TypeError when using 'check-numerics' in combination with 'PolicySavedModelTrigger'

[ULudo]

If I enable tensor numeric checking and initialise a Learner with a PolicySavedModelTrigger I get the following error:

TypeError: Input 'resource' of 'AssignVariableOp' Op has type float32 that does not match expected type of resource.

You can recreate this error if you execute the TensorFlow tutorial 'SAC minitaur with the Actor-Learner API' in Colab and enable tensor numeric checking at the beginning of the file with the following command:

tf.debugging.enable_check_numerics()

I also get this error when I enable dumping debugging information by the following command:

tf.debugging.experimental.enable_dump_debug_info(<path>, tensor_debug_mode="FULL_HEALTH", circular_buffer_size=1)