Sockeye requires:
There are several options for installing Sockeye and its dependencies. Below we list several alternatives and the corresponding instructions.
The easiest way to install is with pip:
> pip install sockeye
If you want to just use Sockeye without extending it, simply install it via
> pip install -r requirements/requirements.txt
> pip install .
after cloning the repository from git.
Developers will be better served by pointing $PYTHONPATH to the root of the git-cloned source.
In an Anaconda environment such as the one provided by the AWS DeepLearning AMI or Azure when using the Data Science Virtual Machine image, users only need to run the following line to install sockeye (on an instance without a GPU):
> conda create -n sockeye python=3.8
> source activate sockeye
> pip install sockeye --no-deps
In order to write training statistics to a Tensorboard event file for visualization, you can optionally install tensorboard
(pip install tensorboard). To visualize these, run the Tensorboard tool with
the logging directory pointed to the training output folder: tensorboard --logdir <model>
In general you can install all optional dependencies from the Sockeye source folder using:
> pip install '.[optional]'
After installation, command line tools such as sockeye-train, sockeye-prepare-data, sockeye-translate, sockeye-average and sockeye-embeddings are available. For example:
> sockeye-train <args>
Equivalently, if the sockeye directory is on your $PYTHONPATH, you can run the modules directly:
> python -m sockeye.train <args>