Release v2.0.1

Overview

Full Changelog: v2.0.0...v2.0.1

Fixed

• Fixed RL device setting in the on policy runner
• Fixes issue with splitting and padding of trajectories for recurrent network architecture training
• Updates wandb and neptune logging by @Mayankm96 in #18

Release v2.0.0

This release adds the following new features to the library:

Added

• Adds empirical normalization for observations and rewards
• Adds logging to Weights and Biases, Neptune
• Adds pre-commit formatter

Fixed

• Fixes issue with splitting and padding of trajectories for recurrent network architecture training

Changed

• Changes the extras key for storing logs. Earlier it was doing extras["episode"], which is now replaced with extras["log"] to make it more generic.
• Modified the config structure to have the class names within their respective algorithm and architecture dictionaries.

Pre-Release v1.0.2

This version corresponds to the original source code for rsl_rl at the point of publication of "Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning" by Rudin et al.

The release contains an optimized version of PPO implementation suited for use with GPU-accelerated simulators such as Isaac Gym.

This is the version of the code compatible with legged_gym.