tensorflow/models · error · ValueError

Evaluation batch size must be divisible by {} times {}

Error message

Evaluation batch size must be divisible by {} times {}

What it means

Error "Evaluation batch size must be divisible by {} times {}" thrown in tensorflow/models.

Source

Thrown at official/recommendation/ncf_input_pipeline.py:146

    strategy: Distribution strategy used for distributed training. If specified,
      used to assert that evaluation batch size is correctly a multiple of total
      number of devices used.

  Returns:
    (training dataset, evaluation dataset, train steps per epoch,
    eval steps per epoch)

  Raises:
    ValueError: If data is being generated online for when using TPU's.
  """
  # NCF evaluation metric calculation logic assumes that evaluation data
  # sample size are in multiples of (1 + number of negative samples in
  # evaluation) for each device. As so, evaluation batch size must be a
  # multiple of (number of replicas * (1 + number of negative samples)).
  num_devices = strategy.num_replicas_in_sync if strategy else 1
  if (params["eval_batch_size"] % (num_devices *
                                   (1 + rconst.NUM_EVAL_NEGATIVES))):
    raise ValueError("Evaluation batch size must be divisible by {} "
                     "times {}".format(num_devices,
                                       (1 + rconst.NUM_EVAL_NEGATIVES)))

  if params["train_dataset_path"]:
    assert params["eval_dataset_path"]

    train_dataset = create_dataset_from_tf_record_files(
        params["train_dataset_path"],
        input_meta_data["train_prebatch_size"],  # pyrefly: ignore[unsupported-operation]
        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"],

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/recommendation/ncf_input_pipeline.py:146 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/33f5f90908cf2f43. Report an issue: GitHub.