|
|
--- |
|
|
license: apache-2.0 |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- accuracy: 1 |
|
|
- recall: 1 |
|
|
- f1: 1 |
|
|
base_model: |
|
|
- distilbert/distilroberta-base |
|
|
tags: |
|
|
- fake-news |
|
|
- text-classification |
|
|
- transformers |
|
|
- distilroberta |
|
|
- nlp |
|
|
- deep-learning |
|
|
- pytorch |
|
|
- huggingface |
|
|
- fine-tuning |
|
|
- misinformation |
|
|
--- |
|
|
**FakeBerta: A Fine-Tuned DistilRoBERTa Model for Fake News Detection** |
|
|
|
|
|
You can check the model's fine-tuning code on my GitHub. |
|
|
|
|
|
Model Overview |
|
|
|
|
|
FakeBerta is a fine-tuned version of DistilRoBERTa for detecting fake news. The model is trained to classify news articles as real (0) or fake (1) using natural language processing (NLP) techniques. |
|
|
Base Model: DistilRoBERTa |
|
|
Task: Fake news classification |
|
|
|
|
|
Example of code using AutoModelForSequenceCalssification: |
|
|
```python |
|
|
from transformers import AutoModelForSequenceClassification, AutoTokenizer |
|
|
import torch |
|
|
|
|
|
model_name = "YerayEsp/FakeBerta" |
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
|
model = AutoModelForSequenceClassification.from_pretrained(model_name) |
|
|
|
|
|
inputs = tokenizer("Breaking: Scientists discover water on Mars!", return_tensors="pt") |
|
|
outputs = model(**inputs) |
|
|
|
|
|
logits = outputs.logits |
|
|
predicted_class = torch.argmax(logits).item() |
|
|
|
|
|
print(f"Predicted class: {predicted_class}") # 0 = Real, 1 = Fake |
|
|
|
|
|
``` |
|
|
Library: Transformers (Hugging Face) |
