Fine Tune already trained NER BERT


I want to fine tune the ner_ontonotes_bert with my data that has the same tags as in Ontonotes data.

I don’t want to train the model from scratch. I just want to fine tune it with some sentences of my context.
Is it possible to do it?


Yes, it is possible.
To fine-tune model on your data you should take ner_ontonotes_bert configuration file and modify data_path in dataset_reader. The data should be in the same format as the original one.
The model will be loaded from the pre-trained checkpoint (specified in save/load_path).

I have specified the data_path in dataset_reader. I wanted to know the file name to be mentioned in save/load_path from the files downloaded in the ner_ontonotes_bert directory.(screenshot attached which shows the files downloaded)


I tried mentioning the ‘checkpoint’ path in the save/load_path. But I am getting the below error:
Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?
** [[{{node save/RestoreV2}}]]**