- Added async_prefetch parameter to SACConfig for improved buffer management.
- Implemented get_iterator method in ReplayBuffer to support asynchronous prefetching of batches.
- Updated learner_server to utilize the new iterator for online and offline sampling, enhancing training efficiency.
- Cached encoder output in select_action method to reduce redundant computations.
- Updated action selection and grasp critic calls to utilize cached encoder features when available.
- Updated SACPolicy to conditionally compute grasp critic losses based on the presence of discrete actions.
- Refactored the forward method to handle grasp critic model selection and loss computation more clearly.
- Adjusted learner server to utilize optimized parameters for grasp critic during training.
- Improved action handling in the ManiskillMockGripperWrapper to accommodate both tuple and single action inputs.
- Cleaned up code formatting for better readability, including consistent spacing and removal of unnecessary blank lines.
- Consolidated continuous action dimension calculation to enhance clarity and maintainability.
- Simplified loss return statements in the forward method to improve code structure.
- Ensured grasp critic parameters are included conditionally based on configuration settings.
- Introduced mock_gripper parameter in ManiskillEnvConfig to enable gripper simulation.
- Added ManiskillMockGripperWrapper to adjust action space for environments with discrete actions.
- Updated SACPolicy to compute continuous action dimensions correctly, ensuring compatibility with the new gripper setup.
- Refactored action handling in the training loop to accommodate the changes in action dimensions.
- Updated SACPolicy to conditionally compute losses for grasp critic based on num_discrete_actions.
- Simplified forward method to return loss outputs as a dictionary for better clarity.
- Adjusted learner_server to handle both main and grasp critic losses during training.
- Ensured optimizers are created conditionally for grasp critic based on configuration settings.
- Removed GraspCriticNetworkConfig class and integrated its parameters into SACConfig.
- Added num_discrete_actions parameter to SACConfig for better action handling.
- Updated SACPolicy to conditionally create grasp critic networks based on num_discrete_actions.
- Enhanced grasp critic forward pass to handle discrete actions and compute losses accordingly.
- Integrated the grasp critic gradient update to the training loop in learner_server
- Added Adam optimizer and configured grasp critic learning rate in configuration_sac
- Added target critics networks update after the critics gradient step
- Enhanced type annotations for variables in the `SACPolicy` class to improve code clarity.
- Updated method calls to use keyword arguments for better readability.
- Streamlined the extraction of batch components, ensuring consistent typing across the class methods.
- Updated the `forward` method in `SACPolicy` to handle loss computation for actor, critic, and temperature models.
- Replaced direct calls to `compute_loss_*` methods with a unified `forward` method in `learner_server`.
- Enhanced batch processing by consolidating input parameters into a single dictionary for better readability and maintainability.
- Removed redundant code and improved documentation for clarity.
- Simplified the `image_features` property to directly iterate over `input_features`.
- Removed unused imports and unnecessary code related to main execution, enhancing clarity and maintainability.
Moved HilSerl env config to configs/env/configs.py
fixes in actor_server and modeling_sac and configuration_sac
added the possibility of ignoring missing keys in env_cfg in get_features_from_env_config function
- Introduced `WrapperConfig` dataclass for environment wrapper configurations.
- Updated `ManiskillEnvConfig` to include a `wrapper` field for enhanced environment management.
- Modified `SACConfig` to return `None` for `observation_delta_indices` and `action_delta_indices` properties.
- Refactored `make_robot_env` function to improve readability and maintainability.
Added support for hil_serl classifier to be trained with train.py
run classifier training by python lerobot/scripts/train.py --policy.type=hilserl_classifier
fixes in find_joint_limits, control_robot, end_effector_control_utils
- Reduced frame rate in `ManiskillEnvConfig` from 400 to 200.
- Enhanced `SACConfig` with new dataclasses for actor, learner, and network configurations.
- Improved input and output feature management in `SACConfig`.
- Refactored `actor_server` and `learner_server` to access configuration properties directly.
- Updated training pipeline to validate configurations and handle dataset repo IDs more robustly.
Add ManiSkill environment configuration and wrappers
- Introduced `VideoRecordConfig` for video recording settings.
- Added `ManiskillEnvConfig` to encapsulate environment-specific configurations.
- Implemented various wrappers for the ManiSkill environment, including observation and action scaling.
- Enhanced the `make_maniskill` function to create a wrapped ManiSkill environment with video recording and observation processing.
- Updated the `actor_server` and `learner_server` to utilize the new configuration structure.
- Refactored the training pipeline to accommodate the new environment and policy configurations.
- Deleted the `find_and_copy_params` function and the `Ensemble` class, as they were deemed unnecessary.
- Cleaned up imports by removing `from_modules` from `tensordict` to enhance code clarity.
- Simplified the assertion in the `Policy` class for better readability.
- Updated SACPolicy to create critic heads using a list comprehension for better readability.
- Simplified the saving and loading of models using `save_model` and `load_model` functions from the safetensors library.
- Introduced `initialize_offline_replay_buffer` function in the learner server to streamline offline dataset handling and replay buffer initialization.
- Enhanced logging for dataset loading processes to improve traceability during training.
- Introduced `grad_clip_norm` parameter in SAC configuration for gradient clipping
- Updated SACPolicy to store temperature as an instance variable for consistent usage
- Modified loss calculations in SACPolicy to utilize the instance temperature
- Enhanced MLP and CriticHead to support a customizable final activation function
- Implemented gradient clipping in the learner server during training steps for both actor and critic
- Added tracking for gradient norms in training information
- Implement `_save_pretrained` method to handle TensorDict state saving
- Add `_from_pretrained` class method for loading SAC policy from files
- Create utility function `find_and_copy_params` to handle parameter copying
- Removed `@torch.no_grad` decorator from Unnormalize forward method
- Added TODO comment for optimizing next action prediction in SAC policy
- Minor formatting adjustment in NaN assertion for log standard deviation
Co-authored-by: Yoel Chornton <yoel.chornton@gmail.com>
- Introduce `storage_device` parameter in SAC configuration and training settings
- Update learner server to use configurable storage device for replay buffer
- Reduce online buffer capacity in ManiSkill configuration
- Modify replay buffer initialization to support custom storage device
- Introduced Ensemble and CriticHead classes for more efficient critic network handling
- Added support for multiple camera inputs in observation encoder
- Optimized image encoding by batching image processing
- Updated configuration for ManiSkill environment with reduced image size and action scaling
- Compiled critic networks for improved performance
- Simplified normalization and ensemble handling in critic networks
Co-authored-by: michel-aractingi <michel.aractingi@gmail.com>
- Fixed big issue in the normalization of the actions in the `forward` function of the critic -- remove the `torch.no_grad` decorator in `normalize.py` in the normalization function
- Fixed performance issue to boost the optimization frequency by setting the storage device to be the same as the device of learning.
Co-authored-by: Adil Zouitine <adilzouitinegm@gmail.com>
AdilZouitine