mirror of
https://github.com/huggingface/lerobot.git
synced 2026-05-16 17:20:05 +00:00
Refactored hilserl config
This commit is contained in:
Michel Aractingi
+82
-42
@@ -56,27 +56,38 @@ pip install -e ".[hilserl]"
|
||||
|
||||
### Understanding Configuration
|
||||
|
||||
The training process begins with proper configuration for the HILSerl environment. The configuration class of interest is `HILSerlRobotEnvConfig` in `lerobot/envs/configs.py`. Which is defined as:
|
||||
The training process begins with proper configuration for the HILSerl environment. The configuration class of interest is `HILSerlRobotEnvConfig` in `lerobot/envs/configs.py`. The configuration is now organized into focused, nested sub-configs:
|
||||
|
||||
<!-- prettier-ignore-start -->
|
||||
```python
|
||||
class HILSerlRobotEnvConfig(EnvConfig):
|
||||
robot: RobotConfig | None = None # Main robot agent (defined in `lerobot/robots`)
|
||||
teleop: TeleoperatorConfig | None = None # Teleoperator agent, e.g., gamepad or leader arm, (defined in `lerobot/teleoperators`)
|
||||
wrapper: EnvTransformConfig | None = None # Environment wrapper settings; check `lerobot/scripts/server/gym_manipulator.py`
|
||||
fps: int = 10 # Control frequency
|
||||
teleop: TeleoperatorConfig | None = None # Teleoperator agent, e.g., gamepad or leader arm
|
||||
processor: HILSerlProcessorConfig # Processing pipeline configuration (nested)
|
||||
dataset: DatasetConfig # Dataset recording/replay configuration (nested)
|
||||
name: str = "real_robot" # Environment name
|
||||
mode: str = None # "record", "replay", or None (for training)
|
||||
device: str = "cuda" # Compute device
|
||||
|
||||
# Nested processor configuration
|
||||
class HILSerlProcessorConfig:
|
||||
control_mode: str = "gamepad" # Control mode
|
||||
observation: ObservationConfig # Observation processing settings
|
||||
image_preprocessing: ImagePreprocessingConfig # Image crop/resize settings
|
||||
gripper: GripperConfig # Gripper control and penalty settings
|
||||
reset: ResetConfig # Environment reset and timing settings
|
||||
inverse_kinematics: InverseKinematicsConfig # IK processing settings
|
||||
reward_classifier: RewardClassifierConfig # Reward classifier settings
|
||||
|
||||
# Dataset configuration
|
||||
class DatasetConfig:
|
||||
repo_id: str | None = None # LeRobot dataset repository ID
|
||||
dataset_root: str | None = None # Local dataset root (optional)
|
||||
task: str = "" # Task identifier
|
||||
num_episodes: int = 10 # Number of episodes for recording
|
||||
episode: int = 0 # episode index for replay
|
||||
device: str = "cuda" # Compute device
|
||||
push_to_hub: bool = True # Whether to push the recorded datasets to Hub
|
||||
pretrained_policy_name_or_path: str | None = None # For policy loading
|
||||
reward_classifier_pretrained_path: str | None = None # For reward model
|
||||
number_of_steps_after_success: int = 0 # For reward classifier, collect more positive examples after a success to train a classifier
|
||||
episode: int = 0 # Episode index for replay
|
||||
push_to_hub: bool = True # Whether to push datasets to Hub
|
||||
fps: int = 10 # Control frequency
|
||||
```
|
||||
<!-- prettier-ignore-end -->
|
||||
|
||||
@@ -133,19 +144,22 @@ Create a configuration file for recording demonstrations (or edit an existing on
|
||||
1. Set `mode` to `"record"`
|
||||
2. Specify a unique `repo_id` for your dataset (e.g., "username/task_name")
|
||||
3. Set `num_episodes` to the number of demonstrations you want to collect
|
||||
4. Set `crop_params_dict` to `null` initially (we'll determine crops later)
|
||||
4. Set `processor.image_preprocessing.crop_params_dict` to `{}` initially (we'll determine crops later)
|
||||
5. Configure `robot`, `cameras`, and other hardware settings
|
||||
|
||||
Example configuration section:
|
||||
|
||||
```json
|
||||
"mode": "record",
|
||||
"repo_id": "username/pick_lift_cube",
|
||||
"dataset_root": null,
|
||||
"task": "pick_and_lift",
|
||||
"num_episodes": 15,
|
||||
"episode": 0,
|
||||
"push_to_hub": true
|
||||
"dataset": {
|
||||
"repo_id": "username/pick_lift_cube",
|
||||
"dataset_root": null,
|
||||
"task": "pick_and_lift",
|
||||
"num_episodes": 15,
|
||||
"episode": 0,
|
||||
"push_to_hub": true,
|
||||
"fps": 10
|
||||
}
|
||||
```
|
||||
|
||||
### Using a Teleoperation Device
|
||||
@@ -191,10 +205,13 @@ The gamepad provides a very convenient way to control the robot and the episode
|
||||
To setup the gamepad, you need to set the `control_mode` to `"gamepad"` and define the `teleop` section in the configuration file.
|
||||
|
||||
```json
|
||||
"teleop": {
|
||||
"type": "gamepad",
|
||||
"use_gripper": true
|
||||
},
|
||||
"teleop": {
|
||||
"type": "gamepad",
|
||||
"use_gripper": true
|
||||
},
|
||||
"processor": {
|
||||
"control_mode": "gamepad"
|
||||
}
|
||||
```
|
||||
|
||||
<p align="center">
|
||||
@@ -216,11 +233,14 @@ The SO101 leader arm has reduced gears that allows it to move and track the foll
|
||||
To setup the SO101 leader, you need to set the `control_mode` to `"leader"` and define the `teleop` section in the configuration file.
|
||||
|
||||
```json
|
||||
"teleop": {
|
||||
"type": "so101_leader",
|
||||
"port": "/dev/tty.usbmodem585A0077921", # check your port number
|
||||
"use_degrees": true
|
||||
},
|
||||
"teleop": {
|
||||
"type": "so101_leader",
|
||||
"port": "/dev/tty.usbmodem585A0077921",
|
||||
"use_degrees": true
|
||||
},
|
||||
"processor": {
|
||||
"control_mode": "leader"
|
||||
}
|
||||
```
|
||||
|
||||
In order to annotate the success/failure of the episode, **you will need** to use a keyboard to press `s` for success, `esc` for failure.
|
||||
@@ -251,7 +271,7 @@ python -m lerobot.scripts.rl.gym_manipulator --config_path src/lerobot/configs/e
|
||||
|
||||
During recording:
|
||||
|
||||
1. The robot will reset to the initial position defined in the configuration file `fixed_reset_joint_positions`
|
||||
1. The robot will reset to the initial position defined in the configuration file `processor.reset.fixed_reset_joint_positions`
|
||||
2. Complete the task successfully
|
||||
3. The episode ends with a reward of 1 when you press the "success" button
|
||||
4. If the time limit is reached, or the fail button is pressed, the episode ends with a reward of 0
|
||||
@@ -310,11 +330,15 @@ observation.images.front: [180, 250, 120, 150]
|
||||
Add these crop parameters to your training configuration:
|
||||
|
||||
```json
|
||||
"crop_params_dict": {
|
||||
"observation.images.side": [180, 207, 180, 200],
|
||||
"observation.images.front": [180, 250, 120, 150]
|
||||
},
|
||||
"resize_size": [128, 128]
|
||||
"processor": {
|
||||
"image_preprocessing": {
|
||||
"crop_params_dict": {
|
||||
"observation.images.side": [180, 207, 180, 200],
|
||||
"observation.images.front": [180, 250, 120, 150]
|
||||
},
|
||||
"resize_size": [128, 128]
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
**Recommended image resolution**
|
||||
@@ -346,23 +370,29 @@ python -m lerobot.scripts.rl.gym_manipulator --config_path src/lerobot/configs/r
|
||||
- **mode**: set it to `"record"` to collect a dataset
|
||||
- **repo_id**: `"hf_username/dataset_name"`, name of the dataset and repo on the hub
|
||||
- **num_episodes**: Number of episodes to record
|
||||
- **number_of_steps_after_success**: Number of additional frames to record after a success (reward=1) is detected
|
||||
- **processor.reset.number_of_steps_after_success**: Number of additional frames to record after a success (reward=1) is detected
|
||||
- **fps**: Number of frames per second to record
|
||||
- **push_to_hub**: Whether to push the dataset to the hub
|
||||
|
||||
The `number_of_steps_after_success` parameter is crucial as it allows you to collect more positive examples. When a success is detected, the system will continue recording for the specified number of steps while maintaining the reward=1 label. Otherwise, there won't be enough states in the dataset labeled to 1 to train a good classifier.
|
||||
The `processor.reset.number_of_steps_after_success` parameter is crucial as it allows you to collect more positive examples. When a success is detected, the system will continue recording for the specified number of steps while maintaining the reward=1 label. Otherwise, there won't be enough states in the dataset labeled to 1 to train a good classifier.
|
||||
|
||||
Example configuration section for data collection:
|
||||
|
||||
```json
|
||||
{
|
||||
"mode": "record",
|
||||
"repo_id": "hf_username/dataset_name",
|
||||
"dataset_root": "data/your_dataset",
|
||||
"num_episodes": 20,
|
||||
"push_to_hub": true,
|
||||
"fps": 10,
|
||||
"number_of_steps_after_success": 15
|
||||
"dataset": {
|
||||
"repo_id": "hf_username/dataset_name",
|
||||
"dataset_root": "data/your_dataset",
|
||||
"num_episodes": 20,
|
||||
"push_to_hub": true,
|
||||
"fps": 10
|
||||
},
|
||||
"processor": {
|
||||
"reset": {
|
||||
"number_of_steps_after_success": 15
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
@@ -422,7 +452,11 @@ To use your trained reward classifier, configure the `HILSerlRobotEnvConfig` to
|
||||
<!-- prettier-ignore-start -->
|
||||
```python
|
||||
env_config = HILSerlRobotEnvConfig(
|
||||
reward_classifier_pretrained_path="path_to_your_pretrained_trained_model",
|
||||
processor=HILSerlProcessorConfig(
|
||||
reward_classifier=RewardClassifierConfig(
|
||||
pretrained_path="path_to_your_pretrained_trained_model"
|
||||
)
|
||||
),
|
||||
# Other environment parameters
|
||||
)
|
||||
```
|
||||
@@ -432,7 +466,13 @@ or set the argument in the json config file.
|
||||
|
||||
```json
|
||||
{
|
||||
"reward_classifier_pretrained_path": "path_to_your_pretrained_model"
|
||||
"processor": {
|
||||
"reward_classifier": {
|
||||
"pretrained_path": "path_to_your_pretrained_model",
|
||||
"success_threshold": 0.7,
|
||||
"success_reward": 1.0
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
|
||||
Reference in New Issue
Block a user