app.py

import gradio as gr
from transformers import AutoModelForSeq2SeqLM, NllbTokenizerFast
import torch

# Load model
model_id = "ClaudBarbara/Open_Access_Khmer"
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
tokenizer = NllbTokenizerFast.from_pretrained(model_id)

def translate(text, direction):
    if direction == "English to Khmer":
        src_lang, tgt_lang = "eng_Latn", "khm_Khmr"
    else:
        src_lang, tgt_lang = "khm_Khmr", "eng_Latn"
    
    tokenizer.src_lang = src_lang
    inputs = tokenizer(text, return_tensors="pt", max_length=512, truncation=True)
    
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            forced_bos_token_id=tokenizer.convert_tokens_to_ids(tgt_lang),
            max_length=512,
            num_beams=4
        )
    
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

demo = gr.Interface(
    fn=translate,
    inputs=[
        gr.Textbox(label="Input Text", lines=5),
        gr.Radio(["English to Khmer", "Khmer to English"], label="Direction", value="English to Khmer")
    ],
    outputs=gr.Textbox(label="Translation", lines=5),
    title="Khmer Legal Bridge",
    description="English-Khmer Legal Translation"
)

demo.launch()