app.py

import gradio as gr
from huggingface_hub import InferenceClient
import os

"""
Copied from inference in colab notebook
"""

from transformers import pipeline

# Load model and tokenizer globally to avoid reloading for every request
model_path = "Mat17892/t5small_enfr_opus"

# translator = pipeline("translation_xx_to_yy", model=model_path)

# def respond(
#     message: str,
#     history: list[tuple[str, str]],
#     system_message: str,
#     max_tokens: int,
#     temperature: float,
#     top_p: float,
# ):
#     message = "translate English to French:" + message

#     response = translator(message)[0]
#     yield response['translation_text']

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, TextIteratorStreamer
import threading

tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForSeq2SeqLM.from_pretrained(model_path)

def respond(
    message: str,
    system_message: str,
    max_tokens: int = 128,
    temperature: float = 1.0,
    top_p: float = 1.0,
):
    # Preprocess the input message
    input_text = system_message + " " + message
    input_ids = tokenizer(input_text, return_tensors="pt").input_ids

    # Set up the streamer
    streamer = TextIteratorStreamer(tokenizer, skip_special_tokens=True)

    # Generate in a separate thread to avoid blocking
    generation_thread = threading.Thread(
        target=model.generate,
        kwargs={
            "input_ids": input_ids,
            "max_new_tokens": max_tokens,
            "do_sample": True,
            "temperature": temperature,
            "top_p": top_p,
            "streamer": streamer,
        },
    )
    generation_thread.start()

    # Stream the output progressively
    generated_text = ""
    for token in streamer:  # Append each token to the accumulated text
        generated_text += token
        yield generated_text


"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""

# Define the interface
with gr.Blocks() as demo:
    gr.Markdown("# Google Translate-like Interface")

    with gr.Row():
        with gr.Column():
            source_textbox = gr.Textbox(
                placeholder="Enter text in English...",
                label="Source Text (English)",
                lines=5,
            )
        with gr.Column():
            translated_textbox = gr.Textbox(
                placeholder="Translation will appear here...",
                label="Translated Text (French)",
                lines=5,
                interactive=False,
            )

    translate_button = gr.Button("Translate")

    with gr.Accordion("Advanced Settings", open=False):
        system_message_input = gr.Textbox(
            value="translate English to French:",
            label="System message",
        )
        max_tokens_slider = gr.Slider(
            minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"
        )
        temperature_slider = gr.Slider(
            minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"
        )
        top_p_slider = gr.Slider(
            minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"
        )

    # Define functionality
    translate_button.click(
        respond,
        inputs=[
            source_textbox,
            system_message_input,
            max_tokens_slider,
            temperature_slider,
            top_p_slider,
        ],
        outputs=translated_textbox,
    )

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