Hello there sagar here anyone know how much time take CRFEntityExtractor to train the #rasa model i have 3000 entity and i stared the model train on 21 December but still now in process
is there any way to make this process faster i have 8gb GPU system config #rasa
Hi @sagar6531,
just to make one thing sure: You have 3000 distinct entities and for those, you provide n samples per each intent? If so, I’d recommend to use a spacy pipeline and to enhance the used spacy model by retraining its custom entities e.g. like described here (Medium) or here (spacy doc).
This way, you could tell Rasa to auto fill the slots by using the spacy-detected entities or to simply extract them if you don’t use slots. In any way, this will reduce the training time of Rasa significantly.
Keep the catastrophical forgetting problem in mind.
If you have 3000 samples containing n custom entities, then most likely there is a problem with your setup since this is not a significant amount of training data. The CRFEntityExtractor uses the scikit-crfsuite which is known to be pretty fast.
Regards Julian
i have 15 intent with 3000 Entity sample statement it take more time i mean Here my pipeline any suggestion in there ?
Hi @sagar6531,
is your dataset a private one? If not, would you mind sending it to me such that I can do some tests with it? The training time you mentioned seem a bit too much imho. Regarding the pipeline you posted, everything seems to be okay with it.
Regards and merry christmas Julian
sure you can do but model training are still running not finish
Hi @sagar6531,
can you please zip your bot and upload it here such than I can test it?
Thanks a lot!
Regards Julian
