b8047
A major bug fix for Apple's Metal framework just dropped for all Llama.cpp users.
Deep Dive
The Llama.cpp team released commit b8047, a critical fix for an Apple Silicon Metal GPU acceleration bug (ACC op). This update is essential for developers and users running AI models on macOS and iOS devices with Apple's M-series chips. The fix ensures stable and correct GPU computation, preventing potential crashes or incorrect outputs from popular open-source models like Llama 3 and Mistral running locally.
Why It Matters
This patch stabilizes local AI inference for millions of Mac users, preventing crashes in popular apps like Ollama and LM Studio.