tensorflow/models · error · ValueError

Shared tf.data service does not support round-robin tf.data

Error message

Shared tf.data service does not support round-robin tf.data service.

What it means

Error "Shared tf.data service does not support round-robin tf.data service." thrown in tensorflow/models.

Source

Thrown at official/core/input_reader.py:539

  ) -> tf.data.Dataset:
    """Potentially distributes a dataset."""
    if self._enable_tf_data_service and input_context:
      if self._enable_round_robin_tf_data_service:
        replicas_per_input_pipeline = input_context.num_replicas_in_sync // (
            input_context.num_input_pipelines)
        base_consumer_index = input_context.input_pipeline_id * (
            replicas_per_input_pipeline)
        num_consumers = input_context.num_input_pipelines * (
            replicas_per_input_pipeline)
        range_dataset = tf.data.Dataset.range(replicas_per_input_pipeline)
        tfds_kwargs = {
            'processing_mode': 'parallel_epochs',
            'service': self._tf_data_service_address,
            'job_name': self._tf_data_service_job_name,
            'num_consumers': num_consumers
        }
        if self._enable_shared_tf_data_service_between_parallel_trainers:
          raise ValueError('Shared tf.data service does not support round-robin'
                           ' tf.data service.')
        dataset = range_dataset.map(lambda i: dataset.apply(  # pylint: disable=g-long-lambda
            tf.data.experimental.service.distribute(
                consumer_index=base_consumer_index + i, **tfds_kwargs)))
        # Use parallel interleave to read multiple batches from a tf.data
        # service worker in parallel.
        dataset = dataset.interleave(
            lambda x: x,
            cycle_length=replicas_per_input_pipeline,
            num_parallel_calls=replicas_per_input_pipeline,
            deterministic=True)
      else:
        tfds_kwargs = {
            'processing_mode': 'parallel_epochs',
            'service': self._tf_data_service_address,
            'job_name': self._tf_data_service_job_name,
        }
        if self._enable_shared_tf_data_service_between_parallel_trainers:

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/core/input_reader.py:539 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of tensorflow/models@e006f5f0d5 (2026-08-24). Data as JSON: /api/errors/cd74b4bdae583651. Report an issue: GitHub.