diskann-benchmark 0.51.0

DiskANN is a fast approximate nearest neighbor search library for high dimensional data
{
  "search_directories":[
    "test_data/disk_index_search"
  ],
  "jobs": [
    {
      "type": "graph-index-build-spherical-quantization",
      "content": {
        "build": {
          "data_type": "float32",
          "data": "disk_index_siftsmall_learn_256pts_data.fbin",
          "distance": "squared_l2",
          "start_point_strategy": "medoid",
          "max_degree": 32,
          "l_build": 50,
          "alpha": 1.2,
          "backedge_ratio": 1.0,
          "num_threads": 8,
          "multi_insert": null
        },
        "search_phase": {
          "search-type": "topk-beta-filter",
          "queries": "disk_index_sample_query_10pts.fbin",
          "groundtruth": "disk_index_10pts_idx_uint32_truth_search_filter_res.bin",
          "beta": 0.5,
          "query_predicates": "query.10.label.jsonl",
          "data_labels": "data.256.label.jsonl",
          "reps": 5,
          "num_threads": [
            8
          ],
          "runs": [
            {
              "search_n": 10,
              "search_l": [
                10,
                20,
                30,
                40,
                50,
                60,
                70,
                80,
                90,
                100
              ],
              "recall_k": 10
            }
          ]
        },
        "seed": 12648430,
        "transform_kind": {
          "padding_hadamard": "same"
        },
        "query_layouts": [
          "four_bit_transposed",
          "same_as_data",
          "full_precision"
        ],
        "num_bits": 1,
        "pre_scale": {
            "some": 0.00390625
        }
      }
    },
    {
      "type": "graph-index-build-spherical-quantization",
      "content": {
        "build": {
          "data_type": "float32",
          "data": "disk_index_siftsmall_learn_256pts_data.fbin",
          "distance": "squared_l2",
          "start_point_strategy": "medoid",
          "max_degree": 32,
          "l_build": 50,
          "alpha": 1.2,
          "backedge_ratio": 1.0,
          "num_threads": 8,
          "multi_insert": null
        },
        "search_phase": {
          "search-type": "topk-beta-filter",
          "queries": "disk_index_sample_query_10pts.fbin",
          "groundtruth": "disk_index_10pts_idx_uint32_truth_search_filter_res.bin",
          "beta": 0.5,
          "query_predicates": "query.10.label.jsonl",
          "data_labels": "data.256.label.jsonl",
          "reps": 5,
          "num_threads": [
            8
          ],
          "runs": [
            {
              "search_n": 10,
              "search_l": [
                10,
                20,
                30,
                40,
                50,
                60,
                70,
                80,
                90,
                100
              ],
              "recall_k": 10
            }
          ]
        },
        "seed": 12648430,
        "transform_kind": {
          "padding_hadamard": "natural"
        },
        "query_layouts": [
          "four_bit_transposed",
          "same_as_data",
          "full_precision"
        ],
        "num_bits": 1,
        "pre_scale": {
            "some": 0.00390625
        }
      }
    },
    {
      "type": "graph-index-build-spherical-quantization",
      "content": {
        "build": {
          "data_type": "float32",
          "data": "disk_index_siftsmall_learn_256pts_data.fbin",
          "distance": "squared_l2",
          "start_point_strategy": "medoid",
          "max_degree": 32,
          "l_build": 50,
          "alpha": 1.2,
          "backedge_ratio": 1.0,
          "num_threads": 8,
          "multi_insert": null
        },
        "search_phase": {
          "search-type": "topk-beta-filter",
          "queries": "disk_index_sample_query_10pts.fbin",
          "groundtruth": "disk_index_10pts_idx_uint32_truth_search_filter_res.bin",
          "beta": 0.5,
          "query_predicates": "query.10.label.jsonl",
          "data_labels": "data.256.label.jsonl",
          "reps": 5,
          "num_threads": [
            8
          ],
          "runs": [
            {
              "search_n": 10,
              "search_l": [
                10,
                20,
                30,
                40,
                50,
                60,
                70,
                80,
                90,
                100
              ],
              "recall_k": 10
            }
          ]
        },
        "seed": 12648430,
        "transform_kind": {
          "padding_hadamard": {
            "override": 64
          }
        },
        "query_layouts": [
          "four_bit_transposed",
          "same_as_data",
          "full_precision"
        ],
        "num_bits": 1,
        "pre_scale": {
            "some": 0.00390625
        }
      }
    },
    {
      "type": "graph-index-build-spherical-quantization",
      "content": {
        "build": {
          "data_type": "float32",
          "data": "disk_index_siftsmall_learn_256pts_data.fbin",
          "distance": "squared_l2",
          "start_point_strategy": "medoid",
          "max_degree": 32,
          "l_build": 50,
          "alpha": 1.2,
          "backedge_ratio": 1.0,
          "num_threads": 8,
          "multi_insert": null
        },
        "search_phase": {
          "search-type": "topk-beta-filter",
          "queries": "disk_index_sample_query_10pts.fbin",
          "groundtruth": "disk_index_10pts_idx_uint32_truth_search_filter_res.bin",
          "beta": 0.5,
          "query_predicates": "query.10.label.jsonl",
          "data_labels": "data.256.label.jsonl",
          "reps": 5,
          "num_threads": [
            8
          ],
          "runs": [
            {
              "search_n": 10,
              "search_l": [
                10,
                20,
                30,
                40,
                50,
                60,
                70,
                80,
                90,
                100
              ],
              "recall_k": 10
            }
          ]
        },
        "seed": 12648430,
        "transform_kind": {
          "padding_hadamard": "natural"
        },
        "query_layouts": [
          "scalar_quantized",
          "same_as_data",
          "full_precision"
        ],
        "num_bits": 2,
        "pre_scale": {
            "some": 0.00390625
        }
      }
    },
    {
      "type": "graph-index-build-spherical-quantization",
      "content": {
        "build": {
          "data_type": "float32",
          "data": "disk_index_siftsmall_learn_256pts_data.fbin",
          "distance": "squared_l2",
          "start_point_strategy": "medoid",
          "max_degree": 32,
          "l_build": 50,
          "alpha": 1.2,
          "backedge_ratio": 1.0,
          "num_threads": 8,
          "multi_insert": null
        },
        "search_phase": {
          "search-type": "topk-beta-filter",
          "queries": "disk_index_sample_query_10pts.fbin",
          "groundtruth": "disk_index_10pts_idx_uint32_truth_search_filter_res.bin",
          "beta": 0.5,
          "query_predicates": "query.10.label.jsonl",
          "data_labels": "data.256.label.jsonl",
          "reps": 5,
          "num_threads": [
            8
          ],
          "runs": [
            {
              "search_n": 10,
              "search_l": [
                10,
                20,
                30,
                40,
                50,
                60,
                70,
                80,
                90,
                100
              ],
              "recall_k": 10
            }
          ]
        },
        "seed": 12648430,
        "transform_kind": {
          "random_rotation": "same"
        },
        "query_layouts": [
          "scalar_quantized",
          "same_as_data",
          "full_precision"
        ],
        "num_bits": 4,
        "pre_scale": {
            "some": 0.00390625
        }
      }
    }
  ]
}