CUDAEXecutionProvider fails for BAAI/bge-m3

[R4ZZ3]

Hi,
I have been trying to get this to run on Runpod for few hours now with different image/template settings.
I have succesfully been able to run this using the torch backend but it seemed to run on CPU based on the runpod monitoring although GPU allocation showed that something was allocated to GPU.
I have tried both cuda 11.8 and 12.1 images. Both dev and stable. Here are my latest template configs:

[R4ZZ3]

Nevermind, running locally with smaller max length + CudaOnnx solved the issue or now the performance is enough to run locally