Install Sample-Factory with Envpool dependencies with PyPI:
EnvPool is a C++-based batched environment pool with pybind11 and thread pool. It has high performance (~1M raw FPS with Atari games, ~3M raw FPS with Mujoco simulator).
We provide examples for envpool for Atari and Mujoco environments. The default parameters provide reasonable training speed, but can be tuning based on your machine configuration to achieve higher throughput.
To train a model with envpool in the
To visualize the training results, use the
Multiple experiments can be run in parallel with the launcher module.
atari_envs is an example launcher script that runs atari envs with 4 seeds.
- Sample-Factory's envpool environments were benchmarked against RL-Games using the same parameters. Sample-Factory achieved the same sample efficiency and wall time in the MuJoCo Ant environment. Additionally, using Envpool for this environment decrease the wall time compared to the default SF parameters without using Envpool.