hf-jobs

Run cloud compute workloads on Hugging Face by executing arbitrary code via UV scripts or custom Docker images. The infrastructure manages hardware selection, timeout configurations, and automated result persistence to the Hub.

Yes, you can run GPU jobs using custom Docker images on Hugging Face. The platform supports deploying containers on CPUs, GPUs, or TPUs for scalable cloud compute and accelerated machine learning tasks.

Handle authentication for batch inference jobs by securely configuring Hugging Face tokens via secrets. This ensures safe Hub interactions and allows automated saving of outputs back to your repository.

Schedule automated machine learning tasks by configuring cloud workloads with specific timeout limits and hardware selection. This approach addresses needs for scalable compute without requiring local setup.

Hugging Face Jobs supports processing large datasets using distributed computing on managed cloud infrastructure. You can execute Python scripts or Docker containers to handle complex data processing and batch jobs.

Use Hugging Face cloud compute for model training to access GPU clusters and TPUs without local hardware setup. It manages authentication, timeouts, and automatically persists training results to the Hub.