arXiv:2210.04243cs.LG
Fine-Tuning Pre-trained Transformers into Decaying Fast Weights
Early discovery. Linked by one identifiable commenter. This paper is outside the main ranking; a mention is not an endorsement.
Abstract
Autoregressive Transformers are strong language models but incur O(T) complexity during per-token generation due to the self-attention mechanism. Recent work proposes kernel-based methods to approximate causal self-attention by replacing it with recurrent formulations with various update rules and feature maps to achieve O(1) time and memory complexity. We explore these approaches and find that they are unnecessarily complex, and propose a simple alternative - decaying fast weights - that runs fast on GPU, outperforms prior methods, and retains 99% of attention's performance for GPT-2. We also show competitive performance on WikiText-103 against more complex attention substitutes.
while this topic has visibility, there is also another relatively unexplored research direction, fine-tuning pretrained transformers into RNNs https://arxiv.org/abs/2210.04243 https://aclanthology.org/2021.emnlp-main.830.pdf
lucidrainsView on HNParent context