Alby M.
@anjohan ever seen something like this? I can't find any relevant information, but this looks like one of those strange internal PyTorch bugs usually resolved by upgrading. Can you try...
Hm I see... one other question, what `nequip` and `pair_nequip` versions are you using, and are you using the same `nequip` version on which you trained the model?
Hm those warnings are normal and are from our code--- generally they can safely be ignored, they are there so that global state is never changed silently to avoid hard-to-debug...
Hm ok I will ask around with the PyTorch people... In the meantime, I wonder if it has something to do with CUDA / CUDA versions? > I have pytorch==1.10.1...
> NEQUIP is using device cuda means it can't have been compiled without CUDA... did you fully remove your `build` directory between different attempts?
I see, so CUDA is the issue (as suspected). Can you try again: ``` import torch m = torch.jit.load("model.pth", map_location="cuda") print(m) ``` as a more relevant Python test?
So I haven't heard anything back from the PyTorch Slack on this... It's possible this comes from your use of CUDA 10.2 (which has been recently deprecated by PyTorch); as...
Hi @xyttyxy , Thanks for your interest in our codes! I think @anjohan fixed this in `pair_allegro`, but maybe the fix never made it here into `pair_nequip`. Thanks for letting...
Should be closed by #52.
That is correct that it's a branch mismatch. If you want to run with NPT, you need to have `StressForceOutput` on your model. (This can be done in `main`, see...