ShannonAI / mrc-for-flat-nested-ner

Code for ACL 2020 paper `A Unified MRC Framework for Named Entity Recognition`

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Inference stage complains BertQueryNER were not initialized from the model checkpoint

Weugene opened this issue · comments

Hello!

After training I run the inference stage and I get the following message:
Some weights of BertQueryNER were not initialized from the model checkpoint at /beegfs/home/e.sharaborin/Diplom/rusbert/sbert_large_nlu_ru and are newly initialized: ['span_embedding.classifier1.bias', 'start_outputs.bias', 'end_outputs.bias', 'span_embedding.classifier2.bias', 'span_embedding.classifier1.weight', 'start_outputs.weight', 'span_embedding.classifier2.weight', 'end_outputs.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.

It seems that in the checkpoint files does not contain the full information about the state?
Best regards,
Evgenii Sharaborin