ez_norm

    description: 'layer ez_norm for configurable component'
    ---

    # ez_norm layer

    `ez_norm` is a custom normalization technique that uses a conv of the input by a linear projection of a style vector.

    ## Optional arguments

        `style` - The name of the style vector to use. Defaults to "w"

    ## input size

    Any 4-d tensor

    ## output size

    Same as input size

    ## syntax

    ```json
      "ez_norm style=[style vector name]"
    ```

    ## examples

    ```json
      "latent name=w",
      ...
      "cat self (ez_norm style=w)"
    ```