magolor / mlhp

Machine Learning Helper Package (pytorch)

Machine Learning Helper Package

Warning: This package is currently underdeveloped and does not guarantee any consistency.

Package Introduction

Package Contents

utils

file_helpers: helper functions for file manipulation, including:

• Partially integrated & improved os and pathlib functions.
• Loading & saving in json and pickle.

view_helpers: helper functions for console debugging output, including:

• Length-restricting view of objects supporting multiple formats.
• Output stream and color setting.
• TQDM wrapping.

Serializable: the essence of mlhp classes: all classes should be designed to be serializable.

log_helpers: a custom logger that does not rely on logging handle (and thus does not conflict with DDP training, for details about this issue, please refer to this post). Notice that we are considering moving this to a new independent Logger module.

helpers: general helper functions, including:

• Daily-used system functions (e.g.: CMD, IP).
• In-program package installation throught PIP and CONDA.
• Miscellaneous

Plotter

A matplotlib.pyplot handle with with wrapping.

Timer

A Timer.

Tracker

A curve tracker supporting updating, finding improvements, logging and plotting.

(developing... )

ModelWrapper

A general model wrapper for arbitrary network, training, evaluating and testing pipeline.

(developing... )