{"id":13740979,"url":"https://github.com/sinhrks/daskperiment","last_synced_at":"2026-02-19T08:48:50.981Z","repository":{"id":138188291,"uuid":"167454153","full_name":"sinhrks/daskperiment","owner":"sinhrks","description":"Reproducibility for Humans: A lightweight tool to perform reproducible machine learning experiment.","archived":false,"fork":false,"pushed_at":"2019-04-24T05:36:47.000Z","size":2329,"stargazers_count":24,"open_issues_count":17,"forks_count":5,"subscribers_count":3,"default_branch":"master","last_synced_at":"2024-11-15T10:44:12.738Z","etag":null,"topics":["dask","machine-learning","reproducibility"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"bsd-3-clause","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/sinhrks.png","metadata":{"files":{"readme":"README.rst","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null}},"created_at":"2019-01-24T23:34:59.000Z","updated_at":"2023-08-28T14:31:27.000Z","dependencies_parsed_at":"2024-01-14T23:56:58.788Z","dependency_job_id":"0b89d382-1f95-4c33-bcb9-6ca0e1a2ba8c","html_url":"https://github.com/sinhrks/daskperiment","commit_stats":null,"previous_names":[],"tags_count":6,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sinhrks%2Fdaskperiment","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sinhrks%2Fdaskperiment/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sinhrks%2Fdaskperiment/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sinhrks%2Fdaskperiment/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/sinhrks","download_url":"https://codeload.github.com/sinhrks/daskperiment/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":253145163,"owners_count":21861198,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["dask","machine-learning","reproducibility"],"created_at":"2024-08-03T04:00:54.160Z","updated_at":"2026-02-19T08:48:50.948Z","avatar_url":"https://github.com/sinhrks.png","language":"Python","funding_links":[],"categories":["Packages","Python"],"sub_categories":[],"readme":"daskperiment\n============\n\n.. image:: https://img.shields.io/pypi/v/daskperiment.svg\n  :target: https://pypi.python.org/pypi/daskperiment/\n.. image:: https://readthedocs.org/projects/daskperiment/badge/?version=latest\n  :target: http://daskperiment.readthedocs.org/en/latest/\n  :alt: Latest Docs\n.. image:: https://travis-ci.org/sinhrks/daskperiment.svg?branch=master\n  :target: https://travis-ci.org/sinhrks/daskperiment\n.. image:: https://codecov.io/gh/sinhrks/daskperiment/branch/master/graph/badge.svg\n  :target: https://codecov.io/gh/sinhrks/daskperiment\n\nOverview\n~~~~~~~~\n\n`daskperiment` is a tool to perform reproducible machine learning experiment.\nIt allows users to define and manage the history of trials\n(given parameters, results and execution environment).\n\nThe package is built on `Dask`, a package for parallel computing with task\nscheduling. Each experiment trial is internally expressed as `Dask` computation\ngraph, and can be executed in parallel.\n\nBenefits\n~~~~~~~~\n\n- Compatibility with standard Python/Jupyter environment (and optionally with standard KVS).\n\n  - No need to set up server applications\n  - No need to registrate on any cloud services\n  - Run on standard / customized Python shells\n\n- Intuitive user interface\n\n  - Few modifications on existing codes are needed\n  - Trial histories are logged automatically (no need to write additional codes for logging)\n  - `Dask` compatible API\n  - Easily accessible experiments history (with `pandas` basic operations)\n  - Less managiment works on Git (no need to make branch per trials)\n  - (Experimental) Web dashboard to manage trial history\n\n- Traceability of experiment related information\n\n  - Trial result and its (hyper) parameters.\n  - Code contexts\n  - Environment information\n\n    - Device information\n    - OS information\n    - Python version\n    - Installed Python packages and its version\n    - Git information\n\n- Reproducibility\n\n  - Check function purity (each step should return the same output for the same inputs)\n  - Automatic random seeding\n\n- Auto saving and loading of previous experiment history\n- Parallel execution of experiment steps\n- Experiment sharing\n\n  - Redis backend\n  - MongoDB backend\n\nFuture Scope\n~~~~~~~~~~~~\n\n- More efficient execution.\n\n  - Omit execution if depending parameters are the same\n  - Distributed execution\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsinhrks%2Fdaskperiment","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsinhrks%2Fdaskperiment","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsinhrks%2Fdaskperiment/lists"}