Template with our basic structure
The repository should follow the next structure:
|_ package_name/
| |_ model/
|_ your_model_script.py
|_ other_code
|_ docs/
|_ fluid/
|_ tests/
|_ Dockerfile
|_ gpu.Dockerfile
|_ Pipfile
|_ Pipefile.lock
|_ setup.py
|_ setup.cfg
In case of implementing an AI model using the framework, be sure to follow the same structure in both the repository
and the apiCall
so the model is loaded correctly.
As an example:
package-name.model.your_model_script.py
The repository works using pipenv to separate development from production packages. To install then:
pipenv install
or
pipenv install --dev
Pipenv will create an environment, uses pip shell
to activate it.
To install individual packages use: pipenv install xxx
so it will be directly
added to the Pipfile.
To run test use: pytest -vv tests
Run the following command to deploy to the desired environment {env}
(which can be ces, integration, staging or production)
fluid-deploy -c fluid/config.json -c fluid/config.{env}.json
If still in development and testing, use -d
so it doesn't require committing changes.
The docker image will be tagged with: dev-yourGithubUser
.
To create a python package, execute the following commands:
python setup.py ${version} sdist bdist_wheel
twine upload --repository-url http://209.133.199.50:8090 dist/* --verbose -u "$USER" -p "$PASS"