This is a part of CNN module of guvi.in, and the code is the same as used in their courses.
A CNN model using CNN built from scratch, as well as using PyTorch's sequential and inbuilt modules, to build a image classifier on CIFAR 10 dataset. The model accuracy on CPU for 64 epochs was Test acc: 53.21, Train acc: 87.58
The same on GPU for 200 epochs was Test acc: 50.13, Train acc: 94.00 The model can be tweaked more by varying the hyperparameters, as well as varying the batch size of the dataset so as to gain maximum accuracy on both train and test data.