ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-06-29 11:16:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

smoke test update

Browse Source
This commit is contained in:
danyao12
2025-02-05 17:22:52 +08:00
parent 8a8dc7f6c3
commit f88ba67e11
2 changed files with 5 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
example/ck_tile/01_fmha/script/smoke_test_bwd_v3.sh
View File

@@ -10,16 +10,18 @@ COMMON_ARGS='-v=1'
set -x
for prec in "fp16" "bf16" ; do
for perm in 0 1 ; do
for hdim in 128 ; do
for hdim in 64 72 96 128 ; do
for v3_atomic_fp32 in 0 1 ; do
for v3_bf16_cvt in 0 1 2 ; do
for mask in 0 1 ; do
$EXE -prec=$prec -b=4 -h=2 -d=$hdim -s=512 -iperm=$perm -operm=$perm -mask=$mask -bwd_v3=1 -v3_atomic_fp32=$v3_atomic_fp32 -mode=0 -kname=$KNAME $COMMON_ARGS
$EXE -prec=$prec -b=1 -h=3 -d=$hdim -s=768 -iperm=$perm -operm=$perm -mask=$mask -bwd_v3=1 -v3_atomic_fp32=$v3_atomic_fp32 -mode=0 -kname=$KNAME $COMMON_ARGS
$EXE -prec=$prec -b=2 -h=4 -h_k=2 -d=$hdim -s=512 -iperm=$perm -operm=$perm -mask=$mask -bwd_v3=1 -v3_atomic_fp32=$v3_atomic_fp32 -v3_bf16_cvt=$v3_bf16_cvt -mode=0 -kname=$KNAME $COMMON_ARGS
$EXE -prec=$prec -b=1 -h=3 -h_k=1 -d=$hdim -s=768 -iperm=$perm -operm=$perm -mask=$mask -bwd_v3=1 -v3_atomic_fp32=$v3_atomic_fp32 -v3_bf16_cvt=$v3_bf16_cvt -mode=0 -kname=$KNAME $COMMON_ARGS
done
done
done
done
done
done
set +x

27
example/ck_tile/01_fmha/script/smoke_test_bwd_v3_xqa.sh
View File

@@ -1,27 +0,0 @@
#!/bin/sh
# TODO: run this script from CK root or build directory
EXE="$(find . -name tile_example_fmha_bwd -type f | head -n 1)"
KNAME=1
export CK_WARMUP=0
export CK_REPEAT=1
COMMON_ARGS='-v=1'
set -x
for prec in "fp16" "bf16" ; do
for perm in 0 1 ; do
for hdim in 72 96 128 ; do
for v3_atomic_fp32 in 0 1 ; do
for v3_bf16_cvt in 0 1 2 ; do
for mask in 0 1 ; do
$EXE -prec=$prec -b=2 -h=4 -h_k=2 -d=$hdim -s=512 -iperm=$perm -operm=$perm -mask=$mask -bwd_v3=1 -v3_atomic_fp32=$v3_atomic_fp32 -v3_bf16_cvt=$v3_bf16_cvt -mode=0 -kname=$KNAME $COMMON_ARGS
$EXE -prec=$prec -b=1 -h=3 -h_k=1 -d=$hdim -s=768 -iperm=$perm -operm=$perm -mask=$mask -bwd_v3=1 -v3_atomic_fp32=$v3_atomic_fp32 -v3_bf16_cvt=$v3_bf16_cvt -mode=0 -kname=$KNAME $COMMON_ARGS
done
done
done
done
done
done
set +x