Skip to content

Gemma-4 MTP: skip pooling/sampling epilogue for MTP graphs (fixes decode crash) + widen DKQ=512 TILE routing to all gqa_ratio#25

Open
PhilEgly wants to merge 2 commits into
AtomicBot-ai:feature/turboquant-kv-cachefrom
PhilEgly:fix/gemma4-mtp-build-pooling-epilogue
Open

Gemma-4 MTP: skip pooling/sampling epilogue for MTP graphs (fixes decode crash) + widen DKQ=512 TILE routing to all gqa_ratio#25
PhilEgly wants to merge 2 commits into
AtomicBot-ai:feature/turboquant-kv-cachefrom
PhilEgly:fix/gemma4-mtp-build-pooling-epilogue

Commits

Commits on Jun 4, 2026

Commits on Jun 6, 2026