classifiers_pytorch

This repository has classifiers designed for different open source datasets using pytorch.

1. FashionMNIST: The final model in this notebook is a 3 layer CNN with dropout, SGD Optimizer and cross entropy loss. I have used early stopping, which can be disabled and I have used "wandb" to log data which is disabled by default but there are instructions in the notebook to do that. I got 92.94% test accuracy (contact me for the pre trained model) when I added data augmentation.
2. Wine data classifier: More details in the folder.
3. MNIST dataset: MNIST classifer which achieves 99.33% test accuracy, used 2 layer convolutional network.