import gradio as gr
from transformers import pipeline

translator = pipeline("translation_en_to_fr")

def translate(text):
result = translator(text)
return result[0]['translation_text']

iface = gr.Interface(fn=translate, inputs="text", outputs="text")
iface.launch()

import gradio as gr
from transformers import pipeline

Load NLLB distill model (English ↔ many, including Malayalam)

If you prefer mBART‑50, change model to "facebook/mbart-50-many-to-many-mmt"

translator = pipeline(
"translation",
model="facebook/nllb-200-distilled-600M",
src_lang="eng_Latn",
tgt_lang="mal_Deva",
max_length=400,
)

LANGS = {
"English → Malayalam": ("eng_Latn", "mal_Deva"),
"Malayalam → English": ("mal_Deva", "eng_Latn"),
}

def translate_text(text, direction):
src, tgt = LANGS[direction]
# Re‑configure pipeline for the chosen direction
t = pipeline(
"translation",
model="facebook/nllb-200-distilled-600M",
src_lang=src,
tgt_lang=tgt,
max_length=400,
num_beams=4,
early_stopping=True,
)
result = t(text)
return result[0]["translation_text"]

demo = gr.Interface(
fn=translate_text,
inputs=[
gr.Textbox(label="Enter text", placeholder="Type here..."),
gr.Radio(
list(LANGS.keys()),
label="Translation direction",
value="English → Malayalam",
),
],
outputs=gr.Textbox(label="Translated text"),
title="English ↔ Malayalam Translator",
description="Translate between English and Malayalam using NLLB‑200.",
examples=[
[
"Hello, how are you?",
"English → Malayalam"
],
[
"ഞാൻ നല്ലതുപോലെ ഇരിക്കുന്നു.",
"Malayalam → English"
],
],
cache_examples=False,
)

demo.launch()