Upload folder using huggingface_hub

qwen-all-v1/checkpoints/steps_35000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ef57bb9149a20bbb0db81b1619fe4ce3447495f97edef1d5ee1f3829d632c14
+size 6817169669

qwen-all-v1/config.json

@@ -10,79 +10,84 @@
   "wandb_project": "starvla-groot-2.5vl-3B",
   "is_debug": false,
   "framework": {
-    "name": "QwenAdapter",
+    "framework_py": "QwenAdapter-0106",
     "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": 14,
-      "state_dim": 14,
-      "action_query_num": 64,
-      "use_pro_version": true,
-      "use_proprio": true,
-      "phase": "Training",
-      "num_actions_chunk": 30
-    },
-    "framework_py": "QwenAdapter-0106"
-  },
-  "datasets": {
-    "vla_data": {
-      "dataset_py": "lerobot_datasets",
-      "data_root_dir": "/data/LumosVLA/LumosVLA/data/1230_data_7_in_1",
-      "data_mix": "train_1230",
-      "per_device_batch_size": 16,
-      "num_workers": 8,
-      "action_type": "abs_ee",
-      "image_size": [
-        224,
-        224
-      ]
+      "base_vlm": "/data/LumosVLA/LumosVLA/playground/Pretrained_models/hub/models--Qwen--Qwen3-VL-2B-Instruct/snapshots/89644892e4d85e24eaac8bacfd4f463576704203"
     }
   },
+  "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": 14,
+    "state_dim": 14,
+    "action_query_num": 64,
+    "use_pro_version": true,
+    "use_proprio": true,
+    "phase": "Training",
+    "num_actions_chunk": 30
+  },
+  "datasets": null,
+  "vla_data": {
+    "dataset_py": "lerobot_datasets",
+    "data_root_dir": "/data/LumosVLA/LumosVLA/data/1230_data_7_in_1",
+    "data_mix": "train_1230",
+    "per_device_batch_size": 16,
+    "num_workers": 8,
+    "action_type": "abs_ee",
+    "image_size": [
+      224,
+      224
+    ]
+  },
   "trainer": {
-    "epochs": 100,
-    "max_train_steps": 50000,
-    "num_warmup_steps": 1000,
-    "save_interval": 5000,
-    "eval_interval": 10000,
-    "learning_rate": {
-      "base": 2.5e-05,
-      "qwen_vl_interface": 1e-05,
-      "action_model": 0.0001
-    },
-    "lr_scheduler_type": "cosine_with_min_lr",
-    "scheduler_specific_kwargs": {
-      "min_lr": 1e-06
-    },
     "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": 1e-08
-    },
-    "is_resume": false,
-    "resume_epoch": null,
-    "resume_step": null,
-    "enable_gradient_checkpointing": true,
-    "enable_mixed_precision_training": true
+    "is_resume": true,
+    "resume_step": 5000
+  },
+  "epochs": 100,
+  "max_train_steps": 50000,
+  "num_warmup_steps": 1000,
+  "save_interval": 5000,
+  "eval_interval": 10000,
+  "learning_rate": {
+    "base": 2.5e-05,
+    "qwen_vl_interface": 1e-05,
+    "action_model": 1e-05
+  },
+  "lr_scheduler_type": "cosine_with_min_lr",
+  "scheduler_specific_kwargs": {
+    "min_lr": 1e-07
+  },
+  "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": 1e-08
   },
+  "is_resume": true,
+  "resume_epoch": null,
+  "resume_step": 35000,
+  "enable_gradient_checkpointing": true,
+  "enable_mixed_precision_training": true,
   "output_dir": "/data/LumosVLA/LumosVLA/results/QwenAdapter-0106"
 }

qwen-all-v1/config.yaml

@@ -8,66 +8,71 @@ wandb_entity: xspadex-nanyang-technological-university-singapore
 wandb_project: starvla-groot-2.5vl-3B
 is_debug: false
 framework:
-  name: QwenAdapter
+  framework_py: QwenAdapter-0106
   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: 14
-    state_dim: 14
-    action_query_num: 64
-    use_pro_version: true
-    use_proprio: true
-    phase: Training
-    num_actions_chunk: 30
-  framework_py: QwenAdapter-0106
-datasets:
-  vla_data:
-    dataset_py: lerobot_datasets
-    data_root_dir: /data/LumosVLA/LumosVLA/data/1230_data_7_in_1
-    data_mix: train_1230
-    per_device_batch_size: 16
-    num_workers: 8
-    action_type: abs_ee
-    image_size:
-    - 224
-    - 224
+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: 14
+  state_dim: 14
+  action_query_num: 64
+  use_pro_version: true
+  use_proprio: true
+  phase: Training
+  num_actions_chunk: 30
+datasets: null
+vla_data:
+  dataset_py: lerobot_datasets
+  data_root_dir: /data/LumosVLA/LumosVLA/data/1230_data_7_in_1
+  data_mix: train_1230
+  per_device_batch_size: 16
+  num_workers: 8
+  action_type: abs_ee
+  image_size:
+  - 224
+  - 224
 trainer:
-  epochs: 100
-  max_train_steps: 50000
-  num_warmup_steps: 1000
-  save_interval: 5000
-  eval_interval: 10000
-  learning_rate:
-    base: 2.5e-05
-    qwen_vl_interface: 1.0e-05
-    action_model: 0.0001
-  lr_scheduler_type: cosine_with_min_lr
-  scheduler_specific_kwargs:
-    min_lr: 1.0e-06
   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: 1.0e-08
-  is_resume: false
-  resume_epoch: null
-  resume_step: null
-  enable_gradient_checkpointing: true
-  enable_mixed_precision_training: true
+  is_resume: true
+  resume_step: 5000
+epochs: 100
+max_train_steps: 50000
+num_warmup_steps: 1000
+save_interval: 5000
+eval_interval: 10000
+learning_rate:
+  base: 2.5e-05
+  qwen_vl_interface: 1.0e-05
+  action_model: 1.0e-05
+lr_scheduler_type: cosine_with_min_lr
+scheduler_specific_kwargs:
+  min_lr: 1.0e-07
+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: 1.0e-08
+is_resume: true
+resume_epoch: null
+resume_step: 35000
+enable_gradient_checkpointing: true
+enable_mixed_precision_training: true
 output_dir: /data/LumosVLA/LumosVLA/results/QwenAdapter-0106