Datasets:

---

Naphula-Archives

/

master_python_list_mergekit_windows

like 0

Follow

Naphula-Archives 13

huggingface_hub is very unstable and requires specific versions for different tasks:

• version 0.34.4: The most stable version for uploading/downloading files. Also works for merging and quanting older models such as Mistral 24B.
• version 1.10.2: Unstable, DO NOT USE. Supposed to be faster at ups/downs but hangs randomly, requiring manual intervention.
• version 1.13.0: Required for merging and quantizing Gemma 4 models.
• version 1.6.0: Unstable, DO NOT USE. Supposed to be faster at ups/downs but hangs randomly, requiring manual intervention.

I tried to use v1.13.0 to upload files and it hung on 99% for several hours. Switched back to v0.34.4 and it works fine.

For transformers, you must use version 4.56.1 for Mistral 24B merging/quants, and version 5.8.0.dev0 for Gemma 4.