wxgkuaile Goto Github PK

This is an algorithm re-implemented in PyTorch, which aims to combine model-based reinforcement learning (MBRL) with normalizing flows for the first time to explore their applications and performance in deep reinforcement learning: When to Trust Your Model: Model-Based Policy Optimization. A Coupled Flow Approach to Imitation Learning

This code is based on a previous paper in the NeurIPS reproducibility challenge that reproduces the result with a tensorflow ensemble model but shows a significant drop in performance with a pytorch ensemble model. This code re-implements the ensemble dynamics model with pytorch and closes the gap.

MuJoCo 1.5 & MuJoCo 2.0

python main_mbpo.py --env_name ${env_name}'-v2' --num_epoch=${num_e} --use_algo 'discriminator'

if you want to quickly reverso context:

sh run.sh

• Official tensorflow implementation: https://github.com/JannerM/mbpo
• Code to the reproducibility challenge paper: https://github.com/jxu43/replication-mbpo