600% memory footprint reduction on Vector Embeddings Quantization github.com 4 points by betto a day ago
Reduced by 600%?