If the user didn't install the tensorbard lib correctly, then the get summary should raise an error.
Current Behavior
Right now the function prints an warning, but if you use the TensorboardLogger, at the end of the training epoch this method will trow an error since self.writer is None.
For long epochs this is a waste of time
Possible Implementation
Either change this with a raise error or add an assertion of the kind
Expected Behavior
If the user didn't install the tensorbard lib correctly, then the get summary should raise an error.
Current Behavior
Right now the function prints an warning, but if you use the TensorboardLogger, at the end of the training epoch this method will trow an error since
self.writer
is None. For long epochs this is a waste of timePossible Implementation
Either change this with a raise error or add an assertion of the kind
in the TensorboardLogger init