Gradio

HF Token (WRITE)

I don't have a token

If no token is provided, the environment variable option and direct download safeguard will be used.

Source HF Repo (e.g. author/model)

Source Filename (e.g. model.safetensors)

Target HF Repo (e.g. author/model)

Target Filename Base (e.g. model-quant)

Format

fp8 int8 rowwise int8-convrot mxfp8

Model Layer Filter

Full precision matrix multiplication

Exclude Layers Regex (Optional)

ℹ️ For more advanced quantization modes, install and use convert-to-quant locally.

Status Log

🤗 Model Quantizer