1from transformers import pipeline
2clf = pipeline("text-classification", model="yolaatar/distilbert-rh-ft", device=0) # CUDA if available
3clf("Bad management, but good work culture")
4# → [{'label': 'LABEL_0', 'score': 0.97}]
5# LABEL_0 → Negative
6# LABEL_1 → Positive
The model inherits biases present in the fine‑tuning data (English web reviews). It may mis‑classify sarcasm, domain‑specific jargon, or texts containing mixed sentiments. Always perform human review when consequences matter.
1from transformers import AutoModelForSequenceClassification, AutoTokenizer
2
3repo = "yolaatar/distilbert-rh-ft"
4model = AutoModelForSequenceClassification.from_pretrained(repo)
5tokenizer = AutoTokenizer.from_pretrained(repo)
1@misc{distilbert_rh_ft_2025,
2 author = {Laatar, Youssef},
3 title = {distilbert-rh-ft: DistilBERT fine‑tuned for binary sentiment classification},
4 year = {2025},
5 howpublished = {\url{https://huggingface.co/yolaatar/distilbert-rh-ft}}
6}
