Transformer Quality in Linear Time
This addresses the computational bottleneck for researchers and practitioners working with long-context language models, though it is incremental in improving existing architectures.
The paper tackles the inefficiency of Transformers in handling long sequences by proposing FLASH, a model that matches the perplexity of improved Transformers while achieving training speedups of up to 12.1× on certain datasets.
We revisit the design choices in Transformers, and propose methods to address their weaknesses in handling long sequences. First, we propose a simple layer named gated attention unit, which allows the use of a weaker single-head attention with minimal quality loss. We then propose a linear approximation method complementary to this new layer, which is accelerator-friendly and highly competitive in quality. The resulting model, named FLASH, matches the perplexity of improved Transformers over both short (512) and long (8K) context lengths, achieving training speedups of up to 4.9$\times$ on Wiki-40B and 12.1$\times$ on PG-19 for auto-regressive language modeling, and 4.8$\times$ on C4 for masked language modeling.