---
id: dora-search
version: "0.1.12"
license: MIT
license_treatment: permissive
maintenance: dormant
---

# dora-search — Easy grid searches for ML.

License: permissive · Maintenance: dormant · Downloads: 177.9K/mo

## What it is and what it does

Dora is a command-line experiment manager designed for machine learning workflows. It lets you define grid searches in pure Python, automatically schedule and deduplicate experiments based on their argument signatures, and monitor progress from the terminal. The tool integrates with argparse, Hydra, and PyTorch Lightning, and can launch experiments locally or remotely via Slurm/Submitit.

You decorate your training script's main function with `@argparse_main` or `@hydra_main`, then create grid files that call a launcher repeatedly with different hyperparameter sets. Dora assigns each experiment a signature for deduplication, manages checkpointing and resumption, and provides a monitoring interface to track metrics and compare runs. It's built for researchers running many related experiments and needing to avoid redundant computation.

Use it for:

- Running systematic hyperparameter sweeps across a cluster without manually scheduling each job
- Automatically resuming interrupted experiments from checkpoints using experiment signatures
- Monitoring multiple concurrent ML training runs and comparing metrics in a terminal table
- Deduplicating experiments so requesting the same hyperparameters twice merges into one run
- Integrating experiment tracking into existing PyTorch Lightning or Hydra-based projects

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Dora is an experiment launching and management tool for machine learning projects that handles grid search scheduling, automatic deduplication of runs, and terminal-based monitoring of experiments.

Yes, but with caution. Dora is well-suited for researchers running grid searches on clusters and needing experiment deduplication and monitoring. However, the package is dormant (last release 2023-05-23, no recent commits), so expect no active maintenance or support for newer dependencies. Install only if your workflow matches its design and you can tolerate potential incompatibilities with recent library versions.

## Install

pip install dora-search
uv add dora-search
poetry add dora-search

## Installing dora-search

Before you install: Installation requires building from source with high friction. The package is dormant—last release was 2023-05-23 with no commits since 2023-10-05—so expect limited maintenance and no active bug fixes or updates.

License in practice: MIT license is permissive, allowing commercial and private use with minimal restrictions, making it suitable for most projects without legal concern.

Quickstart:

pip install dora-search

from dora import argparse_main, get_xp
import argparse

parser = argparse.ArgumentParser()
parser.add_argument('--lr', type=float)

@argparse_main(dir='./outputs', parser=parser)
def main():
    xp = get_xp()
    print(f'Running with config: {xp.cfg}')
    xp.link.push_metrics({'loss': 0.5})

High install friction due to source build requirement; requires Python >=3.7.0

Verify before relying:

- Whether submitit dependency (required for remote job launching) is automatically installed or must be manually configured
- Current compatibility with modern Hydra versions beyond 1.1 mentioned in changelog
- Whether PyTorch Lightning integration still works with recent versions of that library

## Package facts

- License: MIT (permissive)
- Python support: supports_current
- Install friction: high
- Maintenance: dormant
- Downloads: 177.9K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

experiment management ml, grid search scheduling, hyperparameter sweep tool, ml experiment tracking, job deduplication, experiment launcher, ml workflow automation, experiment-tracking, hyperparameter-optimization, distributed-computing

[View on SkillFed](https://skillfed.io/packages/dora-search) · [View on PyPI](https://pypi.org/project/dora-search/)