arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2604.00529cs.LGcs.CL
收藏

MF-QAT:多格式量化感知训练用于弹性推理

MF-QAT: Multi-Format Quantization-Aware Training for Elastic Inference

d-Matrix

Zifei Xu, Sayeh Sharify, Hesham Mostafa

AI总结:

本文提出多格式量化感知训练,使模型在不同量化格式下均表现优异,通过Slice-and-Scale方法实现运行时无损转换,实现弹性精度调整。

摘要:

量化感知训练(QAT)通常针对单一目标数值格式,但实际部署中需根据硬件支持或运行时约束选择数值精度。本文研究多格式QAT,训练模型在多种量化格式下保持鲁棒性。发现多格式QAT可匹配单格式QAT在每个目标精度上的表现,产生一个在不同格式中均表现良好的模型,即使未在训练中见过这些格式。为实现实际部署,我们提出Slice-and-Scale转换过程,将高精度表示转换为低精度格式而无需重新训练。基于此,我们引入一个流程:(i)训练模型使用多格式QAT;(ii)存储单一锚定格式检查点(MXINT8/MXFP8);(iii)在运行时允许即时转换为低MXINT或MXFP格式,而不会产生显著或无额外精度损失。这些组件共同提供了一条实现弹性精度调整的实用路径,允许在不同部署目标上选择运行时格式。

英文摘要:

Quantization-aware training (QAT) is typically performed for a single target numeric format, while practical deployments often need to choose numerical precision at inference time based on hardware support or runtime constraints. We study multi-format QAT, where a single model is trained to be robust across multiple quantization formats. We find that multi-format QAT can match single-format QAT at each target precision, yielding one model that performs well overall across different formats, even formats that were not seen during training. To enable practical deployment, we propose the Slice-and-Scale conversion procedure for both MXINT and MXFP that converts a high-precision representation into lower-precision formats without re-training. Building on this, we introduce a pipeline that (i) trains a model with multi-format QAT, (ii) stores a single anchor format checkpoint (MXINT8/MXFP8), and (iii) allows on-the-fly conversion to lower MXINT or MXFP formats at runtime with negligible-or no-additional accuracy degradation. Together, these components provide a practical path to elastic precision scaling and allow selecting the runtime format at inference time across diverse deployment targets.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑