tf.data.experimental.service.DiserConfig

Configuration class for tf.data service disers.

tf.data.experimental.service.DiserConfig(
    port=0,
    protocol=None,
    work_dir=None,
    fault_tolerant_mode=False,
    worker_addresses=None,
    job_gc_check_interval_ms=None,
    job_gc_timeout_ms=None,
    worker_timeout_ms=None,
    worker_max_concurrent_snapshots=0
)

portSpecifies the port to bind to. A value of 0 indicates that the server may bind to any available port.
protocolThe protocol to use for communicating with the tf.data service, e.g. "grpc".
work_dirA directory to store diser state in. This argument is required for the diser to be able to recover from restarts.
fault_tolerant_modeWhether the diser should write its state to a journal so that it can recover from restarts. Diser state, including registered datasets and created jobs, is synchronously written to the journal before responding to RPCs. If True, work_dir must also be specified.
worker_addressesIf the job uses auto-sharding, it needs to specify a fixed list of worker addresses that will register with the diser. The worker addresses should be in the format "host" or "host:port", where "port" is an integer, named port, or %port% to match any port.
job_gc_check_interval_msHow often the diser should scan through to delete old and unused jobs, in milliseconds. If not set, the runtime will select a reasonable default. A higher value will reduce load on the diser, while a lower value will reduce the time it takes for the diser to garbage collect expired jobs.
job_gc_timeout_msHow long a job needs to be unused before it becomes a candidate for garbage collection, in milliseconds. A value of -1 indicates that jobs should never be garbage collected. If not set, the runtime will select a reasonable default. A higher value will cause jobs to stay around longer with no consumers. This is useful if there is a large gap in time between when consumers read from the job. A lower value will reduce the time it takes to reclaim the resources from expired jobs.
worker_timeout_msHow long to wait for a worker to heartbeat before considering it missing. If not set, the runtime will select a reasonable default.
worker_max_concurrent_snapshotsThe maximum number of snapshots a worker can concurrently process.

portA namedtuple alias for field number 0
protocolA namedtuple alias for field number 1
work_dirA namedtuple alias for field number 2
fault_tolerant_modeA namedtuple alias for field number 3
worker_addressesA namedtuple alias for field number 4
job_gc_check_interval_msA namedtuple alias for field number 5
job_gc_timeout_msA namedtuple alias for field number 6
worker_timeout_msA namedtuple alias for field number 7
worker_max_concurrent_snapshotsA namedtuple alias for field number 8