PUMA Vault
Search
Buscar
Modo oscuro
Modo claro
Explorador
Etiqueta: feed-forward
1 artículo con esta etiqueta.
01 may 2026
Transformer Architecture and Mixture of Experts — Technical Reference for PUMA Models
permanent
transformer
attention
moe
mixture-of-experts
architecture
deepseek
mixtral
llama
scaling
puma-core
research
llm
models
technical-reference
self-attention
feed-forward
positional-encoding