Why Low-Precision Transformer Training Fails: An Analysis on Flash Attention

ICLR 2026 Conference Submission16395 Authors

Keywords: low-precision training, transformer, attention

TL;DR: For the first time, we mechanistically explain why low-precision training with flash attention fails, identifying a vicious cycle of rounding errors and proposing a simple, effective fix.

Abstract: The pursuit of computational efficiency has driven the adoption of low-precision formats for training transformer models. However, this progress is often hindered by notorious training instabilities. This paper provides the first mechanistic explanation for a long-standing and unresolved failure case where training with flash attention in low-precision settings leads to catastrophic loss explosions. Our in-depth analysis reveals that the failure is not a random artifact but caused by two intertwined phenomena: the emergence of similar low-rank representations within the attention mechanism and the compounding effect of biased rounding errors inherent in low-precision arithmetic. We demonstrate how these factors create a vicious cycle of error accumulation that corrupts weight updates, ultimately derailing the training dynamics. To validate our findings, we introduce a minimal modification to the flash attention that mitigates the bias in rounding errors. This simple change stabilizes the training process, confirming our analysis and offering a practical solution to this persistent problem. Code is available at https://anonymous.4open.science/r/why-low-precision-training-fails.

Primary Area: other topics in machine learning (i.e., none of the above)

Submission Number: 16395