Contributing to EvoRL

Install Developer Version

# Step 1: Create a virtual env from conda, venv, pyenv, ...:
# For example, using pyenv-virtualenv:
pyenv virtualenv 3.10 evorl-dev
pyenv activate evorl-dev

# Step 2: Install JAX in advance
# Note: the installation command depends on your platform.
# See https://docs.jax.dev/en/latest/installation.html
pip install -U "jax[cuda12]"

# Step 3: Install the developer version
pip install -e ".[dev]"

# Step 4: post-setup
# This repo is configured with `pre-commit`.
pre-commit install

Debugging

We provide some examples in .vscode/launch.json for debugging in Visual Studio Code.

Documentation

Documentations are written under the docs/ directory. We use Sphinx to construct the documentation. Since we use the MyST-Parser extension, Markdown documents (*.md) all allowed. API References are automatically generated by autodoc2.

If you want to build the documentations in local machine, run the script ./docs/build.sh. Then, the documentation Web files are generated in docs/_build.

Code Style

In general, we follow the Black code style for Python code. For Python documentation, we follow the Google pydoc format.

The following code snippets demonstrate the code style in EvoRL:

class Workflow(AbstractWorkflow):
    """The base class for all Workflows.

    All workflow classes are inherit from this class, and customize by implementing
    """

    def __init__(self, config: DictConfig):
        """Initialize a RLWorkflow instance.

        Args:
            config: the config object.
        """
        self.config = config
        self.recorder = ChainRecorder([])
        self.checkpoint_manager = setup_checkpoint_manager(config)

class ReplayBufferState(PyTreeData):
    """Contains data related to a replay buffer.

    Attributes:
        data: the stored replay buffer data.
        current_index: the pointer used for adding data.
        buffer_size: the current size of the replay buffer.
    """

    data: chex.ArrayTree
    current_index: chex.Array = jnp.zeros((), jnp.int32)
    buffer_size: chex.Array = jnp.zeros((), jnp.int32)

Developers do not need to worry about whether the submitted code satisfy the above code style. Before submitting any commit, the pre-commit will raise warning or errors to remind you what is wrong. Besides, developers can manually execute this process by running

pre-commit run -a