stan-his / DeepFMPO

Code accompanying the paper "Deep reinforcement learning for multiparameter optimization in de novo drug design"
MIT License
67 stars 27 forks source link

Dev/py36+ #1

Closed bp-kelley closed 5 years ago

bp-kelley commented 5 years ago

Hi all. I've updated the 3.6 branch fixing some py3 issues here and there.

The final output is appears to match the paper, so I think it is correct. There were some unexpected issues with encodings being empty that have me nervous, but Show_Epoch.py seems to be valid for the epochs.

Epoch 3988       Mean score: 1.05                Percentage in range: [0.96, 0.74, 0.61, 0.66],  0.39
Epoch 3989       Mean score: 1.07                Percentage in range: [0.92, 0.74, 0.59, 0.68],  0.4
Epoch 3990       Mean score: 1.14                Percentage in range: [0.94, 0.75, 0.58, 0.7],  0.42
Epoch 3991       Mean score: 1.15                Percentage in range: [0.95, 0.79, 0.58, 0.67],  0.43
Epoch 3992       Mean score: 1.08                Percentage in range: [0.95, 0.77, 0.56, 0.66],  0.39
Epoch 3993       Mean score: 1.09                Percentage in range: [0.93, 0.75, 0.58, 0.65],  0.41
Epoch 3994       Mean score: 1.18                Percentage in range: [0.93, 0.76, 0.6, 0.68],  0.44
Epoch 3995       Mean score: 1.22                Percentage in range: [0.95, 0.78, 0.63, 0.68],  0.44
Epoch 3996       Mean score: 1.1                 Percentage in range: [0.95, 0.76, 0.58, 0.69],  0.4
Epoch 3997       Mean score: 1.07                Percentage in range: [0.94, 0.74, 0.6, 0.66],  0.39
Epoch 3998       Mean score: 1.21                Percentage in range: [0.95, 0.79, 0.6, 0.71],  0.45
Epoch 3999       Mean score: 1.01                Percentage in range: [0.94, 0.75, 0.54, 0.65],  0.36
INFO:root:Saving