ikkuna 0.1.0

Ikkuna Neural Network Monitor

<p align=”center”> <img src=”./logo.png” alt=”logo” width=”100”/> </p>

# Ikkuna A tool for monitoring neural network training.

—

Ikkuna provides a framework for adding live training metrics to your PyTorch model with minimal configuration. It is a PubSub framework which allows practitioners to quickly test metrics implemented against a simple API. The following data is provided

• Activations

• Gradients w.r.t weights and biases

• Gradients w.r.t layer outputs

• Weights

• Biases

• Weight updates

• Bias updates

• Metadata such as current step in the training, current labels and current perdictions

Subscribers consume this data and distill it into metrics. Different backends can be used

• Matplotlib

• Tensorboard

# Working with this repository

You should create a conda envorinment from the provided torch.yaml file and pip install -r the provided requirements.txt file. You will also have to install numba for building the documentation until I have the time to figure out how to optionally turn off parts of a doc build.

You should also run python setup.py develop which will install the package with symlinks to this repository. Since all subscribers are setuptools plugins, they are not available unless setup.py is run.

## Documentation The sphinx-generated html documentation is hosted [here](https://peltarion.github.io/ai_ikkuna/).

## Working with the repository/notebooks 1. Clone the repository. 1. cd into the repository. 1. Tell git where to find the configuration information for the iPython Notebooks with this command: git config –add include.path $(pwd)/.gitconfig (The path needs to point to your root git repository where the .gitconfig is stored).

### Adding a new notebook 1. Create a new Jupyter Notebook. 1. Hit Edit -> Edit Notebook Metadata. 1. Add “git”: { “suppress_outputs”: true }, as a top level element to the json metadata. This will be a notification to the git filter that we want to strip the metadata.