b8950
New release fixes parsing edge cases for Gemma 4 across 20+ platforms...
The llama.cpp project, a popular open-source C/C++ implementation of LLM inference, released version b8950 on April 27. This release focuses on enhancing the Gemma 4 model support by adding additional tests to handle parsing edge cases, as detailed in pull request #22420. The commit was signed with GitHub's verified signature (GPG key ID: B5690EEEBB952194), ensuring authenticity.
This version is built for an extensive range of platforms: macOS (Apple Silicon arm64 with optional KleidiAI acceleration, Intel x64, iOS XCFramework), Linux (Ubuntu x64, arm64, s390x CPUs; Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP16 backends), Android arm64, Windows (x64, arm64 CPUs; CUDA 12.4/13.1 DLLs, Vulkan, SYCL, HIP), and openEuler (x86 310p, 910b with ACL Graph; aarch64 310p, 910b with ACL Graph). This broad support ensures that developers can run Gemma 4 inference on virtually any hardware setup.
- Adds additional tests for common/gemma4 to handle parsing edge cases
- Supports 20+ build configurations across macOS, Linux, Windows, Android, and openEuler
- Includes GPU acceleration via Vulkan, ROCm 7.2, CUDA 12/13, and SYCL backends
Why It Matters
Improves Gemma 4 model reliability on diverse hardware, enabling robust local LLM inference.