Skip to content

softplus support for webgpu #27457

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
guschmue merged 2 commits into from
Feb 27, 2026
Merged

softplus support for webgpu #27457

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e603613
softplus support for webgpu
guschmue Feb 25, 2026
c2bdb80
handle overflow for float32 and float16
guschmue Feb 26, 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
28 changes: 27 additions & 1 deletion onnxruntime/core/providers/webgpu/math/unary_elementwise_ops.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -291,5 +291,31 @@ WEBGPU_ELEMENTWISE_KERNEL(LeakyRelu, 16, WebGpuSupportedFloatTypes())
WEBGPU_LU_IMPL(ThresholdedRelu, "select(vec4<x_element_t>(0), a, a > vec4<x_element_t>(uniforms.attr))", "", 1.0f)
WEBGPU_ELEMENTWISE_KERNEL(ThresholdedRelu, 10, WebGpuSupportedFloatTypes())

// For large a, softplus(a) = log(1 + exp(a)) ≈ a. Use a threshold to return a directly,
// avoiding unnecessary exp/log computation and potential overflow.
// PyTorch uses threshold=20 for float32. For float16, exp overflows at ~11.09 so use 11.
class Softplus final : public UnaryElementwise {
public:
Softplus(const OpKernelInfo& info)
: UnaryElementwise{info, "Softplus",
"select("
"select(log(1.0 + exp(a)), a + log(1.0 + exp(-a)), a > x_value_t(0)),"
"a,"
"a > x_value_t(x_element_t(uniforms.attr))"
")",
"",
ShaderUsage::UseValueTypeAlias | ShaderUsage::UseElementTypeAlias} {}

Status ConfigureProgram(const ComputeContext& context, UnaryElementwiseProgram& program) const override {
const auto* input_tensor = context.Input<Tensor>(0);
float threshold = input_tensor->GetElementType() == ONNX_TENSOR_ELEMENT_DATA_TYPE_FLOAT16 ? 11.0f : 20.0f;
program.AddUniformVariables({threshold});
return Status::OK();
}
};

WEBGPU_ELEMENTWISE_VERSIONED_KERNEL(Softplus, 1, 21, WebGpuSupportedFloatTypes())
WEBGPU_ELEMENTWISE_KERNEL(Softplus, 22, WebGpuSupportedFloatTypes())

} // namespace webgpu
} // namespace onnxruntime
} // namespace onnxruntime
4 changes: 4 additions & 0 deletions onnxruntime/core/providers/webgpu/webgpu_execution_provider.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -146,6 +146,8 @@ class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxD
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 16, LeakyRelu);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 10, ThresholdedRelu);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 20, Gelu);
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 1, 21, Softplus);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 22, Softplus);

class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 1, 10, ReduceMax);
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kOnnxDomain, 11, 11, ReduceMax);
Expand Down Expand Up @@ -496,6 +498,8 @@ std::unique_ptr<KernelRegistry> RegisterKernels(bool enable_graph_capture = fals
KERNEL_CREATE_INFO(16, LeakyRelu),
KERNEL_CREATE_INFO(10, ThresholdedRelu),
KERNEL_CREATE_INFO(20, Gelu),
KERNEL_CREATE_INFO_VERSIONED(1, 21, Softplus),
KERNEL_CREATE_INFO(22, Softplus),

// // binary - math
KERNEL_CREATE_INFO_VERSIONED(7, 12, Add),
Expand Down
Loading