OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Keepsake

Keepsake is an open-source Python library for versioning machine-learning experiment files and metadata in Amazon S3 or Google Cloud Storage, with retrieval through a command-line interface or notebook.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Keepsake — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Keepsake when a training workflow needs checkpoints tied to experiment files and metadata, and the team can use Amazon S3 or Google Cloud Storage for storage. Consider its maintenance status when evaluating it for a new workflow.

Who it is for

  • Machine-learning developers who need to associate checkpoints with the code, parameters, training data, dependencies and metrics from a run.
  • Teams that want experiment artifacts stored in their own S3 or Google Cloud Storage bucket.

Documented uses

  • Record experiment parameters and checkpoint metrics alongside code, data, weights and Python dependencies.
  • Retrieve saved code and model weights from a checkpoint to revisit or reproduce a result.
  • List and filter experiments, or compare checkpoints by parameters, dependency versions and metrics.
  • Retrieve experiment results in a notebook for analysis and plotting.
Before you choose

Requirements & limits

  • Requires an Amazon S3 or Google Cloud Storage bucket for the documented storage workflow.
  • The README documents both a CLI workflow and notebook retrieval and analysis.
  • The repository identifies the project as not actively maintained.
  • The repository is public and licensed Apache-2.0.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Add Keepsake tracking to a training script

Record experiment parameters and checkpointed model weights and metrics with Keepsake.

Before you start

  • Python training code using Keepsake; the example also imports PyTorch.
  • An Amazon S3 or Google Cloud Storage bucket for Keepsake files and metadata.

Steps

  1. Install Keepsake with the command below.

    Documented command
    pip install -U keepsake

  2. Import keepsake in the training script and initialize an experiment with keepsake.init(path=".", params={...}).

  3. During training, save the model to a file, then call experiment.checkpoint(path="model.pth", metrics={...}).

Expected outcome

Keepsake tracks experiment code, hyperparameters, training data, weights, metrics, and Python dependencies, storing files and metadata in the configured cloud storage.

Guide limitations

  • The project is not actively maintained.
  • The README does not specify the bucket setup or configuration steps.
Documentation 1 ↗
Source-based practical guide

List experiments from the command line

Filter tracked experiments by a metric using the Keepsake CLI.

Before you start

  • Keepsake CLI and experiment data stored in Amazon S3.

Steps

  1. Run keepsake ls --filter "val_loss<0.2".

  2. Review the listed experiment, host, status, and best-checkpoint fields.

Expected outcome

The CLI displays experiments matching the filter, with their host, status, and best checkpoint.

Guide limitations

  • The example demonstrates filtering by val_loss; other filter syntax is not described.
  • The project is not actively maintained.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
replicate/keepsake · Official homepage title · retained in its source language
How we review sourcesSuggest a correction