How fine can fine-tuning be? Learning efficient language models

Abstract

State-of-the-art performance on language understanding tasks is now achieved with increasingly large networks; the current record holder has billions of parameters. Given a language model pre-trained on massive unlabeled text corpora, only very light supervised fine-tuning is needed to learn a task: the number of fine-tuning steps is typically five orders of magnitude lower than the total parameter count. Does this mean that fine-tuning only introduces small differences from the pre-trained model in the parameter space? If so, can one avoid storing and computing an entire model for each task? In this work, we address these questions by using Bidirectional Encoder Representations from Transformers (BERT) as an example. As expected, we find that the fine-tuned models are close in parameter space to the pre-trained one, with the closeness varying from layer to layer. We show that it suffices to fine-tune only the most critical layers. Further, we find that there are surprisingly many good solutions in the set of sparsified versions of the pre-trained model. As a result, fine-tuning of huge language models can be achieved by simply setting a certain number of entries in certain layers of the pre-trained parameters to zero, saving both task-specific parameter storage and computational cost.

11 pages, 11 figures and 6 tables; accepted to the 23rd International Conference on Artificial Intelligence and Statistics (AISTATS) 2020

References in corpus (13)

• Deep Residual Learning for Image Recognition
• Gaussian Error Linear Units (GELUs)
• BinaryConnect: Training Deep Neural Networks with binary weights during propagations
• The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks
• Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
• To prune, or not to prune: exploring the efficacy of pruning for model compression
• GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding
• Deep Learning Scaling is Predictable, Empirically
• Learning Sparse Neural Networks through $L_0$ Regularization
• Parameter-Efficient Transfer Learning for NLP
• Are All Layers Created Equal?
• Deconstructing Lottery Tickets: Zeros, Signs, and the Supermask
• Superposition of many models into one

Cited by in corpus (1)

• Semi-Siamese Bi-encoder Neural Ranking Model Using Lightweight Fine-Tuning