@tf.function incorrectly handles None (batch) dimension of input tensor

Question

I want to use @tf.function decorator to use auto-graphing feature of tensorflow to convert Python code to tensorflow static graph, so that I can use it