Normalization

12 problems

• Add & Norm (post-norm residual)Medium
• BatchNorm forward (train + eval modes)Medium
• LayerNorm forwardEasy
• Image preprocessing — scale and standardizeMedium
• Dynamic-Tanh (DyT)Medium
• Total LayerNorm parameters in GPT-2Easy
• LayerNorm statistics — per token, over featuresEasy
• Pre-LN vs post-LN blockMedium
• The residual stream grows without LayerNormMedium
• Group normalizationMedium
• Instance normalizationMedium
• RMSNormEasy