Paper ID: 2303.09184
Block-wise Bit-Compression of Transformer-based Models
Gaochen Dong, Wei Chen
With the popularity of the recent Transformer-based models represented by BERT, GPT-3 and ChatGPT, there has been state-of-the-art performance in a range of natural language processing tasks. However, the massive computations, huge memory footprint, and thus high latency of Transformer-based models is an inevitable challenge for the cloud with high real-time requirement. To tackle the issue, we propose BBCT, a method of block-wise bit-compression for transformer without retraining. Our method achieves more fine-grained compression of the whole transformer, including embedding, matrix multiplication, GELU, softmax, layer normalization, and all the intermediate results. As a case, we compress an efficient BERT with the method of BBCT. Our benchmark test results on General Language Understanding Evaluation (GLUE) show that BBCT can achieve less than 1% accuracy drop in most tasks.
Submitted: Mar 16, 2023