Lightweight, lazy experiment tracking

lazyscribe is a lightweight package for model experiment logging. It creates a single JSON file per project, and an experiment is only added to the log when code finishes.

lazyscribe also has functionality for merging projects, paving the way for multiple people to work on a single project.

Installation

To install lazyscribe from PyPI, run

$ python -m pip install lazyscribe

This is the preferred method to install lazyscribe.

Contents

Quickstart

• Tutorials

How-to...

• Create a basic project
• Tag experiments
• Filter experiments within your project
• Log non-global metrics
• Represent the project as a table
• Save and examine experiment artifacts
  • Persistence
  • Loading and validation
• Write a custom artifact handler
  • Building the handler
  • Using the handler
• Version artifacts
  • Using a repository as a standalone structure
  • Promote artifacts from experiments to the repository
• Specify cross-project dependencies
• Log a project to an external file system
• Open existing projects
  • Read-only mode
  • Append mode
  • Editable mode
• Merge project files
  • TL;DR
  • Appending
  • Updating
• Use lazyscribe with Prefect

Reference

• Contributing
  • Types of contributions
  • Get started
• GitHub repository
• lazyscribe
  • lazyscribe package