/ml_floats

A simple tool to compare floating point representations used in Machine Learning

Primary LanguagePython

Compare ML float representations

This is as simple tool to compare the several floating point representations used in machine-learning:

  • float32,
  • float16,
  • bfloat16.
usage: decompose.py [-h] [--type TYPE] number

positional arguments:
  number

optional arguments:
  -h, --help   show this help message and exit
  --type TYPE  [float32|float16|bfloat16]