naver / biobert-pretrained

BioBERT: a pre-trained biomedical language representation model for biomedical text mining
667 stars 88 forks source link

total time require for training #10

Closed AFNANAMIN closed 5 years ago

AFNANAMIN commented 5 years ago

i am running "run_ner" file for training about 7 to 8 hours in colab and model won't show any progress image output of the training is like above is there any error?

jhyuklee commented 5 years ago

Hi,

it seems that the code is running on cpus not on gpus. Otherwise, check hyperparameters like batch size.

(it would be nice to post finetuning related issues in dmis-lab/biobert)

Thanks!

AFNANAMIN commented 5 years ago

can you plz tell the estimated amount of time required in training i am using the same data provided in your github repo

wonjininfo commented 5 years ago

Hi @afnanamin, it depends on the task and the dataset. For NER, training would usually takes 1~20 hours (with gpu; titan xp) depends on the settings and the dataset. It may take longer if you wish to get the best score. Thanks! Wonjin