R2 See, Do - a journey in deep reinforcement learning.
r2seedo
uses hatch
for project management. You'll
need it installed (ideally in an isolated environment) before setting up r2seedo
.
r2seedo
depends on gymnasium
, which
is best installed with conda / mamba, thus hatch-conda
is also recommended.
# Update mamba
mamba update -n base mamba
# Update base environment packages
mamba update -n base --all
# Install hatch & hatchling
mamba install -n base -c conda-forge hatch hatchling
# Install hatch-conda
mamba activate base && python -m pip install hatch-conda
git clone git@github.com:libertininick/r2seedo.git
hatch
will install r2seedo
in development mode along with its development dependencies
inside of a virtual environment managed by hatch
.
# Navigate to root project directory
cd r2seedo
# (optional) if using conda / mamba envs activate the base environment
mamba activate base
# Create default environment
hatch env create
- RL environment packages are best installed in independent virtual environment
- Each RL environment used in
r2seedo
has its own virtual environment configuration defined in pyproject.toml - An environment can be created using
hatch env create <env name>
:
NOTE: Make sure the base environment is activated before creating a new environment: mamba activate base
Environment | Name | Create |
---|---|---|
Atari | gym-atari | hatch env create gym-atari |
Box2D | gym-box2d | hatch env create gym-box2d |
Toy Text | gym-toy_text | hatch env create gym-toy_text |
Run tests and coverage report using hatch
hatch run default:test-cov