Field Detail

• DL4J_RESOURCES_DIR_PROPERTY

  public static final java.lang.String DL4J_RESOURCES_DIR_PROPERTY

  Applicability: Numerous modules, including deeplearning4j-datasets and deeplearning4j-zoo
  Description: Used to set the local location for downloaded remote resources such as datasets (like MNIST) and pretrained models in the model zoo. Default value is set via new File(System.getProperty("user.home"), ".deeplearning4j"). Setting this can be useful if the system drive has limited space/performance, a shared location for all users should be used instead, or if user.home isn't set for some reason.

  See Also:

  Constant Field Values

• DL4J_RESOURCES_BASE_URL_PROPERTY

  public static final java.lang.String DL4J_RESOURCES_BASE_URL_PROPERTY

  Applicability: Numerous modules, including deeplearning4j-datasets and deeplearning4j-zoo
  Description: Used to set the base URL for hosting of resources such as datasets (like MNIST) and pretrained models in the model zoo. This is provided as a fallback in case the location of these files changes; it also allows for (in principle) a local mirror of these files.
  NOTE: Changing this to a location without the same files and file structure as the DL4J resource hosting is likely to break external resource dowloading in DL4J!

  See Also:

  Constant Field Values

• CUSTOM_REGISTRATION_PROPERTY

  public static final java.lang.String CUSTOM_REGISTRATION_PROPERTY

  Applicability: deeplearning4j-nn
  Description: Used for loading legacy format JSON containing custom layers. This system property is provided as an alternative to NeuralNetConfiguration#registerLegacyCustomClassesForJSON(Class[]). Classes are specified in comma-separated format.
  This is required ONLY when ALL of the following conditions are met:
  1. You want to load a serialized net, saved in 1.0.0-alpha or before, AND
  2. The serialized net has a custom Layer, GraphVertex, etc (i.e., one not defined in DL4J), AND
  3. You haven't already called NeuralNetConfiguration#registerLegacyCustomClassesForJSON(Class[])

  See Also:

  Constant Field Values

• CRASH_DUMP_ENABLED_PROPERTY

  public static final java.lang.String CRASH_DUMP_ENABLED_PROPERTY

  Applicability: deeplearning4j-nn
  Description: DL4J writes some crash dumps to disk when an OOM exception occurs - this functionality is enabled by default. This is to help users identify the cause of the OOM - i.e., where native memory is actually consumed. This system property can be used to disable memory crash reporting.

  See Also:

  For configuring the output directory, Constant Field Values

• CRASH_DUMP_OUTPUT_DIRECTORY_PROPERTY

  public static final java.lang.String CRASH_DUMP_OUTPUT_DIRECTORY_PROPERTY

  Applicability: deeplearning4j-nn
  Description: DL4J writes some crash dumps to disk when an OOM exception occurs - this functionality is enabled by default. This system property can be use to customize the output directory for memory crash reporting. By default, the current working directory (System.getProperty("user.dir") or new File("")) will be used

  See Also:

  To disable crash dump reporting, Constant Field Values

• CHART_MAX_POINTS_PROPERTY

  public static final java.lang.String CHART_MAX_POINTS_PROPERTY

  Applicability: deeplearning4j-ui_2.xx
  Description: The DL4J training UI (StatsListener + UIServer.getInstance().attach(ss)) will subsample the number of chart points when a lot of data is present - i.e., only a maximum number of points will be shown on each chart. This is to reduce the UI bandwidth requirements and client-side rendering cost. To increase the number of points in charts, set this property to a larger value. Default: 512 values

  See Also:

  Constant Field Values

• UI_SERVER_PORT_PROPERTY

  public static final java.lang.String UI_SERVER_PORT_PROPERTY

  Applicability: deeplearning4j-play (deeplearning4j-ui_2.xx)
  Description: This property sets the port that the UI will be available on. Default port: 9000. Set to 0 for a random port.

  See Also:

  Constant Field Values

• NTP_SOURCE_SERVER_PROPERTY

  public static final java.lang.String NTP_SOURCE_SERVER_PROPERTY

  Applicability: dl4j-spark_2.xx - NTPTimeSource class (mainly used in ParameterAveragingTrainingMaster when stats collection is enabled; not enabled by default)
  Description: This sets the NTP (network time protocol) server to be used when collecting stats. Default: 0.pool.ntp.org

  See Also:

  Constant Field Values

• NTP_SOURCE_UPDATE_FREQUENCY_MS_PROPERTY

  public static final java.lang.String NTP_SOURCE_UPDATE_FREQUENCY_MS_PROPERTY

  Applicability: dl4j-spark_2.xx - NTPTimeSource class (mainly used in ParameterAveragingTrainingMaster when stats collection is enabled; not enabled by default)
  Description: This sets the NTP (network time protocol) update frequency in milliseconds. Default: 1800000 (30 minutes)

  See Also:

  Constant Field Values

• TIMESOURCE_CLASSNAME_PROPERTY

  public static final java.lang.String TIMESOURCE_CLASSNAME_PROPERTY

  Applicability: dl4j-spark_2.xx - mainly used in ParameterAveragingTrainingMaster when stats collection is enabled; not enabled by default
  Description: This sets the time source to use for spark stats. Default: org.deeplearning4j.spark.time.NTPTimeSource

  See Also:

  Constant Field Values