I used the following to train my model
python train.py --train data/moses_train.csv --name moses
it worked and took about one day to finish 50 epochs. However, when I checked tensorboard, I didn't see learning curves. From the terminal I got:
Opt step 1176000, rec: nan, kl: nan, props mse: 0.00, aff mse: 0.00
NC(=O)CNC(=O)C1(C2=NOC(C3=CC=CC=C3)=C2)CC1 =>
O=C(COC(=O)C1COC2=CC=C(F)C=C2C1)C1CC1 =>
CC1=CC=C(C2=NN3C(=O)N(CC#N)C4=CC=CC=C4C3=N2)C=C1 =>
1.0 valid smiles in batch
fraction of correct characters at reconstruction : 1.533564826240763
CN(C(=O)C1=CN(C2=CC=CC=C2)N=C1)C1=NC2=CC=CC=C2S1 =>
CC(C)CC(NC(N)=O)C(=O)NC(C)C1=CC=CC(N2CCCC2)=C1 =>
CC(OCC1=CC=CC=C1)C(=O)N1CCCC(C(N)=O)C1 =>
1.0 valid smiles in batch
fraction of correct characters at reconstruction : 1.4756944554392248
fraction of correct characters in first valid batch : 1.6493055678438395
[Ep 50/50], batch valid. loss: rec: nan, kl: nan, props mse: 0.00, aff mse: 0.00
I think something was wrong, but I don't know what. I appreciate it if you can help me.
I used the following to train my model python train.py --train data/moses_train.csv --name moses it worked and took about one day to finish 50 epochs. However, when I checked tensorboard, I didn't see learning curves. From the terminal I got:
Opt step 1176000, rec: nan, kl: nan, props mse: 0.00, aff mse: 0.00 NC(=O)CNC(=O)C1(C2=NOC(C3=CC=CC=C3)=C2)CC1 =>
O=C(COC(=O)C1COC2=CC=C(F)C=C2C1)C1CC1 =>
CC1=CC=C(C2=NN3C(=O)N(CC#N)C4=CC=CC=C4C3=N2)C=C1 =>
1.0 valid smiles in batch fraction of correct characters at reconstruction : 1.533564826240763 CN(C(=O)C1=CN(C2=CC=CC=C2)N=C1)C1=NC2=CC=CC=C2S1 =>
CC(C)CC(NC(N)=O)C(=O)NC(C)C1=CC=CC(N2CCCC2)=C1 =>
CC(OCC1=CC=CC=C1)C(=O)N1CCCC(C(N)=O)C1 =>
1.0 valid smiles in batch fraction of correct characters at reconstruction : 1.4756944554392248 fraction of correct characters in first valid batch : 1.6493055678438395 [Ep 50/50], batch valid. loss: rec: nan, kl: nan, props mse: 0.00, aff mse: 0.00
I think something was wrong, but I don't know what. I appreciate it if you can help me.