Umberto Lupo

Thanks @martin-steinegger for your clarification! I don't know much about these things but I lazily wonder if memory mapping could help limit RAM usage. For now, is there any `ColabFold`...

I'm interested in this and may like to help with a PR if @nickbhat is not available. From https://github.com/pytorch/fairseq/blob/8fce12ddd4a0414de5a726af6193cee4893f0a13/fairseq/models/roberta/model.py#L49, it seems that you just follow BERT's random weight initialization, and...

@tomsercu thanks a lot for the clarification!

There are definitely assertion errors in some of these tests too, but I linked to the wrong one!

Thank you @beew. I'm looking into the CI to see if I can reproduce there. Meanwhile, besides unit tests issues, could you clarify what you mean by > plot doesn't...

Thanks @beew! Indeed, this should be the correct solution to @nsauerwald's query.

This PR claimed to have introduced full multimer training already: https://github.com/aqlaboratory/openfold/pull/336 So, did this quietly already happen in August last year?

Concerning @nzrandol's initial report, I'm thinking that the issue may come from https://github.com/deepmind/alphafold/blob/18e12d61314214c51ca266d192aad3cc6619018a/alphafold/common/protein.py#L177 which is involved in the writing of the PDB file on disk. Note that, in the lines...

> I think it is a good idea, but I would not put it directly in main branch. Hmm, I'm not sure I agree. Though I think I see why...

Hi! Perhaps this is related to https://github.com/giotto-ai/giotto-tda/pull/689? Maybe a similar fix can be made here? @rballeba what do you think?