Alby M.
Thanks for the thoughts on that @JackCaoG !
Hi @peastman , I've just brought this branch up-to-date with `develop` and would be very happy to get it merged for the next release. I've left a few comments. >...
Great, thanks @peastman !
Yeah we've seen this before when the there are atom indexes in the `edge_index` that are larger than the number of atoms. (Or more generally, when there are invalid indexes...
> The one thing I haven't done is switch statistics() to use pytorch_runstats, since it doesn't support standard deviation. If that feature is added to it, it should be easy...
> It already uses Welford's algorithm. It's mainly a matter of upstreaming the code to pytorch_runstats, though that will also add some complexity since pytorch_runstats supports features that aren't needed...
Looks good to me, no need to wait on statistics stuff to do a first merge--- I'll get this merged later today. Thanks @peastman !
Merged! If I manage to get around to implementing Welford's into runstats, we can revisit migrating to that and making this a "first class" dataset in terms of feature support...
Hi @JonathanSchmidt1 , Thanks! > Every process seems to get its own wandb log. It's not possible to restart because wandb finds an existing run in each process and then...
@beidouamg this looks like a network error unrelated to the `ddp` branch, but maybe there is a race condition. Have you tried to run without `wandb` enabled?