*   `tf.compat.v1.nn.fused_batch_norm` backprop to `offset` when
            `is_training=False`
        *   `tf.image.adjust_contrast` forward
        *   `tf.nn.depthwise_conv2d` backprop to `filter` when not using cuDNN
            convolution
        *   `tf.image.resize` with `method=ResizeMethod.NEAREST` backprop
        *   `tf.math.bincount` - TODO: confirm exception added

    * It's possible to have arbitrary trees/levels of dependencies with `yield` and exit steps are handled in the correct order automatically.
    * It works by default in Python 3.7 or above. For Python 3.6, it requires the extra backport dependencies:
        * `async-exit-stack`
        * `async-generator`
    * New docs at [Dependencies with `yield`](https://fastapi.tiangolo.com/tutorial/dependencies/dependencies-with-yield/).