CODEWITHSUNDEEP
pytorchbert-language-modelallennlpcrf
Siddharth Tumre
Siddharth Tumre

Reputation: 1

Why the training time of CRF module from allennlp is higher?

I have build a custom Model in pytorch with a BERT + BiLSTM + CRF architecture. For the CRF layer I have used the allennlp's CRF module. Due to the CRF module the training and inference time increases highly.

As far as I know the CRF layer should not increase the training time a lot. Can someone help with this issue.

I have tried training with and without the CRF. It looks like the CRF takes more time.

Upvotes: 0

Views: 55

Answers (0)

Related Questions