tensorflow/models · error · ValueError

Eval batch size {} is not divisible by {}

Error message

Eval batch size {} is not divisible by {}

What it means

Error "Eval batch size {} is not divisible by {}" thrown in tensorflow/models.

Source

Thrown at official/recommendation/data_pipeline.py:425

    self.eval_batch_size = eval_batch_size
    self.num_train_epochs = num_train_epochs
    self.create_data_offline = create_data_offline

    # Training
    if self._train_pos_users.shape != self._train_pos_items.shape:
      raise ValueError(
          "User positives ({}) is different from item positives ({})".format(
              self._train_pos_users.shape, self._train_pos_items.shape))

    (self._train_pos_count,) = self._train_pos_users.shape
    self._elements_in_epoch = (1 + num_train_negatives) * self._train_pos_count
    self.train_batches_per_epoch = self._count_batches(self._elements_in_epoch,
                                                       train_batch_size,
                                                       batches_per_train_step)

    # Evaluation
    if eval_batch_size % (1 + rconst.NUM_EVAL_NEGATIVES):
      raise ValueError("Eval batch size {} is not divisible by {}".format(
          eval_batch_size, 1 + rconst.NUM_EVAL_NEGATIVES))
    self._eval_users_per_batch = int(eval_batch_size //
                                     (1 + rconst.NUM_EVAL_NEGATIVES))
    self._eval_elements_in_epoch = num_users * (1 + rconst.NUM_EVAL_NEGATIVES)
    self.eval_batches_per_epoch = self._count_batches(
        self._eval_elements_in_epoch, eval_batch_size, batches_per_eval_step)

    # Intermediate artifacts
    self._current_epoch_order = np.empty(shape=(0,))
    self._shuffle_iterator = None

    self._shuffle_with_forkpool = not stream_files
    if stream_files:
      self._shard_root = epoch_dir or tempfile.mkdtemp(prefix="ncf_")
      if not create_data_offline:
        atexit.register(tf.io.gfile.rmtree, self._shard_root)
    else:
      self._shard_root = None

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/recommendation/data_pipeline.py:425 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/2e6ebb808de534dd. Report an issue: GitHub.