3 months ago
- Llama.cpp is a widely used method for running LLM models locally.
- Popular LLM servers such as Ollama, lmstudio, and unsloth utilize llama.cpp as their underlying technology.
- Llama.cpp is implemented in C++.
- The author prefers Common Lisp over C++ and believes it can achieve comparable performance with proper declarations.
- A Common Lisp implementation of llama.cpp called llambda.lisp has been created.
- llambda.lisp is available on GitHub at github.com/jrm-code-project/llambda.
- Contributions are sought for features like support for non-gemma architectures, GPU, NPU, and other enhancements.