A python library for supervised learning methods with PyTorch.
This repo has PyTorch implementations of training a ConvNet.
- Transfer Learning for ResNet18 / ResNet34 / ResNet101 / SqueezeNet
- Differntial Learning rate (Applying smaller learning rates on earlier layers & higher on later layers for higher performance)
- SGDR (SGD with Restarts - Accelerates training)
- CLR (Cyclic Learning Rate - Accelerates training (uses cosine annealing instead of linear))
- Training a simple 28x28 network from scratch
- Basic Dataloaders
- Segmentation Dataloaders (WIP)