/tqc_pytorch

Implementation of Truncated Quantile Critics method for continuous reinforcement learning. https://bayesgroup.github.io/tqc/

Primary LanguagePythonMIT LicenseMIT

Instruction

from docker folder:

bash docker_build.sh
bash docker_run.sh
docker exec -it tqc bash

Now you are inside the container, need to install and build mujoco

bash install_mujoco.sh
python
import mujoco_py

Run training:

python main.py --env HalfCheetah-v3 --top_quantiles_to_drop_per_net 0 --max_timesteps 10000000
python main.py --env BackflipCheetah --top_quantiles_to_drop_per_net 0 --max_timesteps 10000000