app.py

import torch
import gradio as gr
from transformers import MBartForConditionalGeneration, MBart50Tokenizer

model = MBartForConditionalGeneration.from_pretrained("ywc1/mbart-finetuned-ja-en-para")
tokenizer = MBart50Tokenizer.from_pretrained("ywc1/mbart-finetuned-ja-en-para")

def translate(text):
    if not text.strip():
        return "Please enter a block of text."
    
    inputs = tokenizer(text, return_tensors="pt", padding=True)

    with torch.no_grad():
        output = model.generate(
            **inputs,
            num_beams=2,
            # max_length=1024,
            early_stopping=True
        )

    return tokenizer.decode(output[0], skip_special_tokens=True)

# Gradio interface
iface = gr.Interface(
    fn=translate,
    inputs=gr.Textbox(lines=7, placeholder="Enter Japanese text"),
    outputs="text",
    title="Japanese → English Academic Translator",
    description="This is a fine-tuned MBart Model that serves to translate paragraphs of Japanese academic text into English. \n You may find the sentece-level translation tool ywc1/ja-en-trans also helpful.",
    flagging_mode="manual",
    flagging_options=["Inaccurate", "Fluency", "Interesting", "Other"],
)

iface.launch()