PyProject Starter

PyProject Starter is a Python project starter kit with docker services and streamlit UI.

This starter builds on top of Timothy Helton's Pyproject Generator: https://github.com/TimothyHelton/pyproject_generator

Getting Started With the New Project

is a fully functioning Python package that may be installed using pip. Docker Images are built into the package and a Makefile provides an easy to call repetitive commands.

PyProject Starter Structure

• applications: Applications directory where new apps can be added
  • streamlit: Streamlit service with sample pages (missing test atm)
    • subpages: Sample subpages for a Streamlit app
• docker: Docker directory
  • mongo_init: Folder with mongo init related files
  • Dockerfile: Dockerfiles for building Docker container images
  • docker-compose.yaml: Yaml file used by Docker Compose to define the services, networks, and volumes for a multi-container application
• docs: Folder used by sphinx for auto-documentation
• sandbox_ai_tutorial: Project main script directory - additional apps are added here
  • configs: Project configuration files
  • test: Project unit tests
• scripts: Folder with setup related scripts

Docker Naming Convention

Service: <project_name> - sandbox_ai_tutorial_python Container: <user_name><project_name> - mthnguyener_sandbox_ai_tutorial_python Image: <project_name> - sandbox_ai_tutorial_python

Makefile Code Completion

It's handy to have code completion when calling targets from the Makefile. To enable this feature add the following to your user profile file.

• On Ubuntu this would be your ~/.profile file.
• On a Mac this would be you ~/.bash_profile file.

complete -W "`grep -oE '^[a-zA-Z0-9_.-]+:([^=]|$)' Makefile | sed 's/[^a-zA-Z0-9_.-]*$//'`" make

Clone the Repository

First, make a local copy of the project. After setting up SSH keys on GitHub call the following command to clone the repository.

git clone <enter_path_to_repo>.git

A directory called sandbox_ai_tutorial will be created where the command was executed. This sandbox_ai_tutorial directory will be referred to as the "package root directory" throughout the project.

Setting Up New Project

1. From current project root directory, run:
   • make new-project
     • Enter the new project name: <project_name>
2. Current project directories and files are created in the new project directory
   • new_project/
3. The new project is created 1-level up from the current project root directory
   • if current project directory is projects/sandbox_ai_tutorial then the new project is created at projects/new_project
4. To add to git:
   • git init
   • git add <new_project_files_and_directories>
   • git commit -m "first commit or any comments"
   • git branch -M main
   • git remote add origin https://github.com/<user_or_organization>/<project>.git
   • git push -u origin main

Initialize the Project

Some functionality of the package is created locally. Run the following command from the package root directory to finish setting up the project.

make getting-started

Jupyter Notebooks

While Jupyter notebooks are not ideal for source code, they can be powerful when applied to path finding and creating training material. The sandbox_ai_tutorial project is capable of creating a Jupyter server in the Python container. Since the package root directory is mounted to the Docker container any changes made on the client will persist on the host and vice versa. For consistency when creating notebooks please store them in the notebooks directory. Call the following commands from the package root directory to start and stop the Jupyter server.

Create a Notebook Server

make notebook

Shutdown a Notebook Server

make notebook-stop-server

Test Framework

The is configured to use the pytest test framework in conjunction with coverage and the YAPF style linter. To run the tests and display a coverage report call the following command from the package root directory.

Test Coverage

make test-coverage

To only run the tests, and not display the coverage, call the following.

Tests

make test

Update Style

To only run the YAPF style linter call this command from the package root directory.

make format-style

Dependencies

Since the sandbox_ai_tutorial utilizes NVIDIA optimized Docker images most of the Python dependencies could be installed using PIP or Conda. The requirements.txt file contains a reference to the specific base image used during development and a list of dependencies.

There is a make target to update the requirements file.

make package-dependencies

Documentation

The package also has an NGINX container to host interactive documentation. Calling the following commands from the package root directory will result in a local web browser displaying the package HTML documentation.

Build Documentation

make docs

View Documentation without Building

make docs-view

Profilers

Before refactoring it's usually a great idea to profile the code. The following methods describe the profilers that are available in the sandbox_ai_tutorial environment, and how to use them.

SNAKEVIZ Execution

To test an entire script just enter the following from the project root directory.

Profile Script

make snakeviz PROFILE_PY=script.py

Memory Profiler

1. Open Jupyter Notebook
2. Load Extension
   • %load_ext memory_profiler
3. Run profiler
   • %memit enter_code_here