tensorflow/models · error · ValueError

TPU training does not support data producer yet. Use pre-pro

Error message

TPU training does not support data producer yet. Use pre-processed data.

What it means

Error "TPU training does not support data producer yet. Use pre-processed data." thrown in tensorflow/models.

Source

Thrown at official/recommendation/ncf_input_pipeline.py:174

        params["batch_size"],
        is_training=True,
        rebatch=False)

    # Re-batch evaluation dataset for TPU Pods.
    # TODO (b/162341937) remove once it's fixed.
    eval_rebatch = (params["use_tpu"] and strategy.num_replicas_in_sync > 8)  # pyrefly: ignore[missing-attribute]
    eval_dataset = create_dataset_from_tf_record_files(
        params["eval_dataset_path"],
        input_meta_data["eval_prebatch_size"],  # pyrefly: ignore[unsupported-operation]
        params["eval_batch_size"],
        is_training=False,
        rebatch=eval_rebatch)

    num_train_steps = int(input_meta_data["num_train_steps"])  # pyrefly: ignore[unsupported-operation]
    num_eval_steps = int(input_meta_data["num_eval_steps"])  # pyrefly: ignore[unsupported-operation]
  else:
    if params["use_tpu"]:
      raise ValueError("TPU training does not support data producer yet. "
                       "Use pre-processed data.")

    assert producer
    # Start retrieving data from producer.
    train_dataset, eval_dataset = create_dataset_from_data_producer(
        producer, params)
    num_train_steps = producer.train_batches_per_epoch
    num_eval_steps = producer.eval_batches_per_epoch

  return train_dataset, eval_dataset, num_train_steps, num_eval_steps

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/recommendation/ncf_input_pipeline.py:174 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/288fed2eadbe350a. Report an issue: GitHub.