Matthias Fey
Matthias Fey
That sounds cool :) We are very happy to take a pull request in on that one, including both the model and an example. Let me know if you have...
Very clean now, thanks for the updates :)
Can you share a bit more information about your system (OS, Python PyTorch version)?
And your PyTorch version?
If you are switching between PyTorch versions, did you make sure to correctly uninstall all existing installations of `torch-cluster` and re-install them? Usually, an error like `Symbol not found: __ZN5torch3jit17parseSchemaOrNameERKNSt3__112basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEEb`...
Interesting problem. The best way to implement this is indeed with a custom CUDA kernel, in which way you can drop the for-loop in option2/3 and avoid the memory blowup...
You can use the wheels from 2.5.0
you can install with `--no-isolation`
Thanks for reporting, will look into it.
Do you mean `indptr = torch.tensor([0, 3, 8])`?