app.py

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

MODEL_LIST = ["internlm/internlm2_5-7b-chat", "internlm/internlm2_5-7b-chat-1m"]
HF_TOKEN = os.environ.get("HF_TOKEN", None)
MODEL_ID = os.environ.get("MODEL_ID", None)
MODEL_NAME = MODEL_ID.split("/")[-1]

TITLE = "<h1><center>internlm2.5-7b-chat</center></h1>"

DESCRIPTION = f"""
<h3>MODEL NOW: <a href="https://hf.co/{MODEL_ID}">{MODEL_NAME}</a></h3>
"""
PLACEHOLDER = """
<center>
<p>InternLM2.5 has open-sourced a 7 billion parameter base model<br> and a chat model tailored for practical scenarios.</p>
</center>
"""


CSS = """
.duplicate-button {
    margin: auto !important;
    color: white !important;
    background: black !important;
    border-radius: 100vh !important;
}
h3 {
    text-align: center;
}
"""

model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID, 
    torch_dtype=torch.float16,
    attn_implementation="flash_attention_2",
    trust_remote_code=True).cuda()
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)

model = model.eval()

@spaces.GPU()
def stream_chat(
    message: str, 
    history: list, 
    temperature: float = 0.8, 
    max_new_tokens: int = 1024, 
    top_p: float = 1.0, 
    top_k: int = 20, 
    penalty: float = 1.2
):
    print(f'message: {message}')
    print(f'history: {history}')
    for resp, history in model.stream_chat(
        tokenizer,
        query = message,
        history = history,
        max_new_tokens = max_new_tokens,
        do_sample = False if temperature == 0 else True,
        top_p = top_p,
        top_k = top_k,
        temperature = temperature,
    ):
        yield resp


chatbot = gr.Chatbot(height=600, placeholder=PLACEHOLDER)

with gr.Blocks(css=CSS, theme="soft") 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,
        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=8192,
                step=1,
                value=1024,
                label="Max New Tokens",
                render=False,
            ),
            gr.Slider(
                minimum=0.0,
                maximum=1.0,
                step=0.1,
                value=1.0,
                label="top_p",
                render=False,
            ),
            gr.Slider(
                minimum=1,
                maximum=20,
                step=1,
                value=20,
                label="top_k",
                render=False,
            ),
            gr.Slider(
                minimum=0.0,
                maximum=2.0,
                step=0.1,
                value=1.2,
                label="Repetition penalty",
                render=False,
            ),
        ],
        examples=[
            ["Help me study vocabulary: write a sentence for me to fill in the blank, and I'll try to pick the correct option."],
            ["What are 5 creative things I could do with my kids' art? I don't want to throw them away, but it's also so much clutter."],
            ["Tell me a random fun fact about the Roman Empire."],
            ["Show me a code snippet of a website's sticky header in CSS and JavaScript."],
        ],
        cache_examples=False,
    )


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