b9515
📦 llama-cppView on GitHub →
✨ 2 features🔧 1 symbols
Summary
This release focuses on internal code cleanup by deduplicating imatrix loading logic into a common loader and reintroducing LLAMA_TRACE with early exit checks for missing quantize metadata.
✨ New Features
- Added back LLAMA_TRACE functionality.
- Early exit implemented when quantize metadata is missing.