app.py

import os
import string
import copy
import gradio as gr
import PIL.Image
import torch
from transformers import BitsAndBytesConfig, pipeline
import re
import time
import random

DESCRIPTION = "# LLaVA 🌋💪 - Now with Arnold Mode!"

model_id = "llava-hf/llava-1.5-7b-hf"
quantization_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_compute_dtype=torch.float16
)
pipe = pipeline("image-to-text", model=model_id, model_kwargs={"quantization_config": quantization_config})

def extract_response_pairs(text):
    turns = re.split(r'(USER:|ASSISTANT:)', text)[1:]
    turns = [turn.strip() for turn in turns if turn.strip()]
    conv_list = []
    for i in range(0, len(turns[1::2]), 2):
        if i + 1 < len(turns[1::2]):
            conv_list.append([turns[1::2][i].lstrip(":"), turns[1::2][i + 1].lstrip(":")])
    return conv_list

def add_text(history, text):
    history = history + [[text, None]]
    return history, ""  # Clear the input field after submission

def arnold_speak(text):
    arnold_phrases = [
        "Come with me if you want to lift!",
        "I'll be back... after my protein shake.",
        "Hasta la vista, baby weight!",
        "Get to da choppa... I mean, da squat rack!",
        "You lack discipline! But don't worry, I'm here to pump you up!"
    ]
    
    text = text.replace(".", "!")  # More enthusiastic punctuation
    text = text.replace("gym", "iron paradise")
    text = text.replace("exercise", "pump iron")
    text = text.replace("workout", "sculpt your physique")
    
    # Add random Arnold phrase to the end
    text += " " + random.choice(arnold_phrases)
    
    return text

def infer(image, prompt, temperature, length_penalty, repetition_penalty, max_length, min_length, top_p):
    try:
        outputs = pipe(images=image, prompt=prompt,
                       generate_kwargs={"temperature": temperature,
                                        "length_penalty": length_penalty,
                                        "repetition_penalty": repetition_penalty,
                                        "max_length": max_length,
                                        "min_length": min_length,
                                        "top_p": top_p})
        inference_output = outputs[0]["generated_text"]
        return inference_output
    except Exception as e:
        print(f"Error during inference: {str(e)}")
        return f"An error occurred during inference: {str(e)}"

def bot(history, text_input, image, temperature, length_penalty, repetition_penalty, max_length, min_length, top_p, arnold_mode):
    if text_input == "":
        yield history + [["Please input text", None]]
        return

    if image is None:
        yield history + [["Please input image or wait for image to be uploaded before clicking submit.", None]]
        return
    
    chat_history = " ".join([item for sublist in history for item in sublist if item is not None])  # Flatten history
    
    system_prompt = "You are a helpful AI assistant. " if not arnold_mode else "You are Arnold Schwarzenegger, the famous bodybuilder and actor. Respond in his iconic style, using his catchphrases and focusing on fitness and motivation."
    
    prompt = f"{system_prompt}\n{chat_history}\nUSER: <image>\n{text_input}\nASSISTANT:"
    
    response = infer(image, prompt, temperature, length_penalty, repetition_penalty, max_length, min_length, top_p)
    
    if arnold_mode:
        response = arnold_speak(response)
    
    history.append([text_input, ""])
    for i in range(len(response)):
        history[-1][1] = response[:i+1]
        time.sleep(0.05)
        yield history

with gr.Blocks() as demo:
    gr.Markdown(DESCRIPTION)
    gr.Markdown("""## LLaVA, one of the greatest multimodal chat models is now available in Transformers with 4-bit quantization! ⚡️
    See the docs here: https://huggingface.co/docs/transformers/main/en/model_doc/llava.""")
    
    chatbot = gr.Chatbot()
    
    with gr.Row():
        image = gr.Image(type="pil")
        with gr.Column():
            text_input = gr.Textbox(label="Chat Input", lines=3)
            arnold_mode = gr.Checkbox(label="Arnold Schwarzenegger Mode")
    
    with gr.Accordion(label="Advanced settings", open=False):
        temperature = gr.Slider(label="Temperature", minimum=0.5, maximum=1.0, value=1.0, step=0.1)
        length_penalty = gr.Slider(label="Length Penalty", minimum=-1.0, maximum=2.0, value=1.0, step=0.2)
        repetition_penalty = gr.Slider(label="Repetition Penalty", minimum=1.0, maximum=5.0, value=1.5, step=0.5)
        max_length = gr.Slider(label="Max Length", minimum=1, maximum=500, value=200, step=1)
        min_length = gr.Slider(label="Minimum Length", minimum=1, maximum=100, value=1, step=1)
        top_p = gr.Slider(label="Top P", minimum=0.5, maximum=1.0, value=0.9, step=0.1)

    with gr.Row():
        clear_button = gr.Button("Clear")
        submit_button = gr.Button("Submit", variant="primary")

    submit_button.click(
        fn=bot,
        inputs=[chatbot, text_input, image, temperature, length_penalty, repetition_penalty, max_length, min_length, top_p, arnold_mode],
        outputs=chatbot
    ).then(
        fn=lambda: "",
        outputs=text_input
    )

    clear_button.click(lambda: ([], None), outputs=[chatbot, image], queue=False)

    examples = [
        ["./examples/baklava.png", "How to make this pastry?"],
        ["./examples/bee.png", "Describe this image."]
    ]
    gr.Examples(examples=examples, inputs=[image, text_input])

if __name__ == "__main__":
    demo.queue(max_size=10).launch(debug=True)