nlp-uoregon / trankit

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

Cuda device instead of use_gpu=True when multiple GPUs are available #84

Open mrgransky opened 5 months ago

mrgransky commented 5 months ago

Given the multilingual pipeline:

import trankit
from trankit import Pipeline
lemmatizer_multi_lingual_pipeline = Pipeline(
            lang='finnish-ftb',
            gpu=True, # <<<<============== using device="cuda:0" ? instead? ==============>>>>
            embedding='xlm-roberta-large', 
            cache_dir='/path/2/cache_dir',
        )
lemmatizer_multi_lingual_pipeline.add('english')
lemmatizer_multi_lingual_pipeline.add('swedish')
lemmatizer_multi_lingual_pipeline.add('danish')
lemmatizer_multi_lingual_pipeline.add('russian')
lemmatizer_multi_lingual_pipeline.add('french')
lemmatizer_multi_lingual_pipeline.add('german')
lemmatizer_multi_lingual_pipeline.set_auto(True)

Instead of gpu=True, Is it possible to specify device="cuda:0" or device="cuda:2" when multiple GPUs are available in a machine?

Cheers,