tobiasheol / ablang2 Goto Github PK

View Code? Open in Web Editor NEW

11.0 11.0 4.0 81 KB

An antibody-specific language model focusing on NGL prediction

License: BSD 3-Clause "New" or "Revised" License

Python 73.57% Jupyter Notebook 26.43%

antibody-design language-model non-germline

ablang2's People

Contributors

Stargazers

Watchers

Forkers

oxpig algw71

ablang2's Issues

can AbLang2 be applied to nanobody?

Hi, Thanks for the great work! Wondering can AbLang2 be applied to nanobody?

What batch size should be used on an A100?

Cannot see any advice on batch size. If we want to improve the speed of obtaining embeddings, what batch sizes would you recommend on an A100?

Also is a dataloader used?

Are unpaired AbLang-2 models available?

Directionality of heavy and light is not clear.

In the example code there is no indication as to whether heavy or light should be first.

Focal loss gamma

Hi, which gamma parameter was used for the Focal loss?

In the paper it is written that sequences were split into train/eval/test via MMSeqs2 clustering with a sequence_identity threshold of 0.95 - could you provide the full set of parameters used for clustering, or were the remaining ones left to be the default? Thanks!

Recommend Projects