mirror of
https://github.com/wassname/ray.git
synced 2026-07-01 19:49:45 +08:00
Sven Mika
165a86f1ab
SAC (both torch and tf versions) are showing issues (crashes) due to numeric instabilities in the SquashedGaussian distribution (sampling + logp after extreme NN outputs). This PR fixes these. Stable MuJoCo learning (HalfCheetah) has been confirmed on both tf and torch versions. A Distribution stability test (using extreme NN outputs) has been added for SquashedGaussian (can be used for any other type of distribution as well).
Ray Documentation
To compile the documentation, run the following commands from this directory. Note that Ray must be installed first.
pip install -r requirements-doc.txt
make html
open _build/html/index.html
To test if there are any build errors with the documentation, do the following.
sphinx-build -W -b html -d _build/doctrees source _build/html