A high-throughput and memory-efficient inference and serving engine for LLMs(Fork for contributing. All changes intended for upstream PRs.)
By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).