Sören Brunk
Sören Brunk
> After doing a bit of research it seems like `nn.Embedding` module is an exception to the norm. There may be other types of pytorch modules that change dtype from...
Perhaps we also need to understand why the Python impl calls `split_with_sizes` in certain cases. We might need to do something similar. https://github.com/pytorch/pytorch/blob/9adfaf880784ec0cf5f085fc3f282cf53650050f/torch/_tensor.py#L770 ```python def split(self, split_size, dim=0): r"""See :func:`torch.split`"""...
Thanks for your interest and apologies @marcelluethi, this is probably due to an issue that came with the update to cuda 12 https://github.com/bytedeco/javacpp-presets/issues/1376 I have a workaround that should get...
Thanks for trying. I definitely need to do more testing on windows although having the same behavior on WSL suggests it might happen always if there's no cuda available. I'll...
Hey @dahlej welcome! Any contribution is helpful so please don't hesitate to participate and ask questions. I want Storch be useful for people no matter what their background or level...
@shadaj @kiendang any thoughts on this?
@marcelluethi thanks for reporting. I'm not sure yet either. We could move the call to `requiresGrad` into in the conversion as a workaround for now.
Thanks @mullerhai for your contribution. Since this is a huge PR, I will need a bit of time to go through it. Thanks for your patience.
This is so amazing @mullerhai!
Let's get CI running again so we can run the tests there. I think we need to update the workflows a bit as the actions are outdated.