app.py

import sys
import time

from importlib.metadata import version

import torch
import gradio as gr

from transformers import MBartForConditionalGeneration, AutoTokenizer

# Config
model_name = "/home/user/app/mbart-large-50-verbalization"
concurrency_limit = 5

device = "cuda" if torch.cuda.is_available() else "cpu"

# Load the model
model = MBartForConditionalGeneration.from_pretrained(
    model_name,
    low_cpu_mem_usage=True,
    device_map=device,
)
model.eval()

tokenizer = AutoTokenizer.from_pretrained(model_name)
tokenizer.src_lang = "uk_XX"
tokenizer.tgt_lang = "uk_XX"

examples = [
    "WP: F-16 навряд чи значно змінять ситуацію на полі бою",
    "Над Україною збили ракету та 7 з 8 Шахедів",
    "Олімпійські ігри-2024. Розклад змагань українських спортсменів на 28 липня",
    "Кампанія Гарріс менш як за тиждень зібрала понад $200 млн",
    "За тиждень Нацбанк продав майже 800 мільйонів доларів на міжбанку",
    "Париж-2024. День 2. Текстова трансляція",
]

title = "Normalize Text for Ukrainian"

# https://www.tablesgenerator.com/markdown_tables
authors_table = """
## Authors

Follow them on social networks and **contact** if you need any help or have any questions:

| <img src="https://avatars.githubusercontent.com/u/7875085?v=4" width="100"> **Yehor Smoliakov** |
|-------------------------------------------------------------------------------------------------|
| https://t.me/smlkw in Telegram                                                                  |
| https://x.com/yehor_smoliakov at X                                                              |
| https://github.com/egorsmkv at GitHub                                                           |
| https://huggingface.co/Yehor at Hugging Face                                                    |
| or use egorsmkv@gmail.com                                                                       |
""".strip()

description_head = f"""
# {title}

## Overview

This space uses https://huggingface.co/skypro1111/mbart-large-50-verbalization model.

Paste the text you want to enhance.
""".strip()

description_foot = f"""
{authors_table}
""".strip()

normalized_text_value = """
Normalized text will appear here.

Choose **an example** below the Normalize button or paste **your text**.
""".strip()

tech_env = f"""
#### Environment

- Python: {sys.version}
""".strip()

tech_libraries = f"""
#### Libraries

- torch: {version('torch')}
- gradio: {version('gradio')}
- transformers: {version('transformers')}
""".strip()


def inference(text, progress=gr.Progress()):
    if not text:
        raise gr.Error("Please paste your text.")

    gr.Info("Starting normalizing", duration=2)

    progress(0, desc="Normalizing...")

    results = []

    sentences = [
        text,
    ]

    for sentence in progress.tqdm(sentences, desc="Normalizing...", unit="sentence"):
        sentence = sentence.strip()

        if len(sentence) == 0:
            continue

        t0 = time.time()

        input_text = "<verbalization>:" + sentence

        encoded_input = tokenizer(
            input_text,
            return_tensors="pt",
            padding=True,
            truncation=True,
            max_length=1024,
        ).to(device)
        output_ids = model.generate(
            **encoded_input, max_length=1024, num_beams=5, early_stopping=True
        )
        normalized_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)

        if not normalized_text:
            normalized_text = "-"

        elapsed_time = round(time.time() - t0, 2)

        normalized_text = normalized_text.strip()
        results.append(
            {
                "sentence": sentence,
                "normalized_text": normalized_text,
                "elapsed_time": elapsed_time,
            }
        )

    gr.Info("Finished!", duration=2)

    result_texts = []

    for result in results:
        result_texts.append(f'> {result["normalized_text"]}')
        result_texts.append("\n")

    sum_elapsed_text = sum([result["elapsed_time"] for result in results])
    result_texts.append(f"Elapsed time: {sum_elapsed_text} seconds")

    return "\n".join(result_texts)


demo = gr.Blocks(
    title=title,
    analytics_enabled=False,
    # theme="huggingface",
    theme=gr.themes.Base(),
)

with demo:
    gr.Markdown(description_head)

    gr.Markdown("## Usage")

    with gr.Row():
        text = gr.Textbox(label="Text", autofocus=True, max_lines=1)
        normalized_text = gr.Textbox(
            label="Normalized text",
            placeholder=normalized_text_value,
            show_copy_button=True,
        )

    gr.Button("Normalize").click(
        inference,
        concurrency_limit=concurrency_limit,
        inputs=text,
        outputs=normalized_text,
    )

    with gr.Row():
        gr.Examples(label="Choose an example", inputs=text, examples=examples)

    gr.Markdown(description_foot)

    gr.Markdown("### Gradio app uses:")
    gr.Markdown(tech_env)
    gr.Markdown(tech_libraries)

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