Spaces:
Running
Running
| # -*- coding: utf-8 -*- | |
| """app.py | |
| Automatically generated by Colab. | |
| Original file is located at | |
| https://colab.research.google.com/drive/1UBKp9Q5WTAiaXvsRK_6KupUXBI3i05wH | |
| """ | |
| from transformers import AutoModelForSeq2SeqLM, AutoTokenizer | |
| import gradio as gr | |
| # Dictionary mapping target languages to Helsinki models | |
| LANG_MODELS = { | |
| "German": "Helsinki-NLP/opus-mt-en-de", | |
| "French": "Helsinki-NLP/opus-mt-en-fr", | |
| "Spanish": "Helsinki-NLP/opus-mt-en-es", | |
| "Italian": "Helsinki-NLP/opus-mt-en-it", | |
| "Japanese": "Helsinki-NLP/opus-mt-en-jap", | |
| } | |
| # Preload models for speed (optional, you can also load on demand) | |
| model_cache = {} | |
| tokenizer_cache = {} | |
| def translate_text_multi(text, target_lang): | |
| model_name = LANG_MODELS.get(target_lang) | |
| if not model_name: | |
| return "Error: language not supported." | |
| # Load model/tokenizer if not already loaded | |
| if model_name not in model_cache: | |
| tokenizer_cache[model_name] = AutoTokenizer.from_pretrained(model_name) | |
| model_cache[model_name] = AutoModelForSeq2SeqLM.from_pretrained(model_name) | |
| tokenizer = tokenizer_cache[model_name] | |
| model = model_cache[model_name] | |
| inputs = tokenizer(text, return_tensors="pt") | |
| outputs = model.generate(**inputs) | |
| translated = tokenizer.decode(outputs[0], skip_special_tokens=True) | |
| return translated | |
| # Gradio UI | |
| with gr.Blocks() as demo: | |
| gr.Markdown("## 🌐 Multi-language Translator") | |
| input_text = gr.Textbox(label="Enter English text", lines=4) | |
| target_language = gr.Dropdown(choices=list(LANG_MODELS.keys()), value="German", label="Target Language") | |
| output_text = gr.Textbox(label="Translation", lines=4) | |
| translate_btn = gr.Button("Translate") | |
| translate_btn.click(fn=translate_text_multi, inputs=[input_text, target_language], outputs=output_text) | |
| demo.launch() |