Zuobai Zhang

Hi, I think the constraint on the protein size is added by ESM actually. I think the best way to deal with big proteins is to get the representations through...

Hi, `tasks.MultipleBinaryClassification` is designed for protein function classification tasks. You can refer to the [GearNet repo](https://github.com/DeepGraphLearning/GearNet) and [this tutorial](https://torchprotein.ai/tutorial_3).

Hi, thx for the question. Here we simply assume that we only consider typical residues and all unknown residues will be treated as glycine. Since we only consider major atoms...

Hi, could you please provide the two versions of torchdrug that you find the default node dimensions are different? This will help us debug. Thanks~ BTW, there is a small...

@KiddoZhu Do you have any idea about this problem?

Hi, it's probably because AlphaFoldDB has updated the latest version while we haven't updated it yet. Could you provide some detailed logs for your problem so that we can debug...

Hi, this is because your targets are str and not transformed to float. One way to deal with this is to use the `utils.literal_eval` function for transformation as in [`data.MoleculeDataset....

Hi, could you share more contexts, e.g., how you define your solver and dataset? Everything works well for me.

Thanks for raising this issue. It seems to be a bug that the `config_dict()` function fails to deal with list arguments in `transforms.Compose`. This has been fixed in https://github.com/DeepGraphLearning/torchdrug/commit/b50884877f8e1185d7500cc9207cc7b3782fb028.

It works for me. Maybe you need to clone the latest repo to fetch the commit and remember to install from the source code.