Show HN: Turboquant.cpp – 학습 없이 임베딩을 1-4비트로 양자화 (400 LoC)

Show HN: Turboquant.cpp – Quantize embeddings to 1-4 bits, no training (400 LoC)