This repo is intended to host a collection of Logan Martel, ML Developer's, personal utilities for ML projects.
- Add basic plotting utilities
- Add basic data io utilities
- Add basic data formatting utilities
- ...
Simply install the latest release from pypi to your active python environment
(equipped with pip
and ideally managed by conda) via: pip install los-ml-tools.
-
Make sure you have all the dependencies needed to build this project. In particular, immediately after git clone, you should be able to successfully build the project from a command-line environment under Python 2.7 via the command:
python setup.py develop -
The recommended IDE for developer consistency is PyCharm. Leveraging PyCharm, it would be ideal for project contributors to ensure that code conforms to PEP 8 standards.
(RECOMMENDATION: Ideally, it is advised to manage Python virtual environments via conda in order to safely segregate module dependencies. In this case, it is recommended to locallize pip installations during conda environment creation, to avoid dependency conflicts, by instantiating the environment with its own pip setup, à la conda create --name custom_venv_name pip. Alternatively, a basic conda venv, equipped with its own localized pip, can be configured from this project directory simply by running conda env create -f environment.yml and then activating the environment (linux example: source activate los-ml-tools)).
Our development workflow will be simple:
Always develop on your own branch. Only merge code to Master via PR's. Only merge code that has no conflicts with master. Only merge code that builds successfully via python setup.py develop.
To leverage the module, simply import any utilities of interest. For example, to apply a plotting utility:
>>> from los_ml_tools import plotting_utils as plotting
>>> plotting.add_single_fold_prc_to_figure(0.95, 0.80, 'blue', 1)
>>> plotting.output_one_fold_prc_roc_results('/tmp/', 'target_class')This code is under the GNU General Public License v3.0.
If you use or modify los-ml-tools, please credit the original author as
- Logan Martel - https://github.com/martelogan