app.py

import streamlit as st
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
import os

token = os.getenv("hf_token")

@st.cache_resource
def load_model():
    model_name = "robzchhangte/bloomz-dv5-with-mztok" 
    # model_name = "robzchhangte/10-vanillagpt2-ft-INS-dv5" 
    tokenizer = AutoTokenizer.from_pretrained("robzchhangte/bloomz-dv5-with-mztok", token=token)
    model = AutoModelForCausalLM.from_pretrained(model_name, token=token)
    return tokenizer, model

tokenizer, model = load_model()

st.title("📝 Mizo Text Generator")

prompt = st.text_area("Enter your prompt (in Mizo):", height=150)

st.text("Example: Lirthei pung nasa chu hmasawnna rah a nih rualin harsatna tam tak..")

generate_button = st.button("Generate Text")

if generate_button and prompt:
    with st.spinner("Generating text..."):
        inputs = tokenizer.encode(prompt, return_tensors='pt')

        outputs = model.generate(
            inputs, 
            max_length=50,
            temperature=0.7,
            top_p=0.9,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id
        )

        generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)

    st.subheader("Generated Text:")
    st.write(generated_text)