[Bug] Version 2.2.0.1 not loading dGPU

[TKan010714]

For some reason, Version 2.2.0.1 doesn't load my dGPU.

Here is the log:

Initialized with provider Cpu on device 0 (requested DXGI adapter index 0, fp16=True).
Provider DirectML not available on device 0: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Provider Cuda not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library
Provider TensorRT not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_TensorRT: Failed to load shared library
Compiled with Cuda: False
Init UI Complete
Provider DirectML not available on device 0: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Init UI

Diagnostic log:

[GpuEnum] starting at 14:43:44.426
[GpuEnum] factory created: 2325625895840
[GpuEnum]   idx=0 hr=0x00000000 success=True adapter=ok
[GpuEnum]   idx=0 desc="NVIDIA GeForce RTX 4060" vendor=0x10DE device=0x2882 vram=8342470656 flags=0 software=False
[GpuEnum]   idx=1 hr=0x00000000 success=True adapter=ok
[GpuEnum]   idx=1 desc="Intel(R) UHD Graphics 770" vendor=0x8086 device=0xA780 vram=2147483648 flags=0 software=False
[GpuEnum]   idx=2 hr=0x00000000 success=True adapter=ok
[GpuEnum]   idx=2 desc="Microsoft Basic Render Driver" vendor=0x1414 device=0x008C vram=0 flags=2 software=True
[GpuEnum]   idx=3 hr=0x887A0002 success=False adapter=null
[GpuEnum] enumeration finished with 2 adapter(s)

In comparison, Version 2.1.0.0 loads my dGPU properly.

Here is the log:

Initialized with provider DirectML on device 1 (requested DXGI adapter index 1).
Failed to initialize provider Cuda on device 0. Message Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_CUDA' in DLL 'onnxruntime'.
Compiled with Cuda: False
Init UI Complete
Init UI

Diagnostic log:

[GpuEnum] starting at 14:43:44.426
[GpuEnum] factory created: 2325625895840
[GpuEnum]   idx=0 hr=0x00000000 success=True adapter=ok
[GpuEnum]   idx=0 desc="NVIDIA GeForce RTX 4060" vendor=0x10DE device=0x2882 vram=8342470656 flags=0 software=False
[GpuEnum]   idx=1 hr=0x00000000 success=True adapter=ok
[GpuEnum]   idx=1 desc="Intel(R) UHD Graphics 770" vendor=0x8086 device=0xA780 vram=2147483648 flags=0 software=False
[GpuEnum]   idx=2 hr=0x00000000 success=True adapter=ok
[GpuEnum]   idx=2 desc="Microsoft Basic Render Driver" vendor=0x1414 device=0x008C vram=0 flags=2 software=True
[GpuEnum]   idx=3 hr=0x887A0002 success=False adapter=null
[GpuEnum] enumeration finished with 2 adapter(s)

[TKan010714]

Thank you for better logging with Version 2.2.0.2. Here are my 2.2.0.2 logs (changed from "Auto" to "TensorRT" a few times + changed from dGPU to iCPU a few times):

Log:

Inference session ready with Cpu in 0.1s.
Building inference session with Cpu — this is where TensorRT builds/loads its engine (first time can be slow) ...
Initialized with provider Cpu on device 0 (requested DXGI adapter index 1, fp16=True).
Provider DirectML not available on device 1: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Provider Cuda not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library
Provider TensorRT not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_TensorRT: Failed to load shared library
Inference session ready with Cpu in 0.0s.
Building inference session with Cpu — this is where TensorRT builds/loads its engine (first time can be slow) ...
Initialized with provider Cpu on device 0 (requested DXGI adapter index 0, fp16=True).
Provider DirectML not available on device 0: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Inference session ready with Cpu in 0.1s.
Building inference session with Cpu — this is where TensorRT builds/loads its engine (first time can be slow) ...
Initialized with provider Cpu on device 0 (requested DXGI adapter index 1, fp16=True).
Provider DirectML not available on device 1: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Provider Cuda not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library
Provider TensorRT not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_TensorRT: Failed to load shared library
Inference session ready with Cpu in 0.0s.
Building inference session with Cpu — this is where TensorRT builds/loads its engine (first time can be slow) ...
Initialized with provider Cpu on device 0 (requested DXGI adapter index 1, fp16=True).
Provider DirectML not available on device 1: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Provider Cuda not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library
Provider TensorRT not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_TensorRT: Failed to load shared library
Inference session ready with Cpu in 0.0s.
Building inference session with Cpu — this is where TensorRT builds/loads its engine (first time can be slow) ...
Initialized with provider Cpu on device 0 (requested DXGI adapter index 1, fp16=True).
Provider DirectML not available on device 1: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Provider Cuda not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library
Provider TensorRT not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_TensorRT: Failed to load shared library
Inference session ready with Cpu in 0.1s.
Building inference session with Cpu — this is where TensorRT builds/loads its engine (first time can be slow) ...
Initialized with provider Cpu on device 0 (requested DXGI adapter index 1, fp16=True).
Provider DirectML not available on device 1: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Provider Cuda not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library
Provider TensorRT not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_TensorRT: Failed to load shared library
Inference session ready with Cpu in 0.1s.
Building inference session with Cpu — this is where TensorRT builds/loads its engine (first time can be slow) ...
Initialized with provider Cpu on device 0 (requested DXGI adapter index 1, fp16=True).
Provider DirectML not available on device 1: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Provider Cuda not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library
Provider TensorRT not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_TensorRT: Failed to load shared library
Inference session ready with Cpu in 0.1s.
Building inference session with Cpu — this is where TensorRT builds/loads its engine (first time can be slow) ...
Initialized with provider Cpu on device 0 (requested DXGI adapter index 1, fp16=True).
Inference session ready with Cpu in 0.1s.
Building inference session with Cpu — this is where TensorRT builds/loads its engine (first time can be slow) ...
Initialized with provider Cpu on device 0 (requested DXGI adapter index 1, fp16=True).
Provider DirectML not available on device 1: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_DML' in DLL 'onnxruntime'.
Provider Cuda not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library
Provider TensorRT not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_TensorRT: Failed to load shared library
Compiled with Cuda: True
Init UI Complete
Provider Cuda not available on device 0: [ErrorCode:Fail] OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library
Init UI

Diagnostic log:

[GpuEnum] starting at 14:22:40.391
[GpuEnum] factory created: 2847238577584
[GpuEnum]   idx=0 hr=0x00000000 success=True adapter=ok
[GpuEnum]   idx=0 desc="Intel(R) UHD Graphics 770" vendor=0x8086 device=0xA780 vram=2147483648 flags=0 software=False
[GpuEnum]   idx=1 hr=0x00000000 success=True adapter=ok
[GpuEnum]   idx=1 desc="NVIDIA GeForce RTX 4060" vendor=0x10DE device=0x2882 vram=8342470656 flags=0 software=False
[GpuEnum]   idx=2 hr=0x00000000 success=True adapter=ok
[GpuEnum]   idx=2 desc="Microsoft Basic Render Driver" vendor=0x1414 device=0x008C vram=0 flags=2 software=True
[GpuEnum]   idx=3 hr=0x887A0002 success=False adapter=null
[GpuEnum] enumeration finished with 2 adapter(s)