I am trying to do some hyperparameter sweeps to find the best neural network configuration. My way of working is that I pick some (e.g. 3) possible parameters and optimize them, and then add others later on once I come up with new possible configurations.
However, in TensorBoard only the hyper parameters from the first run are showing up. Is there a way to set default values for previous runs? Or to specifically select a new group of runs?