GKD: A General Knowledge Distillation Framework for Large-scale Pre-trained Language Model

TL;DR

GKD is a versatile framework enabling efficient knowledge distillation of extremely large-scale pre-trained language models, supporting multiple methods and facilitating deployment on memory-limited hardware.

Contribution

The paper introduces GKD, a flexible framework that allows large-scale PLM distillation with various methods on limited GPU memory, supporting models over 100B parameters.

Findings

Supports distillation of 100B-scale PLMs

Enables use of 25 distillation methods

Operates on 8 NVIDIA A100 GPUs

Abstract

Currently, the reduction in the parameter scale of large-scale pre-trained language models (PLMs) through knowledge distillation has greatly facilitated their widespread deployment on various devices. However, the deployment of knowledge distillation systems faces great challenges in real-world industrial-strength applications, which require the use of complex distillation methods on even larger-scale PLMs (over 10B), limited by memory on GPUs and the switching of methods. To overcome these challenges, we propose GKD, a general knowledge distillation framework that supports distillation on larger-scale PLMs using various distillation methods. With GKD, developers can build larger distillation models on memory-limited GPUs and easily switch and combine different distillation methods within a single framework. Experimental results show that GKD can support the distillation of at least 100B-scale PLMs and 25 mainstream methods on 8 NVIDIA A100 (40GB) GPUs.