Thanks for sending an requirement! Please fill in the following template to help quickly solve your problem.
Flash Attention implementations, copying exponents from vector to Cube
A5
Inputs: BF16, FP32, uint8_t, hiF8, float8_e8m0
Thanks for sending an requirement! Please fill in the following template to help quickly solve your problem.
应用场景
Flash Attention implementations, copying exponents from vector to Cube
期望支持的硬件
A5
期望支持的数据类型、Shape、布局等
Inputs: BF16, FP32, uint8_t, hiF8, float8_e8m0