b9031
📦 llama-cppView on GitHub →
✨ 3 features🔧 2 symbols
Summary
This release focuses on optimizing backend loading by only loading them when necessary, improving initialization efficiency across the common and llama modules.
✨ New Features
- Implemented lazy loading of backends only when required in the common module.
- llama backend initialization now directly calls ggml_backend_load_all().
- Added ggml_backend_load_all() call where llama_backend_init() is not used.