Release 0.2.0

• Update TensorFlow™ dependency to 2.13.0 to use the latest TFL to TOSA legalization
• Models with tfl.custom ops are considered non TOSA compliant
• Add security measures:
  • Add a --sanitizer=[asan|ubsan] option to setup.py to build the Python® package with the Address Sanitizer (ASAN) or Undefined Behavior Sanitizer (UBSAN)
  • Add the Bandit linter
  • Use _FORTIFY_SOURCE=2 on compiled sources
  • Add security counter measures in the Docker™ CI, see docker/README.md for more information
• Add a --nightly option to setup.py for building nightly Python® packages
• Fix TOSA Checker PyPI® homepage links

Release 0.1.0

First release of the TOSA Checker tool. The goal of the tool is to provide an easy way to check if a TensorFlow™ Lite model is compatible with the TOSA specification.

The tool is provided as a Python® package and can be used as follow:

>>> import tosa_checker as tc
>>> checker = tc.TOSAChecker("model.tflite")
>>> checker.is_tosa_compatible()
True

The tool is currently in alpha, the features set is limited and some stability issues may exist.

Future versions may extend the functionalities provided and support for other frameworks will be added.