Pass verbose from Estimator to SparkBackend

[carsonwang]

When I run the following code, currently I can not see the verbose=2 log from horovod.spark.runner like the generated mpirun command. We can create SparkBackend by using this verbose value to see the logs.

keras_estimator = hvd.KerasEstimator(num_proc=args.num_proc,
                                     store=store,
                                     model=model,
                                     optimizer=opt,
                                     loss='mae',
                                     metrics=[exp_rmspe],
                                     custom_objects=CUSTOM_OBJECTS,
                                     feature_cols=all_cols,
                                     label_cols=['Sales'],
                                     validation='Validation',
                                     batch_size=args.batch_size,
                                     epochs=args.epochs,
                                     verbose=2)

[tgaddair]

Hey @carsonwang, thanks for the fix! Could you sign the DCO and rebase your PR off master (to fix unit tests)?

[tgaddair]

LGTM! Thanks!

[carsonwang]

Thanks @tgaddair !