Skip to content

A versatile tool for visualizing entropy loss in TensorFlow-based neural network training, providing insightful scatter plots with annotations.

Notifications You must be signed in to change notification settings

FlyingFathead/neurograph-framework

Repository files navigation

neurograph-framework

neurograph-framework

neurograph-framework is a versatile tool designed to help you visualize entropy loss in TensorFlow-based neural network training. It generates insightful scatter plots with annotations to aid in understanding and analyzing your training progress.

Features

  • Creates a line out of average entropy losses along with scatter plots that display entropy loss over single training iterations. Useful for tracking per-iteration scatter and underlying model trends during training / fine-tuning.
  • Annotations for minimum and maximum loss values as well as per-iteration scatter.
  • Indication of the latest iteration number and average loss value.
  • Overlay warnings in case of missing or outdated data.
  • Customizable to suit various types of iteration data, suitable for all kinds of visualization purposes.
  • Intended to visualize entropy losses as effectively as possible (min/max lines, median, per-iteration scatter etc).

Usage

  1. Clone this repository to your local machine:
git clone https://github.com/FlyingFathead/neurograph-framework/
  1. Navigate to the cloned directory:
cd neurograph-framework/
  1. Install the PyPi requirements with pip install -r requirements.txt

(or, make sure you have these installed):

matplotlib>=3.5.1
Pillow>=9.1.0
numpy>=1.23.5
  1. Run the audit_subprocess.py script to start visualizing your neural network training data:
python audit_subprocess.py setname logs_directory

The plotter graphs are updated every 20 seconds by default.

Happy training and analyzing with neurograph-framework! 📊🧠

About

A terminal-based version of the framework (i.e. for headless training setups): neurograph

My other projects are at: github.com/FlyingFathead/

Releases

No releases published

Packages

No packages published

Languages