Skip to content

Latest commit

 

History

History
67 lines (51 loc) · 1.8 KB

README.md

File metadata and controls

67 lines (51 loc) · 1.8 KB

Online RL

States

SAC

python train_online.py --env_name=HalfCheetah-v2

Pixels

DrQ

MUJOCO_GL=egl XLA_PYTHON_CLIENT_PREALLOCATE=false python train_online_pixels.py --env_name=cheetah-run-v0

Offline RL

States

BC

python train_offline.py --config=configs/offline_config.py:bc --config.model_config.distr=unitstd_normal --env_name=halfcheetah-expert-v2

%BC

python train_offline.py --config=configs/offline_config.py:bc --config.model_config.distr=unitstd_normal --env_name=halfcheetah-medium-expert-v2 --filter_percentile=10

fBC (filtered BC)

python train_offline.py --config=configs/offline_config.py:bc --config.model_config.distr=unitstd_normal --env_name=antmaze-large-play-v2 --filter_threshold=0.5

BC (Autoregressive Policy)

python train_offline.py --config=configs/offline_config.py:bc --config.model_config.distr=ar --env_name=halfcheetah-expert-v2

IQL

AntMaze

python train_offline.py --config=configs/offline_config.py:iql_antmaze --env_name=antmaze-large-play-v2 --eval_interval=100000 --eval_episodes=100

Locomotion

python train_offline.py --config=configs/offline_config.py:iql_mujoco --env_name=halfcheetah-medium-expert-v2 --eval_interval=100000 --eval_episodes=100

Pixels

Collect data

MUJOCO_GL=egl XLA_PYTHON_CLIENT_PREALLOCATE=false python train_online_pixels.py --env_name=cheetah-run-v0 --save_buffer

PixelBC

MUJOCO_GL=egl XLA_PYTHON_CLIENT_PREALLOCATE=false python train_offline_pixels.py --env_name=cheetah-run-v0 --config=configs/offline_pixels_config.py:bc

PixelIQL

MUJOCO_GL=egl XLA_PYTHON_CLIENT_PREALLOCATE=false python train_offline_pixels.py --env_name=cheetah-run-v0 --config=configs/offline_pixels_config.py:iql