thu-ml / SageAttention

Quantized Attention that achieves speedups of 2.1x and 2.7x compared to FlashAttention2 and xformers, respectively, without lossing end-to-end metrics across various models.
BSD 3-Clause "New" or "Revised" License
355 stars 14 forks source link

Notation error in Equation (2) #18

Closed Coco58323 closed 1 week ago

Coco58323 commented 1 week ago
image

The subscripts and superscripts are reversed

jt-zhang commented 1 week ago

Thank you for your suggestion. We will correct this error.