Jason Huang

@connorcoley Thanks for the fast response. To reproduce the high accuracy described in the paper, should I give additional arguments to the command line? For example, should I increase `Maximum...

@connorcoley I can train the model with following command ``` $ python ochem_predict_nn/data/preprocess_candidate_edits_compact.py $ python ochem_predict_nn/main/score_candidates_from_edits_compact.py --hybrid=1 ``` without running: `data/generate_reaction_templates.py` & `data/generate_candidates_edits_fullgrants.py`. #### [1] However, when running predictions, I...

@connorcoley Thanks. With the modified code, I can run predications. I am trying to reproduce high accuracy results with 15k data set. It really take a long long time. Anyway,...

Hi @connorcoley After the prediction, I get similar results as fig4-a in your publication, and have following questions.  [1] In `main/output/a_timstamp_number/probs.dat`, it keeps `true_smiles` & `predicted_smiles`. However, there...

Hello @connorcoley Thanks for the feedback. [1] I get the following error with the `load_weights ` correction ``` 100%|##########################################################################################################################################| 1689/1689 [00:02

Hi @connorcoley I use 3 test-set reactions from `probs.dat`, and the result looks amazing!! Chemists in our team are looking forward to use your work. Thank you for having patience...

Thanks @ttrouill It also works from my site. Looks like: __inference__ uses `hparams` from training, and it is not updated for inference params. https://github.com/tensorflow/nmt/blob/master/nmt/nmt.py#L577 Are you going to create a...