aind-behavior-vr-foraging-curricula

Aind.Behavior.VrForaging.Curricula

A repository of curricula for VR foraging task.

How to run a specific curriculum?

Curricula are modules of the main package: aind_behavior_vr_foraging_curricula.<curriculum_name>.

All curricula are available via the curriculum cli entry point. Available list of commands is printed with the -h flag:

uv run curriculum -h

List available curricula

Available curricula can be listed from the cli using:

uv run curriculum list

Running a curriculum

Curricula can be run using the run subcommand of the curriculum cli entry point.

uv run curriculum run -h

The following arguments are available for the run subcommand:

• --data-directory: Path to the session data directory that will be used to calculate metrics (required)
• --input-trainer-state: Path to a deserialized json file with the current trainer state (required)
• --mute-suggestion: Disables the suggestion output (optional)
• --output-suggestion: A path to save the serialized suggestion (optional)
• --curriculum: The name of the curriculum to run (optional)

For a quick "demo" to ensure everything is working, you can run:

uv run curriculum run --data-directory "demo" --input-trainer-state "foo.json" --curriculum "template"

For real-world applications, you may want to omit the "--curriculum" flag and let the system automatically detect the curriculum from the trainer state.

Style guide

To keep things clear, I suggest the following naming convention:

• Policies should start with p_ (e.g., p_identity_policy)
• Policy transitions should start with pt_
• Stages should start with s_ (e.g., s_stage1)
• Stage transitions should start with st_ and should be named after the stages they transition between (e.g., st_s_stage1_s_stage2)

Define the following modules:

• metrics: Defines (or imports) metrics classes and how to calculate them from data
• stages: Defines the different stages of the VR foraging task. This includes task settings and, optionally, policies
• curriculum: Defines the transitions between the stages and generate entry point to the application

Contributors

Contributions to this repository are welcome! However, please ensure that your code adheres to the recommended DevOps practices below:

Linting

We use ruff as our primary linting tool.

Testing

Attempt to add tests when new features are added. To run the currently available tests, run uv run pytest from the root of the repository.

Lock files

We use uv to manage our lock files and therefore encourage everyone to use uv as a package manager as well.