CODEWITHSUNDEEP
h2oautomlreproducible-research
user1700890
user1700890

Reputation: 7730

H2O AutoML - reproducing results

I am running H2O AutoML on the same date set with the same seed on the same laptop multiple times and getting different predictions from time to time. I wonder if it is possible to get the same numbers on every run. Here is my code:

import h2o
h2o.init()
aml = h2o.automl.H2OAutoML(max_models = 25,
 balance_classes = False, seed = 1)
aml.train(predictors, response, training_frame = h2o_df_train)
aml.leader.predict(h2o_df_test)

It looks like it is possible for GBM, but I wonder if it is possible for whole AutoML run?

My appologies for cross posting, but here is a complete example:

H2O forum

Problem solved

From H2O docs:

"seed: Integer. Set a seed for reproducibility. AutoML can only guarantee reproducibility under certain conditions. H2O Deep Learning models are not reproducible by default for performance reasons, so if the user requires reproducibility, then exclude_algos must contain "DeepLearning". In addition max_models must be used because max_runtime_secs is resource limited, meaning that if the available compute resources are not the same between runs, AutoML may be able to train more models on one run vs another. Defaults to NULL/None."

Upvotes: 0

Views: 78

Answers (0)

Related Questions