Improving Stability of Fine-Tuning Pretrained Language Models via Component-Wise Gradient Norm Clipping
This addresses a practical problem for users of large language models by providing a more stable fine-tuning process, though it is an incremental improvement over existing methods.
The paper tackles the instability in fine-tuning pretrained language models by proposing a component-wise gradient norm clipping method, which improves generalization performance, convergence speed, and training stability.
Fine-tuning over large pretrained language models (PLMs) has established many state-of-the-art results. Despite its superior performance, such fine-tuning can be unstable, resulting in significant variance in performance and potential risks for practical applications. Previous works have attributed such instability to the catastrophic forgetting problem in the top layers of PLMs, which indicates iteratively that fine-tuning layers in a top-down manner is a promising solution. In this paper, we first point out that this method does not always work out due to the different convergence speeds of different layers/modules. Inspired by this observation, we propose a simple component-wise gradient norm clipping method to adjust the convergence speed for different components. Experiment results demonstrate that our method achieves consistent improvements in terms of generalization performance, convergence speed, and training stability. The codebase can be found at https://github.com/yangalan123/FineTuningStability.