Adriano
Adriano
I think that would be right. I haven't been deployed though.
I have the exact same issue, both on WSL and native ubuntu 
> The problem disappears when downgrading to `2023.10.40` I was having this problem before 2023.10.40 on WSL, maybe there is something else going on.
I just realized you are using a custom Cuda Kernel. I imagine this won't be easy to export to ONNX right? Is there a version of the model that uses...
Thank you very much, will try this asap and report back the results.
Exporting the dino models have their own issues (and workarounds) with some ops not being supported so I started trying the resnet50. When running ```python upsampler = torch.hub.load("mhamilton723/FeatUp", 'resnet50').cuda() hr_feats...