app.py

import subprocess
subprocess.run(
    'pip install flash-attn --no-build-isolation', 
    env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, 
    shell=True
)
from threading import Thread
import torch
from PIL import Image
import gradio as gr
import spaces
from transformers import AutoModelForCausalLM, AutoProcessor, TextIteratorStreamer
import os
import time
from huggingface_hub import hf_hub_download



os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"

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 = "<h3><center>MODEL: " + MODEL_NAME + "</center></h3>"

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

model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID,
    torch_dtype=torch.float16,
    low_cpu_mem_usage=True,
    trust_remote_code=True
).to(0)
processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True)
eos_token_id=processor.tokenizer.eos_token_id




@spaces.GPU(queue=False)
def stream_chat(message, history: list, temperature: float, max_new_tokens: int):
    print(message)
    conversation = []
    for prompt, answer in history:
        conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}])
    
    if message["files"]:
        image = Image.open(message["files"][-1])
        conversation.append({"role": "user", "content": f"<|image_1|>\n{message['text']}"})
    else:
        if len(history) == 0:
            raise gr.Error("Please upload an image first.")
        image = None
        conversation.append({"role": "user", "content": message['text']})
    print(conversation)
    
    inputs = processor.tokenizer.apply_chat_template(conversation, tokenize=False, add_generation_prompt=True)
    inputs_ids = processor(inputs, image, return_tensors="pt").to(0)
    streamer = TextIteratorStreamer(processor, **{"skip_special_tokens": True, "skip_prompt": True, 'clean_up_tokenization_spaces':False,}) 

    generate_kwargs = dict(
        streamer=streamer,
        max_new_tokens=max_new_tokens,
        temperature=temperature,
        do_sample=True,
        eos_token_id=eos_token_id,
    )
    if temperature == 0:
        generate_kwargs["do_sample"] = False
    generate_kwargs = {**inputs_ids, **generate_kwargs}
    

    thread = Thread(target=model.generate, kwargs=generate_kwargs)
    thread.start()

    buffer = ""
    for new_text in streamer:
        buffer += new_text
        yield buffer


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)