tensorflow/models · error · ValueError

`teacher_model_init_checkpoint` is not specified.

Error message

`teacher_model_init_checkpoint` is not specified.

What it means

Error "`teacher_model_init_checkpoint` is not specified." thrown in tensorflow/models.

Source

Thrown at official/projects/mobilebert/distillation.py:591

    """Checkpoints for model, stage_id, optimizer for preemption handling."""
    return dict(
        stage_id=self._stage_id,
        volatiles=self._volatiles,
        student_pretrainer=self._student_pretrainer,
        teacher_pretrainer=self._teacher_pretrainer,
        encoder=self._student_pretrainer.encoder_network)

  def initialize(self, model):
    """Loads teacher's pretrained checkpoint and copy student's embedding."""
    # This function will be called when no checkpoint found for the model,
    # i.e., when the training starts (not preemption case).
    # The weights of teacher pretrainer and student pretrainer will be
    # initialized, rather than the passed-in `model`.
    del model
    logging.info('Begin to load checkpoint for teacher pretrainer model.')
    ckpt_dir_or_file = self._task_config.teacher_model_init_checkpoint
    if not ckpt_dir_or_file:
      raise ValueError('`teacher_model_init_checkpoint` is not specified.')

    if tf.io.gfile.isdir(ckpt_dir_or_file):
      ckpt_dir_or_file = tf.train.latest_checkpoint(ckpt_dir_or_file)
    # Makes sure the teacher pretrainer variables are created.
    _ = self._teacher_pretrainer(self._teacher_pretrainer.inputs)
    teacher_checkpoint = tf.train.Checkpoint(
        **self._teacher_pretrainer.checkpoint_items)
    teacher_checkpoint.read(ckpt_dir_or_file).assert_existing_objects_matched()

    logging.info('Begin to copy word embedding from teacher model to student.')
    teacher_encoder = self._teacher_pretrainer.encoder_network
    student_encoder = self._student_pretrainer.encoder_network
    embedding_weights = teacher_encoder.embedding_layer.get_weights()
    student_encoder.embedding_layer.set_weights(embedding_weights)

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/projects/mobilebert/distillation.py:591 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/6e8c7269cb8f07a1. Report an issue: GitHub.