A Framework for ML Research Using PyTorch


Keywords
pytorch
License
MIT
Install
pip install fandak==0.1.3.1

Documentation

PyPI version Travis Build

Fandak: فندک

Coming soon...

Will help you train your models for research.

Installation

Requires Python 3.7 🐍.

pip install fandak

Release Notes

Release Notes

Examples

See examples directory.

Visualizing the effects of hyper-parameters

Sample usage:

python -m fandak.hyper /path/to/root metric1 [metric2] [--exp-name baseline-*] [--params-list path/to/params/list.txt]

If no exp-name is provided, then all are going to be considered. exp-name can be something usable from glob.

We also have to specify the params list. What to have in the figure? Otherwise, we look at the experiments and find the params that are not the same among them. If a param is missing from a config file, then it is replaces with "?".