TensorFlow Core: Low-Level TF API

The TF framework comprises:

• Low-level APIs

• High-level APIs

TF Core is the primary low-level API in TF. Developers use TF Core with Python or some other language to create ML and DL applications.

Here we discuss the main concepts and the components of TF Core.

Tensor

This is a multidimensional data array that has four attributes:

• Rank: Number of dimensions of a tensor.

• Shape: The combined size of a tensor, over entire dimensions.

• Type: The datatype of a tensor.

• Label: The name of a tensor.

The tf.Tensor class creates tensor objects in the following ways:

• By defining constants and variables using tf.constant() and tf.Variable(), respectively.

• By defining placeholders and passing the values to the session.run() (mainly supported by TF v1).

• By converting Python objects (scalars, lists, NumPy arrays and pandas DataFrames) using the tf.convert_to_tensor() function.

Constants

The tf.constant method creates constants, which support integer and float data types.