This website requires JavaScript.
Explore
Help
Sign In
metax-maca
/
FlashMLA
mirror of
https://github.com/MetaX-MACA/FlashMLA.git
Watch
0
Star
0
Fork
You've already forked FlashMLA
1
Code
Issues
Packages
Projects
Releases
Wiki
Activity
d67000b06d
FlashMLA
/
csrc
/
flash_api
History
zhan3916
30f7561775
Performance: Add new tile for small bs
...
Signed-off-by: zhan3916 <zhan3916@gmail.com>
2025-03-25 11:54:33 +08:00
..
flash_api.cpp
Performance: Add new tile for small bs
2025-03-25 11:54:33 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m16n16_bf16_True_True_sm80.cu
Performance: Add new tile for small bs
2025-03-25 11:54:33 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m16n16_bf16_True_True_split_sm80.cu
Performance: Add new tile for small bs
2025-03-25 11:54:33 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m16n16_fp16_True_True_sm80.cu
Performance: Add new tile for small bs
2025-03-25 11:54:33 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m16n16_fp16_True_True_split_sm80.cu
Performance: Add new tile for small bs
2025-03-25 11:54:33 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m32n16_bf16_True_True_sm80.cu
Performance: improve perf by new tile
2025-03-14 19:00:47 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m32n16_bf16_True_True_split_sm80.cu
Performance: improve perf by new tile
2025-03-14 19:00:47 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m32n16_fp16_True_True_sm80.cu
Performance: improve perf by new tile
2025-03-14 19:00:47 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m32n16_fp16_True_True_split_sm80.cu
Performance: improve perf by new tile
2025-03-14 19:00:47 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m64n16_bf16_True_True_sm80.cu
Performance: improve perf by new tile
2025-03-14 19:00:47 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m64n16_bf16_True_True_split_sm80.cu
Performance: improve perf by new tile
2025-03-14 19:00:47 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m64n16_fp16_True_True_sm80.cu
Performance: improve perf by new tile
2025-03-14 19:00:47 +08:00
flash_fwd_splitkv_hdimqk576_hdimv512_m64n16_fp16_True_True_split_sm80.cu
Performance: improve perf by new tile
2025-03-14 19:00:47 +08:00
flash_mla.h
Support splitKV>1
2025-03-06 13:57:41 +08:00