app.py

import requests
import torch
from PIL import Image
from transformers import MllamaForConditionalGeneration, AutoProcessor
import gradio as gr
import spaces



class VisionInstructChat:
    def __init__(self):
        # Initialize the model and processor
        self.model_id = "meta-llama/Llama-3.2-90B-Vision-Instruct"
        self.model = MllamaForConditionalGeneration.from_pretrained(
            self.model_id,
            torch_dtype=torch.bfloat16,
            device_map="auto",
        )
        self.processor = AutoProcessor.from_pretrained(self.model_id)

    # Method to handle the model's response to an image and text input
    @spaces
    def chat_with_model(self, history, image, user_text):
        if image is None or not user_text.strip():
            return history + [["Please upload an image and enter a prompt."]]

        # Prepare messages for the model
        messages = [
            {"role": "user", "content": [
                {"type": "image"},
                {"type": "text", "text": user_text}
            ]}
        ]
        input_text = self.processor.apply_chat_template(messages, add_generation_prompt=True)
        inputs = self.processor(image, input_text, return_tensors="pt").to(self.model.device)

        # Generate response
        output = self.model.generate(**inputs, max_new_tokens=100)
        response = self.processor.decode(output[0])

        # Add user prompt and model response to chat history
        history.append([user_text, response])
        return history

    # Method to reset the chat history
    @spaces
    def reset_chat(self):
        return []

    # Method to create the Gradio interface
    def launch_interface(self):
        with gr.Blocks() as demo:
            gr.Markdown("### Chat with Vision-Instruct Model")

            # Chat history
            chat_history = gr.Chatbot(label="Chat History")

            # Inputs: Image and Text
            with gr.Row():
                with gr.Column(scale=3):
                    image_input = gr.Image(type="pil", label="Upload Image")
                with gr.Column(scale=7):
                    user_input = gr.Textbox(placeholder="Type your message here...", label="Your Prompt")

            # Submit and Clear buttons
            submit_button = gr.Button("Send")
            clear_button = gr.Button("Clear Chat")

            # Button actions
            submit_button.click(fn=self.chat_with_model, inputs=[chat_history, image_input, user_input], outputs=chat_history)
            clear_button.click(fn=self.reset_chat, outputs=chat_history)

        demo.launch()

# Create an instance of the class and launch the interface
chat_app = VisionInstructChat()
chat_app.launch_interface()