Quantization reduces the memory and bandwidth demands of large language model (LLM) inference but can degrade accuracy; offloading higher-precision data can …
机构:Cornell
来源:arXiv 2611.08363 | AI4Papers 论文推荐平台