Description

---

Reimplementation of Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor.

Contributions are welcome. If you find any mistake (very likely) or know how to make it more stable, don't hesitate to send a pull request.

Requirements

---

• mujoco-py
• Plotly
• PyTorch

Run

---

Use the default hyperparameters.

For SAC (Gaussian Policy):

python main.py --algo SAC --env-name HalfCheetah-v2

For SAC (Gaussian Mixture Policy):

python main.py --algo SAC(GMM) --env-name HalfCheetah-v2 --k 4

TODO

---

• Gaussian Policy
• Reparameterization
• Gaussian Mixture Model
• Use 2 Q-functions
• Deterministic Policy
• Soft Actor-Critic (hard target update)
• Evaluate the trained Policy