ImmersiveL

ImmersiveL is a framework and models hub for connecting languages all over the world open and free.

Now, ImmersiveL app is a Chinese-English bidirectional translation framework based on Deepspeed. The primary structure is found within the app directory, composed of a Python 3.8+ environment with Flask, Deepspeed, and PyTorch.

The first model for now are trained on a bloomz model, its license can be found at here. The Apache License are licensed to the derived part of the model and other source code file in this repo.

🌐 Read in Chinese (中文)

Getting Started

1. Clone the Repository and Set Up the Environment

   We recommend using a Python 3.8+ and CUDA 11.8 environment.

   Start by cloning the ImmersiveL repository to your local machine:

   git clone https://github.com/immersive-translate/ImmersiveL.git

   After cloning, navigate to the app directory and install all the necessary packages listed in requirements.txt:

   cd ImmersiveL/app
   pip install -r requirements.txt

2. Run the Application

   If you're using a Linux environment, start the application using Deepspeed with the following command:

   deepspeed --num_gpus 1 app.py

   For Windows users:

   python app_win.py

   Once you see a message similar to * Running on [IP Address], it indicates that the application has successfully started.

3. Test the Deployment

   Now that your application is up and running, test it by translating from Chinese to English:

   curl -X POST -H "Content-Type: application/json" -d '{"text": "欧洲经济增长仍面临较大挑战", "task": "zh2en"}' http://localhost:7000/v1/translate

   If you get a translated result, it indicates that your deployment was successful.

API Endpoints

1. Immersive Translation (/v1/immersive_translate)

Request

• Method: POST
• Content-Type: application/json

{
    "source_lang": "zh-CN",
    "target_lang": "en",
    "text_list": [
        "这是一个测试句子",
        "欧洲经济增长仍面临较大挑战"
    ]
}

• source_lang: Source language code.
• target_lang: Target language code.
• text_list: An array of text strings to be translated.

Response

For the given request:

{
    "translations": [
        {
            "detected_source_lang": "zh-CN",
            "text": "This is a test sentence"
        },
        {
            "detected_source_lang": "zh-CN",
            "text": "Economic growth in Europe continues to face significant challenges"
        }
    ]
}

• translations: An array containing:
  • detected_source_lang: The detected language code of the translated text.
  • text: The translated text.

Language Codes

• zh-CN: Simplified Chinese
• en: English

2. Basic Translation (/v1/translate)

• Method: POST
• Content-Type: application/json

Request Parameters

• text: The text you wish to translate.
• task: Defines the translation direction. Use "zh2en" for Chinese to English, and "en2zh" for English to Chinese.

Response Parameters

• data:
  • translation: The translated output text.
  • info: Additional details about the request, including the model used, original text, etc.