config

Type Members

1. sealed trait CheckpointConfig extends AnyRef

   

   Checkpoint configuration used while training models.

2. case class ClusterConfig(jobSpecs: Map[String, JobConfig]) extends Serializable with Product with Serializable

   

   Represents a cluster as a set of "tasks", organized into "jobs".

   Represents a cluster as a set of "tasks", organized into "jobs".

   A ClusterConfig represents the set of processes that participate in a distributed TensorFlow computation. Every TensorFlow server is constructed in a particular cluster.

   To create a cluster with two jobs and five tasks, you specify the mapping from job names to lists of network addresses (typically hostname-port pairs).

   For example:

   val clusterConfig = ClusterConfig(Map(
     "worker" -> JobConfig.from(
       "worker0.example.com:2222",
       "worker1.example.com:2222",
       "worker2.example.com:2222"),
     "ps" -> JobConfig.from(
       "ps0.example.com:2222",
       "ps1.example.com:2222")))

   Each job may also be specified as a sparse mapping from task indices to network addresses. This enables a server to be configured without needing to know the identity of (for example) all other worker tasks:

   For example:

   val clusterConfig = ClusterConfig(Map(
    "worker" -> JobConfig(1 -> "worker1.example.com:2222"),
    "ps" -> JobConfig.from(
      "ps0.example.com:2222",
      "ps1.example.com:2222")))

   jobSpecs

   Map mapping one or more job names to job configurations.

3. case class JobConfig(tasks: TreeMap[Int, String]) extends Product with Serializable

   

   Job configuration (excluding the job name).

   Job configuration (excluding the job name).

   tasks

   Mapping from task index to the corresponding task network address.

4. case class StepBasedCheckpoints(steps: Int = 1000, maxCheckpointsToKeep: Int = 5, keepCheckpointEveryNHours: Int = 10000) extends CheckpointConfig with Product with Serializable

   

   Checkpoint configuration for step-based checkpoints (i.e., checkpoints every n steps).

   Checkpoint configuration for step-based checkpoints (i.e., checkpoints every n steps).

   steps

   Save checkpoints every this many steps.

   maxCheckpointsToKeep

   Maximum number of recent checkpoint files to keep. As new files are created, older files are deleted. If 0, then all checkpoint files are kept. Defaults to 5 (that is, the 5 most recent checkpoint files are kept).

   keepCheckpointEveryNHours

   Save checkpoints every this many hours. The default value of 10,000 hours effectively disables the feature.

5. case class StepBasedSummaries(steps: Int = 1000) extends SummaryConfig with Product with Serializable

   

   Summary configuration for step-based summaries (i.e., summaries every n steps).

   Summary configuration for step-based summaries (i.e., summaries every n steps).

   steps

   Save summaries every this many steps.

6. sealed trait SummaryConfig extends AnyRef

   

   Summary configuration used while training models.

7. case class TensorBoardConfig(logDir: Path, host: String = "localhost", port: Int = 6006, reloadInterval: Int = 5) extends Product with Serializable

   

   TensorBoard configuration, which can be used when training using Estimators.

   TensorBoard configuration, which can be used when training using Estimators.

   logDir

   Directory containing the logs and summaries that the TensorBoard instance should use.

   host

   Host to use for the TensorBoard service.

   port

   Port to use for the TensorBoard service.

   reloadInterval

   Interval at which the backend reloads more data in seconds.

8. case class TimeBasedCheckpoints(seconds: Int = 600, maxCheckpointsToKeep: Int = 5, keepCheckpointEveryNHours: Int = 10000) extends CheckpointConfig with Product with Serializable

   

   Checkpoint configuration for time-based checkpoints (i.e., checkpoints every n seconds).

   Checkpoint configuration for time-based checkpoints (i.e., checkpoints every n seconds).

   seconds

   Save checkpoints every this many seconds.

   maxCheckpointsToKeep

   Maximum number of recent checkpoint files to keep. As new files are created, older files are deleted. If 0, then all checkpoint files are kept. Defaults to 5 (that is, the 5 most recent checkpoint files are kept).

   keepCheckpointEveryNHours

   Save checkpoints every this many hours. The default value of 10,000 hours effectively disables the feature.

9. case class TimeBasedSummaries(seconds: Int = 600) extends SummaryConfig with Product with Serializable

   

   Summary configuration for time-based summaries (i.e., summaries every n seconds).

   Summary configuration for time-based summaries (i.e., summaries every n seconds).

   seconds

   Save summaries every this many seconds.