shenggan

To align with OpenFold inference where there is no batch dimension, we use `unsqueeze`. If your data is entering evoformer with batch latitude, you can just remove the `unsqueeze/squeeze` operation.

1. we use two platforms in paper, A supercomputer (A100 40 GB) for training experiments and one GPU server (A100 80 GB) for inference experiments. 2. This mismatch precision in...

We can't reproduce the error, you can try to provide an example code and hardware platform information. So that we can assist you to solve the problem.

In some scenarios, such as single sequence inference, mask is not necessary. But in scenarios where **padding** is used, mask is necessary. If you have problems with this kernel, you...

I think you can see the code part of `use_precomputed_alignments`, https://github.com/hpcaitech/FastFold/blob/ad7f0cb5d00975e698b690ecc3c44eddc446c7e1/inference.py#L155-L171 The inference scripts will save the alignment results in `local_alignment_dir`, so you can put your msa search result in...

Feel sorry, this version of FastFold can not support multimer.

From the error message, I think there was an error when loading the weights of alphafold. Can you provide the version of your weights and how you get the weights?

I suppose you need to set `param_path` to the path of a monomer model like `params_model_1.npz` for model_1.

These outputs are as expected. The two pdb files correspond to the protein files before and after relax, respectively.

No save for plddt for now, but you can print by yourself after https://github.com/hpcaitech/FastFold/blob/main/inference.py#L185-L186 And, I think you may need to extend the inference scripts to predict multiple models at...