I needed a model that does reliable matching between English and the 3 Japanese writting system, so I calibrated the compressed one with English/Japanese datasets as well as my private dataset. I included a script in extras, private dataset is not included but generation code is there for reference. Attention layers are left uncompressed for precision and this makes a noticable difference.
Submit in one command
Run it next to your model folder. It makes the torrent, checks your files against Hugging Face, and submits it. You just start seeding and paste your key from your account. It only reads your files and never changes them. Read the script first if you like.
curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo catplusplus/Qwen3-Embedding-4B-NVFP4 ./model-folder