tensorflow/models · error · ValueError

Initializer {} not supported

Error message

Initializer {} not supported

What it means

Error "Initializer {} not supported" thrown in tensorflow/models.

Source

Thrown at official/legacy/xlnet/xlnet_modeling.py:37

import tensorflow as tf, tf_keras
from official.legacy.xlnet import data_utils
from official.nlp.modeling import networks


def gelu(x):
  return tf_keras.activations.gelu(x, approximate=True)


def _get_initializer(flags):
  """Get variable initializer."""
  if flags.init_method == "uniform":
    initializer = tf_keras.initializers.RandomUniform(
        minval=-flags.init_range, maxval=flags.init_range)
  elif flags.init_method == "normal":
    initializer = tf_keras.initializers.RandomNormal(stddev=flags.init_std)
  else:
    raise ValueError("Initializer {} not supported".format(flags.init_method))
  return initializer


def rel_shift(x, klen=-1):
  """Performs relative shift to form the relative attention score."""
  x_size = tf.shape(x)

  x = tf.reshape(x, [x_size[1], x_size[0], x_size[2], x_size[3]])
  x = tf.slice(x, [1, 0, 0, 0], [-1, -1, -1, -1])
  x = tf.reshape(x, [x_size[0], x_size[1] - 1, x_size[2], x_size[3]])
  x = tf.slice(x, [0, 0, 0, 0], [-1, klen, -1, -1])

  return x


def _create_mask(qlen, mlen, dtype=tf.float32, same_length=False):
  """Creates attention mask when single-side context allowed only."""
  attn_mask = tf.ones([qlen, qlen], dtype=dtype)

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/legacy/xlnet/xlnet_modeling.py:37 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/2cbf1a455391f683. Report an issue: GitHub.