ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-15 10:37:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_TILE] fix bug when iperm =0 in fmha fwd (#2820)

Browse Source 
* fix bug when iperm =0 in fmha fwd

* Disable f8 fmha smoke test until fix pr merged

---------

Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com>

[ROCm/composable_kernel commit: 59cb906482]
This commit is contained in:
Haocong WANG
2025-09-16 15:07:10 +08:00
committed by GitHub
parent 9c6cca5bc5
commit 4db9e47cd5
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
example/ck_tile/01_fmha/script/smoke_test_fwd.sh
View File

@@ -97,7 +97,7 @@ run_fp16_appendkv_tests() {
set -x
run_fp16_bf16_tests
run_fp8_tests
# run_fp8_tests
if [ $TEST_APPENDKV -eq 1 ] ; then
run_fp16_appendkv_tests

2
include/ck_tile/ops/fmha/kernel/fmha_fwd_kernel.hpp
View File

@@ -1868,7 +1868,7 @@ struct FmhaFwdKernel
const auto v_dram_naive = make_naive_tensor_view<address_space_enum::global>(
data, // will update this pointer if using paged-kvcache
make_tuple(length, kargs.hdim_v),
make_tuple(kargs.hdim_v, 1),
make_tuple(kargs.stride_v, 1),
number<FmhaPipeline::kAlignmentV>{},
number<1>{});