EduHinglish — MuRIL word-level LID (LoRA)

Token-level language identification for Hinglish text. Labels: HI, EN, NE, UNIV, MIX. Part of the EduHinglish pipeline (K.J. Somaiya College of Engineering).

from transformers import AutoModelForTokenClassification, AutoTokenizer
from peft import PeftModel
base = AutoModelForTokenClassification.from_pretrained("google/muril-base-cased", num_labels=5)
model = PeftModel.from_pretrained(base, "persuasivepost/eduhinglish-muril-lid")
tok = AutoTokenizer.from_pretrained("google/muril-base-cased")