hf-mount

hf-mount is a tool that allows users to mount Hugging Face Buckets and repositories as local filesystems, facilitating on-demand access to models and datasets without the need for downloading. Utilizing either NFS or FUSE backends, it fetches files lazily, making it ideal for resource-limited environments and operations where disk space is a concern, while supporting standard filesystem commands. Its primary use case focuses on read-heavy machine learning workloads, enabling seamless interaction with large models and datasets while avoiding network congestion from unnecessary data transfers.