Om Kumar

@erfanhamdi Were you able to run the whole thing in Colab? Can you please provide the steps that you took?

Rephrasing the issue properly

Yes, the _recursive_decompose function in frag_mol_env.py is meant to stop if it takes more than 1000 recursive iterations, but I removed it in my fork as it was being a...

Thanks a lot for the clarification. Will try creating appropriate fragments and putting them to use.

I created fragments specific to my dataset using BRICS from rdkit, but I have a bunch of questions: 1. Take for instance this molecule:  When BRICS generates the fragments,...

@bengioe has something changed in the code base? Because the colab kernel seems to be crashing at the algo.create_training_data... function. At first I thought, there might be some issues with...

Alright, I was able to resolve the kernel crashing error by going to tools->command palette->Use Fallback Runtime as it seems that Colab has gone through some updates, but still can't...

Thank you for the torch-geometric clarification. And on the null values, yes, I am making sure that the reward values are all positive, and the training steps all look something...

Alright, @bengioe I think I know why these null values are there, there seems to be some mismatch in the way I have indexed the fragments, I need to look...

Hey @kk6398 can you share your outputs over here? You used the demo_colmap.py script without any major changes right? Did you use bundle adjustment as well? Thing is I tried...