🤖FFPA: Extends FA-2/3 via Split-D for large headdims, 1.5x~6×↑🎉 vs SDPA, up to 513~535 TFLOPS🎉 on NVIDIA H200.
-
Updated
Jul 24, 2026 - Python
🤖FFPA: Extends FA-2/3 via Split-D for large headdims, 1.5x~6×↑🎉 vs SDPA, up to 513~535 TFLOPS🎉 on NVIDIA H200.
Autonomous AI Agent Skill for Google AI Edge Gallery - self-evolving, 100% offline
Gemma 4 Agent Skills
The first native Swift inference engine for Gemma 4 on Apple Silicon / iOS
Gemma 4 31B Abliterated — quality-preserving guardrail removal for Google's most capable open model. Apache 2.0. Runs on Apple Silicon via MLX.
Run Gemma 4 locally with Gemma-4 Omni-Desktop. A free, offline AI agent optimized for RTX 30xx & 8GB Macs. Features native vision, voice & OpenClaw support.
Reproducible recipe: serve abliterated Gemma-4-12B (gemma4_unified) at 50-118 tok/s on no-NVLink Blackwell (SM120) via vLLM nightly + ModelOpt FP8/NVFP4 + MTP spec-decode.
PDF redaction that runs entirely on your device. Five detection layers including Gemma 4 contextual PHI. No uploads, no servers, no leaks.
Domain-specialized Gemma 2 27B + Gemma 4 31B for SEC filings — fine-tuned on TPU v6e-8 with PyTorch/XLA FSDPv2, plus a Vertex AI Vector Search RAG demo (69 tickers × 381 filings). Same LoRA recipe, +3.5% / +5.8% BERTScore F1.
On-device LLM client for Android. Fork of Google AI Edge Gallery
Using Google' Gemma 4 Open weights for predicting and analyzing dermatology images, diagnosis and medication
HarmonyOS NEXT native Gemma 4 MNN on-device LLM chat demo
Kotlin Multiplatform engine for running Gemma LLMs on-device on Android via LiteRT-LM — stateful KV-cache chat sessions, resumable model management, function calling. Includes NativeLM, a private Local AI chat app. AGPL-3.0 / commercial.
Run Google's Gemma 4 entirely in your browser via WebGPU. Multimodal chat, E2B vs E4B arena, ONNX conversion toolkit.
DueCare: Agentic LLM safety harness for migrant-worker protection — Gemma 4 Good Hackathon submission
Local/edge agentic coding assistant powered by Gemma 4. Drop-in rulebook from claude-code-pro-pack. Works on Raspberry Pi.
Likas, in Filipino, means nature — yet it also means to evacuate. An AI-powered platform that helps citizens evacuate during disasters caused by nature.
Ash — offline survival assistant for iOS. Gemma 4 E2B/E4B fully on-device (text · image · voice) with RAG-grounded answers over 56 emergency-response packs. Built for the Kaggle Gemma 4 Good Hackathon.
Offline-first, multilingual emergency survival guidance powered by real on-device Gemma 4 inference.
Add a description, image, and links to the gemma-4 topic page so that developers can more easily learn about it.
To associate your repository with the gemma-4 topic, visit your repo's landing page and select "manage topics."