keras-tuner
keras-tuner copied to clipboard
Hyperband tuner stops at first round if number of hyperparameters are small.
If number of hyperparameter combinations are small the Hyperband tuner stops at the first round.
The problem is at
https://github.com/keras-team/keras-tuner/blob/215603d4a728a6e3cab47fb26da0e5322af565bf/keras_tuner/tuners/hyperband.py#L144
if no parameter choice is available, i.e. number of choices are less than self._get_size(), None is returned and the tuner stops with doing the successive halving.
Also have this problem. Any updates when the fix is added to a release?