This repository contains some early experiments combining vectorized environments and value-based methods. See this Google Doc
Prerequisites:
- Python 3.8+
- Poetry
Install dependencies:
poetry install
Train agents:
poetry run python vdqn.py
Train agents with experiment tracking:
poetry run python vdqn.py --track --capture-video
Install dependencies:
poetry install -E atari
Train agents:
poetry run python vdqn_atari.py
poetry run python vc51_atari.py --track --capture-video
Train agents with experiment tracking:
poetry run python vdqn_atari.py
poetry run python vc51_atari.py --track --capture-video
Install dependencies:
poetry install -E pybullet
Train agents:
poetry run python vddpg_continuous_action.py
poetry run python vtd3_continuous_action.py
Train agents with experiment tracking:
poetry run python vddpg_continuous_action.py --track --capture-video
poetry run python vtd3_continuous_action.py --track --capture-video