Esben Jannik Bjerrum

Yes, the pretrained models are exactly the starting point for further finetuning. Pretraining was done with Smiles2smiles tasks with masking and smiles data augmentation, not reaction data. You can train...

Thanks a lot for all you pull request! Unfortunately I'm super busy these final days before Christmas, but hopefully I will find the time to look at them soon

I think this pull-request is not completely there. I think the setting of a fixedbondlength is great, as it keeps the initial molecule decent in size and similar in size....

Thanks for the PR, I'm looking forward to try it out. What icon is missing? The app uses many icons from [https://icons8.com](https://icons8.com), maybe you can find something suitable there? Otherwise...

@jacktday I have been looking at the pull request and tested it out. Cool work. But I have some comments and suggestions. Conceptually the * or R is an atomtype,...

Btw, I quickly made an Icon that sort of matches the other A+ and A- icons, so don't worry about that. They were increase and decrease font from the icons8...

> @EBjerrum Thank you for your feedback and for finding that bug. It looks like rdkit automatically adds the dummy_label "R" to pseudo atoms when converting to mol without my...

> I'm also playing around with using pseudo atoms to create templates. Which simplifies template creation and allows user-created templates. Unfortunately, I'm running into valence errors when adding bonds. >...

Are you working on it and will update your pull request?

No worries, I fully understand that there can be other priorities. BTW, I saw you made an update with some relative imports. Please do that in another branch and pull...