app.py

#!/usr/bin/env python

import os
from threading import Thread
from typing import Iterator
import spaces
import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer

MAX_MAX_NEW_TOKENS = 1024
DEFAULT_MAX_NEW_TOKENS = 512
MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "8192"))

if not torch.cuda.is_available():
    DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU.</p>"

if torch.cuda.is_available():
    model_id = "utter-project/EuroLLM-9B-Instruct"
    tokenizer = AutoTokenizer.from_pretrained(model_id)
    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

@spaces.GPU(duration=30)
def generate(
    message: str,
    chat_history: list[tuple[str, str]],
    max_new_tokens: int = 512,
    temperature: float = 0.06,
    top_p: float = 0.95,
    top_k: int = 40,
    repetition_penalty: float = 1.2,
) -> Iterator[str]:
    
    historical_text = ""
    #Prepend the entire chat history to the message with new lines between each message
    for user, assistant in chat_history:
        historical_text += f"\n{user}\n{assistant}"
        
    if len(historical_text) > 0:
        message = historical_text + f"\n{message}"
    input_ids = tokenizer([message], return_tensors="pt").input_ids
    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
    input_ids = input_ids.to(model.device)

    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        {"input_ids": input_ids},
        streamer=streamer,
        max_new_tokens=max_new_tokens,
        do_sample=True,
        top_p=top_p,
        top_k=top_k,
        temperature=temperature,
        num_beams=1,
        pad_token_id = tokenizer.eos_token_id,
        repetition_penalty=repetition_penalty,
        no_repeat_ngram_size=5,
        early_stopping=False,
    )
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()

    outputs = []
    for text in streamer:
        outputs.append(text)
        yield "".join(outputs)


chat_interface = gr.ChatInterface(
    fn=generate,
    additional_inputs=[
        gr.Slider(
            label="Max new tokens",
            minimum=1,
            maximum=MAX_MAX_NEW_TOKENS,
            step=1,
            value=DEFAULT_MAX_NEW_TOKENS,
        ),
        gr.Slider(
            label="Temperature",
            minimum=0.1,
            maximum=1.2,
            step=0.1,
            value=0.2,
        ),
        gr.Slider(
            label="Top-p (nucleus sampling)",
            minimum=0.05,
            maximum=1.0,
            step=0.05,
            value=0.9,
        ),
        gr.Slider(
            label="Top-k",
            minimum=1,
            maximum=1000,
            step=1,
            value=50,
        ),
        gr.Slider(
            label="Repetition penalty",
            minimum=1.0,
            maximum=2.0,
            step=0.05,
            value=1.2,
        ),
    ],
    stop_btn=None,
    examples=[
        ["Describe the significance of the Eiffel Tower in French culture and history."],
        ["Что такое 'загадочная русская душа' и как это понятие отражается в русской литературе?"],  # Russian: What is the "mysterious Russian soul" and how is this concept reflected in Russian literature?
        ["Jakie są najbardziej znane polskie tradycje bożonarodzeniowe?"],  # Polish: What are the most well-known Polish Christmas traditions?
        ["Welche Rolle spielte die Hanse im mittelalterlichen Europa?"],  # German: What role did the Hanseatic League play in medieval Europe?
        ["日本の茶道の精神と作法について説明してください。"]  # Japanese: Please explain the spirit and etiquette of Japanese tea ceremony.
    ],
)

with gr.Blocks(css="style.css") as demo:
    chat_interface.render()

if __name__ == "__main__":
    demo.queue(max_size=20).launch()