app.py

import gradio as gr
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

MODEL_ID = "openbmb/MiniCPM-o-4_5"

tokenizer = None
model = None

def load_model():
    global tokenizer, model
    if model is not None:
        return

    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)

    # GPU recommandé
    model = AutoModelForCausalLM.from_pretrained(
        MODEL_ID,
        trust_remote_code=True,
        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
        device_map="auto",
    )
    model.eval()

def chat(user_msg, history):
    load_model()

    prompt = user_msg
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

    with torch.no_grad():
        out = model.generate(
            **inputs,
            max_new_tokens=256,
            do_sample=True,
            temperature=0.7,
            top_p=0.9,
        )

    text = tokenizer.decode(out[0], skip_special_tokens=True)
    # On renvoie juste la réponse générée
    return text

demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil"),
    out = gr.Image(type="pil")
btn.click(predict, inputs=inp, outputs=out)

)

if __name__ == "__main__":
    demo.launch()