b9070
📦 llama-cppView on GitHub →
✨ 1 features🐛 3 fixes🔧 1 symbols
Summary
This release introduces Q4_0 MoE GEMM support for OpenCL on Adreno devices and includes several minor cleanup and formatting fixes within the OpenCL backend.
✨ New Features
- Added Q4_0 MoE GEMM support for OpenCL on Adreno GPUs.
🐛 Bug Fixes
- Fixed whitespace issues in OpenCL code.
- Removed unused cl_program definitions in OpenCL.
- Refactored OpenCL #if blocks for improved clarity.