app.py

from PIL import Image
import gradio as gr
import spaces
import os
from huggingface_hub import hf_hub_download
import base64
from llama_cpp import Llama
from llama_cpp.llama_chat_format import Llava15ChatHandler

os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
MODEL_LIST = ["openbmb/MiniCPM-Llama3-V-2_5","openbmb/MiniCPM-Llama3-V-2_5-int4"]
HF_TOKEN = os.environ.get("HF_TOKEN", None)
MODEL_ID = os.environ.get("MODEL_ID")
MODEL_NAME = MODEL_ID.split("/")[-1]

TITLE = "<h1><center>VL-Chatbox</center></h1>"

DESCRIPTION = f'<h3><center>MODEL: <a href="https://hf.co/{MODEL_ID}">{MODEL_NAME}</a></center></h3>'

CSS = """
.duplicate-button {
  margin: auto !important;
  color: white !important;
  background: black !important;
  border-radius: 100vh !important;
}
"""

chat_handler = Llava15ChatHandler.from_pretrained(
    repo_id="openbmb/MiniCPM-Llama3-V-2_5-gguf",
    filename="*mmproj*",
)

llm = Llama.from_pretrained(
  repo_id="openbmb/MiniCPM-Llama3-V-2_5-gguf",
  filename="ggml-model-Q5_K_M.gguf",
  chat_handler=chat_handler,
  n_ctx=4096,
  verbose=True
)

'''
filenames = [
    "*mmproj*",
    "ggml-model-Q5_K_M.gguf"
]

for filename in filenames:
    downloaded_model_path = hf_hub_download(
        repo_id="openbmb/MiniCPM-Llama3-V-2_5-gguf",
        filename=filename,
        local_dir="model"
    )
'''


def image_to_base64_data_uri(file_path):
    with open(file_path, "rb") as img_file:
        base64_data = base64.b64encode(img_file.read()).decode('utf-8')
        return f"data:image/png;base64,{base64_data}"

@spaces.GPU(queue=False)
def stream_chat(message, history: list, temperature: float, max_new_tokens: int):
    print(f'message is - {message}')
    print(f'history is - {history}')
    messages = []

    if message["files"]:
        image = message["files"][-1]
        messages.append({
            "role": "user", 
            "content": [
                {"type": "text", "text": message['text']},
                {"type": "image_url", "image_url":{"url": image}}
            ]
        })
    else:
        if len(history) == 0:
            raise gr.Error("Please upload an image first.")
            image = None
        else:
            image = history[0][0][0]
            for prompt, answer in history:
                if answer is None:
                    messages.extend([{
                        "role": "user", 
                        "content": [
                            {"type": "text", "text": prompt},
                            {"type": "image_url", "image_url": {"url": image}}
                        ]
                    },{
                        "role": "assistant", 
                        "content": ""
                    }])
                else:
                    messages.extend([{
                        "role": "user", 
                        "content": [
                            {"type": "text", "text": prompt},
                            {"type": "image_url", "image_url": {"url": image}}
                        ]
                    }, {
                        "role": "assistant", 
                        "content": answer
                    }])
            messages.append({"role": "user", "content": message['text']})    
    print(f"Messages is -\n{messages}")


    response = llm.create_chat_completion(
        messages = messages,
        temperature=temperature,
        max_tokens=max_new_tokens,
    )

    return response["choices"][0]["text"]


chatbot = gr.Chatbot(height=450)
chat_input = gr.MultimodalTextbox(
    interactive=True, 
    file_types=["image"], 
    placeholder="Enter message or upload file...", 
    show_label=False,

)
EXAMPLES = [
        [{"text": "What is on the desk?", "files": ["./laptop.jpg"]}],
        [{"text": "Where it is?", "files": ["./hotel.jpg"]}],
        [{"text": "Can yo describe this image?", "files": ["./spacecat.png"]}]
]

with gr.Blocks(css=CSS) as demo:
    gr.HTML(TITLE)
    gr.HTML(DESCRIPTION)
    gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")
    gr.ChatInterface(
        fn=stream_chat,
        multimodal=True,
        textbox=chat_input,
        chatbot=chatbot,
        fill_height=True,
        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
        additional_inputs=[
            gr.Slider(
                minimum=0,
                maximum=1,
                step=0.1,
                value=0.8,
                label="Temperature",
                render=False,
            ),
            gr.Slider(
                minimum=128,
                maximum=4096,
                step=1,
                value=1024,
                label="Max new tokens",
                render=False,
            ),
        ],
    ),
    gr.Examples(EXAMPLES,[chat_input])


if __name__ == "__main__":
    demo.queue(api_open=False).launch(show_api=False, share=False)