DefTruth/CUDA-Learn-Notes

Issues

• Include SageAttention Kernel

  #147 opened 6 days ago by jason-huang03
  3

• 🌤🌤 CONTRIBUTE 🎉🎉

  #50 opened 6 days ago by DefTruth
  1

• 你好，关于online safe softmax的速度，貌似并没有明显提升

  #153 opened a month ago by lzcchl
  2

• 关于softmax中的实现的理解，求大佬解惑

  #151 opened a month ago by lzcchl
  2

• 您好，请教一个关于代码中reduce相关的问题

  #6 opened 2 months ago by Ss-shuang123
  3

• __threadfence() 作用

  #7 opened 2 months ago by zbt78
  3

• layer norm实现

  #2 opened 2 months ago by zbt78
  3

• 您好，请问sigmoid算子这里为啥没有考虑指数溢出问题

  #4 opened 3 months ago by Phoenix8215
  2

• resources

  #3 opened 8 months ago by DefTruth
  2