Skip to content

[BugFix]Fix precision issue for LoRA feature #4141

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
paulyu12 merged 5 commits into from
Dec 19, 2025
Merged

[BugFix]Fix precision issue for LoRA feature #4141

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
3e1cb09
[BugFix]This PR aims to fix the precision issue of the LoRA feature i…
hukongyi Nov 12, 2025
a26f016
[BugFix]This PR aims to fix the precision issue of the LoRA feature i…
hukongyi Nov 13, 2025
378b774
[BugFix]This PR aims to fix the precision issue of the LoRA feature i…
hukongyi Nov 13, 2025
2b156bb
310p compile
hukongyi Dec 2, 2025
ee3fafa
Remove non-LoRa operators from the kernel folder in VLLM_ASCEND_CUSTO…
hukongyi Dec 3, 2025
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
4 changes: 4 additions & 0 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,10 @@ set(VLLM_ASCEND_CUSTOM_OP
)

set(VLLM_ASCEND_CUSTOM_OP_EXCLUDE
${KERNEL_FILES}/bgmv_expand.cpp
${KERNEL_FILES}/bgmv_shrink.cpp
${KERNEL_FILES}/sgmv_expand.cpp
${KERNEL_FILES}/sgmv_shrink.cpp
${CMAKE_CURRENT_SOURCE_DIR}/csrc/batch_matmul_transpose/op_kernel/batch_matmul_transpose_kernel.cpp
)

Expand Down
6 changes: 3 additions & 3 deletions csrc/kernels/bgmv_expand.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -342,7 +342,7 @@ class BGMVExpand {

// declare all dtype kernel
BGMV_EXPAND_TYPE_DECLARE(half)
#if (__CCE_AICORE__ >= 220)
#if !defined(__CCE_AICORE__) || (__CCE_AICORE__ >= 220)
BGMV_EXPAND_TYPE_DECLARE(bfloat16_t)
#endif

Expand All @@ -356,8 +356,8 @@ extern void bgmv_expand_impl(AscendType type, void* stream, void* x, void* weigh
bgmv_expand_half<<<blockDim, nullptr, stream>>>(x, weight, indices, indicesSize, yIn, yOut, batchSize, numTokensPerCore,
maxLoRARank, outputHiddenDim, sliceOffset, outputFullDim);
} else if (type == AscendType::BF16) {
#if (__CCE_AICORE__ >= 220)
bgmv_expand_bfloat16_t<<<blockDim, nullptr, stream>>>(x, weight, indices, indicesSize, yIn, yOut, batchSize,
#if !defined(__CCE_AICORE__) || (__CCE_AICORE__ >= 220)
bgmv_expand_bfloat16_t<<<blockDim, nullptr, stream>>>(x, weight, indices, indicesSize, yIn, yOut, batchSize,
numTokensPerCore, maxLoRARank, outputHiddenDim,
sliceOffset, outputFullDim);
#endif
Expand Down
6 changes: 3 additions & 3 deletions csrc/kernels/bgmv_shrink.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -226,7 +226,7 @@ class BGMVShrink {

// declare all dtype kernel
BGMV_SHRINK_TYPE_DECLARE(half)
#if (__CCE_AICORE__ >= 220)
#if !defined(__CCE_AICORE__) || (__CCE_AICORE__ >= 220)
BGMV_SHRINK_TYPE_DECLARE(bfloat16_t)
#endif

Expand All @@ -240,8 +240,8 @@ extern void bgmv_shrink_impl(AscendType type, void* stream, void* x, void* weigh
bgmv_shrink_half<<<blockDim, nullptr, stream>>>(x, weight, indices, indicesSize, y, batchSize, numTokensPerCore,
inputHiddenDim, maxLoRARank, scale);
} else if (type == AscendType::BF16) {
#if (__CCE_AICORE__ >= 220)
bgmv_shrink_bfloat16_t<<<blockDim, nullptr, stream>>>(x, weight, indices, indicesSize, y, batchSize, numTokensPerCore,
#if !defined(__CCE_AICORE__) || (__CCE_AICORE__ >= 220)
bgmv_shrink_bfloat16_t<<<blockDim, nullptr, stream>>>(x, weight, indices, indicesSize, y, batchSize, numTokensPerCore,
inputHiddenDim, maxLoRARank, scale);
#endif
} else {
Expand Down
6 changes: 3 additions & 3 deletions csrc/kernels/sgmv_expand.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -357,7 +357,7 @@ class SGMVExpand {

// declare all dtype kernel
SGMV_EXPAND_TYPE_DECLARE(half)
#if (__CCE_AICORE__ >= 220)
#if !defined(__CCE_AICORE__) || (__CCE_AICORE__ >= 220)
SGMV_EXPAND_TYPE_DECLARE(bfloat16_t)
#endif

Expand All @@ -375,8 +375,8 @@ extern void sgmv_expand_impl(AscendType type, void* stream, void* x, void* weigh
numTokensPerCore, maxLoRARank, outputHiddenDim, sliceOffset,
outputFullDim);
} else if (type == AscendType::BF16) {
#if (__CCE_AICORE__ >= 220)
sgmv_expand_bfloat16_t<<<blockDim, nullptr, stream>>>(x, weight, loraIndices, loraIndicesSize,
#if !defined(__CCE_AICORE__) || (__CCE_AICORE__ >= 220)
sgmv_expand_bfloat16_t<<<blockDim, nullptr, stream>>>(x, weight, loraIndices, loraIndicesSize,
seqLen, seqLenSize, yIn, yOut, batchSize,
numTokensPerCore, maxLoRARank, outputHiddenDim,
sliceOffset, outputFullDim);
Expand Down
6 changes: 3 additions & 3 deletions csrc/kernels/sgmv_shrink.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -242,7 +242,7 @@ class SGMVShrink {

// declare all dtype kernel
SGMV_SHRINK_TYPE_DECLARE(half)
#if (__CCE_AICORE__ >= 220)
#if !defined(__CCE_AICORE__) || (__CCE_AICORE__ >= 220)
SGMV_SHRINK_TYPE_DECLARE(bfloat16_t)
#endif

Expand All @@ -260,8 +260,8 @@ extern void sgmv_shrink_impl(AscendType type, void* stream, void* x, void* weigh
numTokensPerCore, inputHiddenDim, maxLoRARank,
scale);
} else if (type == AscendType::BF16) {
#if (__CCE_AICORE__ >= 220)
sgmv_shrink_bfloat16_t<<<blockDim, nullptr, stream>>>(x, weight, loraIndices, loraIndicesSize,
#if !defined(__CCE_AICORE__) || (__CCE_AICORE__ >= 220)
sgmv_shrink_bfloat16_t<<<blockDim, nullptr, stream>>>(x, weight, loraIndices, loraIndicesSize,
seqLen, seqLenSize,
y, batchSize,
numTokensPerCore, inputHiddenDim, maxLoRARank,
Expand Down
1 change: 1 addition & 0 deletions vllm_ascend/lora/punica_npu.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -255,6 +255,7 @@ def add_lora_embedding(self,
# Embedding layer only need expand op
expand_fun: Callable = (self._expand_prefill
if self.is_prefill else self._expand_decode)
x = x.to(torch.float32)
expand_fun(y, x, lora_b_stacked, add_inputs)

def add_lora_linear(self,
Expand Down
Loading