How to skip trials that end in a wrong model

[fbettcba]

I'm wondering if there is a way to skip trials that end up in a combination of hyperparameters that cause the model to fail building.
For example some combinations end up in errors like: Negative dimension size caused by subtracting 3 from 1 for.
And the process ends with an error: Too many failed attempts to build model.

Is there any way to just ignore the trials that result in error? and continue searching other options?

[haifeng-jin]

Currently, there is no quick solution to this, you may need to make some if statements in the build function, when you use the hps to create the model.

[invernizzi]

It's not an official API, but you can control the number of failed model builds that will be attempted before Keras Tuner gives up.

import keras_tuner as kt

kt.engine.tuner.MAX_FAIL_STREAK = 60