Upload settings

.gitattributes

@@ -36,3 +36,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 proposals/residual_temporal_proposal_k8_alpha0.2.json filter=lfs diff=lfs merge=lfs -text
 manifests/content_fingerprints.json filter=lfs diff=lfs merge=lfs -text
 manifests/valid_anchor_manifest_florence_scratch_v1.json filter=lfs diff=lfs merge=lfs -text
+settings/florence_bridge/frameskip/valid_anchor_manifest.json filter=lfs diff=lfs merge=lfs -text
+settings/florence_robotwin/uniform/valid_anchor_manifest.json filter=lfs diff=lfs merge=lfs -text
+settings/qwen_robotwin/common_start/valid_anchor_manifest.json filter=lfs diff=lfs merge=lfs -text
+settings/qwen_robotwin/residual/valid_anchor_manifest.json filter=lfs diff=lfs merge=lfs -text
+settings/qwen_robotwin/uniform/valid_anchor_manifest.json filter=lfs diff=lfs merge=lfs -text

settings/florence_bridge/frameskip/config.full.yaml

@@ -0,0 +1,116 @@
+run_id: frameskip-adapted-gaussian-tpi-r02-seed-117465
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/external-baselines/frameskip-adapted-gaussian-tpi-r02-20260825
+seed: 117465
+wandb_entity: null
+wandb_project: InfoFlow-StarVLA-BridgeRT
+is_debug: false
+version_id: '0.21'
+framework:
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 1024
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 7
+    state_dim: 7
+    action_horizon: 16
+    repeated_diffusion_steps: 8
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    future_action_window_size: 15
+    past_action_window_size: 0
+  dino:
+    dino_backbone: dinov2_vits14
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/OXE_LEROBOT_DATASET
+    data_mix: bridge_rt_infoflow
+    action_type: delta_ee
+    action_mode: abs
+    precomputed_statistics_path: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-GR00T-Bridge-RT-1/dataset_statistics.json
+    precomputed_statistics_key_map:
+      bridge_orig_lerobot: oxe_bridge
+      fractal20220817_data_lerobot: oxe_rt1
+    CoT_prompt: Your task is {instruction}. To identify the key objects for your task.
+      Locate their bounding boxes in [x1,y1,x2,y2] format.
+    per_device_batch_size: 4
+    load_all_data_for_training: true
+    obs_image_size:
+    - 224
+    - 224
+    video_backend: torchvision_av
+    num_workers: 4
+    balance_dataset_weights: false
+    balance_trajectory_weights: false
+    traceable_sampling: true
+    expected_world_size: 2
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/bridgert/data_provenance.json
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/content_fingerprints.json
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/valid_anchor_manifest_florence_scratch_v1.json
+    valid_anchor_manifest_migration_path: /home/miaochengyu/starVLA-infoflow-codex/external_baselines/frameskip_adapted/preflight/valid_anchor_manifest_migration_20260825.json
+    temporal_proposal_path: null
+    transform_rng_seed: 298226
+trainer:
+  use_deepspeed: true
+  max_train_steps: 100000
+  num_warmup_steps: 5000
+  save_interval: null
+  save_steps:
+  - 20500
+  evaluation_snapshot_interval: 1000
+  eval_interval: 100001
+  repeated_diffusion_steps: 4
+  pretrained_checkpoint: null
+  is_resume: true
+  allow_model_only_resume: false
+  learning_rate:
+    base: 3.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: ''
+  trainable_modules: ''
+  frozen_modules_eval: false
+  logging_frequency: 50
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 8
+  save_format: pt
+  save_final_checkpoint: false
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  resume_accelerate_checkpoint: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-scratch-official-100k/florence-scratch-uniform-seed-42-official100k-mb4-eval1k-20260726/checkpoints/steps_20000_accelerate
+  resume_checkpoint_integrity_manifest: /home/miaochengyu/starVLA-infoflow-codex/protocols/bridgert/common_start_accelerate_integrity_20260730.json
+  resume_checkpoint_integrity_manifest_sha256: 899d2cef3f73191e11f7a1628afc05b60c80af5816b8fd77c1c69e90d5c9ff39
+  branch_parent_seed: 42
+  branch_rng_seed: 117465
+  model_rng_seed: 588643
+  stop_after_steps: 20500
+config_yaml: /home/miaochengyu/starVLA-infoflow-codex/external_baselines/frameskip_adapted/configs/florence_bridge_frameskip_gaussian_r02_seed117465.yaml
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/external-baselines/frameskip-adapted-gaussian-tpi-r02-20260825/frameskip-adapted-gaussian-tpi-r02-seed-117465

settings/florence_bridge/frameskip/config.yaml

@@ -0,0 +1,71 @@
+datasets:
+  vla_data:
+    action_mode: abs
+    balance_dataset_weights: false
+    balance_trajectory_weights: false
+    data_mix: bridge_rt_infoflow
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/OXE_LEROBOT_DATASET
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/content_fingerprints.json
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/bridgert/data_provenance.json
+    dataset_py: lerobot_datasets
+    expected_world_size: 2
+    num_workers: 4
+    per_device_batch_size: 4
+    precomputed_statistics_key_map:
+      bridge_orig_lerobot: oxe_bridge
+      fractal20220817_data_lerobot: oxe_rt1
+    precomputed_statistics_path: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-GR00T-Bridge-RT-1/dataset_statistics.json
+    temporal_proposal_path: null
+    traceable_sampling: true
+    valid_anchor_manifest_migration_path: /home/miaochengyu/starVLA-infoflow-codex/external_baselines/frameskip_adapted/preflight/valid_anchor_manifest_migration_20260825.json
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/valid_anchor_manifest_florence_scratch_v1.json
+    video_backend: torchvision_av
+framework:
+  action_model:
+    action_dim: 7
+    action_horizon: 16
+    action_model_type: DiT-B
+    add_pos_embed: true
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    hidden_size: 1024
+    max_seq_len: 1024
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    num_timestep_buckets: 1000
+    state_dim: 7
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/external-baselines/frameskip-adapted-gaussian-tpi-r02-20260825/frameskip-adapted-gaussian-tpi-r02-seed-117465
+run_id: frameskip-adapted-gaussian-tpi-r02-seed-117465
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/external-baselines/frameskip-adapted-gaussian-tpi-r02-20260825
+seed: 117465
+trainer:
+  freeze_modules: ''
+  gradient_accumulation_steps: 8
+  learning_rate:
+    action_model: 0.0001
+    base: 3.0e-05
+    qwen_vl_interface: 1.0e-05
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 100000
+  num_warmup_steps: 5000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    name: AdamW
+    weight_decay: 1.0e-08
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07

settings/florence_bridge/frameskip/dataset_statistics.json

@@ -0,0 +1,264 @@
+{
+  "oxe_bridge": {
+    "action": {
+      "mean": [
+        5.682967821485363e-05,
+        3.278055373812094e-05,
+        -3.1604828109266236e-05,
+        -3.6026765883434564e-05,
+        -9.757651423569769e-05,
+        6.01583014940843e-05,
+        0.14414729177951813
+      ],
+      "std": [
+        0.004886276229987901,
+        0.006847766456005231,
+        0.006337809696791937,
+        0.014227778584428946,
+        0.015261554008856799,
+        0.038695293138416895,
+        0.3521380562074972
+      ],
+      "max": [
+        0.41691166162490845,
+        0.25864794850349426,
+        0.21218234300613403,
+        3.122201919555664,
+        1.8618112802505493,
+        6.272472858428955,
+        1.0
+      ],
+      "min": [
+        -0.4007510244846344,
+        -0.13874775171279907,
+        -0.22553899884223938,
+        -3.2010786533355713,
+        -1.8618112802505493,
+        -6.279075622558594,
+        0.0
+      ],
+      "q01": [
+        -0.02875255048274994,
+        -0.04170213546603918,
+        -0.026096721179783344,
+        -0.08052874729037285,
+        -0.09249906800687313,
+        -0.20738555490970612,
+        0.0
+      ],
+      "q99": [
+        0.028306663036346436,
+        0.04089853074401617,
+        0.0401805154979229,
+        0.08173403143882751,
+        0.07760760560631752,
+        0.2038465365767479,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        0.07735871523618698,
+        0.007681275252252817,
+        0.016110140830278397,
+        0.0016226600855588913,
+        -0.019300073385238647,
+        0.02691473439335823,
+        0.0,
+        0.17703326046466827
+      ],
+      "std": [
+        0.13736907523153707,
+        0.047840646787850445,
+        0.03799073317180484,
+        0.06560660645733722,
+        0.09093375673607122,
+        0.292728734025679,
+        0.0,
+        0.3536876512395259
+      ],
+      "max": [
+        0.5862360596656799,
+        0.4034728705883026,
+        0.3568263053894043,
+        1.3517684936523438,
+        1.570796251296997,
+        3.141204357147217,
+        0.0,
+        1.1121242046356201
+      ],
+      "min": [
+        -0.04167502000927925,
+        -0.3563207685947418,
+        -0.15537554025650024,
+        -3.141592502593994,
+        -1.4992541074752808,
+        -3.14153790473938,
+        0.0,
+        0.04637829214334488
+      ],
+      "q01": [
+        0.17102622985839844,
+        -0.1698118858039379,
+        -0.055632163770496845,
+        -0.36493172496557236,
+        -0.541871190071106,
+        -1.354275494813919,
+        0.0,
+        0.052190229296684265
+      ],
+      "q99": [
+        0.45322076976299286,
+        0.2354845181107521,
+        0.19489620998501778,
+        0.3780156075954437,
+        0.27568644285202026,
+        1.8500566184520721,
+        0.0,
+        1.0105689764022827
+      ]
+    },
+    "num_transitions": 1305714,
+    "num_trajectories": 53192
+  },
+  "oxe_rt1": {
+    "action": {
+      "mean": [
+        0.0017468753503635526,
+        0.0015664808452129364,
+        -0.00315628363750875,
+        0.010832978412508965,
+        -0.0014390418073162436,
+        0.00022827448265161365,
+        0.13385511934757233
+      ],
+      "std": [
+        0.034738154754287556,
+        0.029950575875189896,
+        0.03716987948198757,
+        0.08027394458720655,
+        0.06586824211259418,
+        0.07296740097742303,
+        0.33989797865933385
+      ],
+      "max": [
+        2.9984593391418457,
+        22.09052848815918,
+        2.7507524490356445,
+        1.570636510848999,
+        1.5321086645126343,
+        1.5691522359848022,
+        1.0
+      ],
+      "min": [
+        -2.0204520225524902,
+        -5.497899532318115,
+        -2.031663417816162,
+        -1.569917917251587,
+        -1.569892168045044,
+        -1.570419430732727,
+        0.0
+      ],
+      "q01": [
+        -0.22453527510166169,
+        -0.14820013284683228,
+        -0.231589707583189,
+        -0.3517994859814644,
+        -0.4193011274933815,
+        -0.43643461108207704,
+        0.0
+      ],
+      "q99": [
+        0.17824687153100965,
+        0.14938379630446405,
+        0.21842354819178575,
+        0.5892666035890578,
+        0.35272657424211445,
+        0.44796681255102094,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        0.1399736851453781,
+        -0.02083534561097622,
+        0.19427375495433807,
+        0.10653364658355713,
+        -0.06201138719916344,
+        0.12378455698490143,
+        0.023165317252278328,
+        0.052437249571084976
+      ],
+      "std": [
+        0.2502841726583427,
+        0.06813404469049046,
+        0.358260305088155,
+        0.2931043604956031,
+        0.2779312663188148,
+        0.33769767306558707,
+        0.09232656626252377,
+        0.16498523970985932
+      ],
+      "max": [
+        1.0534898042678833,
+        0.48018959164619446,
+        1.6896663904190063,
+        1.0,
+        0.9999993443489075,
+        0.9999874830245972,
+        0.9554369449615479,
+        0.9914546012878418
+      ],
+      "min": [
+        -0.4436439275741577,
+        -0.9970501065254211,
+        -0.006579156965017319,
+        0.0,
+        -0.8643477559089661,
+        -0.7079970240592957,
+        -0.7688722014427185,
+        -0.4999994933605194
+      ],
+      "q01": [
+        0.32481380939483645,
+        -0.28334290891885755,
+        0.14107070609927178,
+        0.0,
+        -0.686474204659462,
+        -0.6808923494815826,
+        -0.36045596331357954,
+        -0.454380963742733
+      ],
+      "q99": [
+        0.8750156319141384,
+        0.21247054174542404,
+        1.0727112340927123,
+        1.0,
+        0.9377871316671368,
+        0.9563051050901409,
+        0.45990042358636823,
+        0.7216041100025177
+      ]
+    },
+    "num_transitions": 3786152,
+    "num_trajectories": 87212
+  }
+}

settings/florence_bridge/frameskip/external_resume_certificate.json

@@ -0,0 +1,132 @@
+{
+  "branch_resume_identity": {
+    "config_sha256": "3c6b4479da73caa22b9f37b00f88d319a0657aee51d1e8c313df225dba1c15b0",
+    "identity_schema_version": 2,
+    "manifest_sha256": "a1627df8ff110c7298eaf025ffcd4f48d9ca17117cfcfb3926b962d0bb19c0ca",
+    "model_structure_sha256": "e055e3fa0085edc6a9bf6cdb1605a05e61ae289957e24e37cfe2cfce2dc94c39",
+    "proposal_alpha": null,
+    "proposal_sha256": null,
+    "seed": 117465,
+    "total_batch_size": 64
+  },
+  "completed_steps": 20000,
+  "optimizer_param_groups": [
+    {
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "initial_lr": 1e-05,
+      "lr": 9.407412610098572e-06,
+      "name": "qwen_vl_interface",
+      "weight_decay": 1e-08
+    },
+    {
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "initial_lr": 0.0001,
+      "lr": 9.407412610098571e-05,
+      "name": "action_model",
+      "weight_decay": 1e-08
+    }
+  ],
+  "optimizer_state_entries": 4,
+  "parent_checkpoint": "/mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-scratch-official-100k/florence-scratch-uniform-seed-42-official100k-mb4-eval1k-20260726/checkpoints/steps_20000_accelerate",
+  "parent_checkpoint_commit_sha256": "caaaae0c1c2bc45c0b3b5c0ac8ef6f3bc60a5e189b0bcc9979dad1ba371ef3ad",
+  "parent_checkpoint_files": [
+    {
+      "bytes": 13,
+      "name": "latest",
+      "sha256": "68d0ca762047fd48fdd6a06a5e6465209289c028bfd63f5557a110ff1ebc15ac"
+    },
+    {
+      "bytes": 5514449457,
+      "name": "pytorch_model/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt",
+      "sha256": "42263278dd16848dd50fd159f430bb40efa541d8b62aba5d608531d3af90f1d4"
+    },
+    {
+      "bytes": 5514463601,
+      "name": "pytorch_model/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt",
+      "sha256": "c6e21653b454605b3cb3db62bf9b1bae90ddf63eea1389cc0a0dd351f1f0839c"
+    },
+    {
+      "bytes": 1839089981,
+      "name": "pytorch_model/mp_rank_00_model_states.pt",
+      "sha256": "822f82ffa3ebb7f8e1e9cb15f7c24dafb4054d9fb2b63646fb81ba63172c09b6"
+    },
+    {
+      "bytes": 2154184469,
+      "name": "pytorch_model.pt",
+      "sha256": "c4f39cd339271347b1ab83ddc9c8dae7e07e315bea8cf86812b2e0b6772abd42"
+    },
+    {
+      "bytes": 15017,
+      "name": "random_states_0.pkl",
+      "sha256": "29e1b8cf5838519c544dfe660c71c064b59cd35d8f5c5ae9fa1e63e9cfd3e773"
+    },
+    {
+      "bytes": 15081,
+      "name": "random_states_1.pkl",
+      "sha256": "fed527bb36c43639d19411c7da8f3eb3eca04543ed7eb06f62c1fa978f3bdf31"
+    },
+    {
+      "bytes": 33270,
+      "name": "zero_to_fp32.py",
+      "sha256": "e8a2553b434ac043ec63b022db1e768277acf8d4e70f10a2842db6d238977135"
+    }
+  ],
+  "parent_checkpoint_files_sha256": "223f8085e17142c10429b17ba5be21fee02fb5ae49bd409dd4f13f6b9caab867",
+  "parent_checkpoint_integrity_manifest_sha256": "899d2cef3f73191e11f7a1628afc05b60c80af5816b8fd77c1c69e90d5c9ff39",
+  "parent_config_sha256": "c2ea33c16faf5b550708a494df5600198ef7e8d15bece8dffe83609efa748993",
+  "parent_resume_identity": {
+    "config_sha256": "1c394b6043c323133b0512d3efd761ae8332e5098969864f410b71651aa56f44",
+    "manifest_sha256": "a1627df8ff110c7298eaf025ffcd4f48d9ca17117cfcfb3926b962d0bb19c0ca",
+    "model_structure_sha256": "e055e3fa0085edc6a9bf6cdb1605a05e61ae289957e24e37cfe2cfce2dc94c39",
+    "proposal_alpha": null,
+    "proposal_sha256": null,
+    "seed": 42,
+    "total_batch_size": 64
+  },
+  "proposal_sha256": null,
+  "rank_state_summaries": [
+    {
+      "optimizer_state_entries": 2,
+      "optimizer_state_sha256": "704f68d9f3c71b48c4f4ded7dd46cc6396324adad50611368f6b37f2df38e2f3",
+      "rank": 0,
+      "trainable_state_sha256": "e6b7d1c26f4ecb3d761ca3f1eaa21b7cf4be87def2c5dcef71b2860ff69fb09e"
+    },
+    {
+      "optimizer_state_entries": 2,
+      "optimizer_state_sha256": "96c9f5c983020c067f752234588992862a29feba7cc51804b9baaf3a25dcf98c",
+      "rank": 1,
+      "trainable_state_sha256": "e6b7d1c26f4ecb3d761ca3f1eaa21b7cf4be87def2c5dcef71b2860ff69fb09e"
+    }
+  ],
+  "repeated_diffusion_steps": 4,
+  "rng_streams": {
+    "model_base": 588643,
+    "sampler": 117465,
+    "transform": 298226
+  },
+  "scheduler": {
+    "class": "LambdaLR",
+    "last_epoch": 20000,
+    "last_lr": [
+      9.407412610098572e-06,
+      9.407412610098571e-05
+    ],
+    "step_count": 20001
+  },
+  "scheduler_resume_mode": "reconstructed_from_completed_steps",
+  "schema_version": 1,
+  "status": "exact_external_resume_loaded",
+  "total_batch_size": 64,
+  "trainable_parameter_elements": 919063815,
+  "trainable_parameter_sha256": "f2572c4f7913235cc889bdf27eca7e605bdaa4d68e92c0a32148bbf6decbd998",
+  "trainable_parameter_tensors": 1164,
+  "world_size": 2
+}

settings/florence_bridge/frameskip/initialization_manifest.json

@@ -0,0 +1,6 @@
+{
+  "schema_version": 1,
+  "initialization_seed": 117465,
+  "action_model_sha256": "fbe26e8054c77b62d41353672000d33e47bad277cc0f368bb3f0db571e847682",
+  "action_model_parameter_elements": 142598407
+}

settings/florence_bridge/frameskip/valid_anchor_manifest.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f2123722a291f54ca3414d640f4d9e504ec510521ac5ca620b4745e713cdb8c
+size 101950249

settings/florence_bridge/keyframe/config.full.yaml

@@ -0,0 +1,117 @@
+run_id: keyframe-focused-adapted-seed-117465-20k-to-20500
+run_root_dir: /home/liuyu/student-workspace/outputs/keyframe_focused_adapted/runs
+seed: 117465
+wandb_entity: null
+wandb_project: Keyframe-Focused-adapted-BridgeRT
+is_debug: false
+version_id: '0.21'
+framework:
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 1024
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 7
+    state_dim: 7
+    action_horizon: 16
+    repeated_diffusion_steps: 8
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    future_action_window_size: 15
+    past_action_window_size: 0
+  dino:
+    dino_backbone: dinov2_vits14
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/OXE_LEROBOT_DATASET
+    data_mix: bridge_rt_infoflow
+    action_type: delta_ee
+    action_mode: abs
+    precomputed_statistics_path: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-GR00T-Bridge-RT-1/dataset_statistics.json
+    precomputed_statistics_key_map:
+      bridge_orig_lerobot: oxe_bridge
+      fractal20220817_data_lerobot: oxe_rt1
+    CoT_prompt: Your task is {instruction}. To identify the key objects for your task.
+      Locate their bounding boxes in [x1,y1,x2,y2] format.
+    per_device_batch_size: 4
+    load_all_data_for_training: true
+    obs_image_size:
+    - 224
+    - 224
+    video_backend: torchvision_av
+    num_workers: 4
+    balance_dataset_weights: false
+    balance_trajectory_weights: false
+    traceable_sampling: true
+    expected_world_size: 2
+    dataset_provenance_path: /home/liuyu/student-workspace/code/starVLA-infoflow-codex/protocols/bridgert/data_provenance.json
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/content_fingerprints.json
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/valid_anchor_manifest_florence_scratch_v1.json
+    valid_anchor_manifest_migration_path: null
+    temporal_proposal_path: null
+    transform_rng_seed: 298226
+trainer:
+  use_deepspeed: true
+  max_train_steps: 100000
+  num_warmup_steps: 5000
+  save_interval: null
+  save_steps:
+  - 20500
+  evaluation_snapshot_interval: 1000
+  eval_interval: 100001
+  repeated_diffusion_steps: 4
+  pretrained_checkpoint: null
+  is_resume: true
+  allow_model_only_resume: false
+  learning_rate:
+    base: 3.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: ''
+  trainable_modules: ''
+  frozen_modules_eval: false
+  logging_frequency: 50
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 8
+  save_format: pt
+  save_final_checkpoint: false
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  resume_accelerate_checkpoint: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-scratch-official-100k/florence-scratch-uniform-seed-42-official100k-mb4-eval1k-20260726/checkpoints/steps_20000_accelerate
+  resume_checkpoint_integrity_manifest: /home/liuyu/student-workspace/code/starVLA-infoflow-codex/protocols/bridgert/common_start_accelerate_integrity_20260730.json
+  resume_checkpoint_integrity_manifest_sha256: 899d2cef3f73191e11f7a1628afc05b60c80af5816b8fd77c1c69e90d5c9ff39
+  branch_parent_seed: 42
+  branch_rng_seed: 117465
+  model_rng_seed: 588643
+  sample_loss_weight_key: keyframe_weight
+  stop_after_steps: 20500
+config_yaml: /home/liuyu/student-workspace/code/starVLA-infoflow-codex/external_baselines/keyframe_focused_adapted/configs/keyframe_focused_seed117465_20k_to_20500.yaml
+output_dir: /home/liuyu/student-workspace/outputs/keyframe_focused_adapted/runs/keyframe-focused-adapted-seed-117465-20k-to-20500

settings/florence_bridge/keyframe/config.yaml

@@ -0,0 +1,70 @@
+datasets:
+  vla_data:
+    action_mode: abs
+    balance_dataset_weights: false
+    balance_trajectory_weights: false
+    data_mix: bridge_rt_infoflow
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/OXE_LEROBOT_DATASET
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/content_fingerprints.json
+    dataset_provenance_path: /home/liuyu/student-workspace/code/starVLA-infoflow-codex/protocols/bridgert/data_provenance.json
+    dataset_py: lerobot_datasets
+    expected_world_size: 2
+    num_workers: 4
+    per_device_batch_size: 4
+    precomputed_statistics_key_map:
+      bridge_orig_lerobot: oxe_bridge
+      fractal20220817_data_lerobot: oxe_rt1
+    precomputed_statistics_path: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-GR00T-Bridge-RT-1/dataset_statistics.json
+    temporal_proposal_path: null
+    traceable_sampling: true
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/valid_anchor_manifest_florence_scratch_v1.json
+    video_backend: torchvision_av
+framework:
+  action_model:
+    action_dim: 7
+    action_horizon: 16
+    action_model_type: DiT-B
+    add_pos_embed: true
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    hidden_size: 1024
+    max_seq_len: 1024
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    num_timestep_buckets: 1000
+    state_dim: 7
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+output_dir: /home/liuyu/student-workspace/outputs/keyframe_focused_adapted/runs/keyframe-focused-adapted-seed-117465-20k-to-20500
+run_id: keyframe-focused-adapted-seed-117465-20k-to-20500
+run_root_dir: /home/liuyu/student-workspace/outputs/keyframe_focused_adapted/runs
+seed: 117465
+trainer:
+  freeze_modules: ''
+  gradient_accumulation_steps: 8
+  learning_rate:
+    action_model: 0.0001
+    base: 3.0e-05
+    qwen_vl_interface: 1.0e-05
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 100000
+  num_warmup_steps: 5000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    name: AdamW
+    weight_decay: 1.0e-08
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07

settings/florence_bridge/keyframe/dataset_statistics.json

@@ -0,0 +1,264 @@
+{
+  "oxe_bridge": {
+    "action": {
+      "mean": [
+        5.682967821485363e-05,
+        3.278055373812094e-05,
+        -3.1604828109266236e-05,
+        -3.6026765883434564e-05,
+        -9.757651423569769e-05,
+        6.01583014940843e-05,
+        0.14414729177951813
+      ],
+      "std": [
+        0.004886276229987901,
+        0.006847766456005231,
+        0.006337809696791937,
+        0.014227778584428946,
+        0.015261554008856799,
+        0.038695293138416895,
+        0.3521380562074972
+      ],
+      "max": [
+        0.41691166162490845,
+        0.25864794850349426,
+        0.21218234300613403,
+        3.122201919555664,
+        1.8618112802505493,
+        6.272472858428955,
+        1.0
+      ],
+      "min": [
+        -0.4007510244846344,
+        -0.13874775171279907,
+        -0.22553899884223938,
+        -3.2010786533355713,
+        -1.8618112802505493,
+        -6.279075622558594,
+        0.0
+      ],
+      "q01": [
+        -0.02875255048274994,
+        -0.04170213546603918,
+        -0.026096721179783344,
+        -0.08052874729037285,
+        -0.09249906800687313,
+        -0.20738555490970612,
+        0.0
+      ],
+      "q99": [
+        0.028306663036346436,
+        0.04089853074401617,
+        0.0401805154979229,
+        0.08173403143882751,
+        0.07760760560631752,
+        0.2038465365767479,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        0.07735871523618698,
+        0.007681275252252817,
+        0.016110140830278397,
+        0.0016226600855588913,
+        -0.019300073385238647,
+        0.02691473439335823,
+        0.0,
+        0.17703326046466827
+      ],
+      "std": [
+        0.13736907523153707,
+        0.047840646787850445,
+        0.03799073317180484,
+        0.06560660645733722,
+        0.09093375673607122,
+        0.292728734025679,
+        0.0,
+        0.3536876512395259
+      ],
+      "max": [
+        0.5862360596656799,
+        0.4034728705883026,
+        0.3568263053894043,
+        1.3517684936523438,
+        1.570796251296997,
+        3.141204357147217,
+        0.0,
+        1.1121242046356201
+      ],
+      "min": [
+        -0.04167502000927925,
+        -0.3563207685947418,
+        -0.15537554025650024,
+        -3.141592502593994,
+        -1.4992541074752808,
+        -3.14153790473938,
+        0.0,
+        0.04637829214334488
+      ],
+      "q01": [
+        0.17102622985839844,
+        -0.1698118858039379,
+        -0.055632163770496845,
+        -0.36493172496557236,
+        -0.541871190071106,
+        -1.354275494813919,
+        0.0,
+        0.052190229296684265
+      ],
+      "q99": [
+        0.45322076976299286,
+        0.2354845181107521,
+        0.19489620998501778,
+        0.3780156075954437,
+        0.27568644285202026,
+        1.8500566184520721,
+        0.0,
+        1.0105689764022827
+      ]
+    },
+    "num_transitions": 1305714,
+    "num_trajectories": 53192
+  },
+  "oxe_rt1": {
+    "action": {
+      "mean": [
+        0.0017468753503635526,
+        0.0015664808452129364,
+        -0.00315628363750875,
+        0.010832978412508965,
+        -0.0014390418073162436,
+        0.00022827448265161365,
+        0.13385511934757233
+      ],
+      "std": [
+        0.034738154754287556,
+        0.029950575875189896,
+        0.03716987948198757,
+        0.08027394458720655,
+        0.06586824211259418,
+        0.07296740097742303,
+        0.33989797865933385
+      ],
+      "max": [
+        2.9984593391418457,
+        22.09052848815918,
+        2.7507524490356445,
+        1.570636510848999,
+        1.5321086645126343,
+        1.5691522359848022,
+        1.0
+      ],
+      "min": [
+        -2.0204520225524902,
+        -5.497899532318115,
+        -2.031663417816162,
+        -1.569917917251587,
+        -1.569892168045044,
+        -1.570419430732727,
+        0.0
+      ],
+      "q01": [
+        -0.22453527510166169,
+        -0.14820013284683228,
+        -0.231589707583189,
+        -0.3517994859814644,
+        -0.4193011274933815,
+        -0.43643461108207704,
+        0.0
+      ],
+      "q99": [
+        0.17824687153100965,
+        0.14938379630446405,
+        0.21842354819178575,
+        0.5892666035890578,
+        0.35272657424211445,
+        0.44796681255102094,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        0.1399736851453781,
+        -0.02083534561097622,
+        0.19427375495433807,
+        0.10653364658355713,
+        -0.06201138719916344,
+        0.12378455698490143,
+        0.023165317252278328,
+        0.052437249571084976
+      ],
+      "std": [
+        0.2502841726583427,
+        0.06813404469049046,
+        0.358260305088155,
+        0.2931043604956031,
+        0.2779312663188148,
+        0.33769767306558707,
+        0.09232656626252377,
+        0.16498523970985932
+      ],
+      "max": [
+        1.0534898042678833,
+        0.48018959164619446,
+        1.6896663904190063,
+        1.0,
+        0.9999993443489075,
+        0.9999874830245972,
+        0.9554369449615479,
+        0.9914546012878418
+      ],
+      "min": [
+        -0.4436439275741577,
+        -0.9970501065254211,
+        -0.006579156965017319,
+        0.0,
+        -0.8643477559089661,
+        -0.7079970240592957,
+        -0.7688722014427185,
+        -0.4999994933605194
+      ],
+      "q01": [
+        0.32481380939483645,
+        -0.28334290891885755,
+        0.14107070609927178,
+        0.0,
+        -0.686474204659462,
+        -0.6808923494815826,
+        -0.36045596331357954,
+        -0.454380963742733
+      ],
+      "q99": [
+        0.8750156319141384,
+        0.21247054174542404,
+        1.0727112340927123,
+        1.0,
+        0.9377871316671368,
+        0.9563051050901409,
+        0.45990042358636823,
+        0.7216041100025177
+      ]
+    },
+    "num_transitions": 3786152,
+    "num_trajectories": 87212
+  }
+}

settings/florence_bridge/loss_is/config.full.yaml

@@ -0,0 +1,123 @@
+run_id: loss-is-online-seed-117465-20k-to-20500
+run_root_dir: /home/liuyu/student-workspace/outputs/loss_is_adapted/runs
+seed: 117465
+wandb_entity: null
+wandb_project: Loss-IS-adapted-BridgeRT
+is_debug: false
+version_id: '0.21'
+framework:
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 1024
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 7
+    state_dim: 7
+    action_horizon: 16
+    repeated_diffusion_steps: 8
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    future_action_window_size: 15
+    past_action_window_size: 0
+  dino:
+    dino_backbone: dinov2_vits14
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/OXE_LEROBOT_DATASET
+    data_mix: bridge_rt_infoflow
+    action_type: delta_ee
+    action_mode: abs
+    precomputed_statistics_path: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-GR00T-Bridge-RT-1/dataset_statistics.json
+    precomputed_statistics_key_map:
+      bridge_orig_lerobot: oxe_bridge
+      fractal20220817_data_lerobot: oxe_rt1
+    CoT_prompt: Your task is {instruction}. To identify the key objects for your task.
+      Locate their bounding boxes in [x1,y1,x2,y2] format.
+    per_device_batch_size: 4
+    load_all_data_for_training: true
+    obs_image_size:
+    - 224
+    - 224
+    video_backend: torchvision_av
+    num_workers: 4
+    balance_dataset_weights: false
+    balance_trajectory_weights: false
+    traceable_sampling: true
+    expected_world_size: 2
+    dataset_provenance_path: /home/liuyu/student-workspace/code/starVLA-infoflow-codex/protocols/bridgert/data_provenance.json
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/content_fingerprints.json
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/valid_anchor_manifest_florence_scratch_v1.json
+    valid_anchor_manifest_migration_path: null
+    temporal_proposal_path: null
+    transform_rng_seed: 298226
+trainer:
+  use_deepspeed: true
+  max_train_steps: 100000
+  num_warmup_steps: 5000
+  save_interval: null
+  save_steps:
+  - 20500
+  evaluation_snapshot_interval: 1000
+  eval_interval: 100001
+  repeated_diffusion_steps: 4
+  pretrained_checkpoint: null
+  is_resume: true
+  allow_model_only_resume: false
+  learning_rate:
+    base: 3.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: ''
+  trainable_modules: ''
+  frozen_modules_eval: false
+  logging_frequency: 50
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 8
+  save_format: pt
+  save_final_checkpoint: false
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  resume_accelerate_checkpoint: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-scratch-official-100k/florence-scratch-uniform-seed-42-official100k-mb4-eval1k-20260726/checkpoints/steps_20000_accelerate
+  resume_checkpoint_integrity_manifest: /home/liuyu/student-workspace/code/starVLA-infoflow-codex/protocols/bridgert/common_start_accelerate_integrity_20260730.json
+  resume_checkpoint_integrity_manifest_sha256: 899d2cef3f73191e11f7a1628afc05b60c80af5816b8fd77c1c69e90d5c9ff39
+  branch_parent_seed: 42
+  branch_rng_seed: 117465
+  model_rng_seed: 588643
+  model_rng_trace_key: loss_is_occurrence_id
+  sample_loss_weight_key: loss_is_importance_weight
+  stop_after_steps: 20500
+  loss_is:
+    candidate_factor: 3
+    seed: 117465
+    condition_threshold: 2.0
+    condition_momentum: 0.9
+config_yaml: /home/liuyu/student-workspace/code/starVLA-loss-is-adapted/external_baselines/loss_is_adapted/configs/loss_is_seed117465_20k_to_20500.yaml
+output_dir: /home/liuyu/student-workspace/outputs/loss_is_adapted/runs/loss-is-online-seed-117465-20k-to-20500

settings/florence_bridge/loss_is/config.yaml

@@ -0,0 +1,75 @@
+datasets:
+  vla_data:
+    action_mode: abs
+    balance_dataset_weights: false
+    balance_trajectory_weights: false
+    data_mix: bridge_rt_infoflow
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/OXE_LEROBOT_DATASET
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/content_fingerprints.json
+    dataset_provenance_path: /home/liuyu/student-workspace/code/starVLA-infoflow-codex/protocols/bridgert/data_provenance.json
+    dataset_py: lerobot_datasets
+    expected_world_size: 2
+    num_workers: 4
+    per_device_batch_size: 4
+    precomputed_statistics_key_map:
+      bridge_orig_lerobot: oxe_bridge
+      fractal20220817_data_lerobot: oxe_rt1
+    precomputed_statistics_path: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-GR00T-Bridge-RT-1/dataset_statistics.json
+    temporal_proposal_path: null
+    traceable_sampling: true
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/bridgert/valid_anchor_manifest_florence_scratch_v1.json
+    video_backend: torchvision_av
+framework:
+  action_model:
+    action_dim: 7
+    action_horizon: 16
+    action_model_type: DiT-B
+    add_pos_embed: true
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    hidden_size: 1024
+    max_seq_len: 1024
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    num_timestep_buckets: 1000
+    state_dim: 7
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+output_dir: /home/liuyu/student-workspace/outputs/loss_is_adapted/runs/loss-is-online-seed-117465-20k-to-20500
+run_id: loss-is-online-seed-117465-20k-to-20500
+run_root_dir: /home/liuyu/student-workspace/outputs/loss_is_adapted/runs
+seed: 117465
+trainer:
+  freeze_modules: ''
+  gradient_accumulation_steps: 8
+  learning_rate:
+    action_model: 0.0001
+    base: 3.0e-05
+    qwen_vl_interface: 1.0e-05
+  loss_is:
+    candidate_factor: 3
+    condition_momentum: 0.9
+    condition_threshold: 2.0
+    seed: 117465
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 100000
+  num_warmup_steps: 5000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    name: AdamW
+    weight_decay: 1.0e-08
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07

settings/florence_bridge/loss_is/dataset_statistics.json

@@ -0,0 +1,264 @@
+{
+  "oxe_bridge": {
+    "action": {
+      "mean": [
+        5.682967821485363e-05,
+        3.278055373812094e-05,
+        -3.1604828109266236e-05,
+        -3.6026765883434564e-05,
+        -9.757651423569769e-05,
+        6.01583014940843e-05,
+        0.14414729177951813
+      ],
+      "std": [
+        0.004886276229987901,
+        0.006847766456005231,
+        0.006337809696791937,
+        0.014227778584428946,
+        0.015261554008856799,
+        0.038695293138416895,
+        0.3521380562074972
+      ],
+      "max": [
+        0.41691166162490845,
+        0.25864794850349426,
+        0.21218234300613403,
+        3.122201919555664,
+        1.8618112802505493,
+        6.272472858428955,
+        1.0
+      ],
+      "min": [
+        -0.4007510244846344,
+        -0.13874775171279907,
+        -0.22553899884223938,
+        -3.2010786533355713,
+        -1.8618112802505493,
+        -6.279075622558594,
+        0.0
+      ],
+      "q01": [
+        -0.02875255048274994,
+        -0.04170213546603918,
+        -0.026096721179783344,
+        -0.08052874729037285,
+        -0.09249906800687313,
+        -0.20738555490970612,
+        0.0
+      ],
+      "q99": [
+        0.028306663036346436,
+        0.04089853074401617,
+        0.0401805154979229,
+        0.08173403143882751,
+        0.07760760560631752,
+        0.2038465365767479,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        0.07735871523618698,
+        0.007681275252252817,
+        0.016110140830278397,
+        0.0016226600855588913,
+        -0.019300073385238647,
+        0.02691473439335823,
+        0.0,
+        0.17703326046466827
+      ],
+      "std": [
+        0.13736907523153707,
+        0.047840646787850445,
+        0.03799073317180484,
+        0.06560660645733722,
+        0.09093375673607122,
+        0.292728734025679,
+        0.0,
+        0.3536876512395259
+      ],
+      "max": [
+        0.5862360596656799,
+        0.4034728705883026,
+        0.3568263053894043,
+        1.3517684936523438,
+        1.570796251296997,
+        3.141204357147217,
+        0.0,
+        1.1121242046356201
+      ],
+      "min": [
+        -0.04167502000927925,
+        -0.3563207685947418,
+        -0.15537554025650024,
+        -3.141592502593994,
+        -1.4992541074752808,
+        -3.14153790473938,
+        0.0,
+        0.04637829214334488
+      ],
+      "q01": [
+        0.17102622985839844,
+        -0.1698118858039379,
+        -0.055632163770496845,
+        -0.36493172496557236,
+        -0.541871190071106,
+        -1.354275494813919,
+        0.0,
+        0.052190229296684265
+      ],
+      "q99": [
+        0.45322076976299286,
+        0.2354845181107521,
+        0.19489620998501778,
+        0.3780156075954437,
+        0.27568644285202026,
+        1.8500566184520721,
+        0.0,
+        1.0105689764022827
+      ]
+    },
+    "num_transitions": 1305714,
+    "num_trajectories": 53192
+  },
+  "oxe_rt1": {
+    "action": {
+      "mean": [
+        0.0017468753503635526,
+        0.0015664808452129364,
+        -0.00315628363750875,
+        0.010832978412508965,
+        -0.0014390418073162436,
+        0.00022827448265161365,
+        0.13385511934757233
+      ],
+      "std": [
+        0.034738154754287556,
+        0.029950575875189896,
+        0.03716987948198757,
+        0.08027394458720655,
+        0.06586824211259418,
+        0.07296740097742303,
+        0.33989797865933385
+      ],
+      "max": [
+        2.9984593391418457,
+        22.09052848815918,
+        2.7507524490356445,
+        1.570636510848999,
+        1.5321086645126343,
+        1.5691522359848022,
+        1.0
+      ],
+      "min": [
+        -2.0204520225524902,
+        -5.497899532318115,
+        -2.031663417816162,
+        -1.569917917251587,
+        -1.569892168045044,
+        -1.570419430732727,
+        0.0
+      ],
+      "q01": [
+        -0.22453527510166169,
+        -0.14820013284683228,
+        -0.231589707583189,
+        -0.3517994859814644,
+        -0.4193011274933815,
+        -0.43643461108207704,
+        0.0
+      ],
+      "q99": [
+        0.17824687153100965,
+        0.14938379630446405,
+        0.21842354819178575,
+        0.5892666035890578,
+        0.35272657424211445,
+        0.44796681255102094,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        0.1399736851453781,
+        -0.02083534561097622,
+        0.19427375495433807,
+        0.10653364658355713,
+        -0.06201138719916344,
+        0.12378455698490143,
+        0.023165317252278328,
+        0.052437249571084976
+      ],
+      "std": [
+        0.2502841726583427,
+        0.06813404469049046,
+        0.358260305088155,
+        0.2931043604956031,
+        0.2779312663188148,
+        0.33769767306558707,
+        0.09232656626252377,
+        0.16498523970985932
+      ],
+      "max": [
+        1.0534898042678833,
+        0.48018959164619446,
+        1.6896663904190063,
+        1.0,
+        0.9999993443489075,
+        0.9999874830245972,
+        0.9554369449615479,
+        0.9914546012878418
+      ],
+      "min": [
+        -0.4436439275741577,
+        -0.9970501065254211,
+        -0.006579156965017319,
+        0.0,
+        -0.8643477559089661,
+        -0.7079970240592957,
+        -0.7688722014427185,
+        -0.4999994933605194
+      ],
+      "q01": [
+        0.32481380939483645,
+        -0.28334290891885755,
+        0.14107070609927178,
+        0.0,
+        -0.686474204659462,
+        -0.6808923494815826,
+        -0.36045596331357954,
+        -0.454380963742733
+      ],
+      "q99": [
+        0.8750156319141384,
+        0.21247054174542404,
+        1.0727112340927123,
+        1.0,
+        0.9377871316671368,
+        0.9563051050901409,
+        0.45990042358636823,
+        0.7216041100025177
+      ]
+    },
+    "num_transitions": 3786152,
+    "num_trajectories": 87212
+  }
+}

settings/florence_robotwin/common_start/config.full.yaml

@@ -0,0 +1,107 @@
+run_id: florence-groot-robotwin-xvla3view-official150k-seed42-20260804
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-official150k
+seed: 42
+wandb_entity: null
+wandb_project: InfoFlow-StarVLA-RoboTwin
+is_debug: false
+version_id: '0.21'
+framework:
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+    attn_implementation: eager
+    vl_hidden_dim: 1024
+    selected_view_indices:
+    - 0
+    - 1
+    - 2
+    compile_vision_tower: true
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 1024
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 14
+    state_dim: 0
+    action_horizon: 50
+    repeated_diffusion_steps: 8
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    future_action_window_size: 49
+    past_action_window_size: 0
+  dino:
+    dino_backbone: dinov2_vits14
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    data_mix: robotwin_all_50
+    action_type: abs_qpos
+    action_mode: abs
+    include_state: false
+    sequential_step_sampling: false
+    per_device_batch_size: 2
+    load_all_data_for_training: true
+    image_size:
+    - 224
+    - 224
+    video_backend: pyav
+    num_workers: 4
+    expected_world_size: 2
+trainer:
+  use_deepspeed: true
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  save_interval: null
+  save_steps:
+  - 10000
+  - 11000
+  - 12000
+  - 13000
+  - 14000
+  - 15000
+  - 16000
+  - 17000
+  - 18000
+  - 19000
+  - 20000
+  eval_interval: null
+  repeated_diffusion_steps: 4
+  learning_rate:
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: ''
+  trainable_modules: ''
+  frozen_modules_eval: false
+  logging_frequency: 50
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 16
+  save_format: pt
+  save_final_checkpoint: false
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+config_yaml: /home/miaochengyu/starVLA-infoflow-codex/examples/InfoFlow/train_files/server/florence_groot_robotwin_official150k_2gpu.yaml
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-official150k/florence-groot-robotwin-xvla3view-official150k-seed42-20260804

settings/florence_robotwin/common_start/config.yaml

@@ -0,0 +1,61 @@
+datasets:
+  vla_data:
+    action_mode: abs
+    data_mix: robotwin_all_50
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    dataset_py: lerobot_datasets
+    num_workers: 4
+    per_device_batch_size: 2
+    video_backend: pyav
+framework:
+  action_model:
+    action_dim: 14
+    action_horizon: 50
+    action_model_type: DiT-B
+    add_pos_embed: true
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    hidden_size: 1024
+    max_seq_len: 1024
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    num_timestep_buckets: 1000
+    state_dim: 0
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+    compile_vision_tower: true
+    selected_view_indices:
+    - 0
+    - 1
+    - 2
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-official150k/florence-groot-robotwin-xvla3view-official150k-seed42-20260804
+run_id: florence-groot-robotwin-xvla3view-official150k-seed42-20260804
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-official150k
+seed: 42
+trainer:
+  freeze_modules: ''
+  learning_rate:
+    action_model: 0.0001
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07

settings/florence_robotwin/common_start/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        -0.22423998869955541,
+        1.0729362124204638,
+        0.7750796097517013,
+        -0.3270166309736669,
+        0.061595864617647746,
+        -0.06420705380151048,
+        0.22430071420967582,
+        1.1274866509437562,
+        0.8317736595869064,
+        -0.3638282522559167,
+        -0.024959841337949915,
+        0.053022032135631905,
+        0.6774904182553296,
+        0.659867690205574
+      ],
+      "std": [
+        0.39622931737429135,
+        1.0035904493395147,
+        0.7811611351334437,
+        0.6707485714474851,
+        0.26869082049839216,
+        0.6492019642472783,
+        0.3473367940092324,
+        1.0176226676795117,
+        0.8039915805960643,
+        0.7067249045233707,
+        0.2625964517715169,
+        0.6808759458519373,
+        0.448215115111442,
+        0.45399216711602836
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268352918624878,
+        -5.995517673492431,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.4545854091644288,
+        -5.945669202804565,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        5.016749763488781,
+        3.4792869091033936,
+        4.225283479690551,
+        1.7064905214309687,
+        1.4829185009002717,
+        3.7177388882636553,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.766038613319397,
+        1.4066662788391113,
+        3.5830023288726807,
+        1.0,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        -0.22281882727518681,
+        1.0668349397182468,
+        0.7706326398253446,
+        -0.32523305794224144,
+        0.061181390470028435,
+        -0.06316296239456282,
+        0.2225376246869564,
+        1.1201750242710118,
+        0.8263923591375355,
+        -0.3615948969544843,
+        -0.024702461528722783,
+        0.05203246331657284,
+        0.6793262696266178,
+        0.6616973677277567
+      ],
+      "std": [
+        0.3952287088673167,
+        1.004018577033247,
+        0.7809103811570286,
+        0.6688088363177989,
+        0.2677572381895678,
+        0.6463212583712936,
+        0.3464848832136367,
+        1.018232630032839,
+        0.803879839768273,
+        0.7045612787595747,
+        0.2614341421313462,
+        0.6774619870452645,
+        0.4474922049702762,
+        0.45333916022037923
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268109321594238,
+        -5.912737417221069,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.440324923992157,
+        -5.8623305034637445,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        4.902149534225477,
+        3.4775466918945312,
+        4.224453401565551,
+        1.7064905214309687,
+        1.480757713317871,
+        3.609884884357428,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.7433889162540437,
+        1.4066662788391113,
+        3.581243851184845,
+        1.0,
+        1.0
+      ]
+    },
+    "num_transitions": 6075103,
+    "num_trajectories": 27500
+  }
+}

settings/florence_robotwin/common_start/initialization_manifest.json

@@ -0,0 +1,6 @@
+{
+  "schema_version": 1,
+  "initialization_seed": 42,
+  "action_model_sha256": "915f157d03a58d059818a774e9150aa2028dd054752b513a5b4f3c85313bc383",
+  "action_model_parameter_elements": 141815566
+}

settings/florence_robotwin/uniform/config.full.yaml

@@ -0,0 +1,113 @@
+run_id: florence-robotwin-uniform-resume10k-12k-20260809
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-uniform-resume10k-12k-20260809
+seed: 484370
+wandb_entity: null
+wandb_project: InfoFlow-StarVLA-RoboTwin
+is_debug: false
+version_id: '0.21'
+framework:
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+    attn_implementation: eager
+    vl_hidden_dim: 1024
+    selected_view_indices:
+    - 0
+    - 1
+    - 2
+    compile_vision_tower: true
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 1024
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 14
+    state_dim: 0
+    action_horizon: 50
+    repeated_diffusion_steps: 8
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    future_action_window_size: 49
+    past_action_window_size: 0
+  dino:
+    dino_backbone: dinov2_vits14
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    data_mix: robotwin_all_50
+    action_type: abs_qpos
+    action_mode: abs
+    include_state: false
+    sequential_step_sampling: false
+    per_device_batch_size: 2
+    load_all_data_for_training: true
+    image_size:
+    - 224
+    - 224
+    video_backend: pyav
+    num_workers: 4
+    expected_world_size: 2
+    traceable_sampling: true
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/robotwin/data_provenance.json
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/content_fingerprints.json
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/valid_anchor_manifest.json
+    temporal_proposal_path: null
+    transform_rng_seed: 289726
+trainer:
+  use_deepspeed: true
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  save_interval: null
+  save_steps:
+  - 10500
+  - 11000
+  - 11500
+  - 12000
+  eval_interval: null
+  repeated_diffusion_steps: 4
+  learning_rate:
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: ''
+  trainable_modules: ''
+  frozen_modules_eval: false
+  logging_frequency: 50
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 16
+  save_format: pt
+  save_final_checkpoint: false
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  stop_after_steps: 12000
+  is_resume: true
+  allow_model_only_resume: false
+  resume_accelerate_checkpoint: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-official150k/florence-groot-robotwin-xvla3view-official150k-seed42-20260804/checkpoints/steps_10000_accelerate
+  branch_parent_seed: 42
+  branch_rng_seed: 484370
+  model_rng_seed: 721927
+config_yaml: /home/miaochengyu/starVLA-infoflow-codex/examples/InfoFlow/train_files/server/florence_groot_robotwin_official150k_traceable_10k.yaml
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-uniform-resume10k-12k-20260809/florence-robotwin-uniform-resume10k-12k-20260809

settings/florence_robotwin/uniform/config.yaml

@@ -0,0 +1,68 @@
+datasets:
+  vla_data:
+    action_mode: abs
+    data_mix: robotwin_all_50
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/content_fingerprints.json
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/robotwin/data_provenance.json
+    dataset_py: lerobot_datasets
+    expected_world_size: 2
+    num_workers: 4
+    per_device_batch_size: 2
+    temporal_proposal_path: null
+    traceable_sampling: true
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/valid_anchor_manifest.json
+    video_backend: pyav
+framework:
+  action_model:
+    action_dim: 14
+    action_horizon: 50
+    action_model_type: DiT-B
+    add_pos_embed: true
+    diffusion_model_cfg:
+      cross_attention_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    hidden_size: 1024
+    max_seq_len: 1024
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    num_timestep_buckets: 1000
+    state_dim: 0
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Florence-2-large
+    compile_vision_tower: true
+    selected_view_indices:
+    - 0
+    - 1
+    - 2
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-uniform-resume10k-12k-20260809/florence-robotwin-uniform-resume10k-12k-20260809
+run_id: florence-robotwin-uniform-resume10k-12k-20260809
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-uniform-resume10k-12k-20260809
+seed: 484370
+trainer:
+  freeze_modules: ''
+  gradient_accumulation_steps: 16
+  learning_rate:
+    action_model: 0.0001
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07

settings/florence_robotwin/uniform/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        -0.22423998869955536,
+        1.0729362124204633,
+        0.775079609751701,
+        -0.32701663097366684,
+        0.06159586461764771,
+        -0.06420705380151046,
+        0.22430071420967568,
+        1.1274866509437556,
+        0.831773659586906,
+        -0.3638282522559164,
+        -0.024959841337949897,
+        0.05302203213563182,
+        0.6774904182553295,
+        0.6598676902055738
+      ],
+      "std": [
+        0.39622931737429123,
+        1.003590449339515,
+        0.7811611351334437,
+        0.6707485714474849,
+        0.2686908204983921,
+        0.6492019642472782,
+        0.34733679400923234,
+        1.0176226676795115,
+        0.8039915805960646,
+        0.7067249045233707,
+        0.26259645177151686,
+        0.6808759458519371,
+        0.4482151151114418,
+        0.4539921671160285
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268352918624878,
+        -5.995517673492431,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.4545854091644288,
+        -5.945669202804565,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        5.016749763488781,
+        3.4792869091033936,
+        4.225283479690551,
+        1.7064905214309687,
+        1.4829185009002717,
+        3.7177388882636553,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.766038613319397,
+        1.4066662788391113,
+        3.5830023288726807,
+        1.0,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        -0.22281882727518673,
+        1.0668349397182466,
+        0.770632639825344,
+        -0.3252330579422412,
+        0.06118139047002841,
+        -0.06316296239456277,
+        0.22253762468695631,
+        1.1201750242710113,
+        0.8263923591375351,
+        -0.36159489695448405,
+        -0.024702461528722772,
+        0.05203246331657278,
+        0.6793262696266176,
+        0.6616973677277563
+      ],
+      "std": [
+        0.39522870886731665,
+        1.004018577033247,
+        0.7809103811570285,
+        0.6688088363177986,
+        0.26775723818956776,
+        0.6463212583712935,
+        0.34648488321363663,
+        1.018232630032839,
+        0.8038798397682734,
+        0.7045612787595747,
+        0.26143414213134614,
+        0.6774619870452644,
+        0.44749220497027625,
+        0.45333916022037934
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268109321594238,
+        -5.912737417221069,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.440324923992157,
+        -5.8623305034637445,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        4.902149534225477,
+        3.4775466918945312,
+        4.224453401565551,
+        1.7064905214309687,
+        1.480757713317871,
+        3.609884884357428,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.7433889162540437,
+        1.4066662788391113,
+        3.581243851184845,
+        1.0,
+        1.0
+      ]
+    },
+    "num_transitions": 6075103,
+    "num_trajectories": 27500
+  }
+}

settings/florence_robotwin/uniform/external_resume_certificate.json

@@ -0,0 +1,138 @@
+{
+  "branch_resume_identity": {
+    "config_sha256": "d40666777c1a9b5df6f5f6934614c0fe1b0d9af5d0c24e10d90c1fe4103e4c0f",
+    "identity_schema_version": 2,
+    "manifest_sha256": "dcb728a00da4b4ad7f989fa313d6d3355e13f57efefc7267e9ea092f2ffd8a50",
+    "model_structure_sha256": "2d1c8fd1dfa358c4975ad54934f42c24c8d10f6fcef9e8f7c2585b349b32d96b",
+    "proposal_alpha": null,
+    "proposal_sha256": null,
+    "seed": 484370,
+    "total_batch_size": 64
+  },
+  "completed_steps": 10000,
+  "optimizer_param_groups": [
+    {
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "initial_lr": 1e-05,
+      "lr": 9.97215529648321e-06,
+      "name": "qwen_vl_interface",
+      "weight_decay": 1e-08
+    },
+    {
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "initial_lr": 0.0001,
+      "lr": 9.972155296483209e-05,
+      "name": "action_model",
+      "weight_decay": 1e-08
+    }
+  ],
+  "optimizer_state_entries": 4,
+  "parent_checkpoint": "/mnt/miaochengyu/starvla-infoflow-runtime/runs/florence-robotwin-official150k/florence-groot-robotwin-xvla3view-official150k-seed42-20260804/checkpoints/steps_10000_accelerate",
+  "parent_checkpoint_commit_sha256": "f6f6408ead21ba9cb0d7bbfa26c74733950a504264a26e5b795129ffadff6958",
+  "parent_checkpoint_files": [
+    {
+      "bytes": 1525,
+      "name": "custom_checkpoint_0.pkl",
+      "sha256": "031a5046639dfc967eebbf30b7184dbe2619e26675a74569827aa5edaf6cc85c"
+    },
+    {
+      "bytes": 13,
+      "name": "latest",
+      "sha256": "68d0ca762047fd48fdd6a06a5e6465209289c028bfd63f5557a110ff1ebc15ac"
+    },
+    {
+      "bytes": 5509764721,
+      "name": "pytorch_model/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt",
+      "sha256": "27d9b011b75f40c3d896c9681ee883eeb3d2277f201889faebf0b7fcf5492c52"
+    },
+    {
+      "bytes": 5509778801,
+      "name": "pytorch_model/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt",
+      "sha256": "f657f6f6fb48956b2b406b79511e16b9c2712dc95216b78ba5179818b821d18f"
+    },
+    {
+      "bytes": 1837527613,
+      "name": "pytorch_model/mp_rank_00_model_states.pt",
+      "sha256": "3d66f8b6c0e344c34f06d3b4708c3292cbb8a84ea5a25e650fb64b679a44eb3b"
+    },
+    {
+      "bytes": 2152621885,
+      "name": "pytorch_model.pt",
+      "sha256": "4b6c4f47df7b0f05b8c9287e480d88f2b5e711b4fb6cccbdfffd677ed9bff778"
+    },
+    {
+      "bytes": 15017,
+      "name": "random_states_0.pkl",
+      "sha256": "629d14d87e3cab680e32acec92c80c2578a7fbd1ebc0dee5475e774058f11d41"
+    },
+    {
+      "bytes": 15081,
+      "name": "random_states_1.pkl",
+      "sha256": "3cb98283ee7eaaedcee320e8e3b17e74f2f01a22d9e3b958fa2422a86215c652"
+    },
+    {
+      "bytes": 33270,
+      "name": "zero_to_fp32.py",
+      "sha256": "e8a2553b434ac043ec63b022db1e768277acf8d4e70f10a2842db6d238977135"
+    }
+  ],
+  "parent_checkpoint_files_sha256": "1f8985d0633f16ac3b0381b128cdb4e479c4fdfd67ff15d58736633e4b942bc0",
+  "parent_checkpoint_integrity_manifest_sha256": null,
+  "parent_config_sha256": "d35ab9e43634c7c4a74c5299657df3c17c7d78407dd0fe3d37296a927e353d28",
+  "parent_resume_identity": {
+    "config_sha256": "2440c6327cf5e8c14efc93b800209171c725dae0898181921c62f3490eb8378f",
+    "identity_schema_version": 2,
+    "manifest_sha256": null,
+    "model_structure_sha256": "2d1c8fd1dfa358c4975ad54934f42c24c8d10f6fcef9e8f7c2585b349b32d96b",
+    "proposal_alpha": null,
+    "proposal_sha256": null,
+    "seed": 42,
+    "total_batch_size": 64
+  },
+  "proposal_sha256": null,
+  "rank_state_summaries": [
+    {
+      "optimizer_state_entries": 2,
+      "optimizer_state_sha256": "5a35add2c00249e41e72e1ab16261d87f205bfeaf23386c7263223e4ce168368",
+      "rank": 0,
+      "trainable_state_sha256": "667c6a6a7aa0dc49c32f7bef693f3272efee46064fda5d370d25788291da76a0"
+    },
+    {
+      "optimizer_state_entries": 2,
+      "optimizer_state_sha256": "073f849cd1942522dbfde195bdc29e8ee015e4c5d4be3070381027a13b9784cf",
+      "rank": 1,
+      "trainable_state_sha256": "667c6a6a7aa0dc49c32f7bef693f3272efee46064fda5d370d25788291da76a0"
+    }
+  ],
+  "repeated_diffusion_steps": 4,
+  "rng_streams": {
+    "model_base": 721927,
+    "sampler": 484370,
+    "transform": 289726
+  },
+  "scheduler": {
+    "class": "LambdaLR",
+    "last_epoch": 10000,
+    "last_lr": [
+      9.97215529648321e-06,
+      9.972155296483209e-05
+    ],
+    "step_count": 10001
+  },
+  "scheduler_resume_mode": "accelerate_checkpoint",
+  "schema_version": 1,
+  "status": "exact_external_resume_loaded",
+  "total_batch_size": 64,
+  "trainable_parameter_elements": 918283022,
+  "trainable_parameter_sha256": "2e8f3de6aeee46bca56fab5993103ad30302b45330037cf98f96ffd472f0e9ce",
+  "trainable_parameter_tensors": 1161,
+  "world_size": 2
+}

settings/florence_robotwin/uniform/initialization_manifest.json

@@ -0,0 +1,6 @@
+{
+  "schema_version": 1,
+  "initialization_seed": 484370,
+  "action_model_sha256": "17ab11d253e73f3058c40637d5f31c3462a97cfbf61845148671de47ecca39eb",
+  "action_model_parameter_elements": 141815566
+}

settings/florence_robotwin/uniform/valid_anchor_manifest.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2090f6b9d7e1bc5ef41897436690f031c187d625fa6b71ae0c54f3793660c757
+size 107646168

settings/qwen_robotwin/common_start/config.full.yaml

@@ -0,0 +1,242 @@
+run_id: qwen3vl-groot-robotwin-official150k-adamw8bit-mb5-ga4-seed42-20260810
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-groot-robotwin-official150k
+seed: 42
+wandb_entity: null
+wandb_project: InfoFlow-StarVLA-RoboTwin
+is_debug: false
+version_id: '0.21'
+framework:
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Qwen3-VL-4B-Instruct
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 2048
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 14
+    state_dim: 0
+    action_horizon: 50
+    repeated_diffusion_steps: 8
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      cross_attention_dim: 2560
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    future_action_window_size: 49
+    past_action_window_size: 0
+  dino:
+    dino_backbone: dinov2_vits14
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    data_mix: robotwin_all_50
+    action_type: abs_qpos
+    action_mode: abs
+    include_state: false
+    sequential_step_sampling: false
+    per_device_batch_size: 5
+    load_all_data_for_training: true
+    image_size:
+    - 224
+    - 224
+    video_backend: pyav
+    num_workers: 4
+    expected_world_size: 2
+    traceable_sampling: true
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/robotwin/data_provenance.json
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/content_fingerprints.json
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/valid_anchor_manifest.json
+trainer:
+  use_deepspeed: true
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  save_interval: null
+  save_steps:
+  - 10000
+  - 11000
+  - 12000
+  - 13000
+  - 14000
+  - 15000
+  - 16000
+  - 17000
+  - 18000
+  - 19000
+  - 20000
+  - 21000
+  - 22000
+  - 23000
+  - 24000
+  - 25000
+  - 26000
+  - 27000
+  - 28000
+  - 29000
+  - 30000
+  - 31000
+  - 32000
+  - 33000
+  - 34000
+  - 35000
+  - 36000
+  - 37000
+  - 38000
+  - 39000
+  - 40000
+  - 41000
+  - 42000
+  - 43000
+  - 44000
+  - 45000
+  - 46000
+  - 47000
+  - 48000
+  - 49000
+  - 50000
+  - 51000
+  - 52000
+  - 53000
+  - 54000
+  - 55000
+  - 56000
+  - 57000
+  - 58000
+  - 59000
+  - 60000
+  - 61000
+  - 62000
+  - 63000
+  - 64000
+  - 65000
+  - 66000
+  - 67000
+  - 68000
+  - 69000
+  - 70000
+  - 71000
+  - 72000
+  - 73000
+  - 74000
+  - 75000
+  - 76000
+  - 77000
+  - 78000
+  - 79000
+  - 80000
+  - 81000
+  - 82000
+  - 83000
+  - 84000
+  - 85000
+  - 86000
+  - 87000
+  - 88000
+  - 89000
+  - 90000
+  - 91000
+  - 92000
+  - 93000
+  - 94000
+  - 95000
+  - 96000
+  - 97000
+  - 98000
+  - 99000
+  - 100000
+  - 101000
+  - 102000
+  - 103000
+  - 104000
+  - 105000
+  - 106000
+  - 107000
+  - 108000
+  - 109000
+  - 110000
+  - 111000
+  - 112000
+  - 113000
+  - 114000
+  - 115000
+  - 116000
+  - 117000
+  - 118000
+  - 119000
+  - 120000
+  - 121000
+  - 122000
+  - 123000
+  - 124000
+  - 125000
+  - 126000
+  - 127000
+  - 128000
+  - 129000
+  - 130000
+  - 131000
+  - 132000
+  - 133000
+  - 134000
+  - 135000
+  - 136000
+  - 137000
+  - 138000
+  - 139000
+  - 140000
+  - 141000
+  - 142000
+  - 143000
+  - 144000
+  - 145000
+  - 146000
+  - 147000
+  - 148000
+  - 149000
+  - 150000
+  evaluation_snapshot_interval: 1000
+  evaluation_snapshot_start_step: 10000
+  evaluation_snapshot_end_step: 20000
+  eval_interval: null
+  repeated_diffusion_steps: 4
+  learning_rate:
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: ''
+  trainable_modules: ''
+  frozen_modules_eval: false
+  logging_frequency: 10
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 4
+  save_format: pt
+  save_final_checkpoint: false
+  save_consolidated_model: false
+  optimizer:
+    name: AdamW8bit
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  is_resume: true
+  allow_model_only_resume: false
+config_yaml: /home/miaochengyu/starVLA-infoflow-codex/examples/InfoFlow/train_files/server/qwen3vl_groot_robotwin_official150k_2gpu.yaml
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-groot-robotwin-official150k/qwen3vl-groot-robotwin-official150k-adamw8bit-mb5-ga4-seed42-20260810

settings/qwen_robotwin/common_start/config.yaml

@@ -0,0 +1,64 @@
+datasets:
+  vla_data:
+    action_mode: abs
+    data_mix: robotwin_all_50
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/content_fingerprints.json
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/robotwin/data_provenance.json
+    dataset_py: lerobot_datasets
+    expected_world_size: 2
+    num_workers: 4
+    per_device_batch_size: 5
+    traceable_sampling: true
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/valid_anchor_manifest.json
+    video_backend: pyav
+framework:
+  action_model:
+    action_dim: 14
+    action_horizon: 50
+    action_model_type: DiT-B
+    add_pos_embed: true
+    diffusion_model_cfg:
+      cross_attention_dim: 2560
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    hidden_size: 1024
+    max_seq_len: 1024
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    num_timestep_buckets: 1000
+    state_dim: 0
+  name: QwenGR00T
+  qwenvl:
+    attn_implementation: flash_attention_2
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Qwen3-VL-4B-Instruct
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-groot-robotwin-official150k/qwen3vl-groot-robotwin-official150k-adamw8bit-mb5-ga4-seed42-20260810
+run_id: qwen3vl-groot-robotwin-official150k-adamw8bit-mb5-ga4-seed42-20260810
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-groot-robotwin-official150k
+seed: 42
+trainer:
+  freeze_modules: ''
+  gradient_accumulation_steps: 4
+  learning_rate:
+    action_model: 0.0001
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    name: AdamW8bit
+    weight_decay: 1.0e-08
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07

settings/qwen_robotwin/common_start/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        -0.22423998869955536,
+        1.0729362124204633,
+        0.775079609751701,
+        -0.32701663097366684,
+        0.06159586461764771,
+        -0.06420705380151046,
+        0.22430071420967568,
+        1.1274866509437556,
+        0.831773659586906,
+        -0.3638282522559164,
+        -0.024959841337949897,
+        0.05302203213563182,
+        0.6774904182553295,
+        0.6598676902055738
+      ],
+      "std": [
+        0.39622931737429123,
+        1.003590449339515,
+        0.7811611351334437,
+        0.6707485714474849,
+        0.2686908204983921,
+        0.6492019642472782,
+        0.34733679400923234,
+        1.0176226676795115,
+        0.8039915805960646,
+        0.7067249045233707,
+        0.26259645177151686,
+        0.6808759458519371,
+        0.4482151151114418,
+        0.4539921671160285
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268352918624878,
+        -5.995517673492431,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.4545854091644288,
+        -5.945669202804565,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        5.016749763488781,
+        3.4792869091033936,
+        4.225283479690551,
+        1.7064905214309687,
+        1.4829185009002717,
+        3.7177388882636553,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.766038613319397,
+        1.4066662788391113,
+        3.5830023288726807,
+        1.0,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        -0.22281882727518673,
+        1.0668349397182466,
+        0.770632639825344,
+        -0.3252330579422412,
+        0.06118139047002841,
+        -0.06316296239456277,
+        0.22253762468695631,
+        1.1201750242710113,
+        0.8263923591375351,
+        -0.36159489695448405,
+        -0.024702461528722772,
+        0.05203246331657278,
+        0.6793262696266176,
+        0.6616973677277563
+      ],
+      "std": [
+        0.39522870886731665,
+        1.004018577033247,
+        0.7809103811570285,
+        0.6688088363177986,
+        0.26775723818956776,
+        0.6463212583712935,
+        0.34648488321363663,
+        1.018232630032839,
+        0.8038798397682734,
+        0.7045612787595747,
+        0.26143414213134614,
+        0.6774619870452644,
+        0.44749220497027625,
+        0.45333916022037934
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268109321594238,
+        -5.912737417221069,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.440324923992157,
+        -5.8623305034637445,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        4.902149534225477,
+        3.4775466918945312,
+        4.224453401565551,
+        1.7064905214309687,
+        1.480757713317871,
+        3.609884884357428,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.7433889162540437,
+        1.4066662788391113,
+        3.581243851184845,
+        1.0,
+        1.0
+      ]
+    },
+    "num_transitions": 6075103,
+    "num_trajectories": 27500
+  }
+}

settings/qwen_robotwin/common_start/initialization_manifest.json

@@ -0,0 +1,6 @@
+{
+  "schema_version": 1,
+  "initialization_seed": 42,
+  "action_model_sha256": "f87884ee8b3f2bbc21809d0da733a118e4d95151aa3c3a68ae9f623cd5a7bc4d",
+  "action_model_parameter_elements": 160689934
+}

settings/qwen_robotwin/common_start/valid_anchor_manifest.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2090f6b9d7e1bc5ef41897436690f031c187d625fa6b71ae0c54f3793660c757
+size 107646168

settings/qwen_robotwin/residual/config.full.yaml

@@ -0,0 +1,112 @@
+run_id: qwen3vl-robotwin-residual-resume11k-13k-seed484370-20260815
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-robotwin-residual-resume11k-13k-20260815
+seed: 484370
+wandb_entity: null
+wandb_project: InfoFlow-StarVLA-RoboTwin
+is_debug: false
+version_id: '0.21'
+framework:
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Qwen3-VL-4B-Instruct
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 2048
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 14
+    state_dim: 0
+    action_horizon: 50
+    repeated_diffusion_steps: 8
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      cross_attention_dim: 2560
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    future_action_window_size: 49
+    past_action_window_size: 0
+  dino:
+    dino_backbone: dinov2_vits14
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    data_mix: robotwin_all_50
+    action_type: abs_qpos
+    action_mode: abs
+    include_state: false
+    sequential_step_sampling: false
+    per_device_batch_size: 5
+    load_all_data_for_training: true
+    image_size:
+    - 224
+    - 224
+    video_backend: pyav
+    num_workers: 4
+    expected_world_size: 2
+    traceable_sampling: true
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/robotwin/data_provenance.json
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/content_fingerprints.json
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/valid_anchor_manifest.json
+    temporal_proposal_path: /mnt/miaochengyu/starvla-infoflow-runtime/runs/infoflow-proposals/qwen3vl_robotwin_11k_residual_grid8_gate42f_override_20260814/residual_temporal_proposal.json
+    transform_rng_seed: 289726
+trainer:
+  use_deepspeed: true
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  save_interval: null
+  save_steps:
+  - 11500
+  - 12000
+  - 12500
+  - 13000
+  evaluation_snapshot_interval: 1000
+  evaluation_snapshot_start_step: 10000
+  evaluation_snapshot_end_step: 20000
+  eval_interval: null
+  repeated_diffusion_steps: 4
+  learning_rate:
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: ''
+  trainable_modules: ''
+  frozen_modules_eval: false
+  logging_frequency: 10
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 4
+  save_format: pt
+  save_final_checkpoint: false
+  save_consolidated_model: false
+  optimizer:
+    name: AdamW8bit
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  stop_after_steps: 13000
+  is_resume: true
+  allow_model_only_resume: false
+  resume_accelerate_checkpoint: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-groot-robotwin-official150k/qwen3vl-groot-robotwin-official150k-adamw8bit-mb5-ga4-seed42-20260810/checkpoints/steps_11000_accelerate
+  branch_parent_seed: 42
+  branch_rng_seed: 484370
+  model_rng_seed: 721927
+config_yaml: /home/miaochengyu/starVLA-infoflow-codex/examples/InfoFlow/train_files/server/qwen3vl_groot_robotwin_official150k_2gpu.yaml
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-robotwin-residual-resume11k-13k-20260815/qwen3vl-robotwin-residual-resume11k-13k-seed484370-20260815

settings/qwen_robotwin/residual/config.yaml

@@ -0,0 +1,65 @@
+datasets:
+  vla_data:
+    action_mode: abs
+    data_mix: robotwin_all_50
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/content_fingerprints.json
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/robotwin/data_provenance.json
+    dataset_py: lerobot_datasets
+    expected_world_size: 2
+    num_workers: 4
+    per_device_batch_size: 5
+    temporal_proposal_path: /mnt/miaochengyu/starvla-infoflow-runtime/runs/infoflow-proposals/qwen3vl_robotwin_11k_residual_grid8_gate42f_override_20260814/residual_temporal_proposal.json
+    traceable_sampling: true
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/valid_anchor_manifest.json
+    video_backend: pyav
+framework:
+  action_model:
+    action_dim: 14
+    action_horizon: 50
+    action_model_type: DiT-B
+    add_pos_embed: true
+    diffusion_model_cfg:
+      cross_attention_dim: 2560
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    hidden_size: 1024
+    max_seq_len: 1024
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    num_timestep_buckets: 1000
+    state_dim: 0
+  name: QwenGR00T
+  qwenvl:
+    attn_implementation: flash_attention_2
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Qwen3-VL-4B-Instruct
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-robotwin-residual-resume11k-13k-20260815/qwen3vl-robotwin-residual-resume11k-13k-seed484370-20260815
+run_id: qwen3vl-robotwin-residual-resume11k-13k-seed484370-20260815
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-robotwin-residual-resume11k-13k-20260815
+seed: 484370
+trainer:
+  freeze_modules: ''
+  gradient_accumulation_steps: 4
+  learning_rate:
+    action_model: 0.0001
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    name: AdamW8bit
+    weight_decay: 1.0e-08
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07

settings/qwen_robotwin/residual/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        -0.22423998869955536,
+        1.0729362124204633,
+        0.775079609751701,
+        -0.32701663097366684,
+        0.06159586461764771,
+        -0.06420705380151046,
+        0.22430071420967568,
+        1.1274866509437556,
+        0.831773659586906,
+        -0.3638282522559164,
+        -0.024959841337949897,
+        0.05302203213563182,
+        0.6774904182553295,
+        0.6598676902055738
+      ],
+      "std": [
+        0.39622931737429123,
+        1.003590449339515,
+        0.7811611351334437,
+        0.6707485714474849,
+        0.2686908204983921,
+        0.6492019642472782,
+        0.34733679400923234,
+        1.0176226676795115,
+        0.8039915805960646,
+        0.7067249045233707,
+        0.26259645177151686,
+        0.6808759458519371,
+        0.4482151151114418,
+        0.4539921671160285
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268352918624878,
+        -5.995517673492431,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.4545854091644288,
+        -5.945669202804565,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        5.016749763488781,
+        3.4792869091033936,
+        4.225283479690551,
+        1.7064905214309687,
+        1.4829185009002717,
+        3.7177388882636553,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.766038613319397,
+        1.4066662788391113,
+        3.5830023288726807,
+        1.0,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        -0.22281882727518673,
+        1.0668349397182466,
+        0.770632639825344,
+        -0.3252330579422412,
+        0.06118139047002841,
+        -0.06316296239456277,
+        0.22253762468695631,
+        1.1201750242710113,
+        0.8263923591375351,
+        -0.36159489695448405,
+        -0.024702461528722772,
+        0.05203246331657278,
+        0.6793262696266176,
+        0.6616973677277563
+      ],
+      "std": [
+        0.39522870886731665,
+        1.004018577033247,
+        0.7809103811570285,
+        0.6688088363177986,
+        0.26775723818956776,
+        0.6463212583712935,
+        0.34648488321363663,
+        1.018232630032839,
+        0.8038798397682734,
+        0.7045612787595747,
+        0.26143414213134614,
+        0.6774619870452644,
+        0.44749220497027625,
+        0.45333916022037934
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268109321594238,
+        -5.912737417221069,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.440324923992157,
+        -5.8623305034637445,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        4.902149534225477,
+        3.4775466918945312,
+        4.224453401565551,
+        1.7064905214309687,
+        1.480757713317871,
+        3.609884884357428,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.7433889162540437,
+        1.4066662788391113,
+        3.581243851184845,
+        1.0,
+        1.0
+      ]
+    },
+    "num_transitions": 6075103,
+    "num_trajectories": 27500
+  }
+}

settings/qwen_robotwin/residual/external_resume_certificate.json

@@ -0,0 +1,138 @@
+{
+  "branch_resume_identity": {
+    "config_sha256": "91c26d27b4e350855849912d1438907387518f91f5443c41cd591c4a78787c3c",
+    "identity_schema_version": 2,
+    "manifest_sha256": "dcb728a00da4b4ad7f989fa313d6d3355e13f57efefc7267e9ea092f2ffd8a50",
+    "model_structure_sha256": "7809faa5245147f81b98f3dbd17c7a063a7cb07a21e18af588c748c5925e1407",
+    "proposal_alpha": 0.2,
+    "proposal_sha256": "4e9a9e9d66c2bbbb02ddd8ecec06ad0b514cd299bbfac57bf8c2d44daeb70237",
+    "seed": 484370,
+    "total_batch_size": 40
+  },
+  "completed_steps": 11000,
+  "optimizer_param_groups": [
+    {
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "initial_lr": 1e-05,
+      "lr": 9.959920880962606e-06,
+      "name": "qwen_vl_interface",
+      "weight_decay": 1e-08
+    },
+    {
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "initial_lr": 0.0001,
+      "lr": 9.959920880962605e-05,
+      "name": "action_model",
+      "weight_decay": 1e-08
+    }
+  ],
+  "optimizer_state_entries": 8,
+  "parent_checkpoint": "/mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-groot-robotwin-official150k/qwen3vl-groot-robotwin-official150k-adamw8bit-mb5-ga4-seed42-20260810/checkpoints/steps_11000_accelerate",
+  "parent_checkpoint_commit_sha256": "180f1ed2d2b7a9e6c158974fc9ec951182dfb629f6a2806808ac5639d491ce1b",
+  "parent_checkpoint_files": [
+    {
+      "bytes": 1525,
+      "name": "custom_checkpoint_0.pkl",
+      "sha256": "70788f8a2d7852a70fbb794ab9176d14ad37fb6d1b1d821a368572e1d8524390"
+    },
+    {
+      "bytes": 13,
+      "name": "latest",
+      "sha256": "68d0ca762047fd48fdd6a06a5e6465209289c028bfd63f5557a110ff1ebc15ac"
+    },
+    {
+      "bytes": 13867381301,
+      "name": "pytorch_model/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt",
+      "sha256": "42b6d10bb158b404056983ce88505c32738b29f3a062fe6835d24ce26c521a6a"
+    },
+    {
+      "bytes": 13867381301,
+      "name": "pytorch_model/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt",
+      "sha256": "7b0a1bbc92efffbf056bc40f40e93a94b67ddf3f375dc6ae263bcb048ba6ec63"
+    },
+    {
+      "bytes": 566626,
+      "name": "pytorch_model/zero_pp_rank_0_mp_rank_00_model_states.pt",
+      "sha256": "8a3ef88f9ab0df8197a836dca1d6e2416835f0b3d827cb10a78254b9a969ce7c"
+    },
+    {
+      "bytes": 566498,
+      "name": "pytorch_model/zero_pp_rank_1_mp_rank_00_model_states.pt",
+      "sha256": "67cd03fceb5ee4c2be6affe504f71a48ee10127d6ea3a1f00d05721895ffba53"
+    },
+    {
+      "bytes": 15017,
+      "name": "random_states_0.pkl",
+      "sha256": "636b1c5983ee2f8a5047025027453c9a478c0606389f2f6945b370f39cec9e65"
+    },
+    {
+      "bytes": 15081,
+      "name": "random_states_1.pkl",
+      "sha256": "416b3be53ff049935984cc65385d7dae3b0e5a82df66c9516f61ee9100c41961"
+    },
+    {
+      "bytes": 33270,
+      "name": "zero_to_fp32.py",
+      "sha256": "e8a2553b434ac043ec63b022db1e768277acf8d4e70f10a2842db6d238977135"
+    }
+  ],
+  "parent_checkpoint_files_sha256": "73692dae620e3d4368fe513e8adff4fe54bdb25341b4293e34ee2a6286554f8d",
+  "parent_checkpoint_integrity_manifest_sha256": null,
+  "parent_config_sha256": "4882855d94076fc0ec19a0d511f14b1cbe82860c8b5df3dcc64c5c5f6c0d2321",
+  "parent_resume_identity": {
+    "config_sha256": "1a48ea51cf89ce72a7956aaa29f710bba0326c9b726e52795561b381507d26fa",
+    "identity_schema_version": 2,
+    "manifest_sha256": "dcb728a00da4b4ad7f989fa313d6d3355e13f57efefc7267e9ea092f2ffd8a50",
+    "model_structure_sha256": "7809faa5245147f81b98f3dbd17c7a063a7cb07a21e18af588c748c5925e1407",
+    "proposal_alpha": null,
+    "proposal_sha256": null,
+    "seed": 42,
+    "total_batch_size": 40
+  },
+  "proposal_sha256": "4e9a9e9d66c2bbbb02ddd8ecec06ad0b514cd299bbfac57bf8c2d44daeb70237",
+  "rank_state_summaries": [
+    {
+      "optimizer_state_entries": 4,
+      "optimizer_state_sha256": "e1eea37514fa0478835597d1594b7a43b45cea9adc4ccb79a0de8fb9e10fe9f0",
+      "rank": 0,
+      "trainable_state_sha256": "ca2a19167d59aab693d0a0a1e17ea9adbf9cfd7367dfa7a5c90b04cf2183b717"
+    },
+    {
+      "optimizer_state_entries": 4,
+      "optimizer_state_sha256": "62c4e2908a7e97fe9a1d60642a85968081b55345493a11b9e7f68e49fcdb22f3",
+      "rank": 1,
+      "trainable_state_sha256": "ca2a19167d59aab693d0a0a1e17ea9adbf9cfd7367dfa7a5c90b04cf2183b717"
+    }
+  ],
+  "repeated_diffusion_steps": 4,
+  "rng_streams": {
+    "model_base": 721927,
+    "sampler": 484370,
+    "transform": 289726
+  },
+  "scheduler": {
+    "class": "LambdaLR",
+    "last_epoch": 11000,
+    "last_lr": [
+      9.959920880962606e-06,
+      9.959920880962605e-05
+    ],
+    "step_count": 11001
+  },
+  "scheduler_resume_mode": "accelerate_checkpoint",
+  "schema_version": 1,
+  "status": "exact_external_resume_loaded",
+  "total_batch_size": 40,
+  "trainable_parameter_elements": 32022798,
+  "trainable_parameter_sha256": "1bc504b22e4375be40bbbd8250be5b5f380c14d2fb47ed85deba3604a84b5484",
+  "trainable_parameter_tensors": 957,
+  "world_size": 2
+}

settings/qwen_robotwin/residual/initialization_manifest.json

@@ -0,0 +1,6 @@
+{
+  "schema_version": 1,
+  "initialization_seed": 484370,
+  "action_model_sha256": "1cab43a1e801a94110a20384a9ad6cd76a239fedd9fe17890662da7e22d47e62",
+  "action_model_parameter_elements": 160689934
+}

settings/qwen_robotwin/residual/valid_anchor_manifest.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2090f6b9d7e1bc5ef41897436690f031c187d625fa6b71ae0c54f3793660c757
+size 107646168

settings/qwen_robotwin/uniform/config.full.yaml

@@ -0,0 +1,108 @@
+run_id: qwen3vl-robotwin-uniform-commonstart11k-plus500-seed484370-20260815
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-robotwin-uniform-commonstart11k-plus500-20260815
+seed: 484370
+wandb_entity: null
+wandb_project: InfoFlow-StarVLA-RoboTwin
+is_debug: false
+version_id: '0.21'
+framework:
+  name: QwenGR00T
+  qwenvl:
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Qwen3-VL-4B-Instruct
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 2048
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 14
+    state_dim: 0
+    action_horizon: 50
+    repeated_diffusion_steps: 8
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      cross_attention_dim: 2560
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    future_action_window_size: 49
+    past_action_window_size: 0
+  dino:
+    dino_backbone: dinov2_vits14
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    data_mix: robotwin_all_50
+    action_type: abs_qpos
+    action_mode: abs
+    include_state: false
+    sequential_step_sampling: false
+    per_device_batch_size: 5
+    load_all_data_for_training: true
+    image_size:
+    - 224
+    - 224
+    video_backend: pyav
+    num_workers: 4
+    expected_world_size: 2
+    traceable_sampling: true
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/robotwin/data_provenance.json
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/content_fingerprints.json
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/valid_anchor_manifest.json
+    transform_rng_seed: 289726
+trainer:
+  use_deepspeed: true
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  save_interval: null
+  save_steps:
+  - 11500
+  evaluation_snapshot_interval: 1000
+  evaluation_snapshot_start_step: 10000
+  evaluation_snapshot_end_step: 20000
+  eval_interval: null
+  repeated_diffusion_steps: 4
+  learning_rate:
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: ''
+  trainable_modules: ''
+  frozen_modules_eval: false
+  logging_frequency: 10
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 4
+  save_format: pt
+  save_final_checkpoint: false
+  save_consolidated_model: false
+  optimizer:
+    name: AdamW8bit
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  stop_after_steps: 11500
+  is_resume: true
+  allow_model_only_resume: false
+  resume_accelerate_checkpoint: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-groot-robotwin-official150k/qwen3vl-groot-robotwin-official150k-adamw8bit-mb5-ga4-seed42-20260810/checkpoints/steps_11000_accelerate
+  branch_parent_seed: 42
+  branch_rng_seed: 484370
+  model_rng_seed: 721927
+config_yaml: /home/miaochengyu/starVLA-infoflow-codex/examples/InfoFlow/train_files/server/qwen3vl_groot_robotwin_official150k_2gpu.yaml
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-robotwin-uniform-commonstart11k-plus500-20260815/qwen3vl-robotwin-uniform-commonstart11k-plus500-seed484370-20260815

settings/qwen_robotwin/uniform/config.yaml

@@ -0,0 +1,64 @@
+datasets:
+  vla_data:
+    action_mode: abs
+    data_mix: robotwin_all_50
+    data_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Datasets/RoboTwin
+    dataset_content_fingerprint_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/content_fingerprints.json
+    dataset_provenance_path: /home/miaochengyu/starVLA-infoflow-codex/protocols/robotwin/data_provenance.json
+    dataset_py: lerobot_datasets
+    expected_world_size: 2
+    num_workers: 4
+    per_device_batch_size: 5
+    traceable_sampling: true
+    valid_anchor_manifest_path: /mnt/miaochengyu/starvla-infoflow-runtime/manifests/robotwin/valid_anchor_manifest.json
+    video_backend: pyav
+framework:
+  action_model:
+    action_dim: 14
+    action_horizon: 50
+    action_model_type: DiT-B
+    add_pos_embed: true
+    diffusion_model_cfg:
+      cross_attention_dim: 2560
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+    hidden_size: 1024
+    max_seq_len: 1024
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    num_timestep_buckets: 1000
+    state_dim: 0
+  name: QwenGR00T
+  qwenvl:
+    attn_implementation: flash_attention_2
+    base_vlm: /mnt/miaochengyu/starvla-infoflow-runtime/playground/Pretrained_models/Qwen3-VL-4B-Instruct
+output_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-robotwin-uniform-commonstart11k-plus500-20260815/qwen3vl-robotwin-uniform-commonstart11k-plus500-seed484370-20260815
+run_id: qwen3vl-robotwin-uniform-commonstart11k-plus500-seed484370-20260815
+run_root_dir: /mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-robotwin-uniform-commonstart11k-plus500-20260815
+seed: 484370
+trainer:
+  freeze_modules: ''
+  gradient_accumulation_steps: 4
+  learning_rate:
+    action_model: 0.0001
+    base: 1.0e-05
+    qwen_vl_interface: 1.0e-05
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 150000
+  num_warmup_steps: 5000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    name: AdamW8bit
+    weight_decay: 1.0e-08
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07

settings/qwen_robotwin/uniform/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        -0.22423998869955536,
+        1.0729362124204633,
+        0.775079609751701,
+        -0.32701663097366684,
+        0.06159586461764771,
+        -0.06420705380151046,
+        0.22430071420967568,
+        1.1274866509437556,
+        0.831773659586906,
+        -0.3638282522559164,
+        -0.024959841337949897,
+        0.05302203213563182,
+        0.6774904182553295,
+        0.6598676902055738
+      ],
+      "std": [
+        0.39622931737429123,
+        1.003590449339515,
+        0.7811611351334437,
+        0.6707485714474849,
+        0.2686908204983921,
+        0.6492019642472782,
+        0.34733679400923234,
+        1.0176226676795115,
+        0.8039915805960646,
+        0.7067249045233707,
+        0.26259645177151686,
+        0.6808759458519371,
+        0.4482151151114418,
+        0.4539921671160285
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268352918624878,
+        -5.995517673492431,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.4545854091644288,
+        -5.945669202804565,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        5.016749763488781,
+        3.4792869091033936,
+        4.225283479690551,
+        1.7064905214309687,
+        1.4829185009002717,
+        3.7177388882636553,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.766038613319397,
+        1.4066662788391113,
+        3.5830023288726807,
+        1.0,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        -0.22281882727518673,
+        1.0668349397182466,
+        0.770632639825344,
+        -0.3252330579422412,
+        0.06118139047002841,
+        -0.06316296239456277,
+        0.22253762468695631,
+        1.1201750242710113,
+        0.8263923591375351,
+        -0.36159489695448405,
+        -0.024702461528722772,
+        0.05203246331657278,
+        0.6793262696266176,
+        0.6616973677277563
+      ],
+      "std": [
+        0.39522870886731665,
+        1.004018577033247,
+        0.7809103811570285,
+        0.6688088363177986,
+        0.26775723818956776,
+        0.6463212583712935,
+        0.34648488321363663,
+        1.018232630032839,
+        0.8038798397682734,
+        0.7045612787595747,
+        0.26143414213134614,
+        0.6774619870452644,
+        0.44749220497027625,
+        0.45333916022037934
+      ],
+      "max": [
+        6.45659065246582,
+        4.179152488708496,
+        5.346591472625732,
+        1.7942548990249634,
+        1.8604620695114136,
+        5.43813943862915,
+        7.142920017242432,
+        4.157068729400635,
+        5.672450065612793,
+        1.95806884765625,
+        1.5663840770721436,
+        5.278968811035156,
+        1.0,
+        1.0
+      ],
+      "min": [
+        -7.552278995513916,
+        -0.36354875564575195,
+        -0.18577136099338531,
+        -1.956291913986206,
+        -1.6801013946533203,
+        -7.678869724273682,
+        -8.539926528930664,
+        -0.6294453740119934,
+        -0.07775841653347015,
+        -2.1328067779541016,
+        -2.1285502910614014,
+        -8.228925704956055,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        -7.334665780067444,
+        -5.257390398583084e-07,
+        -2.8215323254698887e-05,
+        -1.841589093208313,
+        -1.2851550579071045,
+        -6.268109321594238,
+        -5.912737417221069,
+        -0.572821855545044,
+        -2.81171942333458e-05,
+        -1.8467556238174438,
+        -1.440324923992157,
+        -5.8623305034637445,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        4.902149534225477,
+        3.4775466918945312,
+        4.224453401565551,
+        1.7064905214309687,
+        1.480757713317871,
+        3.609884884357428,
+        1.258314609527588,
+        2.967935800552368,
+        3.2566452693939207,
+        1.7433889162540437,
+        1.4066662788391113,
+        3.581243851184845,
+        1.0,
+        1.0
+      ]
+    },
+    "num_transitions": 6075103,
+    "num_trajectories": 27500
+  }
+}

settings/qwen_robotwin/uniform/external_resume_certificate.json

@@ -0,0 +1,138 @@
+{
+  "branch_resume_identity": {
+    "config_sha256": "8cf4970a4c7c919c72b15872ea42164783860fd4cecbdc75e30d1766ea349dd3",
+    "identity_schema_version": 2,
+    "manifest_sha256": "dcb728a00da4b4ad7f989fa313d6d3355e13f57efefc7267e9ea092f2ffd8a50",
+    "model_structure_sha256": "7809faa5245147f81b98f3dbd17c7a063a7cb07a21e18af588c748c5925e1407",
+    "proposal_alpha": null,
+    "proposal_sha256": null,
+    "seed": 484370,
+    "total_batch_size": 40
+  },
+  "completed_steps": 11000,
+  "optimizer_param_groups": [
+    {
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "initial_lr": 1e-05,
+      "lr": 9.959920880962606e-06,
+      "name": "qwen_vl_interface",
+      "weight_decay": 1e-08
+    },
+    {
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "initial_lr": 0.0001,
+      "lr": 9.959920880962605e-05,
+      "name": "action_model",
+      "weight_decay": 1e-08
+    }
+  ],
+  "optimizer_state_entries": 8,
+  "parent_checkpoint": "/mnt/miaochengyu/starvla-infoflow-runtime/runs/qwen3vl-groot-robotwin-official150k/qwen3vl-groot-robotwin-official150k-adamw8bit-mb5-ga4-seed42-20260810/checkpoints/steps_11000_accelerate",
+  "parent_checkpoint_commit_sha256": "180f1ed2d2b7a9e6c158974fc9ec951182dfb629f6a2806808ac5639d491ce1b",
+  "parent_checkpoint_files": [
+    {
+      "bytes": 1525,
+      "name": "custom_checkpoint_0.pkl",
+      "sha256": "70788f8a2d7852a70fbb794ab9176d14ad37fb6d1b1d821a368572e1d8524390"
+    },
+    {
+      "bytes": 13,
+      "name": "latest",
+      "sha256": "68d0ca762047fd48fdd6a06a5e6465209289c028bfd63f5557a110ff1ebc15ac"
+    },
+    {
+      "bytes": 13867381301,
+      "name": "pytorch_model/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt",
+      "sha256": "42b6d10bb158b404056983ce88505c32738b29f3a062fe6835d24ce26c521a6a"
+    },
+    {
+      "bytes": 13867381301,
+      "name": "pytorch_model/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt",
+      "sha256": "7b0a1bbc92efffbf056bc40f40e93a94b67ddf3f375dc6ae263bcb048ba6ec63"
+    },
+    {
+      "bytes": 566626,
+      "name": "pytorch_model/zero_pp_rank_0_mp_rank_00_model_states.pt",
+      "sha256": "8a3ef88f9ab0df8197a836dca1d6e2416835f0b3d827cb10a78254b9a969ce7c"
+    },
+    {
+      "bytes": 566498,
+      "name": "pytorch_model/zero_pp_rank_1_mp_rank_00_model_states.pt",
+      "sha256": "67cd03fceb5ee4c2be6affe504f71a48ee10127d6ea3a1f00d05721895ffba53"
+    },
+    {
+      "bytes": 15017,
+      "name": "random_states_0.pkl",
+      "sha256": "636b1c5983ee2f8a5047025027453c9a478c0606389f2f6945b370f39cec9e65"
+    },
+    {
+      "bytes": 15081,
+      "name": "random_states_1.pkl",
+      "sha256": "416b3be53ff049935984cc65385d7dae3b0e5a82df66c9516f61ee9100c41961"
+    },
+    {
+      "bytes": 33270,
+      "name": "zero_to_fp32.py",
+      "sha256": "e8a2553b434ac043ec63b022db1e768277acf8d4e70f10a2842db6d238977135"
+    }
+  ],
+  "parent_checkpoint_files_sha256": "73692dae620e3d4368fe513e8adff4fe54bdb25341b4293e34ee2a6286554f8d",
+  "parent_checkpoint_integrity_manifest_sha256": null,
+  "parent_config_sha256": "4882855d94076fc0ec19a0d511f14b1cbe82860c8b5df3dcc64c5c5f6c0d2321",
+  "parent_resume_identity": {
+    "config_sha256": "1a48ea51cf89ce72a7956aaa29f710bba0326c9b726e52795561b381507d26fa",
+    "identity_schema_version": 2,
+    "manifest_sha256": "dcb728a00da4b4ad7f989fa313d6d3355e13f57efefc7267e9ea092f2ffd8a50",
+    "model_structure_sha256": "7809faa5245147f81b98f3dbd17c7a063a7cb07a21e18af588c748c5925e1407",
+    "proposal_alpha": null,
+    "proposal_sha256": null,
+    "seed": 42,
+    "total_batch_size": 40
+  },
+  "proposal_sha256": null,
+  "rank_state_summaries": [
+    {
+      "optimizer_state_entries": 4,
+      "optimizer_state_sha256": "e1eea37514fa0478835597d1594b7a43b45cea9adc4ccb79a0de8fb9e10fe9f0",
+      "rank": 0,
+      "trainable_state_sha256": "ca2a19167d59aab693d0a0a1e17ea9adbf9cfd7367dfa7a5c90b04cf2183b717"
+    },
+    {
+      "optimizer_state_entries": 4,
+      "optimizer_state_sha256": "62c4e2908a7e97fe9a1d60642a85968081b55345493a11b9e7f68e49fcdb22f3",
+      "rank": 1,
+      "trainable_state_sha256": "ca2a19167d59aab693d0a0a1e17ea9adbf9cfd7367dfa7a5c90b04cf2183b717"
+    }
+  ],
+  "repeated_diffusion_steps": 4,
+  "rng_streams": {
+    "model_base": 721927,
+    "sampler": 484370,
+    "transform": 289726
+  },
+  "scheduler": {
+    "class": "LambdaLR",
+    "last_epoch": 11000,
+    "last_lr": [
+      9.959920880962606e-06,
+      9.959920880962605e-05
+    ],
+    "step_count": 11001
+  },
+  "scheduler_resume_mode": "accelerate_checkpoint",
+  "schema_version": 1,
+  "status": "exact_external_resume_loaded",
+  "total_batch_size": 40,
+  "trainable_parameter_elements": 32022798,
+  "trainable_parameter_sha256": "1bc504b22e4375be40bbbd8250be5b5f380c14d2fb47ed85deba3604a84b5484",
+  "trainable_parameter_tensors": 957,
+  "world_size": 2
+}

settings/qwen_robotwin/uniform/initialization_manifest.json

@@ -0,0 +1,6 @@
+{
+  "schema_version": 1,
+  "initialization_seed": 484370,
+  "action_model_sha256": "1cab43a1e801a94110a20384a9ad6cd76a239fedd9fe17890662da7e22d47e62",
+  "action_model_parameter_elements": 160689934
+}

settings/qwen_robotwin/uniform/valid_anchor_manifest.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2090f6b9d7e1bc5ef41897436690f031c187d625fa6b71ae0c54f3793660c757
+size 107646168