tf.data.experimental.service.WorkerConfig

View source on GitHub

Configuration class for tf.data service dispatchers.

tf.data.experimental.service.WorkerConfig(
    dispatcher_address, worker_address=None, port=0, protocol=None,
    heartbeat_interval_ms=None, dispatcher_timeout_ms=None
)

Fields:

• dispatcher_address: Specifies the address of the dispatcher.
• worker_address: Specifies the address of the worker server. This address is passed to the dispatcher so that the dispatcher can tell clients how to connect to this worker.
• port: Specifies the port to bind to. A value of 0 indicates that the worker can bind to any available port.
• protocol: (Optional.) Specifies the protocol to be used by the server. Defaults to "grpc".
• heartbeat_interval_ms: How often the worker should heartbeat to the dispatcher, in milliseconds. If not set, the runtime will select a reasonable default. A higher value will reduce the load on the dispatcher, while a lower value will reduce the time it takes to reclaim resources from finished jobs.
• dispatcher_timeout_ms: How long, in milliseconds, to retry requests to the dispatcher before giving up and reporting an error. Defaults to 1 hour.