Darts-Conformer: Towards Efficient Gradient-Based Neural Architecture Search For End-to-End ASR
arXiv:2104.02868
Abstract
Neural architecture search (NAS) has been successfully applied to tasks like image classification and language modeling for finding efficient high-performance network architectures. In ASR field especially end-to-end ASR, the related research is still in its infancy. In this work, we focus on applying NAS on the most popular manually designed model: Conformer, and then propose an efficient ASR model searching method that benefits from the natural advantage of differentiable architecture search (Darts) in reducing computational overheads. We fuse Darts mutator and Conformer blocks to form a complete search space, within which a modified architecture called Darts-Conformer cell is found automatically. The entire searching process on AISHELL-1 dataset costs only 0.7 GPU days. Replacing the Conformer encoder by stacking searched cell, we get an end-to-end ASR model (named as Darts-Conformner) that outperforms the Conformer baseline by 4.7\% on the open-source AISHELL-1 dataset. Besides, we verify the transferability of the architecture searched on a small dataset to a larger 2k-hour dataset. To the best of our knowledge, this is the first successful attempt to apply gradient-based architecture search in the attention-based encoder-decoder ASR model.
arXiv commentSubmitted to ASRU 2021
References in corpus (12)
- Neural Architecture Search with Reinforcement Learning
- Neural Architecture Search: A Survey
- Large-Scale Evolution of Image Classifiers
- Conformer: Convolution-augmented Transformer for Speech Recognition
- SNAS: Stochastic Neural Architecture Search
- DARTS+: Improved Differentiable Architecture Search with Early Stopping
- Transformers with convolutional context for ASR
- Understanding and Improving Transformer From a Multi-Particle Dynamic System Point of View
- Streaming Chunk-Aware Multihead Attention for Online End-to-End Speech Recognition
- AutoKWS: Keyword Spotting with Differentiable Architecture Search
- Efficient Neural Architecture Search for End-to-end Speech Recognition via Straight-Through Gradients
- Learned Transferable Architectures Can Surpass Hand-Designed Architectures for Large Scale Speech Recognition