camembert-qfrcola

camembert-qfrcola

CamemBERT-base fine-tuned on QFrCoLA (Beauchemin & Khoury, EMNLP 2025), a Quebec-French corpus of binary linguistic acceptability judgments (0 = ungrammatical, 1 = grammatical/acceptable in Quebec French).

Trained as part of a CSCI 5501 (Deep Learning Applications, Dalhousie University) mini-project quantifying and reducing the Quebec-French acceptability-judgment gap in pretrained French language models. See the accompanying Space demo and project report for zero-shot vs. fine-tuned comparisons.

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

tok = AutoTokenizer.from_pretrained("hineshp6/camembert-qfrcola")
model = AutoModelForSequenceClassification.from_pretrained("hineshp6/camembert-qfrcola")

sentence = "Il fait frette dehors a matin."
inputs = tok(sentence, return_tensors="pt")
probs = torch.softmax(model(**inputs).logits, dim=-1)[0]
print(f"P(acceptable) = {probs[1]:.3f}")

License

Fine-tuned model released under CC-BY-NC-SA 4.0, matching the QFrCoLA dataset license.