Changed num_workers to 8 instead of 16

lower batch size and accumulation argument. Changed misclassified samples to be for last epoch only

Added augmentations as per the original paper

Added graphs for top-1 and top-5 training and test accuracies. Added displaying of misclassified samples.

added table to the training logs to copy to Excel

Added checkpoint and early stopping

Modified code for imagent datase

ResNet 50 base for CIFAR-10