visolex / visobert-hsd

ViSoBERT‑HSD: Hate Speech Detection for Vietnamese Text

Fine‑tuned from uitnlp/visobert on the VN‑HSD unified Vietnamese hate‐speech dataset, combining ViHSD, ViCTSD, and ViHOS.

Model Details

• Base Model : uitnlp/visobert
• Dataset : VN‑HSD (ViSoLex‑HSD unified hate speech corpus)
• Fine‑tuning : HuggingFace Transformers

Hyperparameters

• Batch size: 32
• Learning rate: 3e-5
• Epochs: 100
• Max sequence length: 256

Results

• Accuracy : <INSERT_ACCURACY>
• F1 Score : <INSERT_F1_SCORE>

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("visolex/visobert-hsd")
model = AutoModelForSequenceClassification.from_pretrained("visolex/visobert-hsd")

text = "Hắn ta thật kinh tởm!"
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=256)
logits = model(**inputs).logits
pred = logits.argmax(dim=-1).item()
label_map = {0: "CLEAN", 1: "OFFENSIVE", 2: "HATE"}
print(f"Predicted label: {label_map[pred]}")