An approach to optimize inference of the DIART speaker diarization pipeline

TL;DR

This paper focuses on reducing inference latency in the DIART online speaker diarization system by applying various optimization techniques to its embedding model, balancing speed and accuracy.

Contribution

It evaluates the effectiveness of knowledge distillation, pruning, quantization, and layer fusion in optimizing inference latency for DIART's embedding model.

Findings

Knowledge distillation reduces latency but decreases accuracy.

Quantization and layer fusion improve latency without accuracy loss.

Pruning does not significantly affect latency.

Abstract

Speaker diarization answers the question "who spoke when" for an audio file. In some diarization scenarios, low latency is required for transcription. Speaker diarization with low latency is referred to as online speaker diarization. The DIART pipeline is an online speaker diarization system. It consists of a segmentation and an embedding model. The embedding model has the largest share of the overall latency. The aim of this paper is to optimize the inference latency of the DIART pipeline. Different inference optimization methods such as knowledge distilation, pruning, quantization and layer fusion are applied to the embedding model of the pipeline. It turns out that knowledge distillation optimizes the latency, but has a negative effect on the accuracy. Quantization and layer fusion also have a positive influence on the latency without worsening the accuracy. Pruning, on the other hand, does not improve latency.