Cleanba is CleanRL-style implementation of DeepMind's Sebulba distributed training platform, but with a few different design choices to make distributed RL more reproducible and transparent to use.

Prerequisites:

• Python >=3.8
• Poetry 1.3.2+
• CUDA 11.2+
• CuDNN 8.2+

poetry install
poetry run pip install --upgrade "jax[cuda11_cudnn82]==0.4.8" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
poetry run python cleanba/cleanba_ppo.py --help
poetry run python cleanba/cleanba_ppo.py
poetry run python cleanba/cleanba_impala.py --help
poetry run python cleanba/cleanba_impala.py