app.py

import os
from threading import Thread
from typing import Iterator

import gradio as gr
import spaces
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer

MAX_MAX_NEW_TOKENS = 1024
DEFAULT_MAX_NEW_TOKENS = 256
MAX_INPUT_TOKEN_LENGTH = 1024

DESCRIPTION = """\
# Dicta-IL's dictalm2.0-instruct

dictalm2.0-instruct was introduced in [this Facebook post](https://www.facebook.com/groups/MDLI1/posts/2704204053076959//).

Please, check the [original model card](https://huggingface.co/dicta-il/dictalm2.0-instruct) and [their official blog post](https://dicta.org.il/dicta-lm) for more details.
You can see the other Hebrew models by Dicta-IL [here](https://huggingface.co/dicta-il)

"""

LICENSE = """
<p/>

---
A derivative work of [mistral-7b](https://mistral.ai/news/announcing-mistral-7b/) by Mistral-AI.
The model and space are released under the Apache 2.0 license
"""

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 = "dicta-il/dictalm2.0-instruct"    
    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", torch_dtype=torch.bfloat16, low_cpu_mem_usage=True)
    tokenizer_id = "dicta-il/dictalm2.0-instruct"
    tokenizer = AutoTokenizer.from_pretrained(tokenizer_id)
    tokenizer.use_default_system_prompt = False
        

@spaces.GPU
def generate(
    message: str,
    chat_history: list[tuple[str, str]],
    max_new_tokens: int = 1024,
    temperature: float = 0.6,
    top_p: float = 0.9,
    top_k: int = 50,
    repetition_penalty: float = 1.4,
) -> Iterator[str]:
    conversation = []
    for user, assistant in chat_history:
        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
    conversation.append({"role": "user", "content": message})

    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
    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=True,
    )
    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,
    chatbot=gr.Chatbot(rtl=True, show_copy_button=True),
    textbox=gr.Textbox(text_align = 'right', rtl = True),
    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=4.0,
            step=0.1,
            value=0.3,
        ),
        gr.Slider(
            label="Top-p (nucleus sampling)",
            minimum=0.05,
            maximum=1.0,
            step=0.05,
            value=0.3,
        ),
        gr.Slider(
            label="Top-k",
            minimum=1,
            maximum=1000,
            step=1,
            value=30,
        ),
        gr.Slider(
            label="Repetition penalty",
            minimum=1.0,
            maximum=2.0,
            step=0.05,
            value=1.4,
        ),
    ],
    stop_btn=None,
    examples=[
        ["מתכון לעוגת שוקולד:"],
        ["השלם את הסיפור הקצר הבא:\n האיש האחרון בעולם ישב לבד בחדרו, כשלפתע נשמעה"],
        ["מהי שפת התכנות פייתון?"],
        ["סכם בקצרה את העלילה של סינדרלה"],
        ["שאלה: מהי עיר הבירה של מדינת ישראל?\nתשובה:"],
        ["שאלה: אני ממש עייף, מה כדאי לי לעשות?\nתשובה:"],
    ],
)

with gr.Blocks(css="style.css") as demo:
    gr.Markdown(DESCRIPTION)
    gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
    chat_interface.render()
    gr.Markdown(LICENSE)

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