directmolecularconfgen / dmcg Goto Github PK

Hello, this is Daniel, student of master of Artificial Intelligence at Paris-Saclay University. This is a great work and thanks for sharing the code, however, the results from the paper cannot be reproduced when executing the train command in the readme. Could you provide the precise command to train the model to obtain the paper results, or provide a pretrained model?
Thank you,
Daniel

Please add a license to state how others may use your work.

hi i run the code following your instruction and got the following bugs. do you know how to fix it?

Traceback (most recent call last):
File "train.py", line 359, in
main()
File "train.py", line 190, in main
remove_hs=args.remove_hs,
File "/net/sunlab/psunlab1/molecular_data/graphnn/DMCG/confgen/e2c/dataset.py", line 61, in init
super().init(self.folder, transform, pre_transform)
File "/nethome/tfu42/.conda/envs/dmcg2/lib/python3.7/site-packages/torch_geometric/data/in_memory_dataset.py", line 57, in init
super().init(root, transform, pre_transform, pre_filter)
File "/nethome/tfu42/.conda/envs/dmcg2/lib/python3.7/site-packages/torch_geometric/data/dataset.py", line 88, in init
self._process()
File "/nethome/tfu42/.conda/envs/dmcg2/lib/python3.7/site-packages/torch_geometric/data/dataset.py", line 171, in _process
self.process()
File "/net/sunlab/psunlab1/molecular_data/graphnn/DMCG/confgen/e2c/dataset.py", line 87, in process
self.process_confgf()
File "/net/sunlab/psunlab1/molecular_data/graphnn/DMCG/confgen/e2c/dataset.py", line 336, in process_confgf
data, slices = self.collate(data_list)
File "/nethome/tfu42/.conda/envs/dmcg2/lib/python3.7/site-packages/torch_geometric/data/in_memory_dataset.py", line 116, in collate
add_batch=False,
File "/nethome/tfu42/.conda/envs/dmcg2/lib/python3.7/site-packages/torch_geometric/data/collate.py", line 86, in collate
increment)
File "/nethome/tfu42/.conda/envs/dmcg2/lib/python3.7/site-packages/torch_geometric/data/collate.py", line 128, in _collate
cat_dim = data_list[0].cat_dim(key, elem, stores[0])
TypeError: cat_dim() takes 3 positional arguments but 4 were given

Hey,
In your paper you claimed that the model has 13.29 million parameters but when I run training there are 128 million parameters, can you please explain such a difference? I would not believe that running the model in inference model (without the 3D encoder) would reduce the number of parameters by 10 times.
Cheers,
Carlen

ATOM 1 N LIG 1 -0.435 -0.208 0.637 1.00 0.00 N
ATOM 2 H LIG 1 0.992 -0.005 -0.955 1.00 0.00 H
ATOM 3 C LIG 1 -0.445 0.027 0.067 1.00 0.00 C
ATOM 4 O LIG 1 -0.410 0.813 0.783 1.00 0.00 O
ATOM 5 C LIG 1 -0.986 -0.279 -0.962 1.00 0.00 C
ATOM 6 C LIG 1 -0.428 0.520 -0.648 1.00 0.00 C
ATOM 7 N LIG 1 0.037 0.983 0.805 1.00 0.00 N
ATOM 8 C LIG 1 -0.147 0.998 0.511 1.00 0.00 C
ATOM 9 N LIG 1 -0.649 -0.204 -0.644 1.00 0.00 N
ATOM 10 H LIG 1 -0.185 -0.123 0.495 1.00 0.00 H
ATOM 11 C LIG 1 0.881 -0.881 0.692 1.00 0.00 C
ATOM 12 C LIG 1 -0.270 -0.257 -0.165 1.00 0.00 C
ATOM 13 CL LIG 1 0.403 0.551 0.782 1.00 0.00 CL
ATOM 14 N LIG 1 0.952 -0.736 0.785 1.00 0.00 N
ATOM 15 C LIG 1 -0.955 -0.279 0.533 1.00 0.00 C
ATOM 16 N LIG 1 -0.587 -0.227 0.710 1.00 0.00 N
ATOM 17 C LIG 1 -0.052 -0.521 -0.493 1.00 0.00 C
ATOM 18 C LIG 1 -0.609 0.583 -1.206 1.00 0.00 C
ATOM 19 C LIG 1 0.998 -0.002 -0.276 1.00 0.00 C
ATOM 20 C LIG 1 0.221 -0.265 0.275 1.00 0.00 C
ATOM 21 C LIG 1 0.749 0.101 -0.611 1.00 0.00 C
ATOM 22 C LIG 1 -0.482 -0.747 -0.908 1.00 0.00 C
ATOM 23 C LIG 1 0.704 0.704 -0.239 1.00 0.00 C
ATOM 24 C LIG 1 0.700 -0.547 0.033 1.00 0.00 C
CONECT 1 2 3 9
CONECT 3 4 4 15
CONECT 5 6
CONECT 6 7 7 14
CONECT 7 8
CONECT 8 9 11 11
CONECT 9 10
CONECT 11 12
CONECT 12 13 14 14
CONECT 15 16 19
CONECT 16 17 20
CONECT 17 18
CONECT 18 19
CONECT 20 21 24
CONECT 21 22
CONECT 22 23
CONECT 23 24
END