app.py

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

# Charger ton modèle depuis Hugging Face
@st.cache_resource
def load_model():
    model = AutoModelForCausalLM.from_pretrained(
        "badrben/Open_Ques_Fr_AC4",  # 👈 ton repo HF
        device_map="auto",
        torch_dtype=torch.float16
    )
    tokenizer = AutoTokenizer.from_pretrained("badrben/Open_Ques_Fr_AC4")
    return model, tokenizer

model, tokenizer = load_model()

st.title("💬 Open Questions Francais AC4")
user_input = st.text_area("Pose ta question en français :")

if st.button("Générer"):
    inputs = tokenizer(user_input, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_new_tokens=200)
    st.write(tokenizer.decode(outputs[0], skip_special_tokens=True))