Repository Issue Activity (beta)

#1016 Reduce KV memory for local-attention layers

#1018 main: reduce repeated value loads in Gemma 3 vision attention

#1002 Add experimental W8A8 MatMul with optional QuaRot-style rotation

#982 Add SmolLM2 support — reuse the Qwen3 converter path?

#986 rope_theta is ignored on the dense attention path (hardcoded 10000 / 1000000)

#990 Feature Request / Proposal: "E12B" (Effective 12B) Model Class for Gemma 5 (Optimized for 12GB VRAM / Desktop Users)

#979 Bad build UX: CMake configure is silent for minutes while cloning dependencies

#198 Gemma.cpp hangs on a Gemma 7B model that was finetuned using huggingface peft(QLoRA)

#887 Is Gemma4 supported by gemma.cpp?

#661 T5Gemma support

#888 Support TranslateGemma 4B text-only inference (Gemma 3 4B LM-only, no VIT)

#911 Give a known-good example for https://github.com/google/gemma.cpp/blob/main/API_SERVER_README.md

#904 Fixes

#898 Cpu and environmental errors

#885 LLM output regression from #857 building with Emscripten

#875 Showcase / question: a board-proven offline language runtime on ESP32-C3, and whether this points to a more extreme form of language/runtime co-design

#164 Near-term roadmap

#853 Luminarit‑GreenRAM_V2.1–Server‑Grade

#762 PaliGemma2 ViT Heap Buffer Overflow

#826 Batch mode in GPU

#817 Segmentation fault using gemma 3 with --image_file candy.ppm

#17 [Feature request] Add quantization methods

#778 Build fails on Linux (GCC 15): SentencePiece missing <cstdint> + Highway avx10.2-512 errors

#758 integrate gemma.cpp for Android Galaxy s24 devices

#726 Abort / Assertion-error when loading gemma-3/gemmaCpp/3.0-1b-it-sfp