# replicate/keepsake

Version control for machine learning

Repository: https://github.com/replicate/keepsake
Canonical: https://ross.abutalabs.com/products/keepsake
Homepage: https://keepsake.ai
Language: Python
License: Apache-2.0
License Family: permissive
Topics: version-control, machine-learning
Last push: 2025-02-25T23:52:42+00:00

## Health v2 (maintenance only)
Score: 26/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 8, release rhythm 8, longevity 100
- inputs: {"age_days": 2254, "days_push": 554, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1681, forks 76 (observed 2026-08-28T04:05:21.644950+00:00)

## What it is
Keepsake is a Python library and CLI for version control of machine learning experiments, tracking code, hyperparameters, weights, and metrics to your own Amazon S3 or Google Cloud Storage buckets. It lets you list, filter, diff, and restore experiments and model checkpoints from the command line or notebooks.

## Use cases
- track machine learning experiments with code, params, and metrics
- version model weights in my own S3 bucket
- compare and diff training runs
- restore checkpoints from past experiments
- analyze experiment results in a notebook
- replace a spreadsheet of ML experiment results

## When to choose
- you want experiment tracking with data stored on your own cloud storage with no server
- you need to version model weights alongside code and hyperparameters
- you work in Python with PyTorch, TensorFlow, or scikit-learn and want minimal integration code

## When to avoid
- you need actively maintained software with ongoing support
- you want a hosted experiment tracking dashboard with a web UI
- you don't use Amazon S3 or Google Cloud Storage

## Facets
- artifact type: library
- maturity: abandoned
- function: version-control, machine-learning, cli, logging
- domain: machine-learning, developer-tools, version-control, data-science
- platform: python, cli, cross-platform
- tags: experiment-tracking, model-versioning, mlops, s3, google-cloud-storage, checkpoints

## Member repositories
- replicate/keepsake (main) score 26

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:21.644950+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:41:26.547996+00:00, confidence not recorded.
  - readme: https://github.com/replicate/keepsake (fetched 2026-08-28T04:05:21.644950+00:00, sha 3ff216c756bd)
- Data as of 2026-08-30T08:39:29.467469+00:00.