PyTorch Classification

Minimal and clean training and evaluation codes for baseline performances.

Code Features

• Learning Rate Scheduling is implemented with torch.optim.lr_Scheduler
• Tensorboard visualization is added
• Early stopping is implemented
• Best parameters for validation accuracy is saved
• Confusion matrix for validation set is generated
• Optuna hyper parameter search framework is used to find best parameters

Training and Validation Sets Accuracies for ResNet18