Add new atom type and failed when training about deepcoy HOT 3 CLOSED

Hi @jooewood I was not able to replicate the error message that you posted. I have updated the repo to include a model atom dictionary that includes phosphorus. Instead of --dataset zinc, you should use --dataset zinc_phosphorus when training/generating using DeepCoy. In addition, molecules must be prepared using zinc_phosphorus rather than zinc. See the README in the data directory for more details. Hopefully this solves your issues.

from deepcoy.

Hi @Auroraywood and apologies for the delay in responding to your query. Please could you let me know the error message produced during training?

from deepcoy.

Hi, @fimrie , Here is the error message, My previous account @Auroraywood has forgot email, so I create a new one:

Exception has occurred: AlreadyExistsError

Resource 
__per_step_3/gradients/while/add_11/Enter_grad/ArithmeticOptimizer/AddOpsRewrite_Add/tmp_var/N10tensorflow19TemporaryVariableOp6TmpVarE
	 [[{{node gradients/while/add_11/Enter_grad/ArithmeticOptimizer/AddOpsRewrite_Add/tmp_var}}]]

During handling of the above exception, another exception occurred:

  File "/y/home/zdx/src/deepcoy/GGNN_DeepCoy.py", line 253, in run_epoch
    result = self.sess.run(fetch_list, feed_dict=batch_data)
  File "/y/home/zdx/src/deepcoy/GGNN_DeepCoy.py", line 285, in train
    self.train_data, True)
  File "/y/home/zdx/src/deepcoy/DeepCoy.py", line 1138, in <module>
    model.train()

from deepcoy.