kaze 0.0.15

CLI for Managing The Data Dependency of Deep Learning Projects

What is Kaze?

Have you downloaded a supervised-learning code-block base, and wish someone has specified how to download the dataset needed?

Kaze is a CLI for managing dataset dependencies in ML projects. It is a simple, yet powerful tool for managing dataset dependencies in your project. It is designed to be used in a way that is similar to yarn or npm, but with a focus on the dataset dependency.

Installation

pip install kaze

Example Usage

To download a dataset, you can use the following command:

kaze add -n flowers https://www.robots.ox.ac.uk/\~vgg/data/flowers/102/102flowers.tgz --images $DATASETS/jpg

this will populate your .kaze.yml file with the following:

datasets:
  - name: flowers
    source: https://www.robots.ox.ac.uk/~vgg/data/flowers/102/102flowers.tgz
    images: $DATASETS/jpg

You can also add a dataset from a local file:

kaze add mnist.zip

Usage and Examples

❯ kaze --help
Usage: kaze [OPTIONS] COMMAND [ARGS]...

Options:
  --help  Show this message and exit.

Commands:
  add
  list

kaze list

❯ kaze list
mnist at /Users/ge/kaze_debug/mnist

kaze add ...

❯ kaze add --help
Usage: kaze add [OPTIONS] [SOURCE]

Options:
  -n, --name TEXT
  -o, --path TEXT    target location for the dataset
  -i, --images TEXT  image path
  --labels TEXT      label path
  --voice TEXT       voice path
  --video TEXT       video path
  -q, --quiet        Verbose mode
  -z, --unzip        Decompress the dataset
  -v, --verbose      Verbose mode
  --help             Show this message and exit.