tensorflow/models · error · ValueError

Please specify epsilon for unknown input data type

Error message

Please specify epsilon for unknown input data type

What it means

Error "Please specify epsilon for unknown input data type" thrown in tensorflow/models.

Source

Thrown at official/projects/edgetpu/vision/modeling/custom_layers.py:424

  """
  fqop = _fqop if output_type.is_floating else tf.identity
  safe_axis = axis
  if safe_axis < 0:
    safe_axis = len(input_tensor.shape) + safe_axis
  reduction_size = input_tensor.shape[axis]
  axis_max = tf.math.reduce_max(input_tensor, axis=axis, keepdims=True)
  zero_if_max = tf.subtract(axis_max, input_tensor)
  eps = epsilon if epsilon else 1e-6
  if input_tensor.dtype.is_floating:
    zero_if_max_else_eps = tf.math.minimum(
        _nnapi_scalar(eps, input_tensor.dtype), zero_if_max)
    zero_if_max_else_one = zero_if_max_else_eps * _nnapi_scalar(
        1 / eps, input_tensor.dtype)
  elif input_tensor.dtype.is_integer:
    zero_if_max_else_one = tf.math.minimum(
        _nnapi_scalar(1, input_tensor.dtype), zero_if_max)
  else:
    raise ValueError('Please specify epsilon for unknown input data type')

  # Input type ends here, output type starts here
  zero_if_max_else_one = tf.cast(zero_if_max_else_one, dtype=output_type)
  zero_if_max_else_one = fqop(zero_if_max_else_one)
  one_if_max_else_zero = fqop(
      tf.math.subtract(
          fqop(_nnapi_scalar(1, output_type)), zero_if_max_else_one))
  rev_index = tf.range(reduction_size, 0, -1, dtype=output_type)
  for index in range(safe_axis + 1, len(input_tensor.shape)):
    rev_index = tf.expand_dims(rev_index, axis=index - safe_axis)
  rev_index = fqop(rev_index)
  rev_index_if_max_else_zero = fqop(
      tf.math.multiply(one_if_max_else_zero, rev_index))
  reverse_argmax = fqop(
      tf.math.reduce_max(
          rev_index_if_max_else_zero, axis=axis, keepdims=keepdims, name=name))
  # Final operation obtains name if argmax layer if provided
  return fqop(

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/projects/edgetpu/vision/modeling/custom_layers.py:424 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/e700b8b527afcfc5. Report an issue: GitHub.