search

Hironsan / anago

Bidirectional LSTM-CRF and ELMo for Named-Entity Recognition, Part-of-Speech Tagging and so on.
https://anago.herokuapp.com/
MIT License
1.48k stars 368 forks source link

Stop trainning at epoch 4-5 although setting max epoch equals 20 #57

Closed trangtv57 closed 6 years ago

trangtv57 commented 6 years ago

I am make custom data for my language for ner solution. But when use anago.train with our dataset, its stop at epoch 4-5 and don't show any error for fix?

slouvan commented 6 years ago

I guess it uses early stopping when the F1 does not improve after several epochs.

trangtv57 commented 6 years ago

But when its stop because F1 does not improve, the model should be save and use for predict, or can have notification about this, my training end, and nothing happen. So I can't know extacly behind,

slouvan commented 6 years ago

To save the model for later prediction you can call the save function.

sehoi commented 6 years ago

Are you run training at Jupyter notebook? If so, you should modify trainer.py to take the parameter "verbose=2" (default value is 1 and it error occured when epoch over 3)