Installation

Basics

First install Python >= 3.10, PyTorch >=v.2.0.0, git and conda. We tested the latest JoeyS2T with

• python 3.11

• torch 2.1.2

• torchaudio 2.1.2

• cuda 12.1

Create and activate a conda virtual environment to install the package into:

$ conda -n js2t python=3.11
$ conda activate js2t

Cloning

Then clone JoeyS2T from GitHub and switch to its root directory:

(js2t)$ git clone https://github.com/may-/joeys2t.git
(js2t)$ cd joeys2t

Note

For Windows users, we recommend to doublecheck whether txt files (i.e. test/data/toy/*) have utf-8 encoding.

Installing JoeyNMT

Install JoeyNMT and its requirements:

(js2t)$ python -m pip install -e .

Run the unit tests to make sure your installation is working:

(js2t)$ python -m unittest

Important

When running on GPU you need to manually install the suitable PyTorch version for your CUDA version. For example, you can install PyTorch 2.1.2 with CUDA v12.1 as follows:

$ python -m pip install --upgrade torch==2.1.2 --index-url https://download.pytorch.org/whl/cu121

This is described in the PyTorch installation instructions.

Note

You may need to install extra dependencies (torchaudio backends): ffmpeg, sox, soundfile, etc. Please refer torchaudio documentation for details.

You’re ready to go!