ml_saveload
Save / Load a Spark ML Model Fit
Save / load a ml_model fit.
Usage
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_modelfit.
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.