This topic describes the index build parameters and index search parameters of the CAGRA algorithm.
Index build parameters
Parameter | Type | Default value | Description |
proxima.cagra.index.graph_degree | int | 64 | The number of adjacent edges of a node in the nearest neighbor graph during the query phase. A large value increases the recall rate but decreases performance. |
proxima.cagra.index.intermediate_graph_degree | int | 128 | The number of adjacent edges of a node in the nearest neighbor graph that is created in the graph creation phase. During the query phase, the number is reduced to the one indicated by the proxima.cagra.index.graph_degree parameter. The value of this parameter must be no less than and no more than 1.5 times that of proxima.cagra.index.graph_degree. |
proxima.cagra.index.mode.fp16 | int | 0 | Specifies whether to enable FP16 quantization. To avoid impacts on the recall rate, we recommend that you set this parameter to 1. |
Index search parameters
Parameter | Type | Default value | Description |
proxima.cagra.search.use_batchable_searcher | int | 0 | Specifies whether to enable batch processing. We recommend that you set this parameter to 1 to enable batch processing. This way, the system can process query requests that arrive within a short period of time in a batch. This significantly improves query processing efficiency and query performance. |
proxima.cagra.search.use_gpu_built_index | int | 0 | Specifies whether to use the graph built from the CPU. You can set this parameter only to 0. This indicates that the graph built from the CPU is used. |
proxima.cagra.search.itopk_size | int | 64 | This parameter is related to the size of the candidate set during algorithm iteration. The value must be a multiple of 32 and greater than or equal to the required topK value. If top 100 documents need to be returned, the value of this parameter must be greater than or equal to 100. A large value increases the recall rate but decreases performance. |