Creates a dataset that takes a Bernoulli sample of the contents of another dataset.

    input_dataset, rate, seed, seed2, output_types, output_shapes, name=None

There is no transformation in the Python API for creating this dataset. Instead, it is created as a result of the filter_with_random_uniform_fusion static optimization. Whether this optimization is performed is determined by the experimental_optimization.filter_with_random_uniform_fusion option of


  • input_dataset: A Tensor of type variant.
  • rate: A Tensor of type float32. A scalar representing the sample rate. Each element of input_dataset is retained with this probability, independent of all other elements.
  • seed: A Tensor of type int64. A scalar representing seed of random number generator.
  • seed2: A Tensor of type int64. A scalar representing seed2 of random number generator.
  • output_types: A list of tf.DTypes that has length >= 1.
  • output_shapes: A list of shapes (each a tf.TensorShape or list of ints) that has length >= 1.
  • name: A name for the operation (optional).


A Tensor of type variant.