SmoothQuant and Activation Quantization Quiz

Answers and explanations appear as you take the quiz.

1. What problem does SmoothQuant specifically address in low-precision inference?
   • Extreme outlier values in activation channels
   • Weights that are too small to represent
   • Gradient vanishing during backpropagation
   • Token vocabulary size limits
2. How does SmoothQuant make activations easier to quantize?
   • It clips all activations above a fixed threshold
   • It migrates difficulty by scaling activations down and weights up by the same factor
   • It retrains the model end-to-end in 8-bit
   • It converts activations to floating point only
3. What does the notation W8A8 mean?
   • 8 weight layers and 8 attention heads
   • Weight decay of 8 and alpha of 8
   • 8-bit weights and 8-bit activations
   • 8 workers and 8 accelerators
4. Why does SmoothQuant's scaling add essentially no runtime overhead?
   • It runs on a separate dedicated GPU
   • It only applies during training, never inference
   • Modern CPUs ignore scaling factors
   • The scales are absorbed offline into adjacent weights or fused operations
5. What role does the alpha hyperparameter play in SmoothQuant?
   • It controls how much outlier burden shifts from activations to weights
   • It sets the learning rate for fine-tuning
   • It determines the number of quantization bits
   • It selects which attention heads to prune
6. What result did the SmoothQuant paper report for its evaluated W8A8 large-language-model deployments?
   • A guaranteed 10× speedup on every GPU
   • About 1.56× speedup and roughly 2× memory savings with near-full-precision accuracy
   • No runtime or memory benefit
   • A 50% accuracy drop in every tested model
7. Why does multiplying activation channels by 1/s and corresponding weights by s preserve a linear layer’s product?
   • The output is rounded to the nearest class
   • The model replaces multiplication with addition
   • The reciprocal scale factors cancel in the activation-weight multiplication
   • The weights are set to zero after scaling
8. In SmoothQuant’s scale formula, what does increasing alpha generally do to outlier handling?
   • It removes the model’s weight matrices
   • It turns W8A8 into W16A16
   • It eliminates all quantization error
   • It shifts more scaling burden away from activations and onto weights