search

macanv / BERT-BiLSTM-CRF-NER

Tensorflow solution of NER task Using BiLSTM-CRF model with Google BERT Fine-tuning And private Server services
https://github.com/macanv/BERT-BiLSMT-CRF-NER
4.71k stars 1.26k forks source link

learning rate and decay? #83

Open Rock-L opened 5 years ago

Rock-L commented 5 years ago

hi macan:

when i training the ner model , the loss from 44 to 9.2, but loss decrease very slowly , i think the learning rate decay so small . but i have no way to trace the learning rate , and decay can not be found. do you have any solutions to see this two args?

macanv commented 5 years ago

you can add training hook when you training, just like in line 411 to 414 in bert_base/train/bert_lstm_crf.py About more infomation of hook, you may go to read tensorflow api document.

Rock-L commented 5 years ago

i have tried this way , image but get an error like : image

xmy7216 commented 5 years ago

i have tried this way , image but get an error like : image

Hi, I have the same problem. How do you solve this problem in the end? Thank you very much.