Qwen gr00t fix

examples/pi0/README.md

@@ -147,24 +147,24 @@ Configure the following fields:
 **System settings** (training hyperparameters):
 - `system.batch_size` - Batch size per GPU
 - `system.train_steps` - Total training steps
-- `system.optimizer.name` - Optimizer name (default: `"AdamW"`)
-- `system.optimizer.lr` - Learning rate (default: `2.5e-5`)
-- `system.optimizer.betas` - Optimizer betas (default: `[0.9, 0.95]`)
-- `system.optimizer.eps` - Optimizer epsilon (default: `1.0e-8`)
-- `system.optimizer.weight_decay` - Weight decay (default: `0.01`)
-- `system.scheduler.warmup_steps` - Warmup steps (default: `1000`)
-- `system.scheduler.decay_steps` - Decay steps (default: `30000`)
-- `system.scheduler.decay_lr` - Final learning rate after decay (default: `2.5e-6`)
 - `system.checkpoint.save_checkpoint` - Whether to save checkpoints (default: `true`)
 - `system.checkpoint.save_freq` - Steps between checkpoints (default: `1000`)
-- `system.checkpoint.output_directory` - Checkpoint output directory (default: `${experiment.exp_dir}/ckpt`)
+- `system.checkpoint.output_directory` - Checkpoint output directory (default: `${experiment.exp_dir}`)
 
 **Model settings**:
 - `model.model_name` - Model name: `"pi0"` or `"pi0.5"`
 - `model.checkpoint_dir` - Path to pretrained model (e.g., `/workspace/models/lerobot/pi0_base`)
 - `model.tokenizer_path` - Path to tokenizer (e.g., `/workspace/models/google/paligemma-3b-pt-224`)
 - `model.tokenizer_max_length` - Maximum tokenizer sequence length
 - `model.action_steps` - Number of action steps to predict
+- `model.optimizer.name` - Optimizer name (for example: `"AdamW"`)
+- `model.optimizer.lr` - Learning rate (for example: `2.5e-5`)
+- `model.optimizer.betas` - Optimizer betas (for example: `[0.9, 0.95]`)
+- `model.optimizer.eps` - Optimizer epsilon (for example: `1.0e-8`)
+- `model.optimizer.weight_decay` - Weight decay (for example: `0.01`)
+- `model.optimizer.scheduler.warmup_steps` - Warmup steps (for example: `1000`)
+- `model.optimizer.scheduler.decay_steps` - Decay steps (for example: `30000`)
+- `model.optimizer.scheduler.decay_lr` - Final learning rate after decay (for example: `2.5e-6`)
 
 **Data settings**:
 - `data.data_path` - Path to LeRobot dataset root (e.g., `/workspace/datasets/lerobot/aloha_mobile_cabinet`)
@@ -186,7 +186,7 @@ python run.py --config-path ./examples/pi0/conf --config-name train action=run
 
 Training logs are saved to `outputs/pi0_train/logs/host_0_localhost.output` by default.
 
-Checkpoints are saved to `${experiment.exp_dir}/ckpt` (default: `outputs/pi0_train/ckpt`).
+Checkpoints are saved to `${experiment.exp_dir}/checkpoints` (default: `outputs/pi0_train/checkpoints`).
 
 ### Stop Training
 ```sh

examples/pi0/conf/train/pi0.yaml

@@ -7,20 +7,8 @@ system:
   shuffle: false
   num_workers: 4
 
-  optimizer:
-    name: AdamW
-    lr: 2.5e-5
-    betas: [0.9, 0.95]
-    eps: 1.0e-8
-    weight_decay: 0.01
-
-  scheduler:
-    warmup_steps: 1000
-    decay_steps: 30000
-    decay_lr: 2.5e-6
-
   checkpoint:
-    output_directory: ${experiment.exp_dir}/ckpt
+    output_directory: ${experiment.exp_dir}
     # Whether to save checkpoint
     save_checkpoint: true
     # Number of steps between checkpoints
@@ -36,6 +24,17 @@ model:
   tokenizer_max_length: 48
   action_steps: 50
 
+  optimizer:
+    name: AdamW
+    lr: 2.5e-5
+    betas: [0.9, 0.95]
+    eps: 1.0e-8
+    weight_decay: 0.01
+    scheduler:
+      warmup_steps: 1000
+      decay_steps: 30000
+      decay_lr: 2.5e-6
+
 data:
   # Path to the training data
   data_path: /workspace/datasets/lerobot/aloha_mobile_cabinet

examples/pi0_5/README.md

@@ -154,24 +154,24 @@ Configure the following fields:
 **System settings** (training hyperparameters):
 - `system.batch_size` - Batch size per GPU
 - `system.train_steps` - Total training steps
-- `system.optimizer.name` - Optimizer name (default: `"AdamW"`)
-- `system.optimizer.lr` - Learning rate (default: `2.5e-5`)
-- `system.optimizer.betas` - Optimizer betas (default: `[0.9, 0.95]`)
-- `system.optimizer.eps` - Optimizer epsilon (default: `1.0e-8`)
-- `system.optimizer.weight_decay` - Weight decay (default: `0.01`)
-- `system.scheduler.warmup_steps` - Warmup steps (default: `1000`)
-- `system.scheduler.decay_steps` - Decay steps (default: `30000`)
-- `system.scheduler.decay_lr` - Final learning rate after decay (default: `2.5e-6`)
 - `system.checkpoint.save_checkpoint` - Whether to save checkpoints (default: `true`)
 - `system.checkpoint.save_freq` - Steps between checkpoints (default: `1000`)
-- `system.checkpoint.output_directory` - Checkpoint output directory (default: `${experiment.exp_dir}/ckpt`)
+- `system.checkpoint.output_directory` - Checkpoint output directory (default: `${experiment.exp_dir}`)
 
 **Model settings**:
 - `model.model_name` - Model name: `"pi0.5"`
 - `model.checkpoint_dir` - Path to pretrained model (e.g., `/workspace/models/lerobot/pi05_base`)
 - `model.tokenizer_path` - Path to tokenizer (e.g., `/workspace/models/google/paligemma-3b-pt-224`)
 - `model.tokenizer_max_length` - Maximum tokenizer sequence length (default: `200` for pi0.5)
 - `model.action_steps` - Number of action steps to predict
+- `model.optimizer.name` - Optimizer name (for example: `"AdamW"`)
+- `model.optimizer.lr` - Learning rate (for example: `2.5e-5`)
+- `model.optimizer.betas` - Optimizer betas (for example: `[0.9, 0.95]`)
+- `model.optimizer.eps` - Optimizer epsilon (for example: `1.0e-8`)
+- `model.optimizer.weight_decay` - Weight decay (for example: `0.01`)
+- `model.optimizer.scheduler.warmup_steps` - Warmup steps (for example: `1000`)
+- `model.optimizer.scheduler.decay_steps` - Decay steps (for example: `30000`)
+- `model.optimizer.scheduler.decay_lr` - Final learning rate after decay (for example: `2.5e-6`)
 
 **Data settings**:
 - `data.data_path` - Path to LeRobot dataset root (e.g., `/workspace/datasets/lerobot/aloha_mobile_cabinet`)
@@ -193,7 +193,7 @@ python run.py --config-path ./examples/pi0_5/conf --config-name train action=run
 
 Training logs are saved to `outputs/pi0_5_train/logs/host_0_localhost.output` by default.
 
-Checkpoints are saved to `${experiment.exp_dir}/ckpt` (default: `outputs/pi0_5_train/ckpt`).
+Checkpoints are saved to `${experiment.exp_dir}/checkpoints` (default: `outputs/pi0_5_train/checkpoints`).
 
 ### Stop Training
 ```sh

examples/pi0_5/conf/train/pi0_5.yaml

@@ -7,20 +7,8 @@ system:
   shuffle: false
   num_workers: 4
 
-  optimizer:
-    name: AdamW
-    lr: 2.5e-5
-    betas: [0.9, 0.95]
-    eps: 1.0e-8
-    weight_decay: 0.01
-
-  scheduler:
-    warmup_steps: 1000
-    decay_steps: 30000
-    decay_lr: 2.5e-6
-
   checkpoint:
-    output_directory: ${experiment.exp_dir}/ckpt
+    output_directory: ${experiment.exp_dir}
     # Whether to save checkpoint
     save_checkpoint: true
     # Number of steps between checkpoints
@@ -36,6 +24,17 @@ model:
   tokenizer_max_length: 200
   action_steps: 50
 
+  optimizer:
+    name: AdamW
+    lr: 2.5e-5
+    betas: [0.9, 0.95]
+    eps: 1.0e-8
+    weight_decay: 0.01
+    scheduler:
+      warmup_steps: 1000
+      decay_steps: 30000
+      decay_lr: 2.5e-6
+
 data:
   # Path to the training data
   data_path: /workspace/datasets/lerobot/aloha_mobile_cabinet