MiniCPM icon indicating copy to clipboard operation
MiniCPM copied to clipboard
verified publisher icon OpenBMB

Due to Flashattention, inference cannot be performed on v100

Open WenXIN-AI opened this issue 1 year ago • 1 comments

Description / 描述

FlashAttention only supports Ampere GPUs or newer.

Case Explaination / 案例解释

Due to Flashattention, inference cannot be performed on v100

WenXIN-AI avatar Sep 28 '24 15:09 WenXIN-AI

done with caseid 592863

saurabh12453 avatar Oct 16 '24 09:10 saurabh12453

Hi, you can use eager mode for inference.

zh-zheng avatar Jun 07 '25 06:06 zh-zheng