LEAP: Linear Explainable Attention in Parallel for causal language modeling with O(1) path length, and O(1) inference
Primary LanguageJupyter NotebookCreative Commons Zero v1.0 UniversalCC0-1.0