Project template for a microcosm-based ML service using SageMaker.

1. Install cookiecutter:

   pip install cookiecutter
   

2. Change to an appropriate working directory.

3. Run cookiecutter against this repository:

   cookiecutter https://github.com/globality-corp/cookiecutter-microcosm-sagemaker
   

4. Answer the questions.

5. Change into the {{ repository_name }} directory

6. [Globality only] Run the globality-build setup steps.

7. Initialize a new git project:

   git init
   git add .
   git commit -m "Initial commit"
   

When experimenting with changes to this cookiecutter, you may find the scripts/instantiate.sh script to be useful. It will instantiate the cookiecutter with the default parameters, first removing any previous instantiations. You can then cd into the instantiated directory to run linting and unit tests.

Below describes one possible local development flow.

Open two shells. In the first shell, no virtualenv is active, and the cwd is the top level of this cookiecutter repo. Run

./scripts/instantiate.sh

In the second shell,

cd ml-papaya-extractor
pyenv virtualenv 3.7.2 ml-papaya-extractor
pyenv shell ml-papaya-extractor
pip install flake8 flake8-print flake8-logging-format flake8-isort flake8-quotes mypy

During development, make all changes to the cookiecutter template, not to the instantiated test directory. Then every time you'd like to test a change, in the first shell, run

./scripts/instantiate.sh

Then in the second shell, to run unit tests, run

cd .. && cd ml-papaya-extractor && pip install -e '.[test]' && python setup.py nosetests

The initial cd commands are because the instantiate step above clobbers the ml-papaya-extractor directory.

Similarly, to run linting, run

cd .. && cd ml-papaya-extractor && flake8 .