app.py

from transformers import AutoTokenizer, AutoModelForCausalLM, TextIteratorStreamer
from threading import Thread
import gradio as gr
import json
import subprocess
import os


def install_vllm_from_patch():
    script_path = "./install.sh"
    if not os.path.exists(script_path):
        print(f"Error: install.sh is not exist!")
        return False

    try:
        print(f"begin run install.sh")
        result = subprocess.run(
            ["bash", script_path],
            check=True,
            stdout=subprocess.PIPE,
            stderr=subprocess.PIPE,
            text = True,
            timeout = 300
        )
        print(f"result: {result}")
        return True
    except Exception as e:
        print(f"Error: {str(e)}")
        return False

#install vllm from patch file
#install_vllm_from_patch()


# load model and tokenizer
model_name = "inclusionAI/Ling-mini-2.0"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True
).eval()

def respond(
    message,
    history: list[dict[str, str]],
    system_message,
    max_tokens,
#    temperature,
#    top_p
):
    """
    For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
    """
    #client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")

    if len(system_message) == 0:
        system_message = "## 你是谁\n\n我是百灵(Ling)，一个由蚂蚁集团(Ant Group) 开发的AI智能助手"

    messages = [{"role": "system", "content": system_message}]

    messages.extend(history)

    messages.append({"role": "user", "content": message})

    print(f"system_prompt: {json.dumps(messages, ensure_ascii=False, indent=2)}")

    text = tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True
    )

    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)

    model_inputs = tokenizer([text], return_tensors="pt", return_token_type_ids=False).to(model.device)

    print(f"max_new_tokens={max_tokens}")
    model_inputs.update(
        dict(max_new_tokens=max_tokens,
                             streamer = streamer,
#                             temperature = 0.7,
#                             top_p = 1,
#                             presence_penalty = 1.5,
             )
    )

    # Start a separate thread for model generation to allow streaming output
    thread = Thread(
        target=model.generate,
        kwargs=model_inputs,
    )
    thread.start()

    # Accumulate and yield text tokens as they are generated
    acc_text = ""
    for text_token in streamer:
        acc_text += text_token  # Append the generated token to the accumulated text
        yield acc_text  # Yield the accumulated text

    # Ensure the generation thread completes
    thread.join()


"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
chatbot = gr.ChatInterface(
    respond,
    type="messages",
    additional_inputs=[
        gr.Textbox(value="", label="System message"),
        gr.Slider(minimum=1, maximum=32000, value=512, step=1, label="Max new tokens"),
   #     gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
   #     gr.Slider(
   #         minimum=0.1,
   #         maximum=1.0,
   #         value=0.95,
   #         step=0.05,
   #         label="Top-p (nucleus sampling)",
    #    ),
    ],
)

with gr.Blocks() as demo:
#    with gr.Sidebar():
#        gr.LoginButton()
    chatbot.render()


if __name__ == "__main__":
    demo.launch()