With your TensorFlow environment configured, this chapter introduces the essential building blocks for constructing and training machine learning models. We will cover the primary data structure, the Tensor, along with key operations and concepts like automatic differentiation.

Specifically, you will learn to:

• Define and understand TensorFlow Tensors, including their attributes like shape (e.g., $(batch\_size, features)$) and data type ($dtype$).
• Perform various operations on Tensors, from basic arithmetic to matrix manipulations.
• Use tf.Variable to manage mutable state, critical for holding model parameters that change during training.
• Calculate gradients using automatic differentiation via tf.GradientTape, the mechanism behind optimizing models.
• Convert Python functions into optimized TensorFlow graphs for improved performance using tf.function.

Mastering these concepts provides the necessary foundation for effectively using TensorFlow's higher-level APIs, which we will cover subsequently.