This chapter provides a foundational comparison between TensorFlow and PyTorch, focusing on their core mechanics. If you are accustomed to TensorFlow, this is where we begin to map your existing knowledge to the PyTorch environment.

You will learn to:

• Distinguish TensorFlow's static computation graphs from PyTorch's dynamic, define-by-run approach.
• Compare tf.Tensor with torch.Tensor, including their creation, attributes, and common operations.
• Understand PyTorch's automatic differentiation mechanism, autograd, in relation to TensorFlow's tf.GradientTape.
• Examine how both frameworks integrate with NumPy for efficient data handling.
• Manage computations across different devices, such as CPUs and GPUs, in PyTorch.

We will cover the essential differences and similarities, building a solid base for working with PyTorch. The chapter concludes with practical exercises to apply these core concepts.