search

Graylab / IgFold

Fast, accurate antibody structure prediction from deep learning on massive set of natural antibodies
Other
325 stars 61 forks source link

IgFold.py RuntimeError #33

Closed px172 closed 1 year ago

px172 commented 1 year ago

OS: ubuntu 22.04.2 LTS NVIDIA Driver Version: 525.85.12
CUDA Version: 12.0 Python 3.8.16

Ran the demo code of "Antibody structure prediction from sequence" and got #30 error. After modifying the AntiBERTyRunner.py suggested in #30, another error occurs: `

The code, data, and weights for this work are made available for non-commercial use
(including at commercial entities) under the terms of the JHU Academic Software License
Agreement. For commercial inquiries, please contact dmalon11[at]jhu.edu.
License: https://github.com/Graylab/IgFold/blob/main/LICENSE.md

Loading 4 IgFold models... Using device: cuda:0 Loading /home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/igfold/trained_models/IgFold/igfold_1.ckpt... Loading /home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/igfold/trained_models/IgFold/igfold_2.ckpt... Loading /home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/igfold/trained_models/IgFold/igfold_3.ckpt... Loading /home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/igfold/trained_models/IgFold/igfold_5.ckpt... Successfully loaded 4 IgFold models. Loaded AntiBERTy model. Traceback (most recent call last): File "demo.py", line 16, in igfold.fold( File "/home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/igfold/IgFoldRunner.py", line 106, in fold model_out = fold( File "/home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/igfold/utils/folding.py", line 206, in fold model_out = model(model_in) File "/home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl return forward_call(*input, kwargs) File "/home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/igfold/model/IgFold.py", line 248, in forward str_nodes = self.str_node_transform(bert_feats) File "/home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl return forward_call(*input, *kwargs) File "/home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/torch/nn/modules/container.py", line 204, in forward input = module(input) File "/home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl return forward_call(input, kwargs) File "/home/px172/anaconda3/envs/IgFold38/lib/python3.8/site-packages/torch/nn/modules/linear.py", line 114, in forward return F.linear(input, self.weight, self.bias) RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument mat1 in method wrapper_addmm)

`

Investigated IgFold.py and made a modification:

move bert_feats to the same device before the str_node_transform operation.

`

    bert_feats = bert_feats.to(self.device)
    str_nodes = self.str_node_transform(bert_feats)

`

After the fix, the prediction is generated.

jeffreyruffolo commented 1 year ago

Thanks for pointing this out! This now fixed in the AntiBERTy repo.