Tensorflow-model-remediation

Latest version: v0.1.7.1

Safety actively analyzes 626403 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

0.1.7

Major Features and Improvements
* Adding Counterfactual Logit Pairing to the TensorFlow Model Remediation library.

Bug fixes and other small changes

Breaking changes

Deprecations

0.1.6

Major Features and Improvements
* Adding a new MinDiff loss of Adjusted MMD Loss

Bug fixes and other small changes
* `build_min_diff_dataset` now supports SparseTensors.
* Adding an enable/disable feature for summary_histogram.
* Moving TensorFlow Model Remediation to support Python 3 only.

Breaking changes

Deprecations

0.1.5

Major Features and Improvements

* Add support for multiple applications of MinDiff within MinDiffModel. This
includes:

* Utils for validating and manipulating structures for MinDiff applications
* Added functionality in input utils to enable packing multiple sets of
MinDiff data into one dataset.
* Changes to MinDiffModel to support multiple applications of MinDiff in a
single instance.

Bug fixes and other small changes
* Remove protobuf and tensorflow_model_analysis dependencies.
* `pack_min_diff_data` now conserves shape or original inputs.
* Checks an orginal models args before passing `mask` to the call function.

Breaking changes
* (Minor) Change default name of MinDiffLoss instance to be snake_case.

Deprecations

0.1.4

Major Features and Improvements

Bug fixes and other small changes

* Add support for MinDiffModel subclasses that are also subclasses of TF
Functional class.
* Add uci related utils for colabs. This is unrelated to the actual
package api but is used in our tutorials and guides.

Breaking changes

Deprecations

0.1.3

Major Features and Improvements

Bug fixes and other small changes

* Implement `get_config` and `from_config` methods for losses, kernels and
MinDiffModel register classes as keras objects.

Breaking changes

* (Minor) Change output of `test_step` to not include `min_diff_loss` metric if
`min_diff_data` is not included.
* (Minor) Ensure that MinDiffModel metric has a unique name. In the case of a
collision, the name will be changed to "min_diff_loss_N" for the lowest N
(starting at 1) that makes it unique.

Deprecations

0.1.2

Major Features and Improvements

* Initial Model Remediation Release.

Bug fixes and other changes

* N/A

Breaking changes

* N/A

Deprecations

* N/A

Links

Releases

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.