lamdx4/phobert-vi-moderation

phoBERT Vietnamese Moderation (binary)

How to use

from transformers import AutoTokenizer, AutoModelForSequenceClassification
repo = "lamdx4/phobert-vi-moderation"
tok  = AutoTokenizer.from_pretrained(repo, use_fast=False)
mdl  = AutoModelForSequenceClassification.from_pretrained(repo)
probs = mdl(**tok("bình luận ...", return_tensors="pt", truncation=True, max_length=192)).logits.softmax(-1)
print({"valid": float(probs[0,0]), "invalid": float(probs[0,1])})