search

thunlp / PL-Marker

Source code for "Packed Levitated Marker for Entity and Relation Extraction"
MIT License
257 stars 34 forks source link

Applying New datasets #28

Closed Jay0412 closed 2 years ago

Jay0412 commented 2 years ago

I fixed error thx a lot :)

YeDeming commented 2 years ago

The loss is very low now. What's the results when you test on the training set?

Jay0412 commented 2 years ago

I fixed error thx a lot :)

YeDeming commented 2 years ago

The training loss is very low means that the model can get a high performance on the training set, but it didn't. Maybe you should check this.
Sorry, I am busy for the EMNLP submission now and have no time to check this dataset in details.