TFDS now supports the Croissant 🥐 format! Read the documentation to know more.

tfds.benchmark

View source on GitHub

Benchmarks any iterable (e.g tf.data.Dataset).

tfds.benchmark(
 ds: Iterable[Any], *, num_iter: Optional[int] = None, batch_size: int = 1
) -> BenchmarkResult

Used in the notebooks

Used in the tutorials

Usage:

ds = tfds.load('mnist', split='train')
ds = ds.batch(32).prefetch(buffer_size=tf.data.AUTOTUNE)
tfds.benchmark(ds, batch_size=32)

Reports:

  • Total execution time
  • Setup time (first warmup batch)
  • Number of examples/sec

Args

ds Dataset to benchmark. Can be any iterable. Note: The iterable will be fully consumed.
num_iter Number of iteration to perform (iteration might be batched)
batch_size Batch size of the dataset, used to normalize iterations

Returns

statistics The recorded statistics, for eventual post-processing

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2024年04月26日 UTC.