mirror of
https://github.com/saymrwulf/onnxruntime.git
synced 2026-07-29 20:14:01 +00:00
[js/webgpu] enable f16 for concat (#18528)
### Description With this PR `realesrgan-t64-f16` models becomes 32.8 ms from 1052.55 ms. Now the whole model run on jsep.
This commit is contained in:
parent
81a763a9eb
commit
ac8598a837
1 changed files with 8 additions and 4 deletions
|
|
@ -12,7 +12,8 @@ ONNX_OPERATOR_VERSIONED_KERNEL_EX(
|
|||
1, 3,
|
||||
kJsExecutionProvider,
|
||||
(*KernelDefBuilder::Create())
|
||||
.TypeConstraint("T", {DataTypeImpl::GetTensorType<float>(),
|
||||
.TypeConstraint("T", {DataTypeImpl::GetTensorType<MLFloat16>(),
|
||||
DataTypeImpl::GetTensorType<float>(),
|
||||
DataTypeImpl::GetTensorType<int32_t>()}),
|
||||
Concat);
|
||||
|
||||
|
|
@ -22,7 +23,8 @@ ONNX_OPERATOR_VERSIONED_KERNEL_EX(
|
|||
4, 10,
|
||||
kJsExecutionProvider,
|
||||
(*KernelDefBuilder::Create())
|
||||
.TypeConstraint("T", {DataTypeImpl::GetTensorType<float>(),
|
||||
.TypeConstraint("T", {DataTypeImpl::GetTensorType<MLFloat16>(),
|
||||
DataTypeImpl::GetTensorType<float>(),
|
||||
DataTypeImpl::GetTensorType<int32_t>()}),
|
||||
Concat);
|
||||
|
||||
|
|
@ -32,7 +34,8 @@ ONNX_OPERATOR_VERSIONED_KERNEL_EX(
|
|||
11, 12,
|
||||
kJsExecutionProvider,
|
||||
(*KernelDefBuilder::Create())
|
||||
.TypeConstraint("T", {DataTypeImpl::GetTensorType<float>(),
|
||||
.TypeConstraint("T", {DataTypeImpl::GetTensorType<MLFloat16>(),
|
||||
DataTypeImpl::GetTensorType<float>(),
|
||||
DataTypeImpl::GetTensorType<int32_t>()}),
|
||||
Concat);
|
||||
|
||||
|
|
@ -42,7 +45,8 @@ ONNX_OPERATOR_KERNEL_EX(
|
|||
13,
|
||||
kJsExecutionProvider,
|
||||
(*KernelDefBuilder::Create())
|
||||
.TypeConstraint("T", {DataTypeImpl::GetTensorType<float>(),
|
||||
.TypeConstraint("T", {DataTypeImpl::GetTensorType<MLFloat16>(),
|
||||
DataTypeImpl::GetTensorType<float>(),
|
||||
DataTypeImpl::GetTensorType<int32_t>()}),
|
||||
Concat);
|
||||
|
||||
|
|
|
|||
Loading…
Reference in a new issue