SAC

Reproduce SAC with PARL

Based on PARL, the SAC algorithm of deep reinforcement learning has been reproduced, reaching the same level of indicators as the paper in Mujoco benchmarks.

Paper: SAC in Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor

Mujoco games introduction

PARL currently supports the open-source version of Mujoco provided by DeepMind, so users do not need to download binaries of Mujoco as well as install mujoco-py and get license. For more details, please visit Mujoco

Benchmark result

• Each experiment was run three times with different seeds

How to use

Dependencies:

• python3.7+
• parl>=2.1.1
• paddlepaddle>=2.0.0
• gym>=0.26.0
• mujoco-py>=2.2.2

Start Training:

Train

# To train for HalfCheetah-v4(default),Hopper-v4,Walker2d-v4,Ant-v4
# --alpha 0.2(default)
python train.py --env [ENV_NAME]

# To reproduce the performance of Humanoid-v4
python train.py --env Humanoid-v4 --alpha 0.05