Upload folder using huggingface_hub

qwenvl-eep0108/checkpoints/steps_5000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb5525ac67a9e861feaa2d405abd626df367ac79ad4c8994690426e9ebdb76fb
+size 6867598474

qwenvl-eep0108/config.json

@@ -0,0 +1,87 @@
+{
+  "run_id": "QwenAdapter-0108-eep",
+  "run_root_dir": "/data/LumosVLA/LumosVLA/results",
+  "seed": 42,
+  "trackers": [
+    "jsonl",
+    "wandb"
+  ],
+  "wandb_entity": "xspadex-nanyang-technological-university-singapore",
+  "wandb_project": "qwenadapter-eep-0108",
+  "is_debug": false,
+  "framework": {
+    "name": "QwenAdapter",
+    "qwenvl": {
+      "base_vlm": "/data/LumosVLA/LumosVLA/playground/Pretrained_models/hub/models--Qwen--Qwen3-VL-2B-Instruct/snapshots/89644892e4d85e24eaac8bacfd4f463576704203",
+      "attn_implementation": "flash_attention_2",
+      "vl_hidden_dim": 2048
+    },
+    "action_model": {
+      "action_model_type": "VLA-Adapter",
+      "hidden_dim": 2048,
+      "action_dim": 20,
+      "state_dim": 20,
+      "action_query_num": 64,
+      "use_pro_version": true,
+      "use_proprio": true,
+      "phase": "Training",
+      "num_actions_chunk": 30
+    },
+    "framework_py": "QwenAdapter-0108-eep"
+  },
+  "datasets": {
+    "vla_data": {
+      "dataset_py": "lerobot_datasets",
+      "data_root_dir": "/data/LumosVLA/LumosVLA/data/0106_data_10_in_1",
+      "data_mix": "train_0107eep",
+      "per_device_batch_size": 32,
+      "num_workers": 4,
+      "action_type": "abs_ee",
+      "image_size": 224
+    }
+  },
+  "trainer": {
+    "epochs": 100,
+    "max_train_steps": 15000,
+    "num_warmup_steps": 1000,
+    "save_interval": 2500,
+    "eval_interval": 100000,
+    "learning_rate": {
+      "base": 5e-05,
+      "qwen_vl_interface": 1e-05,
+      "action_model": 0.0002
+    },
+    "lr_scheduler_type": "polynomial",
+    "scheduler_specific_kwargs": {
+      "lr_end": 1e-06,
+      "power": 1.0
+    },
+    "freeze_modules": "qwen_vl_interface",
+    "loss_scale": {
+      "vla": 1.0,
+      "vlm": 0.05
+    },
+    "max_grad_norm": 1.0,
+    "warmup_ratio": 0.1,
+    "weight_decay": 0.0,
+    "logging_frequency": 10,
+    "gradient_clipping": 1.0,
+    "gradient_accumulation_steps": 1,
+    "optimizer": {
+      "name": "AdamW",
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "weight_decay": 0
+    }
+  },
+  "is_resume": false,
+  "resume_epoch": null,
+  "resume_step": null,
+  "enable_gradient_checkpointing": true,
+  "enable_mixed_precision_training": true,
+  "wandb_run_name": "v1",
+  "output_dir": "/data/LumosVLA/LumosVLA/results/QwenAdapter-0108-eep"
+}

qwenvl-eep0108/config.yaml

@@ -0,0 +1,73 @@
+run_id: QwenAdapter-0108-eep
+run_root_dir: /data/LumosVLA/LumosVLA/results
+seed: 42
+trackers:
+- jsonl
+- wandb
+wandb_entity: xspadex-nanyang-technological-university-singapore
+wandb_project: qwenadapter-eep-0108
+is_debug: false
+framework:
+  name: QwenAdapter
+  qwenvl:
+    base_vlm: /data/LumosVLA/LumosVLA/playground/Pretrained_models/hub/models--Qwen--Qwen3-VL-2B-Instruct/snapshots/89644892e4d85e24eaac8bacfd4f463576704203
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  action_model:
+    action_model_type: VLA-Adapter
+    hidden_dim: 2048
+    action_dim: 20
+    state_dim: 20
+    action_query_num: 64
+    use_pro_version: true
+    use_proprio: true
+    phase: Training
+    num_actions_chunk: 30
+  framework_py: QwenAdapter-0108-eep
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /data/LumosVLA/LumosVLA/data/0106_data_10_in_1
+    data_mix: train_0107eep
+    per_device_batch_size: 32
+    num_workers: 4
+    action_type: abs_ee
+    image_size: 224
+trainer:
+  epochs: 100
+  max_train_steps: 15000
+  num_warmup_steps: 1000
+  save_interval: 2500
+  eval_interval: 100000
+  learning_rate:
+    base: 5.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0002
+  lr_scheduler_type: polynomial
+  scheduler_specific_kwargs:
+    lr_end: 1.0e-06
+    power: 1.0
+  freeze_modules: qwen_vl_interface
+  loss_scale:
+    vla: 1.0
+    vlm: 0.05
+  max_grad_norm: 1.0
+  warmup_ratio: 0.1
+  weight_decay: 0.0
+  logging_frequency: 10
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 1
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 0
+is_resume: false
+resume_epoch: null
+resume_step: null
+enable_gradient_checkpointing: true
+enable_mixed_precision_training: true
+wandb_run_name: v1
+output_dir: /data/LumosVLA/LumosVLA/results/QwenAdapter-0108-eep

qwenvl-eep0108/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        0.23282504230737688,
+        -0.05165965789929033,
+        0.21386985182762147,
+        -91.47970352172852,
+        20.121428585052488,
+        42.44750967025757,
+        0.23724629282951357,
+        0.08887637909501792,
+        0.21939275562763214,
+        45.390340900421144,
+        29.264903831481938,
+        -119.05479431152344,
+        0.03987310770899058,
+        0.03885028753429652
+      ],
+      "std": [
+        0.09953932969670672,
+        0.1263747541518203,
+        0.07284480907274637,
+        133.6138685971796,
+        18.383623797067543,
+        155.94565103777856,
+        0.10225355728400369,
+        0.08295887480529306,
+        0.06472772608178973,
+        158.29364115016972,
+        16.00381071103632,
+        97.27639744542212,
+        0.026943868199832993,
+        0.02694874948904217
+      ],
+      "max": [
+        0.5771363973617554,
+        0.5499705076217651,
+        1.1769417524337769,
+        179.99998474121094,
+        89.99874114990234,
+        179.9999237060547,
+        0.5673798322677612,
+        0.5647112131118774,
+        0.6844663023948669,
+        179.99998474121094,
+        89.99887084960938,
+        179.99996948242188,
+        0.05999999865889549,
+        0.05999999865889549
+      ],
+      "min": [
+        -0.15116086602210999,
+        -0.5200218558311462,
+        0.050602782517671585,
+        -180.0,
+        -20.672367095947266,
+        -179.99993896484375,
+        -0.0248020701110363,
+        -0.29858818650245667,
+        0.06846695393323898,
+        -179.99998474121094,
+        -13.50222110748291,
+        -179.9997100830078,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        0.05501473695039749,
+        -0.3686788082122803,
+        0.0868777185678482,
+        -179.78305053710938,
+        -9.416550636291504,
+        -179.69680786132812,
+        0.056286852806806564,
+        -0.18694551289081573,
+        0.10547168552875519,
+        -179.8378448486328,
+        -0.19658219814300537,
+        -179.51170349121094,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        0.47519516944885254,
+        0.4694986939430237,
+        0.46994128823280334,
+        179.755126953125,
+        89.57238006591797,
+        179.736328125,
+        0.48799216747283936,
+        0.3558615446090698,
+        0.43878117203712463,
+        179.83042907714844,
+        89.67033386230469,
+        179.49917602539062,
+        0.05999999865889549,
+        0.05999999865889549
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        0.24175260961055758,
+        -0.049470676161581656,
+        0.21608835756778721,
+        -1.413518315553665,
+        0.428703111410141,
+        0.7577013611793519,
+        0.24074497371912004,
+        0.08981470651924611,
+        0.21937729418277743,
+        0.8831917621195315,
+        0.6019748389720918,
+        -2.0126270771026613,
+        0.04148903861641884,
+        0.03994188271462917
+      ],
+      "std": [
+        0.09280708088090048,
+        0.12742504567945281,
+        0.0744079447762679,
+        2.5498722585501863,
+        0.3793943873796246,
+        2.808145353959299,
+        0.09863927576384764,
+        0.0815411369762193,
+        0.06492154445492224,
+        2.805201274470533,
+        0.31938597693868265,
+        1.884995382965943,
+        0.025358022526895105,
+        0.026140522839142637
+      ],
+      "max": [
+        0.5393329858779907,
+        0.48442599177360535,
+        0.6967939734458923,
+        3.1415927410125732,
+        1.5707963705062866,
+        3.1415927410125732,
+        0.5489550232887268,
+        0.44545501470565796,
+        0.6456500291824341,
+        3.1415927410125732,
+        1.5707963705062866,
+        3.1415927410125732,
+        0.06159999966621399,
+        0.05998999997973442
+      ],
+      "min": [
+        -0.10788500308990479,
+        -0.5191159844398499,
+        0.05315700173377991,
+        -3.1410515308380127,
+        -1.5302174091339111,
+        -3.1410515308380127,
+        -0.02319999970495701,
+        -0.26581600308418274,
+        0.06903599947690964,
+        -3.1410515308380127,
+        -0.23581144213676453,
+        -3.1410515308380127,
+        0.00014000000373926014,
+        7.000000186963007e-05
+      ],
+      "q01": [
+        0.056127000600099564,
+        -0.36912801861763,
+        0.08751751482486725,
+        -3.137159585952759,
+        -0.16739033162593842,
+        -3.1360411643981934,
+        0.056008998304605484,
+        -0.15914200246334076,
+        0.10651859641075134,
+        -3.138237953186035,
+        -0.005713686812669039,
+        -3.131770372390747,
+        0.0004199999966658652,
+        7.000000186963007e-05
+      ],
+      "q99": [
+        0.4720889925956726,
+        0.4121397137641907,
+        0.4779260456562042,
+        3.1415927410125732,
+        1.5365502834320068,
+        3.1415927410125732,
+        0.4841437041759491,
+        0.35178446769714355,
+        0.42531776428222656,
+        3.1415927410125732,
+        1.5435858964920044,
+        3.1415927410125732,
+        0.059780001640319824,
+        0.05992000177502632
+      ]
+    },
+    "num_transitions": 885334,
+    "num_trajectories": 881
+  }
+}