Fast Forwarding Low-Rank Training
This work addresses efficiency for researchers and practitioners finetuning large language models, representing an incremental improvement over existing parameter-efficient methods like LoRA.
The paper tackles the problem of high computational costs in finetuning pretrained language models by proposing Fast Forward, an optimization strategy that alternates between regular steps and stages repeating the most recent step until loss stops improving, achieving up to an 87% reduction in FLOPs and 81% reduction in train time without compromising performance.
Parameter efficient finetuning methods like low-rank adaptation (LoRA) aim to reduce the computational costs of finetuning pretrained Language Models (LMs). Enabled by these low-rank settings, we propose an even more efficient optimization strategy: Fast Forward, a simple and effective approach to accelerate large segments of training. In a Fast Forward stage, we repeat the most recent optimizer step until the loss stops improving on a tiny validation set. By alternating between regular optimization steps and Fast Forward stages, Fast Forward provides up to an 87\% reduction in FLOPs and up to an 81\% reduction in train time over standard SGD with Adam. We validate Fast Forward by finetuning various models on different tasks and demonstrate that it speeds up training without compromising model performance. Additionally, we analyze when and how to apply Fast Forward.