ptq

Installation and usage

This skill should be used when the user asks to "quantize a model", "run PTQ", "post-training quantization", "NVFP4 quantization", "FP8 quantization", "INT8 quantization", "INT4 AWQ", "quantize LLM", "quantize MoE", "quantize VLM", or needs to produce a quantized HuggingFace or TensorRT-LLM checkpoint from a pretrained model using ModelOpt.

skills use ptq