Beginning Anomaly Detection Using

 Intro to PyTorch: A Simple Classifier Model

Now that you have a better idea of what a CNN is and how a classifier model looks like in 

Keras, let’s jump straight into implementing a CNN in PyTorch.

Figure 3-53.  The output of running the code in Figure 

3-52

Chapter 3   IntroduCtIon to deep LearnIng

---

112

PyTorch doesn’t abstract everything to the extent that Keras does, so there’s a bit 

more syntax involved. If you would like to explore this framework further, check out 

Appendix B, where we cover the basics of PyTorch, its functionality, and apply it to the 

models that you will explore in Chapter 

7

.

Just like in Keras, however, you start by importing the necessary modules and 

defining your hyperparameters (Figure 

3-54

).

In PyTorch, you must specify to torch that you want to use the GPU if it exists. In 

Keras, since you are using tensorflow-gpu as the back end (what Keras runs on top of), it 

is expected that you have a GPU, CUDA, and cuDNN installed.

Now configure your hyperparameters (Figure 

3-55

).