Open prashant-saxena opened 8 months ago
If you're using a GPU, could you please also include the version of CUDA you're using? print(torch.version.cuda)
will give that to you.
Also, have you posted your question on any ONNX or PyTorch forums as well? If so, could you include links to those posts here too please?
Windows 10 Python 3.10.0 torch 2.1.2 gfpgan 1.3.8 onnx 1.14.0 onnxruntime 1.15.0
Errors:
I'm getting GFPGANv1.3.onnx model but not sure about these errors. I recently started diving in model writing and training. I think the initial warnings about data type are because of gfpgan 1.3.8. It was last released on Sep 16, 2022, and there must be some inconsistencies between torch 2.1.2 (Dec 15, 2023).
When I try to create onnx runtime using:
again I'm getting an error which the first stage was stating:
I would appreciate it if someone pointed out the solution for these errors and more importantly shed some light about this conversion process. Earlier I was under the impression that this conversion would be straight forward and can be done by some external script but I believe I was wrong.