Ryan Kortvelesy

Okay, I just did. Is it possible/straightforward to migrate the rest of the library to core? On Wed, Jul 13, 2022 at 3:37 PM cpuhrsch ***@***.***> wrote: > Hello @Acciorocketships...

In the meantime I have implemented a workaround using torch.unique and torch.scatter_reduce_

My implementation doesn't use a mask over a dense tensor, if that's what you're asking. This issue is no longer blocking my work.