Edge-ASR: Towards Low-Bit Quantization of Automatic Speech Recognition Models

TL;DR

This paper benchmarks advanced post-training quantization methods on edge ASR models, revealing that 3-bit quantization can maintain high accuracy, thus enabling efficient deployment on resource-limited devices.

Contribution

It provides a comprehensive evaluation of state-of-the-art PTQ methods on leading edge-ASR models, offering insights into quantization trade-offs for edge deployment.

Findings

3-bit quantization can preserve accuracy with advanced PTQ techniques

Quantization impacts model accuracy, memory, and computational efficiency

Benchmark results guide optimal quantization configurations for edge ASR

Abstract

Recent advances in Automatic Speech Recognition (ASR) have demonstrated remarkable accuracy and robustness in diverse audio applications, such as live transcription and voice command processing. However, deploying these models on resource-constrained edge devices (e.g., IoT device, wearables) still presents substantial challenges due to strict limits on memory, compute and power. Quantization, particularly Post-Training Quantization (PTQ), offers an effective way to reduce model size and inference cost without retraining. Despite its importance, the performance implications of various advanced quantization methods and bit-width configurations on ASR models remain unclear. In this work, we present a comprehensive benchmark of eight state-of-the-art (SOTA) PTQ methods applied to two leading edge-ASR model families, Whisper and Moonshine. We systematically evaluate model performances (i.e., accuracy, memory I/O and bit operations) across seven diverse datasets from the open ASR leader-board, analyzing the impact of quantization and various configurations on both weights and activations. Built on an extension of the LLM compression toolkit, our framework integrates edge-ASR models, diverse advanced quantization algorithms, a unified calibration and evaluation data pipeline, with detailed analysis tools. Our results characterize the trade-offs between efficiency and accuracy, demonstrating that even $3$-bit quantization can succeed on high capacity models when using advanced PTQ techniques. These findings provide valuable insights for optimizing ASR models on low-power, always-on edge devices.