Skip to content

[Pipeline] Fix Python packaging pipeline failures #29200

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
hariharans29 merged 16 commits into from
Jun 23, 2026
+31 −5
Merged

[Pipeline] Fix Python packaging pipeline failures #29200

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
231d70b
Debug
hariharans29 Jun 20, 2026
6bba3dd
Add debug file
hariharans29 Jun 20, 2026
d5bab99
More debug
hariharans29 Jun 20, 2026
8d1fccd
Add right test exclusion when the test runs on non-AVX512 hosts
hariharans29 Jun 20, 2026
86302dd
Fixes
hariharans29 Jun 20, 2026
f26fe93
More debug
hariharans29 Jun 20, 2026
8b8dc5f
More debug 2
hariharans29 Jun 20, 2026
118ba2b
Debug 3
hariharans29 Jun 20, 2026
dd744c9
Debug 5
hariharans29 Jun 20, 2026
f2ea65b
Debug 6
hariharans29 Jun 20, 2026
412a5b4
Debug 7
hariharans29 Jun 20, 2026
ce168d2
Debug 8
hariharans29 Jun 20, 2026
492a5f7
PR feedback
hariharans29 Jun 21, 2026
5bf8ca7
Re-work PR feedback
hariharans29 Jun 21, 2026
17d323c
PR feedback 2
hariharans29 Jun 22, 2026
808d2f2
Merge remote-tracking branch 'origin' into hari/py_pipeline_fix
hariharans29 Jun 23, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 13 additions & 5 deletions cmake/onnxruntime_mlas.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -796,11 +796,6 @@ else()
${MLAS_SRC_DIR}/intrinsics/avx2/qdwconv_avx2.cpp
${MLAS_SRC_DIR}/intrinsics/avx2/saturation_check_avx2.cpp
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx2.cpp
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit.h
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit.cpp
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit_blklen64.h
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit_blklen128.h
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit_blklen32.h
${MLAS_SRC_DIR}/sqnbitgemm_lut_kernel_avx2.h
${MLAS_SRC_DIR}/sqnbitgemm_lut_kernel_avx2.cpp
${MLAS_SRC_DIR}/rotary_embedding_kernel_avx2.h
Expand Down Expand Up @@ -847,9 +842,22 @@ endif()
${MLAS_SRC_DIR}/x86_64/QgemmU8X8KernelAvx512Core.S
${MLAS_SRC_DIR}/x86_64/ConvSymKernelAvx512Core.S
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512.cpp
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit.h
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit.cpp
Comment thread
hariharans29 marked this conversation as resolved.
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit_blklen64.h
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit_blklen128.h
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit_blklen32.h
)
set_source_files_properties(${mlas_platform_srcs_avx512core} PROPERTIES COMPILE_FLAGS "-mfma -mavx512vnni -mavx512bw -mavx512dq -mavx512vl")

# Strip -mavx512vnni from the W2 scalar oracle / pack-helper TU so the
Comment thread
hariharans29 marked this conversation as resolved.
# compiler cannot autovectorize its int8 dot-product loops to vpdpbusd.
# Needed because this helper runs at model load on AVX-512-only
# (non-VNNI) hosts via the AVX-512 W2 dispatch. TU is pure C++ -- no
# AVX-512 intrinsics inside.
set_source_files_properties(${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512_2bit.cpp PROPERTIES
COMPILE_FLAGS "-mfma -mavx512bw -mavx512dq -mavx512vl")

set(mlas_platform_srcs_avx512vnni
${MLAS_SRC_DIR}/sqnbitgemm_kernel_avx512vnni.cpp
${MLAS_SRC_DIR}/qkv_quant_kernel_avx512vnni.cpp
Expand Down
18 changes: 18 additions & 0 deletions onnxruntime/test/mlas/unittest/test_sqnbitgemm_2bit_gemm.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -310,6 +310,9 @@ void RunW2Case(size_t M, size_t N, size_t K, bool WithBias, uint32_t seed,
// handler in a follow-up.
//
TEST(MlasSq2BitTest, Scalar_BlkLen64) {
if (!GetMlasPlatform().Avx512Supported_) {
GTEST_SKIP() << "AVX-512 not available on this host";
}
struct Shape {
size_t M, N, K;
};
Expand Down Expand Up @@ -351,6 +354,9 @@ TEST(MlasSq2BitTest, Scalar_BlkLen64) {
// Same coverage with per-block non-default zero points.
//
TEST(MlasSq2BitTest, Scalar_BlkLen64_WithZeroPoints) {
if (!GetMlasPlatform().Avx512Supported_) {
GTEST_SKIP() << "AVX-512 not available on this host";
}
struct Shape {
size_t M, N, K;
};
Expand Down Expand Up @@ -806,6 +812,9 @@ constexpr struct {
} // namespace

TEST(MlasSq2BitTest, Scalar_BlkLen128) {
if (!GetMlasPlatform().Avx512Supported_) {
GTEST_SKIP() << "AVX-512 not available on this host";
}
for (uint32_t seed : {0xC0FFEEu, 0xBADC0DEu}) {
for (const auto& s : kSimdShapes_BlkLen128) {
for (bool bias : {false, true}) {
Expand All @@ -819,6 +828,9 @@ TEST(MlasSq2BitTest, Scalar_BlkLen128) {
}

TEST(MlasSq2BitTest, Scalar_BlkLen128_WithZeroPoints) {
if (!GetMlasPlatform().Avx512Supported_) {
GTEST_SKIP() << "AVX-512 not available on this host";
}
for (uint32_t seed : {0xC0FFEEu, 0xBADC0DEu}) {
for (const auto& s : kSimdShapes_BlkLen128) {
for (bool bias : {false, true}) {
Expand Down Expand Up @@ -1160,6 +1172,9 @@ constexpr struct {
} // namespace

TEST(MlasSq2BitTest, Scalar_BlkLen32) {
if (!GetMlasPlatform().Avx512Supported_) {
GTEST_SKIP() << "AVX-512 not available on this host";
}
for (uint32_t seed : {0xC0FFEEu, 0xBADC0DEu}) {
for (const auto& s : kSimdShapes_BlkLen32) {
for (bool bias : {false, true}) {
Expand All @@ -1173,6 +1188,9 @@ TEST(MlasSq2BitTest, Scalar_BlkLen32) {
}

TEST(MlasSq2BitTest, Scalar_BlkLen32_WithZeroPoints) {
if (!GetMlasPlatform().Avx512Supported_) {
GTEST_SKIP() << "AVX-512 not available on this host";
}
for (uint32_t seed : {0xC0FFEEu, 0xBADC0DEu}) {
for (const auto& s : kSimdShapes_BlkLen32) {
for (bool bias : {false, true}) {
Expand Down
Loading