FastAST: Accelerating Audio Spectrogram Transformer via Token Merging and Cross-Model Knowledge Distillation
This work addresses the problem of real-time, resource-efficient audio analysis for applications in audio classification, though it is incremental as it builds on existing AST methods.
The paper tackles the challenge of optimizing efficiency in Audio Spectrogram Transformer (AST) models for audio classification without compromising accuracy, by introducing FastAST, which integrates Token Merging (ToMe) and Cross-Model Knowledge Distillation (CMKD) to increase throughput with minimal accuracy impact and even improve accuracy compared to AST while maintaining faster inference speeds.
Audio classification models, particularly the Audio Spectrogram Transformer (AST), play a crucial role in efficient audio analysis. However, optimizing their efficiency without compromising accuracy remains a challenge. In this paper, we introduce FastAST, a framework that integrates Token Merging (ToMe) into the AST framework. FastAST enhances inference speed without requiring extensive retraining by merging similar tokens in audio spectrograms. Furthermore, during training, FastAST brings about significant speed improvements. The experiments indicate that FastAST can increase audio classification throughput with minimal impact on accuracy. To mitigate the accuracy impact, we integrate Cross-Model Knowledge Distillation (CMKD) into the FastAST framework. Integrating ToMe and CMKD into AST results in improved accuracy compared to AST while maintaining faster inference speeds. FastAST represents a step towards real-time, resource-efficient audio analysis.