SAC

This project provides an implementation of the Soft Actor-Critic (SAC) algorithm for solving continuous control tasks using environments from OpenAI Gym (e.g., MuJoCo). SAC is an off-policy actor-critic algorithm that combines maximum entropy reinforcement learning with function approximation for efficient and stable training.

See: https://gymnasium.farama.org/environments/mujoco/

Run local mlflow

mlflow ui

Train

python sac_train.py
python sac_train.py --profile ant
python sac_train.py --profile humanoid

Test

python sac_eval.py
python sac_eval.py --profile ant
python sac_eval.py --profile humanoid
python sac_eval.py --profile ant --model_uri 'runs:/55db85ebf343496783f5f2b88389b604/policy_net_episode_1100'