search

PaddlePaddle / PaddleCustomDevice

PaddlePaddle custom device implementaion. (『飞桨』自定义硬件接入实现)
Apache License 2.0
68 stars 142 forks source link

[MLU] optimize range kernel; flash_attn kernel #1190

Closed ShawnNew closed 4 months ago

ShawnNew commented 4 months ago
  1. make range kernel a function, which uses cnnlArange_v2
  2. use range function in flash_attn
paddle-bot[bot] commented 4 months ago

Thanks for your contribution!