Upload folder using huggingface_hub

qwen-0126v/checkpoints/steps_32000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:945903208ae5588e598efe5b3eca4764efa4e5819a5830fb16da9930bb5e60d4
+size 6867599621

qwen-0126v/config.json

@@ -0,0 +1,90 @@
+{
+  "run_id": "QwenAdapter-0126v-eep",
+  "run_root_dir": "/data/LumosVLA/LumosVLA/results",
+  "seed": 42,
+  "trackers": [
+    "jsonl",
+    "wandb"
+  ],
+  "wandb_entity": "xspadex-nanyang-technological-university-singapore",
+  "wandb_project": "qwenvl-0126v-eep",
+  "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"
+  },
+  "datasets": {
+    "vla_data": {
+      "dataset_py": "lerobot_datasets",
+      "data_root_dir": "/data/LumosVLA/LumosVLA/data/starvla_intern",
+      "data_mix": "train_internAdapter",
+      "per_device_batch_size": 32,
+      "num_workers": 4,
+      "action_type": "abs_ee",
+      "image_size": 256,
+      "video_backend": "pyav"
+    }
+  },
+  "trainer": {
+    "epochs": 100,
+    "max_train_steps": 55000,
+    "num_warmup_steps": 1000,
+    "save_interval": 1000,
+    "eval_interval": 100000,
+    "learning_rate": {
+      "base": 4e-06,
+      "qwen_vl_interface": 4e-06,
+      "action_model": 1e-05
+    },
+    "lr_scheduler_type": "cosine_with_min_lr",
+    "scheduler_specific_kwargs": {
+      "min_lr": 5e-07
+    },
+    "freeze_modules": "qwen_vl_interface.model.model.language_model",
+    "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": true,
+    "resume_step": 27000
+  },
+  "is_resume": false,
+  "resume_epoch": null,
+  "resume_step": null,
+  "enable_gradient_checkpointing": true,
+  "enable_mixed_precision_training": true,
+  "report_to": "wandb",
+  "wandb_run_name": "v1",
+  "output_dir": "/data/LumosVLA/LumosVLA/results/QwenAdapter-0126v-eep"
+}

qwen-0126v/config.yaml

@@ -0,0 +1,50 @@
+datasets:
+  vla_data:
+    data_mix: train_internAdapter
+    data_root_dir: /data/LumosVLA/LumosVLA/data/starvla_intern
+    dataset_py: lerobot_datasets
+    per_device_batch_size: 32
+    video_backend: pyav
+framework:
+  action_model:
+    action_dim: 20
+    action_query_num: 64
+    hidden_dim: 2048
+    num_actions_chunk: 30
+    phase: Training
+    state_dim: 20
+    use_pro_version: true
+    use_proprio: true
+  name: QwenAdapter
+  qwenvl:
+    base_vlm: /data/LumosVLA/LumosVLA/playground/Pretrained_models/hub/models--Qwen--Qwen3-VL-2B-Instruct/snapshots/89644892e4d85e24eaac8bacfd4f463576704203
+    vl_hidden_dim: 2048
+output_dir: /data/LumosVLA/LumosVLA/results/QwenAdapter-0126v-eep
+run_id: QwenAdapter-0126v-eep
+run_root_dir: /data/LumosVLA/LumosVLA/results
+seed: 42
+trainer:
+  eval_interval: 100000
+  freeze_modules: qwen_vl_interface.model.model.language_model
+  gradient_accumulation_steps: 1
+  gradient_clipping: 1.0
+  is_resume: true
+  learning_rate:
+    action_model: 1.0e-05
+    base: 4.0e-06
+    qwen_vl_interface: 4.0e-06
+  logging_frequency: 10
+  lr_scheduler_type: cosine_with_min_lr
+  max_train_steps: 55000
+  num_warmup_steps: 1000
+  optimizer:
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 0
+  save_interval: 1000
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+wandb_entity: xspadex-nanyang-technological-university-singapore
+wandb_project: qwenvl-0126v-eep

qwen-0126v/dataset_statistics.json

@@ -0,0 +1,218 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        0.23971196915954351,
+        -0.07325393698192784,
+        0.2039592359215021,
+        -1.5161913484334946,
+        0.38040905725210905,
+        0.9461286216974258,
+        0.03827051632106304,
+        0.23643459752202034,
+        0.09655647608451545,
+        0.20731287077069283,
+        1.2944340095855296,
+        0.5062535796314478,
+        -1.8863605856895447,
+        0.03741371212527156
+      ],
+      "std": [
+        0.09051698658774786,
+        0.11684912922379904,
+        0.07275034909773587,
+        2.488221130855096,
+        0.34500947795229847,
+        2.7312296948562444,
+        0.02671876115137093,
+        0.09143749808186113,
+        0.07616377902910691,
+        0.05979330309838163,
+        2.6559062722556184,
+        0.30711328684858114,
+        2.077428273369966,
+        0.02728443102320894
+      ],
+      "max": [
+        0.5393329858779907,
+        0.48442599177360535,
+        0.6967939734458923,
+        3.1415927410125732,
+        1.5707963705062866,
+        3.1415927410125732,
+        0.06384000182151794,
+        0.5489550232887268,
+        0.45939499139785767,
+        0.6456500291824341,
+        3.1415927410125732,
+        1.5707963705062866,
+        3.1415927410125732,
+        0.060339998453855515
+      ],
+      "min": [
+        -0.10788500308990479,
+        -0.5191159844398499,
+        0.05315700173377991,
+        -3.1410515308380127,
+        -1.5302174091339111,
+        -3.1410515308380127,
+        0.00014000000373926014,
+        -0.02319999970495701,
+        -0.26581600308418274,
+        0.06903599947690964,
+        -3.1410515308380127,
+        -0.6304129362106323,
+        -3.1410515308380127,
+        0.0
+      ],
+      "q01": [
+        0.044102057814598083,
+        -0.36912801861763,
+        0.08692353963851929,
+        -3.137159585952759,
+        -0.20070484280586243,
+        -3.137392044067383,
+        0.0003499999875202775,
+        0.053725481033325195,
+        -0.15914200246334076,
+        0.10173498094081879,
+        -3.138237953186035,
+        -0.11282174289226532,
+        -3.1348206996917725,
+        7.000000186963007e-05
+      ],
+      "q99": [
+        0.4720889925956726,
+        0.4121397137641907,
+        0.4779260456562042,
+        3.1415927410125732,
+        1.5365502834320068,
+        3.1415927410125732,
+        0.059780001640319824,
+        0.4840700030326843,
+        0.34832507371902466,
+        0.42531776428222656,
+        3.1415927410125732,
+        1.5435858964920044,
+        3.1415927410125732,
+        0.05992000177502632
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        false
+      ]
+    },
+    "state": {
+      "mean": [
+        0.23971196915954351,
+        -0.07325393698192784,
+        0.2039592359215021,
+        -1.5161913484334946,
+        0.38040905725210905,
+        0.9461286216974258,
+        0.03827051632106304,
+        0.23643459752202034,
+        0.09655647608451545,
+        0.20731287077069283,
+        1.2944340095855296,
+        0.5062535796314478,
+        -1.8863605856895447,
+        0.03741371212527156
+      ],
+      "std": [
+        0.09051698658774786,
+        0.11684912922379904,
+        0.07275034909773587,
+        2.488221130855096,
+        0.34500947795229847,
+        2.7312296948562444,
+        0.02671876115137093,
+        0.09143749808186113,
+        0.07616377902910691,
+        0.05979330309838163,
+        2.6559062722556184,
+        0.30711328684858114,
+        2.077428273369966,
+        0.02728443102320894
+      ],
+      "max": [
+        0.5393329858779907,
+        0.48442599177360535,
+        0.6967939734458923,
+        3.1415927410125732,
+        1.5707963705062866,
+        3.1415927410125732,
+        0.06384000182151794,
+        0.5489550232887268,
+        0.45939499139785767,
+        0.6456500291824341,
+        3.1415927410125732,
+        1.5707963705062866,
+        3.1415927410125732,
+        0.060339998453855515
+      ],
+      "min": [
+        -0.10788500308990479,
+        -0.5191159844398499,
+        0.05315700173377991,
+        -3.1410515308380127,
+        -1.5302174091339111,
+        -3.1410515308380127,
+        0.00014000000373926014,
+        -0.02319999970495701,
+        -0.26581600308418274,
+        0.06903599947690964,
+        -3.1410515308380127,
+        -0.6304129362106323,
+        -3.1410515308380127,
+        0.0
+      ],
+      "q01": [
+        0.044102057814598083,
+        -0.36912801861763,
+        0.08692353963851929,
+        -3.137159585952759,
+        -0.20070484280586243,
+        -3.137392044067383,
+        0.0003499999875202775,
+        0.053725481033325195,
+        -0.15914200246334076,
+        0.10173498094081879,
+        -3.138237953186035,
+        -0.11282174289226532,
+        -3.1348206996917725,
+        7.000000186963007e-05
+      ],
+      "q99": [
+        0.4720889925956726,
+        0.4121397137641907,
+        0.4779260456562042,
+        3.1415927410125732,
+        1.5365502834320068,
+        3.1415927410125732,
+        0.059780001640319824,
+        0.4840700030326843,
+        0.34832507371902466,
+        0.42531776428222656,
+        3.1415927410125732,
+        1.5435858964920044,
+        3.1415927410125732,
+        0.05992000177502632
+      ]
+    },
+    "num_transitions": 1509415,
+    "num_trajectories": 1361
+  }
+}