app.py

import gradio as gr
from huggingface_hub import InferenceClient
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Load both translation models from Hugging Face
# English to Moroccan Arabic (Darija)
tokenizer_eng_to_darija = AutoTokenizer.from_pretrained("Saidtaoussi/AraT5_Darija_to_MSA")
model_eng_to_darija = AutoModelForSeq2SeqLM.from_pretrained("Saidtaoussi/AraT5_Darija_to_MSA")

# Moroccan Arabic (Darija) to Modern Standard Arabic (MSA)
tokenizer_darija_to_msa = AutoTokenizer.from_pretrained("lachkarsalim/Helsinki-translation-English_Moroccan-Arabic")
model_darija_to_msa = AutoModelForSeq2SeqLM.from_pretrained("lachkarsalim/Helsinki-translation-English_Moroccan-Arabic")

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
    translation_choice: str,
):
    """
    Responds to the input message by selecting the translation model based on the user's choice.
    """
    messages = [{"role": "system", "content": system_message}]

    for val in history:
        if val[0]:  # User message
            messages.append({"role": "user", "content": val[0]})
        if val[1]:  # Assistant message
            messages.append({"role": "assistant", "content": val[1]})

    # Append the user message
    messages.append({"role": "user", "content": message})

    # Initialize the response variable
    response = ""

    # Translate based on the user's choice
    if translation_choice == "Moroccan Arabic to MSA":
        # Translate Moroccan Arabic (Darija) to Modern Standard Arabic
        inputs = tokenizer_darija_to_msa(message, return_tensors="pt", padding=True)
        outputs = model_darija_to_msa.generate(inputs["input_ids"], num_beams=5, max_length=512, early_stopping=True)
        response = tokenizer_darija_to_msa.decode(outputs[0], skip_special_tokens=True)

    elif translation_choice == "English to Moroccan Arabic":
        # Translate English to Moroccan Arabic (Darija)
        inputs = tokenizer_eng_to_darija(message, return_tensors="pt", padding=True)
        outputs = model_eng_to_darija.generate(inputs["input_ids"], num_beams=5, max_length=512, early_stopping=True)
        response = tokenizer_eng_to_darija.decode(outputs[0], skip_special_tokens=True)

    return response


# Gradio interface setup
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)"),
        gr.Dropdown(
            label="Choose Translation Direction",
            choices=["English to Moroccan Arabic", "Moroccan Arabic to MSA"],
            value="English to Moroccan Arabic"
        ),
    ],
)

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