Views
No views yet
google/flan-t5-small fine-tuné sur le dataset GonzaloA/fake_news
pour détecter les fausses informations (fake news).1from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
2
3tokenizer = AutoTokenizer.from_pretrained("<Daniloking>/flan-t5-fakenews-detector")
4modele = AutoModelForSeq2SeqLM.from_pretrained("<Daniloking>/flan-t5-fakenews-detector")
5
6declaration = "classify as true or fake: The government reduced taxes by 10%."
7inputs = tokenizer(declaration, return_tensors="pt")
8outputs = modele.generate(**inputs, max_new_tokens=8)
9print(tokenizer.decode(outputs[0], skip_special_tokens=True))| Métrique | Avant FT | Après FT |
|---|---|---|
| Accuracy | 0.020 | 0.980 |
| F1 | 0.032 | 0.980 |