TensorRT-LLMs/cpp/kernels/fmha_v2/test_sm90_configs.sh
qsang-nv 0fd59d64ab
infra: open source fmha v2 kernels (#4185)
* add fmha repo

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* fix format

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* fix code style

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* fix header

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* fix header kernel_traits.h

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* add .gitignore file

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* add SLIDING_WINDOW_ATTENTION

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* fix style

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* fix format

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* update setup.py

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

* update build_wheel.py

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>

---------

Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com>
Signed-off-by: qsang-nv <200703406+qsang-nv@users.noreply.github.com>
2025-05-15 10:56:34 +08:00

46 lines
2.3 KiB
Bash
Executable File

#!/bin/sh
set -e
set -x
#bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 64 -b 3 -v 2 -h 2
#bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 64 -b 3 -v 2 -h 2 -scale-bmm1 0.25
#bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 64 -b 3 -v 2 -h 2 -use-tma
#bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 64 -b 3 -v 2 -h 2 -scale-bmm1 0.25 -use-tma
####################################################################################################
# H G M M A F P 1 6 . F P 1 6
bin/fmha.exe -v 0 -runs 1 -s 512 -d 64 -min-s 1 -b 1 -h 1
bin/fmha.exe -v 0 -runs 1 -s 384 -d 64 -min-s 1 -b 1 -h 1
bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 1 -b 1 -h 1
bin/fmha.exe -v 0 -runs 1 -s 128 -d 64 -min-s 1 -b 1 -h 1
bin/fmha.exe -v 0 -runs 1 -s 512 -d 64 -min-s 1 -b 1 -h 1 -ignore-b1opt
bin/fmha.exe -v 0 -runs 1 -s 384 -d 64 -min-s 1 -b 1 -h 1 -ignore-b1opt
bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 1 -b 1 -h 1 -ignore-b1opt
bin/fmha.exe -v 0 -runs 1 -s 128 -d 64 -min-s 1 -b 1 -h 1 -ignore-b1opt
####################################################################################################
# Q G M M A E 4 M 3 . F P 3 2
bin/fmha.exe -v 0 -runs 1 -s 512 -d 64 -min-s 1 -b 1 -h 1 -e4m3
bin/fmha.exe -v 0 -runs 1 -s 384 -d 64 -min-s 1 -b 1 -h 1 -e4m3
bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 1 -b 1 -h 1 -e4m3
bin/fmha.exe -v 0 -runs 1 -s 128 -d 64 -min-s 1 -b 1 -h 1 -e4m3
bin/fmha.exe -v 0 -runs 1 -s 512 -d 64 -min-s 1 -b 1 -h 1 -e4m3 -ignore-b1opt
bin/fmha.exe -v 0 -runs 1 -s 384 -d 64 -min-s 1 -b 1 -h 1 -e4m3 -ignore-b1opt
bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 1 -b 1 -h 1 -e4m3 -ignore-b1opt
bin/fmha.exe -v 0 -runs 1 -s 128 -d 64 -min-s 1 -b 1 -h 1 -e4m3 -ignore-b1opt
####################################################################################################
# I G M M A I N T 8 . I N T 3 2
bin/fmha.exe -v 0 -runs 1 -s 512 -d 64 -min-s 1 -b 1 -h 1 -int8
bin/fmha.exe -v 0 -runs 1 -s 384 -d 64 -min-s 1 -b 1 -h 1 -int8
bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 1 -b 1 -h 1 -int8
bin/fmha.exe -v 0 -runs 1 -s 128 -d 64 -min-s 1 -b 1 -h 1 -int8
bin/fmha.exe -v 0 -runs 1 -s 512 -d 64 -min-s 1 -b 1 -h 1 -int8 -ignore-b1opt
bin/fmha.exe -v 0 -runs 1 -s 384 -d 64 -min-s 1 -b 1 -h 1 -int8 -ignore-b1opt
bin/fmha.exe -v 0 -runs 1 -s 256 -d 64 -min-s 1 -b 1 -h 1 -int8 -ignore-b1opt
bin/fmha.exe -v 0 -runs 1 -s 128 -d 64 -min-s 1 -b 1 -h 1 -int8 -ignore-b1opt