Illustration drawn by RepoMasters from the repo’s ID and its language color. Credits
PrototypeCommon

humanrouter/ddtree-mlx

Tree-based speculative decoding for Apple Silicon (MLX). ~10-15% faster than DFlash on code, ~1.5x over autoregressive. First MLX port with custom Metal kernels for hybrid model support.

Owner
humanrouter (personal account)
Stars
Stars: 145
Forks
Forks: 13
Language
Python
License
None
Created
2026
Topics
apple-siliconinferencellmmlxspeculative-decoding
Tier
Prototype (100 to 249 ★) What this tier means
View on GitHub

Nearby in Python

All Python repos