Sir how to know that validation perplexity in OpenNMT ,when my training period is stop any situation.
Validation perplexity is printed in the logs and are part of the model filename that is saved after each epoch.
Otherwise see here:
sir i have a doubt i am taking 80% of training corpus , 20% of validation src/tgt i.e differece for the training corpus and also testing corpus is difference it is correct.
This was already covered in the forum. You could just use:
- 2000 for validating
- 2000 for testing
- the rest for training
i am studying after Applying but i want to ask my training src/tgt corpus is 1 million and validation src/tgt corpus 0.2 million and also testing 0.2 million, validation is different as compression to training also testing is it correct?
Sorry, I have some trouble to understand what your question or issue is. Could you be clearer?
Thanks very much!