search

yyDing1 / GNER

Rethinking Negative Instances for Generative Named Entity Recognition
Apache License 2.0
45 stars 2 forks source link

Training Data Preperation #8

Closed parzival1l closed 5 days ago

parzival1l commented 1 week ago

Hello @yyDing1 ,

First of all, this paper was a great read. Thank you for that. I am also working the zero-shot prediction of NERs. Currently I am trying to replicate the same data format to IT llama3.1-8B-Instruct model. I tried making individual calls to gner_dataset.py but have been unsuccessful, if you could provide some instructions on creating the data for training it would be of a huge help.

yyDing1 commented 1 week ago

Thank you for your interest in our work.

Below is an example of calling gner_dataset.py https://github.com/yyDing1/GNER/blob/16d30d5ab3300741d019f09baa26dca4cb351e16/src/run.py#L218-L225

I hope this is helpful. If you continue to experience issues, could you provide any specific code snippets you’ve tried or any error messages you encountered?

parzival1l commented 1 week ago

Thanks for your response, I tried the same in debug mode - from run.py as recommended but I'm getting an error from the load_dataset call. My datasets package version is '2.21.0' image

yyDing1 commented 1 week ago

It seems that higher versions of datasets do not support this. You could try downgrading the datasets library to version 2.17 or 2.18, as mentioned in https://github.com/yyDing1/GNER/issues/1.

parzival1l commented 5 days ago

Thanks a lot, it worked.