Onnx Runtime EntryPointNotFoundException: OrtGetApiBase in Unity Application.

[ledbetterj1atwit]

Describe the issue

Im using ML.net and The Onnx Runtime (3.0.1, 1.17.1 respectively) to load an Onnx model for inference from within a Unity(2021.3.23f1) application. The packages were installed from NuGet using NuGetForUnity which places the packages in <project dir>/Assets/Packages

Whenever I try to apply the model (mlContext.Transforms.ApplyOnnxModel(outputColumnNames, inputColumnNames, modelPathS);) I get

EntryPointNotFoundException: OrtGetApiBase assembly:<unknown assembly> type:<unknown type> member:(null)
Microsoft.ML.OnnxRuntime.NativeMethods..cctor () (at <db8a5557c9b54ae7905ea2adba430b5c>:0)
Rethrow as TypeInitializationException: The type initializer for 'Microsoft.ML.OnnxRuntime.NativeMethods' threw an exception.
Microsoft.ML.OnnxRuntime.SessionOptions..ctor () (at <db8a5557c9b54ae7905ea2adba430b5c>:0)
Microsoft.ML.Transforms.Onnx.OnnxModel..ctor (System.String modelFile, System.Nullable`1[T] gpuDeviceId, System.Boolean fallbackToCpu, System.Boolean ownModelFile, System.Collections.Generic.IDictionary`2[TKey,TValue] shapeDictionary, System.Int32 recursionLimit, System.Nullable`1[T] interOpNumThreads, System.Nullable`1[T] intraOpNumThreads) (at <23755ae4f3b243f8bebd1b959e7493a8>:0)
Microsoft.ML.Transforms.Onnx.OnnxTransformer..ctor (Microsoft.ML.Runtime.IHostEnvironment env, Microsoft.ML.Transforms.Onnx.OnnxTransformer+Options options, System.Byte[] modelBytes) (at <23755ae4f3b243f8bebd1b959e7493a8>:0)
Microsoft.ML.Transforms.Onnx.OnnxTransformer..ctor (Microsoft.ML.Runtime.IHostEnvironment env, System.String[] outputColumnNames, System.String[] inputColumnNames, System.String modelFile, System.Nullable`1[T] gpuDeviceId, System.Boolean fallbackToCpu, System.Collections.Generic.IDictionary`2[TKey,TValue] shapeDictionary, System.Int32 recursionLimit, System.Nullable`1[T] interOpNumThreads, System.Nullable`1[T] intraOpNumThreads) (at <23755ae4f3b243f8bebd1b959e7493a8>:0)
Microsoft.ML.Transforms.Onnx.OnnxScoringEstimator..ctor (Microsoft.ML.Runtime.IHostEnvironment env, System.String[] outputColumnNames, System.String[] inputColumnNames, System.String modelFile, System.Nullable`1[T] gpuDeviceId, System.Boolean fallbackToCpu, System.Collections.Generic.IDictionary`2[TKey,TValue] shapeDictionary, System.Int32 recursionLimit, System.Nullable`1[T] interOpNumThreads, System.Nullable`1[T] intraOpNumThreads) (at <23755ae4f3b243f8bebd1b959e7493a8>:0)
Microsoft.ML.OnnxCatalog.ApplyOnnxModel (Microsoft.ML.TransformsCatalog catalog, System.String[] outputColumnNames, System.String[] inputColumnNames, System.String modelFile, System.Nullable`1[T] gpuDeviceId, System.Boolean fallbackToCpu) (at <23755ae4f3b243f8bebd1b959e7493a8>:0)
SicknessPredictor+PredictJob.Execute () (at Assets/SicknessPredictor.cs:353)
Unity.Jobs.IJobExtensions+JobStruct`1[T].Execute (T& data, System.IntPtr additionalPtr, System.IntPtr bufferRangePatchData, Unity.Jobs.LowLevel.Unsafe.JobRanges& ranges, System.Int32 jobIndex) (at <f712b1dc50b4468388b9c5f95d0d0eaf>:0)

I should add I'm not super familiar with C# development

To reproduce

1. Create a Unity Project
2. Download the NuGet packages (Microsoft.ML 3.0.1, Microsoft.ML.OnnxRuntime 1.17.1, Microsoft.ML.OnnxTransformer 3.0.1) and place them and their dependencies in <project dir>/Assets/Packages
3. Create an MLContext and call .Transforms.ApplyOnnxModel() or another method that uses the Onnx Runtime.

Urgency

This is for a thesis project I was hoping to defend on April 15th but I have a backup solution in mind if I cannot use the Onnx Runtime.

Platform

Windows

OS Version

10 Pro 22H2 19045.4170

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

1.17.1

ONNX Runtime API

C#

Architecture

X64

Execution Provider

Default CPU

Execution Provider Library Version

No response

[github-actions]

This issue has been automatically marked as stale due to inactivity and will be closed in 30 days if no further activity occurs. If further support is needed, please provide an update and/or more details.

[AIceDog]

Same problem, do you solve it?

[ledbetterj1atwit]

No, I've had no progress.
Ended up running my model in python using TensorFlow and sending data back and forth with sockets(slow, bad)