ArGTClass
is a bloomz
based classification model, finetuned to categorize a comprehensive spectrum
of fourteen distinct subjects that are Religion,
Finance and Economics, Politics, Medical, Cul-
ture, Sports, Science and Technology, Anthro-
pology and Sociology, Art and Literature, Edu-
cation, History, Language and Linguistics, Law,
as well as Philosophy in Arabic.
For more details, check out our paper
Finetuning code in the following notebook:
Full classification example (CPU)
from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
tokenizer = AutoTokenizer.from_pretrained("dru-ac/ArGTClass")
model = AutoModelForSequenceClassification.from_pretrained("dru-ac/ArGTClass")
text = " .قصفت إسرائيل مستشفى المعمداني في مدينة غزة، والذي خلف مئات الشهداء والجرحى"
inputs = tokenizer(text, return_tensors= 'pt')
outputs = model(**inputs)
ind = outputs.logits.argmax(dim=-1)[0]
predicted_class = model.config.id2label[ind.item()]
Full classification example (GPU)
from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
tokenizer = AutoTokenizer.from_pretrained("dru-ac/ArGTClass")
model = AutoModelForSequenceClassification.from_pretrained("dru-ac/ArGTClass", device_map = 'auto')
text = " .قصفت إسرائيل مستشفى المعمداني في مدينة غزة، والذي خلف مئات الشهداء والجرحى"
inputs = tokenizer(text, return_tensors= 'pt').to("cuda")
outputs = model(**inputs)
ind = outputs.logits.argmax(dim=-1)[0]
predicted_class = model.config.id2label[ind.item()]
Pipeline example (CPU & GPU)
from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained("dru-ac/ArGTClass")
model = AutoModelForSequenceClassification.from_pretrained("dru-ac/ArGTClass", device_map = 'auto')
classifier = pipeline("text-classification", model=model, tokenizer= tokenizer)
text = " .قصفت إسرائيل مستشفى المعمداني في مدينة غزة، والذي خلف مئات الشهداء والجرحى"
classifier(text)
- Downloads last month
- 10
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.