wangchenghuidream/rnnmgm

RNNMGM

1. data preprocess

   you can use the dataset int data/

   or using utils/augmentation.py for data preprocess

2. train

   generation model(launcher_of_clm.py)

1. pretain:
	train_clm(data_path='data/Ds_9.csv', smi_idx=0, model_name='pt', epochs=30, fq_saving=5)
2. fine-tune:
	train_clm(data_path='data/Dm.csv', smi_idx=0, model_name='tl', epochs=30, fq_saving=5)

​ predictive model(launcher_of_sm.py)

train_predictor('data/Dm.csv', pretrain_path, prop_idx=11, epochs=100, k=10, enum_smi=100, patience=3, fq_saving=5)
note: pretrain_path should be the path of model parameters saved in the training process of pretrained generation model  

3 . generation

run the generate() or valid_generate() in launcher_of_clm.py

4.Score

run the score() in launcher_of_sm.py