app.py

import torch
import gradio as gr
from transformers import MarianMTModel, MarianTokenizer

model_name = "ywc1/marian-finetuned-ja-en"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)

def translate(text):
    if not text.strip():
        return "Please enter some 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 (no length limit)"),
    outputs="text",
    title="Japanese → English Academic Translator",
    description="Translation powered by a fine-tuned MarianMT model. This tool works best for sentence-by-sentence translation.",
    flagging_mode="manual",
    flagging_options=["Inaccurate", "Fluency", "Interesting", "Other"],
)

iface.launch()