When using keras, it writes all executions in same graph 'training'

[dingus221]

Directory path is hardcoded as 'training', right?
training_log_dir = os.path.join(log_dir, 'training')

[PRaezUC3M]

I just rewrote the init method for that. It's a really simple fix but it helps a lot when comparing executions.

class TensorBoardColabCallback(TensorBoard):
    def __init__(self, tbc=None, write_graph=True, name=None, **kwargs):
        # Make the original `TensorBoard` log to a subdirectory 'training'

        if tbc is None:
            return

        log_dir = tbc.get_graph_path()

        training_log_dir = os.path.join(log_dir, 'training_{}'.format(name))
        super(TensorBoardColabCallback, self).__init__(training_log_dir, **kwargs)

        # Log the validation metrics to a separate subdirectory
        self.val_log_dir = os.path.join(log_dir, 'validation_{}'.format(name))