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,
    history: list[tuple[str, str]],
    system_message: str,
    max_tokens: int = 128,
    temperature: float = 1.0,
    top_p: float = 1.0,
):
    # Preprocess the input message
    input_text = "translate English to French: " + 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
    for token in streamer:
        yield token


"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)


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