Hands-On GPU Computing with Python

The two primary features of PyTorch

Now that we've got an overview of PyTorch, let's explore further and discuss its nitty-gritty. The two primary features of PyTorch are as follows:

• It assists in tensor computation (like NumPy) with strong GPU acceleration (similar to CuPy, as discussed in Chapter 8, Working with Anaconda, CuPy, and Numba for GPUs).
• PyTorch is essential for deep neural networks built on a tape-based automatic differentiation (as discussed in the previous section) system.

An epoch is one complete training cycle on a training dataset. Once every sample in the set is seen, the second epoch begins. The AutoGrad module in PyTorch implements automatic differentiation. It is helpful in building neural networks and can save time ...