app.py

# Import libraries
import gradio as gr
import torch
from PIL import Image
from transformers import AutoModel, AutoTokenizer
import spaces

device="cuda"

# Load the model and tokenizer
model = AutoModel.from_pretrained('openbmb/MiniCPM-Llama3-V-2_5', trust_remote_code=True, torch_dtype=torch.float16)
model = model.to(device='cuda')
tokenizer = AutoTokenizer.from_pretrained('openbmb/MiniCPM-Llama3-V-2_5', trust_remote_code=True)
model.eval()

# Define a function to generate a response
@spaces.GPU
def generate_response(image, question):
    msgs = [{'role': 'user', 'content': question}]
    res = model.chat(
        image=image,
        msgs=msgs,
        tokenizer=tokenizer,
        sampling=True,
        temperature=0.7,
        stream=True
    )
    generated_text = ""
    for new_text in res:
        generated_text += new_text
    return generated_text

# Create the footer with links
footer = """
<div style="text-align: center; margin-top: 20px;">
    <a href="https://www.linkedin.com/in/pejman-ebrahimi-4a60151a7/" target="_blank">LinkedIn</a> |
    <a href="https://github.com/arad1367/Visual_QA_MiniCPM-Llama3-V-2_5_GradioApp" target="_blank">GitHub</a> |
    <a href="https://arad1367.pythonanywhere.com/" target="_blank">Live demo of my PhD defense</a>
    <br>
    Made with 💖 by Pejman Ebrahimi
</div>
"""

# Create a Gradio interface using gr.Blocks
with gr.Blocks(theme='abidlabs/dracula_revamped') as demo:
    gr.Markdown("Visual Question Answering - Complete chart and image analysis")
    gr.Markdown("Input an image and a question related to the image to receive a response.")
    image_input = gr.Image(type="pil", label="Image")
    question_input = gr.Textbox(label="Question")
    output_text = gr.Textbox(label="Response")
    image_input.change(generate_response, inputs=[image_input, question_input], outputs=output_text)
    gr.HTML(footer)

# Launch the app
demo.launch(debug=True)