sentence-transformers 5.6.1 fixes Flash Attention embedding bug in XLM-R models

tomaarsen · x · 2026-07-23

The sentence-transformers library has released version 5.6.1, available via pip install sentence-transformers==5.6.1.

This update fixes an issue affecting XLM-R based models (such as bge-m3 and multilingual-e5) and other RoBERTa-family checkpoints, but only when Flash Attention was explicitly requested. The default padded execution path and all-mpnet-base-v2 were never affected.

Related event: Sentence Transformers 5.6.1 Fixes Flash Attention Embedding Regression(3 posts)→

Original post →

More from coding & agent

coding & agent channel →