How can I implement early stopping. I'm working on NER and after a particular epoch the loss stops to decrease there I want to apply early stop. Kindly help me out.
question
Aditya715
All 3 comments
Hi @Aditya715, by default flair uses an early stopping mechanism:
Reduce learning rate and weight decay when a metric has stopped improving.
So when a learning rate of < 0.0001 is reached, training will be stopped :)
I hope this helps!
stefan-it
on 9 Jul 2019
Perhaps we should add the stopping criterion as a parameter, i.e. let users specify at which point the learning rate becomes to small.
alanakbik
on 10 Jul 2019
👍3
Closed by PR #871 - thanks again @stefan-it!
alanakbik
on 16 Jul 2019
Was this page helpful?
0 / 5 - 0 ratings
Related issues
aschmu
路
3Comments
stefan-it
路
3Comments
mittalsuraj18
路
3Comments
shoarora
路
3Comments
happypanda5
路
3Comments
Most helpful comment
Perhaps we should add the stopping criterion as a parameter, i.e. let users specify at which point the learning rate becomes to small.