HW13 Network Compression

Task description:

• Network Compression: Make your model smaller without losing the performance.
  
• Knowledge Distillation - Train a small model(student model) with teacher model, small model can learn better.
  
• Use Depthwise and Pointwise Convolution to reduce the model parameters.(parameters<10k)
  
• Train an image classifier to classify the kind of food.
  
• Network pruning
  

---

Some methods to be used:

• Residual learning
• move some validation set to training set --> train : 12796, valid : 500
• Using data augmentation like HW3, instead of putting all Data augmentation methods in begin, put one by one in gradually.

---

How to Reproduce the best result

Open the Google Colab
Upload the「ML2022Spring_HW13_ipynb」的副本

Replenishment

• You can't directly run the code, download the data is required.
• Watch out the file directory, and have to load the checkpoint successfully.