chenhongmin.will 8b939854d8 enable scale

2025-02-28 20:07:32 +08:00

..

cutlass @ afa1772203

Initial commit

2025-02-24 09:20:23 +08:00

flash_api.cpp

change to use per_tensor

2025-02-26 10:21:09 +08:00

flash_fwd_mla_bf16_sm90.cu

update gmem

2025-02-25 09:45:19 +08:00

flash_fwd_mla_fp8_sm90.cu

update gmem

2025-02-25 09:45:19 +08:00

flash_fwd_mla_kernel.h

enable scale

2025-02-28 20:07:32 +08:00

flash_mla_utils.cu

fix compile

2025-02-25 21:52:11 +08:00

flash_mla.h

update fp8 api

2025-02-26 08:33:25 +08:00

fp8_transpose_v.h

use fa'3 transv

2025-02-28 14:54:44 +08:00

named_barrier.h

add transv barrier

2025-02-26 17:57:00 +08:00

softmax.h

enable scale

2025-02-28 20:07:32 +08:00

static_switch.h

Initial commit

2025-02-24 09:20:23 +08:00

utils.h

use mm1's Aregs instead of mma0's Cregs

2025-02-27 11:59:17 +08:00