Freaks City

Model Structure The basic unit consists of token embedding with positional encoding, encoder, and decoder. Encoder: Self-attention layer with skip connections and layer normalization, followed by a feed-forward network (FFN) with skip connections and layer normalization. Decoder: Self-attention layer with skip connections and layer normalizati ...

Attention Mechanisms Multi-Head Attention The attention mechanism computes: The scaling factor \(\sqrt{d_k}\) prevents large inner product values that could cause gradient instability. Assuming Q and K elements have mean 0 and variance \(\sigma^2\), the variance of \(QK^T\) grows with \(d_k\). Scaling by \(\sqrt{d_k}\) maintains stable varianc ...