TODO: Project Intro
Install Anaconda, f1tenth_gym, and project dependencies. Check gym 0.21 issue for further details. Configuration is subject to change.
conda create -n ese-650-fp python=3.8 pip=21 setuptools=65.5.0 # create a new conda env under the root dir
conda activate ese-650-fp
pip install -e . # install setup.py for f1tenth_gym
pip install psutil packaging # install ignored ipykernel dependencies
pip install -r requirements.txt # install other dependencies
# conda deactivate # exit ese-650-fp env
# conda remove -n ese-650-fp --all # remove ese-650-fp envInstall pytorch and Cuda. Please check official installation for your platforms.
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118 # for Derek's PCTest your environment.
python test.py # check if cuda is available
python main.py # check if f1tenth_gym is workingTo run cleanrl for studying, config a new env for the cloned repo. Please check the repo for detailed instruction.
conda create -n clean-rl python=3.8
conda activate clean-rl
pip install -r requirements/requirements.txt # for cartpole only
python cleanrl/ppo.py --seed 1 --env-id CartPole-v0 --total-timesteps 50000 --capture_video # cd ./videos for visualized results
tensorboard --logdir runs # open another terminal to see the training processTrain your model using
python ppo_continuous.py --total-timesteps 50000 # for simplicity
tensorboard --logdir runs # enable visualization in another terminalTest your model using
python inference.py