Question use IL2CPP build Universal Windows Platform, but every time I obtain a bunch of error.

It can be built normally on the Windows platform, and an error will be reported on the UWP platform.
use IL2CPP build Universal Windows Platform, but every time I obtain a bunch of error.
who can help me.

the error message:

Exception: Unity.IL2CPP.Building.BuilderFailedException: Build failed with 0 successful nodes and 1 failed ones
Annotation: IL2CPP_CodeGen E:/unityWorksprce/HELLO 3D3/Library/Il2cppBuildCache/UWP/ARM64/buildstate/artifacts/il2cpp_conv_xk0o.traceevents
Cmdline: "C:\Program Files\Unity\Hub\Editor\2021.3.14f1c1\Editor\Data\il2cpp\build\deploy\il2cpp.exe" --convert-to-cpp --directory="E:/unityWorksprce/HELLO 3D3/Temp/StagingArea/Data/Managed" --data-folder="E:\unityWorksprce\HELLO 3D3\Library\Il2cppBuildCache\UWP\ARM64\il2cppOutput\Data" --generatedcppdir="E:/unityWorksprce/HELLO 3D3/Library/Il2cppBuildCache/UWP/ARM64/il2cppOutput" --symbols-folder="E:\unityWorksprce\HELLO 3D3\Library\Il2cppBuildCache\UWP\ARM64\il2cppOutput\Symbols" --emit-null-checks --enable-array-bounds-check --code-generation-option=EnableInlining --stats-output-dir="E:/unityWorksprce/HELLO 3D3/Library/Il2cppBuildCache/UWP/ARM64/il2cppStats" --dotnetprofile=unityaot-win32 --cachedirectory="E:/unityWorksprce/HELLO 3D3/Library/Il2cppBuildCache/UWP/ARM64" --profiler-report --profiler-output-file="E:/unityWorksprce/HELLO 3D3/Library/Il2cppBuildCache/UWP/ARM64/buildstate/artifacts/il2cpp_conv_xk0o.traceevents"
ExitCode: -1
Stdout:
Error: IL2CPP error for method 'System.ValueTuple`2<Tensorflow.Tensor,Tensorflow.Tensor> Tensorflow.image_ops_impl::non_max_suppression_padded_v2(Tensorflow.Tensor,Tensorflow.Tensor,Tensorflow.Tensor,System.Single,System.Single,System.Boolean,System.Boolean,System.Int32)' in assembly 'E:\unityWorksprce\HELLO 3D3\Temp\StagingArea\Data\Managed\Tensorflow.Binding.dll'
System.ArgumentException: System.Object is not an array type (Parameter 'typeReference')
at Unity.IL2CPP.ArrayUtilities.ArrayElementTypeOf(ResolvedTypeInfo typeReference)
at Unity.IL2CPP.MethodBodyWriter.ProcessInstruction(Node node, InstructionBlock block, ResolvedInstruction ins)
at Unity.IL2CPP.MethodBodyWriter.GenerateCodeRecursive(Node node, ReadOnlyDictionary`2 instructionBlocks)
at Unity.IL2CPP.MethodBodyWriter.Generate()
at Unity.IL2CPP.CodeWriters.CodeWriterExtensions.WriteMethodWithMetadataInitialization(IGeneratedMethodCodeWriter writer, String methodSignature, Action`2 writeMethodBody, String uniqueIdentifier, MethodReference methodRef, Boolean writingMethodBody)
at Unity.IL2CPP.MethodWriter.WriteMethodDefinition(AssemblyWriteContext context, IGeneratedMethodCodeWriter writer, MethodReference method)
at Unity.IL2CPP.SourceWriter.WriteTypesMethods(SourceWritingContext context, IGeneratedMethodCodeWriter writer, TypeWritingInformation& writingInformation, NPath filePath, Boolean writeMarshalingDefinitions)
at Unity.IL2CPP.SourceWriters.SourceWriterBase`2.WriteItem(StreamWorkItemData`2 data)
at Unity.IL2CPP.Contexts.Scheduling.Streams.FileLevelParallelStreamManager`3.WorkerWriteItemsToFile(WorkItemData`2 data)
at Unity.IL2CPP.Contexts.Scheduling.PhaseWorkScheduler`1.WorkerLoop(Object data)

at il2cpp.Program.DoRun(String[] args, RuntimePlatform platform, Il2CppCommandLineArguments il2CppCommandLineArguments, BuildingOptions buildingOptions, Boolean throwExceptions)

 

It appears to be a case of IL2CPP stripping of TensorFlow types.

Is you project using Reflection? If so, you'll need to override IL2CPP byte stripping, covered in this documentation page.

 

Tensor tensorNum = graph.OperationByName("num_detections");
Tensor tensorBoxes = graph.OperationByName("detection_boxes");
Tensor tensorScores = graph.OperationByName("detection_scores");
Tensor tensorClasses = graph.OperationByName("detection_classes");
Tensor imgTensor = graph.OperationByName("image_tensor");
Tensor[] outTensorArr = new Tensor[] { tensorNum, tensorBoxes, tensorScores, tensorClasses };
var results = sess.run(outTensorArr, new FeedItem(imgTensor, imgArr));

run method:
public virtual NDArray[] run(object fetches, params FeedItem[] feed_dict);


may be the problem is,run method need object type, and call it use Tensor[] type.
IL2CPP don't support this writing method,but it can be executed normally in unity.
In this case, do I need to modify the source code of TensorFlow to solve the problem,Pass exact Parameter type.

 

Code (CSharp):

1. Tensor tensorNum = graph.OperationByName("num_detections");
2. Tensor tensorBoxes = graph.OperationByName("detection_boxes");
3. Tensor tensorScores = graph.OperationByName("detection_scores");
4. Tensor tensorClasses = graph.OperationByName("detection_classes");
5. Tensor imgTensor = graph.OperationByName("image_tensor");
6. Tensor[] outTensorArr = new Tensor[] { tensorNum, tensorBoxes, tensorScores, tensorClasses };
7. var results = sess.run(outTensorArr, new FeedItem(imgTensor, imgArr));
8.  
9. run method:
10. public virtual NDArray[] run(object fetches, params FeedItem[] feed_dict);

may be the problem is,run method need object type, and call it use Tensor[] type.
IL2CPP don't support this writing method,but it can be executed normally in unity.
In this case, do I need to modify the source code of TensorFlow to solve the problem,Pass exact Parameter type.

 

It can be built normally on the Windows platform, and an error will be reported on the UWP platform.

 

Did you solved this?

 

Hello, did someone knows how to solve it?