开发环境配置

conda create -n GAN python=3.7.12

activate GAN

pip install paddlepaddle==1.8.5 parl==1.4 gym==0.18.0 atari-py==0.2.6 rlschool==0.3.1 protobuf==3.19.6 metagym==0.1.1 matplotlib==3.3.4 numpy==1.19.5 scikit-learn==0.24.2 make_env==0.0.7 pyglet==1.5.0 mock==3.0.5 -i https://mirrors.aliyun.com/pypi/simple

pip install paddlepaddle-gpu==1.8.5.post107 parl==1.4 gym==0.18.0 atari-py==0.2.6 rlschool==0.3.1 protobuf==3.19.6 metagym==0.1.1 matplotlib==3.3.4 numpy==1.19.5 scikit-learn==0.24.2 make_env==0.0.7 pyglet==1.5.0 mock==3.0.5 -i https://mirrors.aliyun.com/pypi/simple

conda create -n test python=3.9
activate test

python -m pip install paddlepaddle-gpu==2.5.2.post120 -f https://www.paddlepaddle.org.cn/whl/windows/mkl/avx/stable.html
pip install numpy==1.23.5 parl==2.2.1 gym==0.26.2 pygame -i https://mirrors.aliyun.com/pypi/simple

from rlschool import make_env

env = make_env('LiftSim')
observation = env.reset()
action = [2, 0, 4, 0, 7, 0, 10, 0]
for i in range(100):
    env.render()    # use render to show animation
    next_obs, reward, done, info = env.step(action)

xparl start --port 8010
python xxx/LiftSim_baseline/A2C/train.py