PyTorch Mini Projects

A collection of small-scale projects that helped me learn the basics of the PyTorch framework

Project #1: Simple Neural Network on Fashion MNIST dataset

What I learned:

• Linear Layers
• Activation functions
• Optimizers
• The low-level logic of Network hidden layers
• Creating a model class
• Handling the GPU
• Neural Network training pipeline
• Neural Network performance evaluation

Project #2: Residual Neural Networks on CIFAR10 dataset

What I learned:

• Data transformations
• Residual blocks
• Logic of ResNets
• Adam Optimizer
• One cycle policy learning rate scheduler
• Tuning ResNets

Project 3: DCGANs on MNIST dataset

What I learned:

• Objectives of Discriminators and Generators
• Deconvolutions
• Using latent tensors or noise to generate fake outputs
• Loss function of the Generator
• Training pipeline of DCGANs
• Tuning DCGANs

DCGAN Result: