Hugging Face and SkyPilot announced an integration allowing users to mount Hugging Face Buckets and Hub repositories as storage backends directly in SkyPilot jobs across multiple cloud providers. The setup removes egress costs for reading models and datasets, letting compute run wherever GPU capacity is available while pulling from the same centralized storage. It relies on Xet’s content-defined chunking to deduplicate checkpoints and model variants, with benchmarks showing roughly 170 MB/s checkpoint write speeds across AWS, GCP, and Lambda infrastructure.
