import gradio as gr
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
import torch

TASK = "translation"
MODEL = "facebook/nllb-200-distilled-600M"
LANGS = ["Kinyarwanda","English","French","Swahili","Luganda","Lingala"]
LANGUAGES={
    "english": "eng_Latn",
    "kinyarwanda": "kin_Latn",
    "swahili": "swh_Latn",
    "luganda": "lug_Latn",
    "lingala": "lin_Latn",
    "french": "fra_Latn"
}

device = 0 if torch.cuda.is_available() else -1
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL)
tokenizer = AutoTokenizer.from_pretrained(MODEL)

def choose_language(language): 
    result = LANGUAGES.get(language.lower())
    if result:
        return result        
    return

def translate(text, source_lang, target_lang, max_length=400):
    """
    Translate text from source language to target language
    """
    src_lang = choose_language(source_lang)
    tgt_lang= choose_language(target_lang)
    if src_lang==None:
        return "Error: the source langage is incorrect"
    elif tgt_lang==None:
        return "Error: the target language is incorrect"

    translation_pipeline = pipeline(TASK,
                                    model=model,
                                    tokenizer=tokenizer,
                                    src_lang=src_lang,
                                    tgt_lang=tgt_lang,
                                    max_length=max_length,
                                    device=device)
    result = translation_pipeline(text)
    return result[0]['translation_text']


gradio_ui= gr.Interface(
    fn=translate,
    title="NLLB 600M Translation Demo",
    inputs= [
        gr.components.Textbox(label="Text"),
        gr.components.Dropdown(label="Source Language", choices=LANGS),
        gr.components.Dropdown(label="Target Language", choices=LANGS),
        gr.components.Slider(8, 400, value=400, step=8, label="Max Length")
    ],
    outputs=gr.outputs.Textbox(label="Translated text")
)

gradio_ui.launch()