Note that while epochs commonly consist of a single window, they can also contain multiple windows if ``.window()`` is used or there are multiple ``.repeat()`` calls.
Transformations prior to the call to ``.repeat()`` will be cached. However, note that the initial read will not be cached unless there is a subsequent transformation or ``.fully_executed()`` call. Transformations made to the DatasetPipeline after the repeat will always be executed once for each repetition of the Dataset.
For example, in the following pipeline, the ``map(func)`` transformation only occurs once. However, the random shuffle is applied to each repetition in the pipeline. However, if we omitted the map transformation, then the pipeline would re-read from the base data on each repetition.
Result caching only applies if there are *transformation* stages prior to the pipelining operation. If you ``repeat()`` or ``window()`` a Dataset right after the read call (e.g., ``ray.data.read_parquet(...).repeat()``), then the read will still be re-executed on each repetition. This optimization saves memory, at the cost of repeated reads from the datasource. To force result caching in all cases, use ``.fully_executed().repeat()``.
Similar to how you can ``.split()`` a Dataset, you can also split a DatasetPipeline with the same method call. This returns a number of DatasetPipeline shards that share a common parent pipeline. Each shard can be passed to a remote task or actor.
**Code**:
..code-block:: python
# Create a pipeline that loops over its source dataset indefinitely.
1. It allows you to focus on the training logic and automatically handles distributed setup for your framework of choice (PyTorch, Tensorflow, or Horovod).
2. It has out of the box fault-tolerance and elastic training
3. And it comes with support for standard ML tools and features that practitioners love such as checkpointing and logging.
**Code**
..code-block:: python
def train_func():
# This is a dummy train function just iterating over the dataset shard.
# You should replace this with your training logic.
Sometimes, you may want to change the structure of an existing pipeline. For example, after generating a pipeline with ``ds.window(k)``, you may want to repeat that windowed pipeline ``n`` times. This can be done with ``ds.window(k).repeat(n)``. As another example, suppose you have a repeating pipeline generated with ``ds.repeat(n)``. The windowing of that pipeline can be changed with ``ds.repeat(n).rewindow(k)``. Note the subtle difference in the two examples: the former is repeating a windowed pipeline that has a base window size of ``k``, while the latter is re-windowing a pipeline of initial window size of ``ds.num_blocks()``. The latter may produce windows that span multiple copies of the same original data if ``preserve_epoch=False`` is set:
