b379c824ad70cf62186b7683b2f57408d9e583",
-    "tests/adaptive_mask/test_quality_policy.py": "36cab8d346c3a5858b3d01064c66d8be2094f82c1fd90ceefa1ef0c59aeb74fc",
-    "tests/adaptive_mask/test_reporting.py": "7b4ca8fc420811fc50693df2f3a5956eb775c2256faae89049a2e0a053b9f0e7",
-    "tests/adaptive_mask/test_state_data.py": "f6269ba9fa4db5ced8e766616bf521c35f69d184618b0ca0a7c25f6978622d58",
-    "tests/adaptive_mask/test_training_control_flow.py": "3bb6bbf7a29d41ec36e47834441e38009a8035a370cb33bc5410519e06f64482",
-    "tests/critical_windows/test_contracts_metrics.py": "6c549378fe88fd94b7c74d62c063b9c9f707e3566472b91d957544db41c98502",
-    "tests/critical_windows/test_evidence.py": "6c5bc4fd71cc64920be771f11a222ac2bb3ae5ac6a4f9387c59b143b01c6ddbf",
-    "tests/critical_windows/test_inference_contract.py": "e50a8af35c8afda1ae634f5b3a60843f31c8f5b67afc8a5e96164341d01e962a",
-    "tests/critical_windows/test_packing.py": "524a2dde8de5f130dbba43ff0068a1da88b3eedb762d11179dc5d3cc8a3ab64e",
-    "tests/critical_windows/test_score.py": "e6476d108bbf342466d737da50c57e44f3d06fb1d363a46983584126ebe7d747",
-    "tests/critical_windows/test_selection.py": "19a173032b7a0b0afdd0d1c81f0780b282eacebad28fdb8bb8d28200f33aa36e",
-    "tests/critical_windows/test_staged_integration.py": "5057203be75fd58debdbe07b87a41a5e2b19fe68dcc7fe1e6a2d7450582648d4",
-    "tests/critical_windows/test_staged_review.py": "294e4cce568baa6e12eea00ab94b95a2cdbce3e8c83292a332f48b4651caa322",
-    "tests/flexible_mask/test_branch_best.py": "c654dc19650b9ff782e60dea7684d39513d949b8fc57d4a08d1cef2db4b8f4c0",
-    "tests/flexible_mask/test_branch_integrity.py": "c26921e3fe9735c2124026b5bd962cb75fe66865299a34f3a87e7c6fa395d4d0",
-    "tests/flexible_mask/test_calibration_batch.py": "6f301271ffabf355adc3440c3403a3eb36716007b75809a92d039ac7ee5de208",
-    "tests/flexible_mask/test_checkpoint_contracts.py": "e81c44855d1e75050d30d059f244a137ea22a1ee97fa72f9ce1a09e01a9abe3f",
-    "tests/flexible_mask/test_deferred_metrics.py": "11bf7c6b0a52a7c1faf91b4d740eda0dc50954e18a4ce88e3ba46a7d0e256884",
-    "tests/flexible_mask/test_diagnostic_indices.py": "f0acd2eabbec17c028ac54092cb4e38fe7f3a3f2278376188c6ce64cdc2f7b77",
-    "tests/flexible_mask/test_distributed_bootstrap.py": "599f12d24ba4350400b8500e4f99ecc454268c3cf1cec65f4978d5e6ac52fccd",
-    "tests/flexible_mask/test_early_recovery.py": "73b4fa0bfe6bfc29a6d69f4f5b2aed579335939a8dcaf49e0f94af548e42b58c",
-    "tests/flexible_mask/test_execution_optimizations.py": "7ec90d1dbe6f5d3c870ab8bec6845ff8340ad1e7b5c84e499f4f1cace8f0b813",
-    "tests/flexible_mask/test_flexible_contracts.py": "a124723c6ccf69c020ac292d2ea962260aa7f8c4c40f4a057fc71e8fc02d25f3",
-    "tests/flexible_mask/test_lossless_checkpoint.py": "c32052278ec3e75e77183d12a54d827c4dc8851ba7594f64c431d5407fa51745",
-    "tests/flexible_mask/test_no_kv.py": "05fec14834fa98d750a25a517229276de929ad2a0c3c73fe44b3cc8807be0c61",
-    "tests/flexible_mask/test_pressure_continuation.py": "af6f766ee98cb2834db22cadfc5007ecd5084f189b32cb68bce297dcc17abe70",
-    "tests/flexible_mask/test_recovery_extension.py": "123ee62a7eb3be2b70daececfbcf8cf693fcf339f43406c338145299941c0c4f",
-    "tests/flexible_mask/test_recovery_subset.py": "0b381c979631633be0d6794b5d06e6d59459665a46f09c34323bca2366fd1ea3",
-    "tests/flexible_mask/test_resident_teacher.py": "a3e0652e6f5c4f732a165c4488033e88199db2e1190a2b4a5a9e56dd4a3142f7",
-    "tests/flexible_mask/test_speed_review.py": "45877539f9f7c6a6b343305831cf47e8c99577d2eba76b0505e95d60f82f0a4f",
-    "tests/flexible_mask/test_throughput_continuation.py": "e835c70061bdfc60b488914c3605352125bce48e56a5f7682f5b7483a6bdd081",
-    "tests/full_mask/test_contracts.py": "678db4c71703498fab3b01e977ac89fc655daa3d64d91a46ebc2dce3c1c46b49",
-    "tests/full_mask/test_curated_inputs.py": "47e4b63a896b20c863578a36d36d59ff07fdc06e141247a8f0eb51f0b9960f14",
-    "tests/full_mask/test_curation.py": "32964ac87ce900e8dc9dd3c27064420fb206bed42e42197fcc3bfe2d92407633",
-    "tests/full_mask/test_curation_v3.py": "49567dd73c9b0f35d1c931a5b9fea95659b5af744f81dd56a11acbae9cd1e85a",
-    "tests/full_mask/test_curation_v3_integration.py": "383f657797e9178d4e4d63cd6d081b6b22487d1181ffc3ca2b8715ab44309549",
-    "tests/full_mask/test_curation_v3_probe.py": "ded6f849b183fe2e457a0224c0807a9ce6d210032a7bbd8e7c4f246c700b5ed7",
-    "tests/full_mask/test_curation_v3_quality.py": "6a62ff8e3a7346038b92a9ebf75616bef3d6ebd20ddacea6521e90421d23a35d",
-    "tests/full_mask/test_mask_algebra.py": "e053f5eb699efda323d9a533a09a45a5c57d311e0b4525c25c2a4731923df7ab",
-    "tests/full_mask/test_readiness_monitor.py": "1d5777010e28801aa29670b2cb6363875529ad6cc40435d9c18307e1a4ac5005",
-    "tests/mask_search/test_search_readiness.py": "15eca541947806aa898869fd32540102e6fb509d6b47dc801cb4aed5db1668aa"
-  },
-  "identity": "790bffc9464fa7a026aa43b217aec2033d186895cad643166daa0bcc01914d5c"
-}