CLI¤
The jaxnasium command bootstraps a new project and can copy algorithm or training
templates into an existing one. It is provided by
create-rl-app; Jaxnasium registers the console script and the files that add can copy.
The easiest way to run it without installing anything first is uv:
uvx jaxnasium init <projectname>
# or, equivalently:
uvx jaxnasium <projectname>
init¤
Creates a new project directory: a uv package, a train.py, and (optionally) an
environment template and a local copy of an algorithm.
jaxnasium init my-project
jaxnasium init my-project -y
jaxnasium init my-project --algorithm sac --no-env-template --environment Pendulum-v1
| Flag | Default | Meaning |
|---|---|---|
-y / --yes |
off | Accept every default without prompting |
--env-template / --no-env-template |
prompt (yes) |
Copy a custom Environment template into the package |
--algorithm-source / --no-algorithm-source |
prompt (no) |
Copy the algorithm source into the project instead of importing it |
--algorithm {ppo,sac,dqn,pqn} |
ppo |
Algorithm wired into train.py |
--environment <id> |
CartPole-v1 |
Environment id for jym.make(...) when no env template is included |
add¤
Copies a registered file or bundle into the current directory (or into a path you pass). Use this inside an existing project.
jaxnasium add ppo
jaxnasium add sweep
jaxnasium add mlp architectures/mlp.py
| Name | What it copies |
|---|---|
starter |
train.py and example_env.py |
ppo / sac / dqn / pqn |
The algorithm file, plus agent_networks.py if it is not already there |
sweep |
sweep.py, a multi-environment / multi-algorithm search |
bronet / cnn / mlp |
The matching architecture module |
The registry lives in src/jaxnasium/create-rl-app.toml if you want to see or extend what add can copy.