YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

mlxtorepo

Convert any Hugging Face model to MLX format on your Mac and publish it as a public repo under your own HF account β€” one script, one command.

./convert_and_upload.sh                 # interactive (prompts: model repo + quant)
./convert_and_upload.sh --model Qwen/Qwen3-8B --quant q8
./convert_and_upload.sh --dry-run       # preview the full plan, touch nothing

Produces e.g. jesusoctavioas/<ModelName>-mlx-q4 on the Hub.

What it does (v6)

  1. Creates + activates a venv (~/mlxtorepo/.venv)
  2. Installs huggingface_hub[cli], hf-transfer, mlx-lm
  3. Authenticates (reuses ~/.cache/huggingface/token, or prompts inside the script)
  4. Checks disk space (230 GB fresh / 100 GB when reusing a download)
  5. Downloads the model (hf-transfer, resumable, reuses existing)
  6. Converts to MLX (-q --q-bits N, or fp16) β€” quant 2–8
  7. Writes a README card (base_model, tags, usage snippet)
  8. Uploads to your namespace (overwrites if the repo exists)
  9. Shows where files live + interactive cleanup menu (keep-all by default)

Files

File What
convert_and_upload.sh the pipeline (v1–v6 changelog in its header)
AGENTS.md start here if you're an agent β€” commands, rules, gotchas
PROMPTS.md 11 reusable prompt templates from real sessions
SETUP.md setup, dependencies, API keys, disk budget
RETROSPECTIVE.md lessons learned, issues + fixes, decisions
CONVERSATION.md full decision history of the project
GUIDE.md human step-by-step guide

Requirements

  • macOS (Apple Silicon), python3, ~100–230 GB free disk
  • Hugging Face account + write token (the script prompts for it if missing)
  • mlx-lm supports quant q2–q8 + fp16; model archs up to what mlx-lm supports (e.g. qwen4_exp is not supported by mlx-lm yet β€” see RETROSPECTIVE.md)

Stable release

Tag v6-stable β€” current. Check the banner the script prints on every run: convert_and_upload.sh β€” version: 6

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support