Johannes Gäßler
e495d1e748
* CUDA: fix Gemma E4B MTP FlashAttention * remove unused template declaration
12 lines
398 B
Plaintext
12 lines
398 B
Plaintext
// This file has been autogenerated by generate_cu_files.py, do not edit manually.
|
|
|
|
#include "../fattn-mma-f16.cuh"
|
|
|
|
DECL_FATTN_MMA_F16_CASE(64, 64, 4, 2);
|
|
DECL_FATTN_MMA_F16_CASE(80, 80, 4, 2);
|
|
DECL_FATTN_MMA_F16_CASE(96, 96, 4, 2);
|
|
DECL_FATTN_MMA_F16_CASE(112, 112, 4, 2);
|
|
DECL_FATTN_MMA_F16_CASE(128, 128, 4, 2);
|
|
DECL_FATTN_MMA_F16_CASE(256, 256, 4, 2);
|
|
DECL_FATTN_MMA_F16_CASE(512, 512, 4, 2);
|