Schedule Visualization
0.001
Initial LR
0.0001
Final LR
0.00001
Min LR
0.001
Max LR
Learning Rate vs Epoch
Epoch 0
Epoch 50
Epoch 100
Schedule Comparison
Add schedules to compare
PyTorch Implementation
import torch.optim as optim from torch.optim.lr_scheduler import StepLR optimizer = optim.Adam(model.parameters(), lr=0.001) scheduler = StepLR(optimizer, step_size=30, gamma=0.1) for epoch in range(100): train(...) scheduler.step()
💡 Best Practices
- Step decay works well for image classification tasks
- Use warmup when training with large batch sizes
- Monitor validation loss to avoid premature decay
LR Values at Key Epochs
| Epoch | Learning Rate | % of Initial | Log10(LR) |
|---|