Torch.save() error only in release mode

Hi everyone, I am using libtorch-1.8.2 with cuda 11.1 and visual studio 2022.
When I try to save a simple tensor it works just fine in debug mode and it fails in release. I think I linked everything ok since it’s the only difference between the two and I can do inference/training on both just fine.

The error always takes me to serialize.h (line 71).
If anyone has any idea/suggestion I would really appreciate it.

Thanks!

Nevermind, the issue was a mismatch with debug/release dlls.