Problem on Prediction

[LTJer]

I am trying running a prediction job through the whole "surface_DOS_dataset" and the "Surf_model.pth" in the 'saved_models' folder

command:
python main.py --data_path='../surface_DOS_dataset/' --model_path='Surf_model.pth' --run_mode='Predict'

The modifications in config file:

LMBTR_descriptor: "False" 
SOAP_descriptor: "False"

error message:

File "main.py", line 508, in <module>
    main()
  File "main.py", line 301, in main
    dataset, config["Training"]["loss"], config["Job"]
  File "/u/au/sa/liutsungwei/scratch/DOS/MatDeepLearn_DOS/matdeeplearn/training/training.py", line 663, in predict
    dos_out, scaling = model(data)
  File "/u/au/sa/liutsungwei/scratch/conda/envs/MDL/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/u/au/sa/liutsungwei/scratch/DOS/MatDeepLearn_DOS/matdeeplearn/models/dospredict.py", line 114, in forward
    out = self.conv_list[i](out, data.edge_index, data.edge_attr)
  File "/u/au/sa/liutsungwei/scratch/conda/envs/MDL/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/u/au/sa/liutsungwei/scratch/DOS/MatDeepLearn_DOS/matdeeplearn/models/dospredict.py", line 253, in forward
    out = self.propagate(edge_index, x=x, edge_attr=edge_attr, size=size)
  File "/u/au/sa/liutsungwei/scratch/conda/envs/MDL/lib/python3.7/site-packages/torch_geometric/nn/conv/message_passing.py", line 454, in propagate
    out = self.aggregate(out, **aggr_kwargs)
  File "/u/au/sa/liutsungwei/scratch/conda/envs/MDL/lib/python3.7/site-packages/torch_geometric/nn/conv/message_passing.py", line 578, in aggregate
    return self.aggr_module(inputs, index, ptr=ptr, dim_size=dim_size,
  File "/u/au/sa/liutsungwei/scratch/conda/envs/MDL/lib/python3.7/site-packages/torch/nn/modules/module.py", line 948, in __getattr__
    type(self).__name__, name))
AttributeError: 'GC_block' object has no attribute 'aggr_module'

Is there anything wrong with my input? or command?

Thanks a lot

[vxfung]

Looks like it may be something wrong on the pytorch geometric end - perhaps the version is not right. What is your current version?

[LTJer]

Here are versions of torch relevant packages from my env

torch                1.8.0
torch-cluster        1.6.0
torch-geometric      2.2.0
torch-scatter        2.1.0
torch-sparse         0.6.12
torch-spline-conv    1.2.1
torchvision          0.9.0

[vxfung]

The package was tested on torch-geometric 1.7.0 so you may try to see if it helps if you downgrade torch-geometric to the appropriate version. A version of the code which is compatible with the newest torch-geometric is being developed, however it may need some more time to complete.

[LTJer]

Thanks for your reply and patience.
Unfortunately, after I tried several different methods, my system could not accommodate the geometric 1.7.0.
Looking forward to the new version update.

Best~