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

TL;DR

This paper introduces a multi-format quantization-aware training method that enables a single model to perform well across various numerical precisions and supports elastic inference with on-the-fly format conversion.

Contribution

It proposes a novel multi-format QAT approach and a practical conversion pipeline for elastic inference, allowing dynamic precision scaling during deployment.

Findings

Multi-format QAT matches single-format performance at each precision.

The Slice-and-Scale conversion enables format changes without re-training.

The pipeline supports runtime format selection with minimal accuracy loss.

Abstract

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.