Managing dependencies with runtime_env

Ray runtime_env installs pip packages and ships a working directory to the cluster at run time. You add a dependency without rebuilding a container image, which keeps interactive development fast.

Inject dependencies interactively

Pass runtime_env to ray.init(). Ray installs the packages and uploads the working directory to the cluster before your code runs.

import ray

ray.init(runtime_env={
    "pip": ["pandas==2.2.2", "scikit-learn"],
    "working_dir": "./src",
})

Inject dependencies for a submitted job

For a job you submit from the command line, pass the same environment with --working-dir and --runtime-env-json.

ray job submit \
    --address sagemaker_ray://my-cluster/my-namespace \
    --working-dir ./src \
    --runtime-env-json '{"pip": ["pandas==2.2.2", "scikit-learn"]}' \
    -- python my-script.py

For the full set of runtime_env fields, including conda environments and environment variables, see Ray runtime environments in the Ray documentation.