다음을 통해 공유


Deprecated, discontinued, or changed features in Machine Learning Server

Important

This content is being retired and may not be updated in the future. The support for Machine Learning Server will end on July 1, 2022. For more information, see What's happening to Machine Learning Server?

This article notifies you about pending status changes to features and function libraries, such as deprecated and discontinued functions in the custom packages and libraries installed with the product.

Machine Learning Server 9.4

There are no announcements for package or function deprecation, discontinuation, or changes to existing functionality.

To learn more about this release, see What's New for feature announcements and Known issues for descriptions and possible workarounds to known problems found in this release.

Machine Learning Server 9.3

There are no announcements for package or function deprecation, discontinuation, or changes to existing functionality.

To learn more about this release, see What's New for feature announcements and Known issues for descriptions and possible workarounds to known problems found in this release.

Machine Learning Server 9.2.1

RevoScaleR Package

Function Status Replacement
RxHadoopMR Deprecated RxSpark
RxHadoopMR-class Deprecated RxSpark-class
RxInTeradata Deprecated None. The only in-database compute context is SQL Server.
RxInTeradata-class Deprecated None. The only in-database compute context is SQL Server.

For more information, see discontinued RevoScaleR functions and deprecated RevoScaleR functions.

Microsoft R Server 9.1.0

RevoScaleR Package

Function Status Replacement
rxGetNodes Deprecated rxGetAvailableNodes
RxHpcServer Deprecated RxSpark or RxHadoopMR
rxImportToXdf Deprecated rxImport
rxDataStepXdf Deprecated rxDataStep
rxDataFrameToXdf Deprecated rxDataStep
rxXdfToDataFrame Deprecated rxDataStep
rxSortXdf Deprecated rxSort
rxGetVarInfoXdf Discontinued rxGetVarInfo)
rxGetInfoXdf Discontinued rxGetInfo)

For more information, see discontinued RevoScaleR functions and deprecated RevoScaleR functions.

RevoMods Package

In RevoMods functions are discontinued (all were intended for use solely by the R Productivity Environment discontinued in Microsoft R Server 8.0.3):

  • ? (use the standard R ?, previously masked)
  • q (use the standard R q function, previously masked)
  • quit (use the standard R quit function, previously masked)
  • revoPlot (use the standard R plot function)
  • revoSource (use the standard R source function)

Microsoft R Server 9.0.1

Changed in this release

  • R Server is built on Microsoft R Open 3.3.2 (R 3.3.2).
  • R packages added to MRO include curl, Jsonlite, and R6.
  • R packages added to R Server include mrupdate (for automated update checks) and CompatibilityAPI (to check for feature compatibility between versions).
  • DeployR (now referred to as operationalization) no longer supports MongoDB. Use PostgreSQL, SQL Server, or Azure SQL Database instead.

Deprecated or Discontinued

  • RevoScaleR deprecated functions:

    • RxHpcServer (no replacement)
    • rxReadXdf (use rxImport or rxDataStep)
  • RevoScaleR discontinued functions:

    • (none)

Bug Fixes

  • Not available.

Known Issues

See here: https://docs.microsoft.com/machine-learning-server/resources-known-issues

Microsoft R Server 8.0.5

Deprecated or Discontinued

  • The R Productivity Environment (RPE) IDE for Revolution R Enterprise:

    • The RPE is now defunct. We recommend use of alternative IDEs.
    • These R packages that support the RPE have been removed: revoIpe, pkgXMLBuilder, XML, and RevoRpeConnector.
  • RevoScaleR changes:

    • These RevoScaleR functions are now deprecated:

      • rxImportToXdf (use rxImport)
      • rxDataStepXdf (use rxDataStep)
      • rxXdfToDataFrame (use rxDataStep)
      • rxDataFrameToXdf (use rxDataStep)
    • These RevoScaleR functions are now defunct:

      • rxgetvarinfo (use rxGetVarInfo)
      • rxGetInfoXdf (us rxGetInfo)
      • the covariance argument (only) to the rxLinMod function

Changed in this release

  • A new internal variable, .rxPredict is available inside transformation functions to indicate that the data is being processed from a prediction rather than a model estimation.

  • Microsoft R documentation has been moved from the product distribution to MSDN. The “doc” directories in the RevoScaleR and RevoPemaR packages have been removed, as has the top-level R Server “doc” directory.

  • Microsoft R licenses and Third Party Notices files are now included in the new MicrosoftR package. The Revo.home() function now points to the location of this directory, and Revo.home(“licenses”) points to the “licenses” directory within. The Revo.home(“doc”) component is now defunct.

  • An opt-in telemetry feature allows you to anonymously help improve Microsoft R Server by enabling us to gather data on the R Server functions you use, operating system, R version, and RevoScaleR version. Turn it on using the rxPrivacyControl function in RevoScaleR.

Bug Fixes

  • rxKmeans now works with an RxHadoopMR compute context and an RxTextData data source.

  • When writing to an RxTextData data source using rxDataStep, specifying firstRowIsColNames to FALSE in the output data source will now correctly result in no column names being written.

  • When writing to an RxTextData data source using rxDataStep, setting quoteMark to "" in the output data source will now result in no quote marks written around character data.

  • When writing to an RxTextData data source using rxDataStep, setting missingValueString to "" in the output data source will now result in empty strings for missing values.

  • Using rxImport, if quotedDelimiters was set to TRUE, transforms could not be processed.

  • rxImport of RxTextData now reports a warning instead of an error if mismatched quotes are encountered.

  • When using rxImport and appending to an .xdf file, a combination of the use of newName and newLevels for different variables in colInfo could result in newLevels being ignored, resulting in a different set of factor levels for the appended data.

  • When using rxPredict, with an in-formula transformation of the dependent variable, an error was given if the variable used in the transformation was not available in the prediction data set.

  • Hadoop bug fix for incompatibility when using both HA and Kerberos on HDP.

  • Deployr default grid node and any additional nodes should have 20 slots by default

  • DeployR was generating very large (~52GB) catalina.out log files.

  • When running scripts in the DeployR Repository Manager's Test tab, any numeric values set to 0 were ignored and not included as part of the request.

Known Issues

See here: https://docs.microsoft.com/machine-learning-server/resources-known-issues

Microsoft R Server 8.0.0

Known Issues

See here: https://go.microsoft.com/fwlink/?LinkID=717964&clcid=0x409