propagate() got an unexpected keyword argument 'distance'

[Dengsongquanch]

When I run the run.sh in 3_train in example, there occurs the error below:
Traceback (most recent call last):
File "/public/home/yika/miniforge3/envs/deepH3/bin/deeph-train", line 8, in
sys.exit(main())
File "/public/home/yika/miniforge3/envs/deepH3/lib/python3.9/site-packages/deeph/scripts/train.py", line 20, in main
kernel.train(train_loader, val_loader, test_loader)
File "/public/home/yika/miniforge3/envs/deepH3/lib/python3.9/site-packages/deeph/kernel.py", line 513, in train
train_losses = self.kernel_fn(train_loader, 'TRAIN')
File "/public/home/yika/miniforge3/envs/deepH3/lib/python3.9/site-packages/deeph/kernel.py", line 658, in kernel_fn
output = self.model(
File "/public/home/yika/miniforge3/envs/deepH3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/public/home/yika/miniforge3/envs/deepH3/lib/python3.9/site-packages/deeph/model.py", line 633, in forward
atom_fea, edge_fea = self.mp1(atom_fea0, edge_idx, edge_fea0, batch, distance, edge_vec)
File "/public/home/yika/miniforge3/envs/deepH3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/public/home/yika/miniforge3/envs/deepH3/lib/python3.9/site-packages/deeph/model.py", line 406, in forward
atom_fea = self.cgconv(atom_fea, edge_idx, edge_fea, batch, distance)
File "/public/home/yika/miniforge3/envs/deepH3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/public/home/yika/miniforge3/envs/deepH3/lib/python3.9/site-packages/deeph/model.py", line 145, in forward
out = self.propagate(edge_index, x=x, edge_attr=edge_attr, distance=distance, size=size)
TypeError: propagate() got an unexpected keyword argument 'distance'

the version of my pyg is 2.6.1,should I make it lower than 2.6.1?

[mzjb]

Please consider using the package versions specified in the README (https://github.com/mzjb/DeepH-pack?tab=readme-ov-file#python).

[Dengsongquanch]

Please consider using the package versions specified in the README (https://github.com/mzjb/DeepH-pack?tab=readme-ov-file#python).

thank you. However, on our computer cluster, cuda is version 11.8, and pytorch1.9.1 can only find cuda11.1. Can a higher version of cuda be compatible with a lower version of pytorch?

[mzjb]

Yes, a higher system CUDA version (like 11.8) can be compatible with a lower version of PyTorch (like 1.9.1 with CUDA 11.1), as PyTorch typically uses its own built-in CUDA libraries.

[duguex]

Follow the instructions in the README is the best way. But if you want to run the code in the latest version of packages, you can follow the instructions of pyg and issue 9551, 9772 of pytorch-geometric.

Actually, you can change the comment of /your_conda_path/envs/deeph/lib/python3.12/site-packages/deeph/model.py to run the code in the latest version of packages.

def forward(self, x: Union[torch.Tensor, PairTensor], edge_index: Adj,
            edge_attr: OptTensor, batch, distance, size: Size = None) -> torch.Tensor:
    """"""
    if isinstance(x, torch.Tensor):
        x: PairTensor = (x, x)

    # propagate_type: (x: PairTensor, edge_attr: OptTensor, distance: Optional[Tensor], size: Optional[Tensor])
    ......

By the way, here is the package version I used to run the code:

# Name                    Version                   Build  Channel
deeph                     0.2.2                    pypi_0    pypi
e3nn                      0.4.4                    pypi_0    pypi
h5py                      3.12.1          nompi_py312hd203070_103    conda-forge
numpy                     2.2.1           py312h7e784f5_0    conda-forge
pathos                    0.3.0           py312h06a4308_0  
psutil                    5.9.0           py312h5eee18b_0  
pymatgen                  2024.11.13      py312hc0a28a1_0    conda-forge
python                    3.12.8          h9e4cc4f_1_cpython    conda-forge
pytorch                   2.4.1           cpu_mkl_py312hf535c18_100    conda-forge
pytorch-geometric         2.6.1           py312_torch_2.4.0_cpu    rusty1s
scipy                     1.14.1          py312h62794b6_2    conda-forge
tensorboard               2.17.0          py312h06a4308_0  
tensorboard-data-server   0.7.0           py312h52d8a92_1  
torch-scatter             2.1.1           py312hd7e3cc5_9    conda-forge
torchaudio                2.4.1           cpu_py312hd217f58_2    conda-forge
torchvision               0.19.1          cpu_py312h5ad8046_2    conda-forge