metadata
license: cc0-1.0
datasets:
- kairaamilanii/cyberbullying-indonesia
language:
- id
metrics:
- accuracy
- confusion_matrix
base_model:
- indolem/indobertweet-base-uncased
pipeline_tag: text-classification
This model is based on a BERT model trained with a few bullying detection datasets. It is trained exclusively in the Indonesian language.
from transformers import BertTokenizer, AutoModelForSequenceClassification
model_path = 'kairaamilanii/IndoBERT-Bullying-Classifier'
tokenizer = BertTokenizer.from_pretrained(model_path)
model = AutoModelForSequenceClassification.from_pretrained(model_path)
text = "KOK JELEK BANGET SIH" # Example text for prediction
inputs = tokenizer(text, return_tensors="pt")
with torch.no_grad():
outputs = model(**inputs)
predicted_class = torch.argmax(outputs.logits, dim=-1).item()
print(f"Predicted class: {predicted_class}")
if predicted_class == 1:
print("Prediction: Bullying")
else:
print("Prediction: Non-bullying")
example output:
[{'Predicted class': 1, 'Prediction': Bullying}]