---
pretty_name: hf_clean_cache snapshot
tags:
- model-cache
- reproducibility
- archive
---

# hf_clean_cache snapshot

This repository archives a local Hugging Face model cache for reproducibility. It is a cache snapshot, not a newly created training or evaluation dataset.

## Contents

The files named hf_clean_cache.tar.part-000, hf_clean_cache.tar.part-001, and so on are consecutive parts of one uncompressed tar archive. The archive contains the top-level hf_clean_cache/ directory, including hidden cache metadata, lock files, original model files, and internal symbolic links.

To reconstruct, verify, and restore the directory after downloading all parts and SHA256SUMS into one directory:

~~~bash
cat hf_clean_cache.tar.part-* > hf_clean_cache.tar
sha256sum -c SHA256SUMS
tar -xf hf_clean_cache.tar -C /path/to/parent
~~~

Archive size: 44,991,477,760 bytes in 12 parts. SHA-256 of the reconstructed archive: ad52af04a8471fa9bfa1b8794bfa415739026d768f96841dd343e78b450bcb4d. SHA256SUMS contains the checksum of each part and the reconstructed archive.

## Sources and terms

The cache includes files from these model repositories or locally named downloads:

- [Efficient-Large-Model/Sana_1600M_512px_diffusers](https://huggingface.co/Efficient-Large-Model/Sana_1600M_512px_diffusers)
- [google/flan-t5-xl](https://huggingface.co/google/flan-t5-xl)
- [google/siglip-so400m-patch14-384](https://huggingface.co/google/siglip-so400m-patch14-384)
- openai_clip_vit_base_patch32, openai_clip_vit_large_patch14
- laion_clip_vit_l_14, google_siglip_base_patch16_224

This snapshot does not grant a new license for the included models. Consult the original model repositories and their terms before reuse.