Embedding Quantization: 25-45x retrieval speedup, 32x or 4x less memory usage

4 points| cubie | 1 year ago |huggingface.co