I am trying to distribute hyperparameter tuning using hyperopt on a tensorflow.keras model. I am using sparkTrials in my fmin:

spark_trials = SparkTrials(parallelism=4)

...

best_hyperparam = fmin(fn=CNN_HOF, 

                 space=space, 

                 algo=tpe.suggest, 

                 max_evals=tuner_max_evals,

                  trials=spark_trials)

but I am receiving this error:

TypeError: cannot pickle '_thread.lock' object

the only way the code is working is skipping the trials passing by commenting out the line trials=spark_trials which means there would be no distributed tuning.

Any idea how can I fix this?

@Tian Tan​ 

@Sara Dooley​