メモ: torch.nn.functional.kl_div
torch.nn.functional.kl_divのメモ
Notes
RoPE Implementation
How RoPE is implemented in practice.
メモ: PCA
PCAの基礎と実装
LogitLens without bias
Removing bias term from LogitLens.
RMSNorm and LayerNorm
Relationship between RMSNorm and LayerNorm and rewritten LayerNorm using centering matrix.
