Skip to content

[QNN-EP] Restrict zero pads validation check to NPU backend #26534

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
yuslepukhin merged 1 commit into from
Dec 18, 2025
Merged

[QNN-EP] Restrict zero pads validation check to NPU backend #26534

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
48 changes: 25 additions & 23 deletions onnxruntime/core/providers/qnn/builder/opbuilder/pad_op_builder.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -191,8 +191,9 @@ Status PadOpBuilder::ProcessAttributesAndOutputs(QnnModelWrapper& qnn_model_wrap
bool has_negative = std::any_of(tensor_data, tensor_data + size, [](int64_t item) { return item < 0; });
bool has_positive = std::any_of(tensor_data, tensor_data + size, [](int64_t item) { return item > 0; });

// Zero padding value gives 3110 error on QNN.
if (!has_positive && !has_negative) {
// Zero padding value gives 3110 error on QNN on NPU.
const bool is_npu_backend = IsNpuBackend(qnn_model_wrapper.GetQnnBackendType());
if (is_npu_backend && !has_positive && !has_negative) {
return ORT_MAKE_STATUS(ONNXRUNTIME, FAIL, "Got QNN invalid zero only padding value.");
}

Expand Down Expand Up @@ -236,7 +237,7 @@ Status PadOpBuilder::ProcessAttributesAndOutputs(QnnModelWrapper& qnn_model_wrap
qnn_model_wrapper.AddParamWrapper(std::move(mode_param));

QnnParamWrapper pad_amount_param(node_unit.Index(), node_unit.Name(), QNN_OP_PAD_PARAM_PAD_AMOUNT,
std::move(pad_amount_dim), std::move(pad_amount));
std::move(pad_amount_dim), std::vector<uint32_t>(pad_amount));
Comment thread
yuslepukhin marked this conversation as resolved.
param_tensor_names.push_back(pad_amount_param.GetParamTensorName());
qnn_model_wrapper.AddParamWrapper(std::move(pad_amount_param));

Expand All @@ -245,24 +246,26 @@ Status PadOpBuilder::ProcessAttributesAndOutputs(QnnModelWrapper& qnn_model_wrap
ORT_RETURN_IF_ERROR(ProcessConstantValue(qnn_model_wrapper, param_tensor_names, node_unit, inputs[2]));
} // constant_value

std::vector<uint32_t> pad_output_shape;
if (has_negative) {
if (!has_negative) {
// Non-negative pads maps directly onto QNN pad.
ORT_RETURN_IF_ERROR(ProcessOutputs(qnn_model_wrapper, node_unit,
std::move(input_names),
std::move(param_tensor_names),
logger, do_op_validation, GetQnnOpType(node_unit.OpType())));
} else {
// If the pads contain negative values, we will need to use a StridedSlice, possibly preceded by a Pad.
// mixed pad values => Pad + StridedSlice
// non-positive pad values => StridedSlice

std::vector<uint32_t> pad_output_shape;
for (uint32_t i = 0; i < input_shape.size(); ++i) {
pad_output_shape.push_back(input_shape[i] + pad_amount[2 * i] + pad_amount[2 * i + 1]);
}
}

std::string pad_output_name = utils::GetUniqueName(node_unit, "_Pad_output");
// Step 1. Add pad if has_positive
if (has_positive) {
// Only positive.
if (!has_negative) {
ORT_RETURN_IF_ERROR(ProcessOutputs(qnn_model_wrapper, node_unit,
std::move(input_names),
std::move(param_tensor_names),
logger, do_op_validation, GetQnnOpType(node_unit.OpType())));
// Mixed sign pad value.
} else {
std::string pad_output_name = utils::GetUniqueName(node_unit, "_Pad_output");

if (has_positive) {
// Step 1. Add Pad to handle any positive values
TensorInfo input_info = {};
ORT_RETURN_IF_ERROR(qnn_model_wrapper.GetTensorInfo(node_unit.Inputs()[0], input_info));

Expand All @@ -274,19 +277,18 @@ Status PadOpBuilder::ProcessAttributesAndOutputs(QnnModelWrapper& qnn_model_wrap
std::vector<uint32_t>(pad_output_shape));
ORT_RETURN_IF_NOT(qnn_model_wrapper.AddTensorWrapper(std::move(pad_output)),
"Failed to add Pad output tensor.");

ORT_RETURN_IF_NOT(qnn_model_wrapper.CreateQnnNode(pad_name,
QNN_OP_PACKAGE_NAME_QTI_AISW,
GetQnnOpType(node_unit.OpType()),
std::move(input_names),
std::vector<std::string>(input_names),
Comment thread
yuslepukhin marked this conversation as resolved.
{pad_output_name},
std::move(param_tensor_names),
do_op_validation),
"Failed to add Pad node.");
}
}

// Step 2. Add Slice if has_negative
if (has_negative) {
// Step 2. Add Slice to handle the negative values
TensorInfo output_info = {};
ORT_RETURN_IF_ERROR(qnn_model_wrapper.GetTensorInfo(node_unit.Outputs()[0], output_info));

Expand All @@ -302,7 +304,7 @@ Status PadOpBuilder::ProcessAttributesAndOutputs(QnnModelWrapper& qnn_model_wrap
output_info.quant_param.Copy(),
std::vector<uint32_t>(output_shape));
ORT_RETURN_IF_NOT(qnn_model_wrapper.AddTensorWrapper(std::move(org_output)),
"Failed to add Pad output tensor.");
"Failed to add StridedSlice output tensor.");

// Create Slice param
const size_t input_rank = input_shape.size();
Expand Down Expand Up @@ -340,7 +342,7 @@ Status PadOpBuilder::ProcessAttributesAndOutputs(QnnModelWrapper& qnn_model_wrap
{org_output_name},
{slice_param_tensor_name},
do_op_validation),
"Failed to add Pad node.");
"Failed to add StridedSlice node.");
}

return Status::OK();
Expand Down
Loading