This is a collection of deep reinforcement methods built to solve partially-observable environments.
git clone [email protected]:mweiss17/jumping_quadrupeds.gitcd jumping_quadrupedspip install -e .
There are several methods implemented in the jumping_quadrupeds, including PPO, DRQ-v2, SPR, and an ETH world-model method. In order to use each, you need to specify the environment, the agent, and the training parameters.
python3 scripts/rl-tests/01-train.py experiments/ppo-car --inherit templates/base --macro templates/agents/ppo.yml --config.use_wandb True