Your first step is to install the Core Engine, which comes bundled as a SDK and CLI in a good old
pip package. Please note that we only support Python >= 3.5, so adjust your pip accordingly.
Also, we highly encourage you to install the Core Engine on a virtual environment. We install dependencies like
Tensorflow that might cause your base installations to be overridden.
We like to use virtualenvwrapper to manage our Python virtual environments.
When you're set with your environment, run:
pip install cengine
This would install both the CLI and Python SDK to your machine. Once the installation is completed, you can check whether the installation was successful through:
If you would like to learn more about the current release, please visit the PyPi homepage.
The Core Engine handles orchestration of the workloads created by the platform, but the actual execution happens on your infrastructure. To set this up, we have detailed instructions for Google Cloud Platform, Amazon Web Services and Microsoft Azure. Please follow the corresponding guides for your preferred setup.
Deploying on a private cloud is also possible. Please contact us for more details!
In order to get the Tensoflow Model Analysis evaluation visualizations to work, you must also run:
jupyter nbextension install --py --symlink tensorflow_model_analysisjupyter nbextension enable --py tensorflow_model_analysis
:::important You might be required to run this as
sudo if you're not working on a virtualenv. :::
:::warning If you encounter a
File already exists in database error error after the first command, this is most likely due to a known bug with one of our dependencies, namely PyArrow. Unfortuanetely, this will cause some features to not work on your machine (namely pipeline evaluate and compare).
We are aware of this issue and are working hard to fix it. The next release (> 0.0.15) of the cengine CLI will fix this issue. :::
For Bash, add this to ~/.bashrc:
eval "$(_CENGINE_COMPLETE=source_bash cengine)"
For Zsh, add this to ~/.zshrc:
eval "$(_CENGINE_COMPLETE=source_zsh cengine)"
For Fish, add this to ~/.config/fish/completions/foo-bar.fish:
eval (env _CENGINE_COMPLETE=source_fish cengine)