tf.io.read_file | TensorFlow v2.16.1 (original) (raw)

tf.io.read_file

Stay organized with collections Save and categorize content based on your preferences.

Reads the contents of file.

View aliases

Compat aliases for migration

tf.compat.v1.io.read_file, tf.compat.v1.read_file

tf.io.read_file(
    filename, name=None
)

Used in the notebooks

Used in the guide	Used in the tutorials
tf.data: Build TensorFlow input pipelines	Simple audio recognition: Recognizing keywords pix2pix: Image-to-image translation with a conditional GAN Transfer learning with YAMNet for environmental sound classification Adversarial example using FGSM Neural style transfer

This operation returns a tensor with the entire contents of the input filename. It does not do any parsing, it just returns the contents as they are. Usually, this is the first step in the input pipeline.

Example:

with open("/tmp/file.txt", "w") as f: f.write("asdf") `` 4 tf.io.read_file("/tmp/file.txt") <tf.Tensor: shape=(), dtype=string, numpy=b'asdf'>

Example of using the op in a function to read an image, decode it and reshape the tensor containing the pixel data:

@tf.function def load_image(filename): raw = tf.io.read_file(filename) image = tf.image.decode_png(raw, channels=3) # the `print` executes during tracing. print("Initial shape: ", image.shape) image.set_shape([28, 28, 3]) print("Final shape: ", image.shape) return image

Args
filename	string. filename to read from.
name	string. Optional name for the op.

Returns
A tensor of dtype "string", with the file contents.