Question: How to continue to make new trials from previous trials?

Hello, Since the selection of the optimal parameters can be very time consuming. Often libraries for Bayesian selection of optimal parameters of machine learning models allow us to save the study object. In order to be able to load it in the future and continue the selection of the optimal parameters further from the very same place where we stopped.

for example optuna library

import optuna
from optuna.samplers import TPESampler

study = optuna.create_study(direction="maximize", sampler=TPESampler(n_startup_trials=50))
study.optimize(objective_0, n_trials=100)
joblib.dump(study, 'save_path') # to save hyperparameter object
study = joblib.load(tuna_path) # to restore hyperparameter object
study.optimize(objective_0, n_trials=100) # to continue for a new 100 trials from previous 100 trials (from 100 to 200 trial)

AutoKeras has similar functionality ?

for example

clf = ak.StructuredDataClassifier(
    overwrite=True,
    max_trials=3) # 3 trials
# Feed the tensorflow Dataset to the classifier.
clf.fit(train_set, epochs=10) # fit

Then save clf  object like
clf.save_to_file('....')
Then 
clf.load_from_file.
Then 
clf.fit_for_next_trials from previous ?

[haifeng-jin]

You don’t need to create a new object. You just run the same .py file again. If you want to create a new object, please make sure all the configs remains the same as the previous one.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.