tf.data.experimental.at

Returns the element at a specific index in a datasest.

tf.data.experimental.at(
    dataset, index
)

Currently, random access is supported for the following tf.data operations:

• tf.data.Dataset.from_tensor_slices,
• tf.data.Dataset.from_tensors,
• tf.data.Dataset.shuffle,
• tf.data.Dataset.batch,
• tf.data.Dataset.shard,
• tf.data.Dataset.map,
• tf.data.Dataset.range,
• tf.data.Dataset.zip,
• tf.data.Dataset.skip,
• tf.data.Dataset.repeat,
• tf.data.Dataset.list_files,
• tf.data.Dataset.SSTableDataset,
• tf.data.Dataset.concatenate,
• tf.data.Dataset.enumerate,
• tf.data.Dataset.parallel_map,
• tf.data.Dataset.prefetch,
• tf.data.Dataset.take,

• tf.data.Dataset.cache (in-memory only)

  Users can use the cache operation to enable random access for any dataset, even one comprised of transformations which are not on this list. E.g., to get the third element of a TFDS dataset:

  ds = tfds.load("mnist", split="train").cache()
  elem = tf.data.Dataset.experimental.at(ds, 3)