suriyadeepan/practical_seq2seq

Unable to train the model

Opened this issue · 2 comments

I tried training your model for around 20,000 iterations on a GPU but when I tested it on the testing data and the unseen data, it is always printing "I don't know".Can someone please help me out?

You've faced a common problem in seq2seq learning. Based on my experience,it's probably because the learning rate is too small for the model to get out of the 'trap' of same response.Try to use bigger learning rate like 0.2 or even 1.

What is the code to train the model.