tf.keras.layers.Pipeline

View source on GitHub

Applies a series of layers to an input.

Inherits From: Layer, Operation

tf.keras.layers.Pipeline(
    layers, name=None
)

This class is useful to build a preprocessing pipeline, in particular an image data augmentation pipeline. Compared to a Sequential model, Pipeline features a few important differences:

• It's not a Model, just a plain layer.
• When the layers in the pipeline are compatible with tf.data, the pipeline will also remain tf.data compatible. That is to say, the pipeline will not attempt to convert its inputs to backend-native tensors when in a tf.data context (unlike a Sequential model).

Example:

from keras import layers
preprocessing_pipeline = layers.Pipeline([
    layers.AutoContrast(),
    layers.RandomZoom(0.2),
    layers.RandomRotation(0.2),
])

# `ds` is a tf.data.Dataset
preprocessed_ds = ds.map(
    preprocessing_pipeline,
    num_parallel_calls=4,
)

Methods

from_config

View source

@classmethod
from_config(
    config
)

Creates an operation from its config.

This method is the reverse of get_config, capable of instantiating the same operation from the config dictionary.

if "dtype" in config and isinstance(config["dtype"], dict):
    policy = dtype_policies.deserialize(config["dtype"])