ml_saveload: Save / Load a Spark ML Model Fit

Description Usage Arguments Details

Description

Save / load a ml_model fit.

Usage

1
2
3
ml_load(sc, file, meta = ml_load_meta(file))

ml_save(model, file, meta = ml_save_meta(model, file))

Arguments

sc

A spark_connection.

file

The path where the Spark model should be serialized / deserialized.

meta

The path where the R metadata should be serialized / deserialized. Currently, this must be a local filesystem path. Alternatively, this can be an R function that saves / loads the metadata object.

model

A ml_model fit.

Details

These functions are currently experimental and not yet ready for production use. Unfortunately, the training summary information for regression fits (linear, logistic, generalized) are currently not serialized as part of the model fit, and so model fits recovered through ml_load will not work with e.g. fitted, residuals, and so on. Such fits should still be suitable for generating predictions with new data, however.


sparklyr documentation built on Aug. 13, 2017, 9:02 a.m.