Roshan Rao
Roshan Rao
Thanks again for your work on this. I've been playing around with it and I think the decreased residue ID condition does solve it for homomers, but unfortunately sometimes PDB...
@nickbhat I believe you did the encoding of SS3 data - can you answer this?
Hi, I'm not currently planning on adding the Bepler model since I personally do not have the time to devote to getting everything working and playing nicely with the rest...
That seems plausible. We've moved to pytorch lightning in an internal version of this code, which sidesteps some of the version issues. We are looking into cleaning that up and...
For now, the new version of TAPE has `torch>=1.0,
Hi! Do you specifically want to re-implement bert-base, or just a transformer? I have code to train a version of ESM-1b [here](https://github.com/rmrao/msa-transformer/blob/main/train_esm.py). This code scales better and will also result...
The models are trained on Pfam, which is released under LGPL. I'm not sure about the other datasets. Academically, the appropriate thing to do is to cite the corresponding paper...
I'll try to do this at some point - working on a couple of other things. If you want to submit a pull request, I'm happy to merge it. Otherwise...
Oh oops. Shouldn't really matter - predicting the start and stop tokens is easy enough for the model. I'll update the code when I get the chance.
Oops - thanks for pointing it out!