Debugger V2 is inactive because no data is available.
To use the debugger,
  1. Add the following line to the beginning of your program:
    tf.debugging.experimental.enable_dump_debug_info( logdir, tensor_debug_mode="FULL_HEALTH", circular_buffer_size=-1 )
  2. Re-run the program.
Auto-alerts for problems found
Integrated debugging to trace problems to their causes
Link log to code
The log directory must contain TensorFlow Debugger (V2) data. tf.debugging.experimental.enable_dump_debug_info() will collect tensor data, graph structures, the associated stack traces, and source code to the specificed directory logdir as the instrumented TensorFlow program executes.
See documentation of the Python API of Debugger V2.
See here for other TensorFlow debugging APIs.