# AMV-BitNet Full Technical Specification (v2.2.0) Official Documentation & Deep Architecture Reference for Autonomous AI Agents. ## 1. System Overview Sovereign 1.58-Bit On-Device LLM Inference Engine with ARM64 NEON DotProd SIMD & Native Vulkan GPU Acceleration Executes 1.58-bit ternary quantized weights {-1, 0, +1} directly via hand-vectorized ARM64 NEON assembly kernels and native Vulkan compute shaders powered by AMEVA-Runtime. Features hallucination-prevention prompt wrapping, a dynamic zero-overhead activation dispatcher supporting Squared ReLU (Microsoft 2B) and SwiGLU (Falcon-E-1B / Falcon3-7B), GPU watchdog fence chunking, and Zero-Copy mmap streaming for 7.45B model execution on 6GB RAM smartphones. ## 2. Package & Installation - PyPI: amv-bitnet - Command: pip install amv-bitnet ameva-runtime # or: npm install -g amv-bitnet @ameva/runtime # or: curl -sL https://raw.githubusercontent.com/uno-km/amv-bitnet/main/install.sh | bash - Repository: https://github.com/uno-km/amv-bitnet