What's Changed * Introduce `TestCase` and ensure mixed precision works with all layers by james77777778 in https://github.com/james77777778/keras-aug/pull/137
What's Changed * Add `ToDType` and fix bugs in `RandAugment` and `TrivialAugmentWide` by james77777778 in https://github.com/james77777778/keras-aug/pull/136
What's Changed * Update guides and README by james77777778 in https://github.com/james77777778/keras-aug/pull/132 * Fix `app.py` by james77777778 in https://github.com/james77777778/keras-aug/pull/133 * Add `p` argument to `RandAugment` and `TrivialAugmentWide` and fix bug in `RandomApply` by james77777778 in https://github.com/james77777778/keras-aug/pull/134 * Update `app.py` by james77777778 in https://github.com/james77777778/keras-aug/pull/135
Now, KerasAug is compatible with Keras 3 and `tf.data.Dataset` with all backends!
Note that the new APIs largly follow `torchvision` and the correctness of all layers has been verified through unit tests.
0.5.8
What's Changed * Fix streamlit requirements by james77777778 in https://github.com/james77777778/keras-aug/pull/124 * Fix `tf.where` inconsistent dtype bug by james77777778 in https://github.com/james77777778/keras-aug/pull/126