tensorflow/models · error · ValueError

likelihood={self.likelihood} only support univariate logits.

Error message

likelihood={self.likelihood} only support univariate logits.Got logits dimension: {logits.shape[-1]}

What it means

Error "likelihood={self.likelihood} only support univariate logits.Got logits dimension: {logits.shape[-1]}" thrown in tensorflow/models.

Source

Thrown at official/nlp/modeling/layers/gaussian_process.py:343

            shape=(gp_feature_dim, gp_feature_dim),
            dtype=self.dtype,
            initializer=tf_keras.initializers.Identity(self.ridge_penalty),
            trainable=False,
            aggregation=tf.VariableAggregation.ONLY_FIRST_REPLICA))
    self.built = True

  def make_precision_matrix_update_op(self,
                                      gp_feature,
                                      logits,
                                      precision_matrix):
    """Defines update op for the precision matrix of feature weights."""
    if self.likelihood != 'gaussian':
      if logits is None:
        raise ValueError(
            f'"logits" cannot be None when likelihood={self.likelihood}')

      if logits.shape[-1] != 1:
        raise ValueError(
            f'likelihood={self.likelihood} only support univariate logits.'
            f'Got logits dimension: {logits.shape[-1]}')

    batch_size = tf.shape(gp_feature)[0]
    batch_size = tf.cast(batch_size, dtype=gp_feature.dtype)

    # Computes batch-specific normalized precision matrix.
    if self.likelihood == 'binary_logistic':
      prob = tf.sigmoid(logits)
      prob_multiplier = prob * (1. - prob)
    elif self.likelihood == 'poisson':
      prob_multiplier = tf.exp(logits)
    else:
      prob_multiplier = 1.

    gp_feature_adjusted = tf.sqrt(prob_multiplier) * gp_feature
    precision_matrix_minibatch = tf.matmul(
        gp_feature_adjusted, gp_feature_adjusted, transpose_a=True)

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/nlp/modeling/layers/gaussian_process.py:343 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/36f089121a872d7f. Report an issue: GitHub.