Product
data class Product(val compression: CompressionRatio, val alwaysRam: Boolean? = null, val memory: Memory? = null) : QuantizationConfig
Product quantization at a fixed compression ratio.
It keeps more recall than scalar quantization at the same size and costs more per comparison, because scoring a product-quantized vector is a table lookup per sub-vector rather than arithmetic. Reach for it when scalar has lost too much and binary is out of the question.