PARL/examples/TD3

• Reproduce TD3 with PARL
• • Mujoco games introduction
  • Benchmark result
• How to use
• • Dependencies:
  • Start Training:

Reproduce TD3 with PARL

Based on PARL, the TD3 algorithm of deep reinforcement learning has been reproduced, reaching the same level of indicators as the paper in Mujoco benchmarks.

Include following approaches:

• Clipped Double Q-learning
• Target Networks and Delayed Policy Update
• Target Policy Smoothing Regularization

TD3 in
Addressing Function Approximation Error in Actor-Critic Methods

Mujoco games introduction

Please see here to know more about Mujoco games.

Benchmark result

How to use

Dependencies:

• python3.5+
• paddlepaddle>=1.6.1
• parl
• gym
• mujoco-py>=1.50.1.0

Start Training:

# To train an agent for HalfCheetah-v2 game
python train.py

# To train for different game and different loss type
# python train.py --env [ENV_NAME]