hqq-quantization

Compresses language models with HQQ without calibration data, checks generation quality, and saves the quantized model.

When to use it

Use for calibration-free model quantization, fast compression, HuggingFace or vLLM deployment, quantized fine-tuning, or very low-bit experiments.

Give it a model-compression request; it provides code to quantize a model, test generation, and save the model and tokenizer.