Hugging Face Xet Cache

Why clean Hugging Face Xet Cache?

• Old Xet chunk data from earlier repository revisions keeps consuming disk space even after you no longer need those versions, which shows up as a large ~/.cache/huggingface/xet folder
• A transfer cache filled with rarely reused pack fragments reduces the benefit of deduplication, so new downloads still take time while local storage keeps growing
• Stale transfer metadata can lead to repeated chunk verification or re-fetch attempts, which users notice as downloads that appear to stall or restart parts of a fetch
• When the cache lives on a nearly full system drive, Xet-backed pulls may fail mid-transfer with no-space errors while model and dataset storage elsewhere looks fine
• Shared workstations and CI runners often accumulate leftover Xet chunks from many repositories, causing unexpectedly high disk usage between jobs and slower cleanup of build agents
• Removing transfer caches older than seven days clears temporary chunk state that is no longer helping current workloads, trading a slower next download for recovered disk space
• The cache directory can become fragmented after many incremental writes and deletions, so filesystem scans and backup jobs take longer because they must traverse thousands of small cached objects

Common questions about Hugging Face Xet Cache

Download Kudu and reclaim your disk space.

Available on Windows, macOS, and Linux. No account required, no feature gates, no telemetry without consent. All cleaning targets are open source and community-auditable.