Skip to content
Kitbay

huggingface-trackio

Apache-2.0Updated on Nov 11, 202611,334GitHub stars

A Hugging Face skill for Trackio, its experiment-tracking library.

What the skill teaches

A Hugging Face skill for Trackio, its experiment-tracking library. It covers three jobs: logging metrics from training code with the Python API (including TRL's report_to option), raising alerts for training problems with optional webhooks, and pulling logged metrics and alerts back out through the CLI as JSON for analysis. Runs can sync to a Hugging Face Space to give a live dashboard that survives after a cloud instance shuts down; the skill warns that automatically created Spaces are public unless marked private. Each job has its own reference file.

Capabilities

  • Log training metrics with the Trackio Python API
  • Fire alerts for training issues
  • Retrieve logged metrics and alerts as JSON from the CLI
  • Sync runs to a Space dashboard