python3.13 Restoring with GPEN-BFR: cublasLt64_12.dll not found

[tester4488]

First, confirm

• I have read the instruction carefully
• I have searched the existing issues
• I have updated the extension to the latest version

What happened?

with python 3.13 and Restoring with GPEN-BFR:

[ReActor] 09:55:56 - STATUS - Restoring with GPEN-BFR-512.onnx | Face Size is set to 512 2025-11-02 09:55:58.5010506 [E:onnxruntime:Default, provider_bridge_ort.cc:2251 onnxruntime::TryGetProviderInfo_CUDA] E:\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1844 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : Error loading "E:\env\venv313\Lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll" which depends on "cublasLt64_12.dll" which is missing. (Error 126: "Impossibile trovare il modulo specificato.")

with python 3.12 works.

Steps to reproduce the problem

any wf with only facerestore or faceswap node

Sysinfo

windows 10, nvidia rtx4060

Relevant console log

[ReActor] 09:55:56 - STATUS - Restoring with GPEN-BFR-512.onnx | Face Size is set to 512
2025-11-02 09:55:58.5010506 [E:onnxruntime:Default, provider_bridge_ort.cc:2251 onnxruntime::TryGetProviderInfo_CUDA] E:\_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1844 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : Error loading "E:\env\venv313\Lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll" which depends on "cublasLt64_12.dll" which is missing. (Error 126: "Impossibile trovare il modulo specificato.")

Additional information

No response

[Buumcode]

venv313

Error loading "E:\env\venv313\Lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll"

You need to install onnxruntime-gpu in venv313.

1/ Activate your venv
2/ Do following steps:

pip uninstall onnxruntime
pip uninstall onnxruntime-gpu

pip install onnxruntime-gpu

DO NOT install onnxruntime <-- never! (It's force you to work on CPU, not GPU).
Check your Reactor uses GPU -- not CPU!

[tester4488]

same:

`(e:\env\venv313) C:\Users\mc>python -m pip uninstall onnxruntime onnxruntime-gpu
Found existing installation: onnxruntime 1.23.2
Uninstalling onnxruntime-1.23.2:
Would remove:
e:\env\venv313\lib\site-packages\onnxruntime-1.23.2.dist-info*
e:\env\venv313\lib\site-packages\onnxruntime*
e:\env\venv313\scripts\onnxruntime_test.exe
Would not remove (might be manually added):
e:\env\venv313\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll
e:\env\venv313\lib\site-packages\onnxruntime\capi\onnxruntime_providers_tensorrt.dll
Proceed (Y/n)? y
Successfully uninstalled onnxruntime-1.23.2
Found existing installation: onnxruntime-gpu 1.23.2
Uninstalling onnxruntime-gpu-1.23.2:
Would remove:
e:\env\venv313\lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll
e:\env\venv313\lib\site-packages\onnxruntime\capi\onnxruntime_providers_tensorrt.dll
e:\env\venv313\lib\site-packages\onnxruntime_gpu-1.23.2.dist-info*
Proceed (Y/n)? y
Successfully uninstalled onnxruntime-gpu-1.23.2

(e:\env\venv313) C:\Users\mc>python -m pip install onnxruntime-gpu
Collecting onnxruntime-gpu
Downloading onnxruntime_gpu-1.23.2-cp313-cp313-win_amd64.whl.metadata (5.6 kB)
Requirement already satisfied: coloredlogs in e:\env\venv313\lib\site-packages (from onnxruntime-gpu) (15.0.1)
Requirement already satisfied: flatbuffers in e:\env\venv313\lib\site-packages (from onnxruntime-gpu) (25.9.23)
Requirement already satisfied: numpy>=1.21.6 in e:\env\venv313\lib\site-packages (from onnxruntime-gpu) (1.26.4)
Requirement already satisfied: packaging in e:\env\venv313\lib\site-packages (from onnxruntime-gpu) (25.0)
Requirement already satisfied: protobuf in e:\env\venv313\lib\site-packages (from onnxruntime-gpu) (6.33.0)
Requirement already satisfied: sympy in e:\env\venv313\lib\site-packages (from onnxruntime-gpu) (1.14.0)
Requirement already satisfied: humanfriendly>=9.1 in e:\env\venv313\lib\site-packages (from coloredlogs->onnxruntime-gpu) (10.0)
Requirement already satisfied: pyreadline3 in e:\env\venv313\lib\site-packages (from humanfriendly>=9.1->coloredlogs->onnxruntime-gpu) (3.5.4)
Requirement already satisfied: mpmath<1.4,>=1.1.0 in e:\env\venv313\lib\site-packages (from sympy->onnxruntime-gpu) (1.3.0)
Downloading onnxruntime_gpu-1.23.2-cp313-cp313-win_amd64.whl (244.5 MB)
---------------------------------------- 244.5/244.5 MB 58.0 MB/s 0:00:04
Installing collected packages: onnxruntime-gpu
Successfully installed onnxruntime-gpu-1.23.2
`
and:

`[ReActor] 07:12:39 - STATUS - Restoring with GPEN-BFR-1024.onnx | Face Size is set to 1024
2025-11-03 07:12:41.0253257 [E:onnxruntime:Default, provider_bridge_ort.cc:2251 onnxruntime::TryGetProviderInfo_CUDA] E:_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1844 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : Error loading "E:\env\venv313\Lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll" which depends on "cublasLt64_12.dll" which is missing. (Error 126: "Impossibile trovare il modulo specificato.")

Prompt executed in 13.85 seconds
`

[Buumcode]

cublasLt64_12.dll is missing

Here is adress to my file:

C:\ProgramData\miniconda3\Lib\site-packages\torch\lib\cublasLt64_12.dll

You need to reinstall torch. Ask ChatGPT what to do.

[tester4488]

i have
torch-2.10.0.dev20251101+cu130.dist-info

Not a big problem, I wanted to point this out and understand why, now I understand. Thanks.

[Buumcode]

i have

May be System don't know abouth that - and you need target something for the PATH? What ChatGPT says? Ask him.

[tester4488]

i have

May be System don't know abouth that - and you need target something for the PATH? What ChatGPT says? Ask him.

onnxruntime_providers_cuda.dll" which depends on "cublasLt64_12.dll" which is missing

I think the wheel is not compatible with cuda13 yet