submitit

What it is and what it does

Submitit is a Python 3.8+ library that simplifies job submission to Slurm clusters by wrapping the submission process and providing a concurrent.futures-compatible executor interface. You define a Python function, submit it via an executor, and retrieve results and logs programmatically—all from within Python rather than writing shell scripts. It handles job tracking, stdout/stderr capture, and can requeue preempted jobs.

The package is designed for researchers and engineers who want to launch compute tasks on HPC clusters without leaving Python. It trades off some of the distributed scheduling sophistication of other frameworks for lower-level control: each submitted task becomes an individual cluster job with its own logs and lifecycle, making it suitable for medium-to-large tasks rather than many small parallel operations.

Use it for:

• Submit long-running Python functions to a Slurm cluster and poll for results without manual job tracking
• Prototype distributed code locally using concurrent.futures, then swap to submitit executor for cluster execution
• Capture and inspect individual job logs and stderr for debugging failed cluster tasks
• Implement checkpoint-and-requeue workflows for preemptible cluster environments
• Run hyperparameter searches or batch experiments where each trial is a separate cluster job

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Submitit wraps Slurm job submission and monitoring, letting you submit Python functions to a cluster and retrieve results, logs, and job status from within Python code.

Yes, if you regularly submit jobs to Slurm clusters from Python code. The low install friction, permissive license, stable maturity, and zero known vulnerabilities make it a safe choice. The aging maintenance status is a minor concern for a stable library but worth monitoring if you need active feature development.

Install

submitit on PyPI

pip install submitit

uv add submitit

poetry add submitit

Installing submitit

Before you install

Low install friction with only 2 runtime dependencies (cloudpickle and typing_extensions). Maintenance status is aging—last release was 240 days ago—but the package is stable.

License in practice

MIT license is permissive; you can use submitit in proprietary projects without restriction, though you must include the license notice.

Quickstart

pip install submitit

import submitit

def add(a, b):
    return a + b

executor = submitit.AutoExecutor(folder="log_test")
executor.update_parameters(timeout_min=1, slurm_partition="dev")
job = executor.submit(add, 5, 7)
output = job.result()

Slurm cluster must be accessible; requires Python 3.8 or later

Verify before relying

• Whether Slurm must be installed on the submission machine or only on cluster nodes
• Compatibility with non-Slurm cluster schedulers beyond the plugin system mentioned
• Performance characteristics for large numbers of concurrent job submissions

Package facts

Evidence: submitit-1.5.4-py3-none-any.whl

Tags

More Distributed Computing packages