Towards Fine-tuning Pre-trained Language Models with Integer Forward and Backward Propagation
This work addresses the problem of reducing computational costs for fine-tuning pre-trained models, which is incremental as it extends integer methods to backward propagation.
The authors tackled the computational and energy inefficiency of fine-tuning large language models like BERT by using integer arithmetic for both forward and backward propagation, achieving performance matching floating-point baselines with 16-bit integers and a 3.1-point average drop with 8-bit integers on benchmarks.
The large number of parameters of some prominent language models, such as BERT, makes their fine-tuning on downstream tasks computationally intensive and energy hungry. Previously researchers were focused on lower bit-width integer data types for the forward propagation of language models to save memory and computation. As for the backward propagation, however, only 16-bit floating-point data type has been used for the fine-tuning of BERT. In this work, we use integer arithmetic for both forward and back propagation in the fine-tuning of BERT. We study the effects of varying the integer bit-width on the model's metric performance. Our integer fine-tuning uses integer arithmetic to perform forward propagation and gradient computation of linear, layer-norm, and embedding layers of BERT. We fine-tune BERT using our integer training method on SQuAD v1.1 and SQuAD v2., and GLUE benchmark. We demonstrate that metric performance of fine-tuning 16-bit integer BERT matches both 16-bit and 32-bit floating-point baselines. Furthermore, using the faster and more memory efficient 8-bit integer data type, integer fine-tuning of BERT loses an average of 3.1 points compared to the FP32 baseline.