logp accuracy error.

[owenustc]

Hi, I repeat to execute your LogP module with two test. But, in first test, I only get 0.2413 for valid SMILES when I used retrained generator and predictor to do reinforcement learning. And my retrained generator also get 0.8876497315159025 for drug-like region and 0.7263 for valid SMILES before reinforcement learning. I don't know why my generator get the low valid SMILES after Reinforcement learning.
And in second test, I get 0.7698 for valid SMILES and 0.9664848012470771 for drug-like region when I used your checkpoint of generator. Do you have some tricks for generator training?

[owenustc]

I use the default configuration of the code to run generator training.

[owenustc]

My generator got a high drug-like region, about 0.99, but very lower valid SIMILES, only 0.27. I just to re-execute your generator re-training code. Could you help me?

[isayev]

@owenustc just filter out invalid SMILES if you need more samples.

Also, training RL models require some experience and practice; look up for the learning rates and gradients. Better take RL class if you are not sure what to change.

[owenustc]

@isayev，In your opinion, do you think the generator with lower valid SIMILES is right? I take the RL class of your code to do RL training for generator. And do you meet this situation that the generator valid SMILES lower than 0.3 after RL trainning? Because I use your checkpoint of generator to do RL training and the valid SIMILES is more than 0.7, I guess the checkpoint training of your generator model could have some tricks.