b8169
📦 llama-cppView on GitHub →
✨ 2 features🐛 1 fixes🔧 1 symbols
Summary
This release enhances ggml performance by fixing AMX issues and introducing batched support, leading to faster perplexity calculation times.
✨ New Features
- Added batched support to ggml operations.
- Enabled AMX support in ggml.
🐛 Bug Fixes
- Fixed issues related to AMX usage, resulting in performance improvements and removal of CPU_REPACK memory usage when AMX is active.