---
title: "Migrating Your GitHub CI to Hugging Face Jobs"
description: "Trackio's GitHub Actions CI was limited by generic hosted machines, maintenance-related delays, and lack of readily available GPU runners, while the project needed CPU checks and CUDA-based tests. The setup keeps GitHub Actions as the CI coordinator and uses a dispatcher Space plus the huggingface/jobs-actions bridge to launch ephemeral GitHub runners inside Hugging Face Jobs, selected through labels such as hf-jobs-cpu-upgrade and hf-jobs-t4-small. The guide covers duplicating and configuring the dispatcher, creating a GitHub App, supplying runner credentials and an HF token, changing runs-on, choosing Docker images, and validating runs and logs; Playwright and NVIDIA CUDA images are used for CPU and GPU examples. In Trackio, the CPU job took 1m10s versus a 1m40s GitHub ubuntu-latest baseline, while the GPU check passed in 45s and cost less than a cent at the t4-small rate; logs were available through both systems."
---

# Migrating Your GitHub CI to Hugging Face Jobs

[Hugging Face](https://yomu.fyi/company/hugging-face) · Abubakar Abid · Jun 9, 2026

**Type:** Tutorial

## Summary

Trackio's GitHub Actions CI was limited by generic hosted machines, maintenance-related delays, and lack of readily available GPU runners, while the project needed CPU checks and CUDA-based tests. The setup keeps GitHub Actions as the CI coordinator and uses a dispatcher Space plus the huggingface/jobs-actions bridge to launch ephemeral GitHub runners inside Hugging Face Jobs, selected through labels such as hf-jobs-cpu-upgrade and hf-jobs-t4-small. The guide covers duplicating and configuring the dispatcher, creating a GitHub App, supplying runner credentials and an HF token, changing runs-on, choosing Docker images, and validating runs and logs; Playwright and NVIDIA CUDA images are used for CPU and GPU examples. In Trackio, the CPU job took 1m10s versus a 1m40s GitHub ubuntu-latest baseline, while the GPU check passed in 45s and cost less than a cent at the t4-small rate; logs were available through both systems.

## Context

Trackio wanted reliable CPU CI for basic unit tests and frontend checks, plus GPU CI for tests requiring actual CUDA hardware. The article also says GitHub-hosted runners use generic machines, can be slow or unavailable during maintenance, and do not make GPU access readily available to most open-source projects.

## Approach / What changed

A dispatcher Docker Space receives GitHub workflow\_job.queued webhooks, verifies them, checks for hf-jobs-\* labels, mints a short-lived GitHub runner registration token, and starts a Hugging Face Job on matching hardware. The Job boots an ephemeral GitHub Actions runner, which executes the workflow and reports status back to GitHub. Setup involves creating the dispatcher, installing a GitHub App, adding credentials and tokens, changing runs-on labels, and selecting suitable Docker images.

## Takeaways

- Changing a workflow from runs-on: ubuntu-latest to labels such as hf-jobs-cpu-upgrade or hf-jobs-t4-small routes the job to Hugging Face Jobs through the dispatcher.
- Trackio's CPU job using the Microsoft Playwright image ran in 1m10s, compared with 1m40s for the GitHub ubuntu-latest baseline, an improvement of about 30%.
- Trackio's GPU check ran on a t4-small Hugging Face Job, passed in 45 seconds, and cost less than a cent for that duration; Hugging Face Job logs can also be fetched with the CLI.

**Tags:** [CI/CD](https://yomu.fyi/topic/ci-cd), [Docker](https://yomu.fyi/topic/docker), [Machine Learning](https://yomu.fyi/topic/machine-learning), [Performance](https://yomu.fyi/topic/performance)

- Source: [Hugging Face](https://huggingface.co/blog/github-ci-hf-jobs)
- Source URL: https://huggingface.co/blog/github-ci-hf-jobs
- Ingested by Yomu: 2026-08-29T01:50:11.502Z

[Read original post](https://huggingface.co/blog/github-ci-hf-jobs)