saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-30 20:18:08 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

[WebNN EP] Add outputDataType option for the ArgMax/ArgMin ops (#21385)

Browse source 
### Description
WebNN spec introduces a new option: `outputDataType` to `argMax` and
`argMin` ops, it's default value is `int32`, we should explicitly set it
to `int64` for WebNN EP.

Spec CR: "Add outputDataType to argmin/argmax"
https://github.com/webmachinelearning/webnn/pull/730
This commit is contained in:
Wanming Lin 2024-07-23 02:56:09 +08:00 committed by GitHub
parent a6c5e2cd20
commit 17e9ea6235
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 5 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
onnxruntime/core/providers/webnn/builders/impl/argmax_min_op_builder.cc
View file

@ -50,6 +50,11 @@ Status ArgMaxMinOpBuilder::AddToModelBuilderImpl(ModelBuilder& model_builder,
options.set("axes", axes);
options.set("keepDimensions", keep_dims == 1);
options.set("selectLastIndex", select_last_index == 1);
// TODO: use WebNN's opSupportLimits API to check the backend's supported output data types.
// If the backend doesn't support int64 output, we should use default int32 output data type
// then do a type casting (int32 -> int64) for the output. Refer to the CoreML EP for how to
// support int64 output.
options.set("outputDataType", "int64");
emscripten::val output = emscripten::val::object();
const auto& op_type = node.OpType();