app.py

import gradio as gr
import time
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
from flores200_codes import flores_codes


def load_models():
    # build model and tokenizer
    model_name_dict = {
        "nllb-distilled-600M": "facebook/nllb-200-distilled-600M",
        "nllb-distilled-1.3B": "facebook/nllb-200-distilled-1.3B",
        # "nllb-1.3B": "facebook/nllb-200-1.3B",
        # "nllb-3.3B": "facebook/nllb-200-3.3B",
    }

    model_dict = {}

    for call_name, real_name in model_name_dict.items():
        print("\tLoading model: %s" % call_name)
        model = AutoModelForSeq2SeqLM.from_pretrained(real_name)
        tokenizer = AutoTokenizer.from_pretrained(real_name)
        model_dict[call_name + "_model"] = model
        model_dict[call_name + "_tokenizer"] = tokenizer

    return model_dict


def translation(model_name, source, target, text):
    start_time = time.time()
    source = flores_codes[source]
    target = flores_codes[target]

    model = model_dict[model_name + "_model"]
    tokenizer = model_dict[model_name + "_tokenizer"]

    translator = pipeline(
        "translation",
        model=model,
        tokenizer=tokenizer,
        src_lang=source,
        tgt_lang=target,
    )

    # sentence-wise translation
    sentences = text.split("\n")
    translated_sentences = []
    for sentence in sentences:
        translated_sentence = translator(sentence, max_length=400)[0][
            "translation_text"
        ]
        translated_sentences.append(translated_sentence)
    output = "\n".join(translated_sentences)

    end_time = time.time()

    # output = translator(text, max_length=400)
    # full_output = output
    # output = output[0]["translation_text"]
    result = {
        "inference_time": end_time - start_time,
        "source": source,
        "target": target,
        "result": output,
        # "full_output": full_output,
    }
    return result, output


if __name__ == "__main__":
    print("\tinit models")

    global model_dict

    model_dict = load_models()

    # define gradio demo
    lang_codes = list(flores_codes.keys())
    inputs = [
        gr.inputs.Radio(
            [
                "nllb-distilled-600M",
                "nllb-distilled-1.3B",
                # "nllb-1.3B",
                # "nllb-3.3B"
            ],
            label="NLLB Model",
            default="nllb-distilled-1.3B",
        ),
        gr.inputs.Dropdown(lang_codes, default="Najdi Arabic", label="Source"),
        gr.inputs.Dropdown(lang_codes, default="English", label="Target"),
        gr.inputs.Textbox(lines=5, label="Input text"),
    ]

    outputs = [
        gr.outputs.JSON(label="Metadata"),
        gr.outputs.Textbox(label="Output text"),
    ]

    title = "NLLB (No Language Left Behind) demo"

    demo_status = "Demo is running on CPU"
    description = f"""Using NLLB model, details: https://github.com/facebookresearch/fairseq/tree/nllb.

    {demo_status}"""
    examples = [
        ["nllb-distilled-1.3B", "Najdi Arabic", "English", "جلست اطفال"],
        [
            "nllb-distilled-600M",
            "Najdi Arabic",
            "English",
            "شد للبيع طابقين مع شرع له نظيف حق غمارتين",
        ],
    ]

    gr.Interface(
        translation,
        inputs,
        outputs,
        title=title,
        description=description,
        examples=examples,
        examples_per_page=50,
    ).launch()