search

smiles724 / InstructMol

8 stars 1 forks source link

question #2

Closed rjyh2876 closed 1 year ago

rjyh2876 commented 1 year ago

Thanks your answer! 可以解释一下下面图面里代码运行的问题吗。 image 感谢！

smiles724 commented 1 year ago

Hi, thanks for your inquiry. Which sort of algorithm did you intend to reproduce? I suppose you try to implement a conventional machine learning method such as Transformer and LSTM. However, I am afraid that I have deleted or changed some parts of those codes, and the codebase is mainly designed for GNNs. If you desire to run algorithms that are not based on graphs, please refer to the original MOLECULEACE github: https://github.com/molML/MoleculeACE

However, if you strongly recommend me to update the code so that it supports all categories of approaches, please let me know.

rjyh2876 commented 1 year ago

sorry I try to python the teacher_main.py.But when it turn to 2/5fold, it turn wrong. it said keyerror :'mask' I find in the 2 fold, it add the mask wrongly?

smiles724 commented 1 year ago

sorry I try to python the teacher_main.py.But when it turn to 2/5fold, it turn wrong. it said keyerror :'mask' I find in the 2 fold, it add the mask wrongly?

Thanks for the question. The problem is raised due to the modification of the x_train and y_train. Please change the splitting code to the following line. BTW, we are gonna update the codebase in the near future. Please wait for the update. x_tr_fold, y_tr_fold = [copy.deepcopy(x_train[i]) for i in split['train_idx']], [copy.deepcopy(y_train[i]) for i in split['train_idx']] x_val_fold, y_val_fold = [copy.deepcopy(x_train[i]) for i in split['val_idx']], [copy.deepcopy(y_train[i]) for i in split['val_idx']]