davda54 / CEM-RL

Combining Evolutionary Algorithms and deep RL in various ways

CEM-RL

Pytorch implementation of CEM-RL: https://arxiv.org/pdf/1810.01222.pdf

To reproduce the results of the paper:

Without importance mixing:

python es_grad.py --env ENV_NAME --use_td3 --output OUTPUT_FOLDER

With importance mixing:

python es_grad_im.py --env ENV_NAME --use_td3 --output OUTPUT_FOLDER

TD3:

python distributed.py --env ENV_NAME --use_td3 --output OUTPUT_FOLDER