## Untrusted inputs during training and prediction

TensorFlow supports a wide range of input data formats. For example it can
process images, audio, videos, and text. There are several modules specialized
in taking those formats, modifying them, and/or converting them to intermediate
formats that can be processed by TensorFlow.

These modifications and conversions are handled by a variety of libraries that

TFCI_BAZEL_TARGET_SELECTING_CONFIG_PREFIX=linux_arm64
# Note: this is not set to "--cpu", because that changes the package name
# to tensorflow_cpu. These ARM builds are supposed to have the name "tensorflow"
# despite lacking Nvidia CUDA support.
TFCI_BUILD_PIP_PACKAGE_ARGS="--repo_env=WHEEL_NAME=tensorflow"
TFCI_DOCKER_ENABLE=1
TFCI_DOCKER_IMAGE=gcr.io/tensorflow-sigs/build-arm64:tf-2-16-multi-python
TFCI_DOCKER_PULL_ENABLE=1

 * Improvements and fixes in Keras loss masking:
    * Whether you represent a ragged tensor as a `tf.RaggedTensor` or using [keras masking](https://www.tensorflow.org/guide/keras/masking_and_padding), the returned loss values should be the identical to each other. In previous versions Keras may have silently ignored the mask.