khannaekta commented on a change in pull request #443: DL: Add training for multiple models URL: https://github.com/apache/madlib/pull/443#discussion_r327879849
########## File path: src/ports/postgres/modules/deep_learning/test/madlib_keras_predict.sql_in ########## @@ -274,12 +274,15 @@ FROM cifar10_predict; -- Predict with correctly shaped data, must go thru. -- Update output_summary table to reflect -- class_values, num_classes and model_arch_id for shaped data -UPDATE keras_saved_out -SET model_arch = (SELECT model_arch from model_arch where model_id = 3); -UPDATE keras_saved_out_summary -SET model_arch_id = 3, - num_classes = 3, - class_values = ARRAY[0,NULL,NULL]::INTEGER[]; +DROP TABLE IF EXISTS keras_saved_out, keras_saved_out_summary; Review comment: The `UPDATE` query was using a model with different num_classes and setting it to 3 in the summary table. This caused the following predict tests to fail due to the new assertion that got added in `deserialize_as_nd_weights()`. The code and tests do as expected, to fail if the number of elements in the model_weights and model from model_arch do not match, but the setup wasn't correct, therefore updated it to call fit first. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
