Method compresses diffusion language models to run faster and cheaper

Original: STaR-Quant: State-Time Consistent Post-Training Quantization for Diffusion Large Language Models

Source: arxiv.org ↗