A Compression-Compilation Framework for On-mobile Real-time BERT Applications
This work addresses the challenge of real-time NLP applications on resource-constrained mobile devices, representing an incremental improvement in optimization techniques for mobile deployment.
The paper tackles the problem of deploying BERT models on mobile devices by proposing a compression-compilation co-design framework that balances accuracy and latency, achieving up to 7.8x speedup compared to TensorFlow-Lite with minimal accuracy loss and enabling real-time applications like Question Answering and Text Generation with latencies as low as 45ms.
Transformer-based deep learning models have increasingly demonstrated high accuracy on many natural language processing (NLP) tasks. In this paper, we propose a compression-compilation co-design framework that can guarantee the identified model to meet both resource and real-time specifications of mobile devices. Our framework applies a compiler-aware neural architecture optimization method (CANAO), which can generate the optimal compressed model that balances both accuracy and latency. We are able to achieve up to 7.8x speedup compared with TensorFlow-Lite with only minor accuracy loss. We present two types of BERT applications on mobile devices: Question Answering (QA) and Text Generation. Both can be executed in real-time with latency as low as 45ms. Videos for demonstrating the framework can be found on https://www.youtube.com/watch?v=_WIRvK_2PZI