Model Compression Quiz

5 questions Pass: 70% +25 pts (50 with Pro)

Quiz covering Advanced Training

What you'll be asked

  1. Which of the following model compression techniques reduces the number of bits required to represent each weight in a neural network?
  2. In the context of Knowledge Distillation, what is the primary role of the 'Teacher' model?
  3. When applying unstructured pruning to a neural network, what is the most significant practical trade-off encountered?
  4. What is the primary benefit of Weight Sharing (or Weight Clustering) in model compression?
  5. When performing Post-Training Quantization (PTQ), why is a representative calibration dataset required?

Answers and explanations come after you submit.