Skip to content

fantinuoli/Local-NMT

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Run a pre-trained Huggingface Machine Translation engine with UI on your local computer

Installation

You need to have Python (3.8 or 3.9), Tensorflow, Torch and any VirtualEnv tool (in this example pipenv) installed on your machine. Please refer to the respective installation pages should you have any problem during installation.

The following supposes you are using pipenv as virtual environment management, but it will work with any other (just substitute pipenvwith the usual way you install modules):

  • Clone the repository on your local machine (click on the code button and download as zip file. Unzip it in a known location.)
  • open Terminal/Shell
  • navigate to App directory
  • pipenv shell
  • pipenv install

This will install the required dependencies: flask, transformers, torch, sentencepiece. You can also install these modules manually with pipenv installand the name of the module.

Note: I tested this on a macOS with Intel processor. I have not been able to install all dependencies on macOS with the new ARM/M1 processors. On macOS you may need to install XCode (you can install it freely from App Store). On Windows the same procedure should work out-of-the-box.

Pre-trained translation models

Refer to the Huggingface page for a list of all pre-trained Helsinki MT models. See the language code of the model in the last two parts of the model name (for example opus-mt-it-en for Italian to English).

Run

  • modify the desired language combination directly in app.py, where source_langis the language code for the source and target_lang for the target language (as in the model name). The corresponding language model will be automatically downloaded.
  • open Terminal/Shell
  • navigate to App directory
  • pipenv shell
  • python app.py
  • open Browser and copy and paste URL indicated in prompt (typically: http://127.0.0.1:5000/)

User Interface

alt text

License

MIT

About

Deployable MT engine

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published