I'm trying to view the conceptual graph of a fairly complex TensorFlow model in TensorBoard. However, the option is greyed out. (I have no issue viewing the op graph).

My understanding is that in TensorBoard, the keras tag is necessary to view the conceptual graph. However, there are no tags at all when I look at the "Graph" tab on TensorBoard.

All I've written for the callback is the below block. There is no additional custom TensorBoard code:

tensorboard_callback = tf.keras.callbacks.TensorBoard(
        log_dir='tensorboard_logs',
        write_graph=True,
        histogram_freq=1,
    )

The model is multivariate with a custom loss function, which I double checked inherits from tf.keras.losses.Loss. From what I can tell (this isn't a repo I wrote), all other custom code also inherits from the respective Keras class.

Additionally, the code base uses TensorFlow 2.13.0, Keras 2.13.1, and TensorBoard 2.13.0.

How can I ensure that the keras tag shows up on TensorBoard so I can view the model's conceptual graph?