search

nlp-uoregon / trankit

Trankit is a Light-Weight Transformer-based Python Toolkit for Multilingual Natural Language Processing
Apache License 2.0
725 stars 99 forks source link

How to download resources manually #14

Closed eliasyin closed 3 years ago

eliasyin commented 3 years ago

Due to the network is poor, when I run the code

from trankit import Pipeline
p = Pipeline('english')

It's very slow and often break off. 41.6M/1.12G [07:26<3:17:34, 90.6kB/s] So, I wander if there are methods that I can manually download the resources.

minhhdvn commented 3 years ago

Hi @eliasyin, This part is for downloading xlm-roberta-base pretrained model which is provided by the transformers library. You can look at this discussion to see how to manually download it. Once the xlm-roberta-base model files are downloaded, please put them at ./cache/trankit/xlm-roberta-base which is the default cache directory. Thanks.

eliasyin commented 3 years ago

Thank you @minhhdvn