Releases: bethgelab/foolbox
Releases · bethgelab/foolbox
Version 3.0.3
Fixes a bug in the BrendelBethgeAttack
and updated Numba to silence warnings.
Version 3.0.2
Fixes a bug in the BrendelBethgeAttack
(thanks @AidanKelley)
Version 3.0.1
Bug fixes
- type annotations are now correctly exposed using
py.typed
(file was missing in MANIFEST) - TransformBoundsWrapper now correctly handles
data_format
(thanks @zimmerrol)
Version 3.0.0
New Features
Foolbox 3 aka Foolbox Native has been rewritten from scratch with performance in mind. All code is running natively in PyTorch, TensorFlow and JAX, and all attacks have been rewritten with real batch support.
Version 3.0.0b1
New Features
- added
foolbox.gradient_estimators
- improved attack hyperparameter documentation
Version 3.0.0b0
Foolbox 3 aka Foolbox Native has been rewritten from scratch with performance in mind. All code is running natively in PyTorch, TensorFlow and JAX, and all attacks have been rewritten with real batch support.
Warning: This is a pre-release beta version. Expect breaking changes.
Version 2.4.0
New Features
- fixed PyTorch model gradients (fixes DeepFool with batch size > 1)
- added support for TensorFlow 2.0 and newer (Graph and Eager mode)
- refactored the tests
- support for the latest
randomgen
version
Version 2.3.0
New Features
- new
EnsembleAveragedModel
(thanks to @zimmerrol) - new
foolbox.utils.flatten
- new
foolbox.utils.atleast_kd
- new
foolbox.utils.accuracy
PyTorchModel
now always warns if model is in train mode, not just once- batch support for
ModelWithEstimatedGradients
Bug fixes
- fixed dtype when using Adam PGD with a PyTorch model
- fixed CW attack hyperparameters
Version 2.2.0
New Features
- support for Foolbox extensions using the
foolbox.ext
namespace
Version 2.1.0
New Features
- New
foolbox.models.JAXModel
class to support JAX models (https://github.com/google/jax) - The
preprocessing
argument of models now supports aflip_axis
key to support common preprocessing operations like RGB to BGR in a nice way. This builds on the ability to pass dicts topreprocessing
introduced in Foolbox 2.0.
Bug fixes and improvements
- Fixed a serious bug in the
LocalSearchAttack
(thanks to @duoergun0729) foolbox.utils.samples
now warns if samples are repeatedfoolbox.utils.sampels
now uses PNGs instead of JPGs (except for ImageNet)- Other bug fixes
- Improved docstrings
- Improved docs