论文标题
动态模型大小的统一级联编码器ASR模型
A Unified Cascaded Encoder ASR Model for Dynamic Model Sizes
论文作者
论文摘要
在本文中，我们提出了一个动态的级联编码器自动语音识别（ASR）模型，该模型统一了不同部署方案的模型。此外，该模型可以显着降低模型大小和功耗而不会损失质量。也就是说，使用动态级联编码器模型，我们探索了三种技术，以最大程度地提高每个模型大小的性能：1）在共享编码器时为每个子模型使用单独的解码器； 2）使用漏斗 - 提高编码器效率； 3）平衡因果关系的大小，以提高质量并适合部署限制。总体而言，与级联的编码器模型相比，提出的大中等模型的尺寸较小30％，并将功耗降低了33％。统一大型，中和小型模型的三重大小模型可实现37％的总尺寸减少，而质量损失最小，同时大大减少了拥有单独模型的工程工作。
In this paper, we propose a dynamic cascaded encoder Automatic Speech Recognition (ASR) model, which unifies models for different deployment scenarios. Moreover, the model can significantly reduce model size and power consumption without loss of quality. Namely, with the dynamic cascaded encoder model, we explore three techniques to maximally boost the performance of each model size: 1) Use separate decoders for each sub-model while sharing the encoders; 2) Use funnel-pooling to improve the encoder efficiency; 3) Balance the size of causal and non-causal encoders to improve quality and fit deployment constraints. Overall, the proposed large-medium model has 30% smaller size and reduces power consumption by 33%, compared to the baseline cascaded encoder model. The triple-size model that unifies the large, medium, and small models achieves 37% total size reduction with minimal quality loss, while substantially reducing the engineering efforts of having separate models.
