feat(gemma4): feature-complete backend with DFlash + MTP + sparse-FA decode (supersedes PR #175 skeleton) - #193

Closed

dusterbloom wants to merge 4 commits into

Luce-Org:mainfrom

dusterbloom:feature/gemma4-feature-complete

Commits on May 14, 2026

feat(gemma4): feature-complete backend with DFlash + MTP + sparse-FA decode (supersedes PR #175 skeleton)
dusterbloom
committed
b0b4b7e
refactor(gemma4): rename use_pflash → use_sparse_fa (clarify ggml_flash_attn_sparse is not the PFlash product)
dusterbloom
committed
9b311d0

Commits on May 15, 2026