algos.yaml

float:
  any:
    sptag:
      docker-tag: ann-benchmarks-sptag
      module: ann_benchmarks.algorithms.sptag
      constructor: Sptag
      base-args: ["@metric"]
      run-groups:
        sptag:
          args: [['BKT', 'KDT']]
          query-args: [[100, 200, 400, 1000, 2000, 4000]]
    DolphinnPy:
      disabled: true
      docker-tag: ann-benchmarks-dolphinn # Docker tag
      module: ann_benchmarks.algorithms.dolphinnpy # Python class
      constructor: DolphinnPy # Python class name
      run-groups:
        base:
          args: [[10, 50, 100, 200, 1000, 2000]]
    faiss-lsh:
      disabled: true
      docker-tag: ann-benchmarks-faiss
      module: ann_benchmarks.algorithms.faiss
      constructor: FaissLSH
      base-args: ["@metric"]
      run-groups:
        base:
          # When @args is a list, the result is the Cartesian product of all of
          # the things it contains; entries that aren't a list will be treated
          # as lists of length one.
          args: [[32, 64, 128, 256, 512, 1024, 2048, 4096]]
          # This run group will produce eight algorithm instances:
          # FaissLSH(32), FaissLSH(64), and so on up to FaissLSH(4096).
    faiss-ivf:
      docker-tag: ann-benchmarks-faiss
      module: ann_benchmarks.algorithms.faiss
      constructor: FaissIVF
      base-args: ["@metric"]
      run-groups:
        base:
          args: [[32,64,128,256,512,1024,2048,4096,8192]]
          query-args: [[1, 5, 10, 50, 100, 200]]
    faiss-gpu:
      disabled: true
      docker-tag: ann-benchmarks-faiss
      module: ann_benchmarks.algorithms.faiss_gpu
      constructor: FaissGPU
      run-groups:
        base:
          args: [[400, 1024, 4096, 8192, 16384],
                 [1, 10, 40, 100, 200]]
    hnswlib:
      docker-tag: ann-benchmarks-hnswlib
      module: ann_benchmarks.algorithms.hnswlib
      constructor: HnswLib
      base-args: ["@metric"]
      run-groups:
        M-4:
          arg-groups:
            - {"M": 4,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-8:
          arg-groups:
            - {"M": 8,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-12:
          arg-groups:
            - {"M": 12,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-16:
          arg-groups:
            - {"M": 16,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-24:
          arg-groups:
            - {"M": 24,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-36:
          arg-groups:
            - {"M": 36,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-48:
          arg-groups:
            - {"M": 48,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-64:
          arg-groups:
            - {"M": 64,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-96:
          arg-groups:
            - {"M": 96,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]

    hnsw(faiss):
     docker-tag: ann-benchmarks-faiss
     module: ann_benchmarks.algorithms.faiss_hnsw
     constructor: FaissHNSW
     base-args: ["@metric"]
     run-groups:
       M-4:
         arg-groups:
           - {"M": 4,  "efConstruction": 500}
         query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
       M-8:
         arg-groups:
           - {"M": 8,  "efConstruction": 500}
         query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
       M-12:
         arg-groups:
           - {"M": 12,  "efConstruction": 500}
         query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
       M-16:
         arg-groups:
           - {"M": 16,  "efConstruction": 500}
         query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
       M-24:
         arg-groups:
           - {"M": 24,  "efConstruction": 500}
         query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
       M-36:
         arg-groups:
           - {"M": 36,  "efConstruction": 500}
         query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
       M-48:
         arg-groups:
           - {"M": 48,  "efConstruction": 500}
         query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
       M-64:
         arg-groups:
           - {"M": 64,  "efConstruction": 500}
         query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
       M-96:
         arg-groups:
           - {"M": 96,  "efConstruction": 500}
         query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]

    flann:
     docker-tag: ann-benchmarks-flann
     module: ann_benchmarks.algorithms.flann
     constructor: FLANN
     base-args: ["@metric"]
     run-groups:
       flann:
         args: [[0.2, 0.5, 0.7, 0.8, 0.9, 0.95, 0.97]]
    annoy:
      docker-tag: ann-benchmarks-annoy
      module: ann_benchmarks.algorithms.annoy
      constructor: Annoy
      base-args: ["@metric"]
      run-groups:
        annoy:
          args: [[100, 200, 400]]
          query-args: [[100, 200, 400, 1000, 2000, 4000, 10000, 20000, 40000,
                      100000, 200000, 400000]]
          # This run group produces 3 algorithm instances -- Annoy("angular",
          # 100), Annoy("angular", 200), and Annoy("angular", 400) -- each of
          # which will be used to run 12 different queries.
    milvus:
      docker-tag: ann-benchmarks-milvus
      module: ann_benchmarks.algorithms.milvus
      constructor: Milvus
      base-args: ["@metric"]
      run-groups:
        milvus:
          args: [['IVF_FLAT', 'IVF_SQ8'], [100, 300, 1000, 3000, 10000, 30000]]  # nlist
          query-args: [[1, 3, 10, 30, 100, 300]]  # nprobe (should be <= nlist)
    nearpy:
      disabled: true
      docker-tag: ann-benchmarks-nearpy
      module: ann_benchmarks.algorithms.nearpy
      constructor: NearPy
      base-args: ["@metric"]
      run-groups:
        nearpy:
          args: [[10, 12, 14, 16], [5, 10, 20, 40]]
        extra:
          args: [16, [5, 10, 15, 20, 25, 30, 40]]
    n2:
      docker-tag: ann-benchmarks-n2
      module: ann_benchmarks.algorithms.n2
      constructor: N2
      base-args: ["@metric"]
      run-groups:
        M-4:
          arg-groups:
            - {"M": 4,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-8:
          arg-groups:
            - {"M": 8,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-12:
          arg-groups:
            - {"M": 12,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-16:
          arg-groups:
            - {"M": 16,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-24:
          arg-groups:
            - {"M": 24,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-36:
          arg-groups:
            - {"M": 36,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-48:
          arg-groups:
            - {"M": 48,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-64:
          arg-groups:
            - {"M": 64,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-96:
          arg-groups:
            - {"M": 96,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]] 

    bruteforce:
      disabled: true
      docker-tag: ann-benchmarks-sklearn
      module: ann_benchmarks.algorithms.bruteforce
      constructor: BruteForce
      base-args: ["@metric"]
      run-groups:
        empty:
          args: []
    bruteforce-blas:
      docker-tag: ann-benchmarks-sklearn
      module: ann_benchmarks.algorithms.bruteforce
      constructor: BruteForceBLAS
      base-args: ["@metric"]
      run-groups:
        empty:
          args: []
    dummy-algo-st:
      disabled: true
      docker-tag: ann-benchmarks-sklearn
      module: ann_benchmarks.algorithms.dummy_algo
      constructor: DummyAlgoSt
      base-args: ["@metric"]
      run-groups:
        empty:
          args: []
    dummy-algo-mt:
      disabled: true
      docker-tag: ann-benchmarks-sklearn
      module: ann_benchmarks.algorithms.dummy_algo
      constructor: DummyAlgoMt
      base-args: ["@metric"]
      run-groups:
        empty:
          args: []
    ball:
      disabled: true
      docker-tag: ann-benchmarks-sklearn
      module: ann_benchmarks.algorithms.balltree
      constructor: BallTree
      base-args: ["@metric"]
      run-groups:
        ball:
          args: &treeargs [[10, 20, 40, 100, 200, 400, 1000]]
    kd:
      docker-tag: ann-benchmarks-sklearn
      module: ann_benchmarks.algorithms.kdtree
      constructor: KDTree
      base-args: ["@metric"]
      run-groups:
        ball:
          args: *treeargs
    BallTree(nmslib):
      docker-tag: ann-benchmarks-nmslib
      module: ann_benchmarks.algorithms.nmslib
      constructor: NmslibReuseIndex
      base-args: ["@metric", "vptree"]
      run-groups:
        base:
          # When @args is a dictionary, algorithm instances will be generated
          # by taking the Cartesian product of all of its values.
          arg-groups:
                - {"tuneK": 10, "desiredRecall": [0.99, 0.97, 0.95, 0.9, 0.85, 0.8,
                    0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1]}
                - False
          # This run group produces thirteen algorithm instances:
          # NmslibNewIndex("angular", "vptree", {"tuneK": 10,
          # "desiredRecall": 0.99}), NmslibNewIndex("angular", "vptree",
          # {"tuneK": 10, "desiredRecall": 0.97}), and so on up to
          # NmslibNewIndex("angular", "vptree", {"tuneK": 10, "desiredRecall":
          # 0.1}).
    pynndescent:
      docker-tag: ann-benchmarks-pynndescent
      module: ann_benchmarks.algorithms.pynndescent
      constructor: PyNNDescent
      base-args: ["@metric"]
      run-groups:
        NN-10-20:
          arg-groups:
            - {"n_neighbors": [10, 20], "diversify_prob": [1.0],
               "pruning_degree_multiplier":[1.5, 2.0], "leaf_size": 32}
          query-args: [[0.0, 0.02, 0.04, 0.06, 0.08, 0.10, 0.12, 0.14, 0.16]]
        NN-40-80:
          arg-groups:
            - {"n_neighbors": [40, 80], "diversify_prob": [0.0, 1.0],
               "pruning_degree_multiplier":[2.0, 2.5], "leaf_size": 64}
          query-args: [[0.0, 0.04, 0.08, 0.12, 0.16, 0.20, 0.24, 0.28, 0.32]]
    NGT-panng:
      docker-tag: ann-benchmarks-ngt
      module: ann_benchmarks.algorithms.panng_ngt
      constructor : PANNG
      base-args : ["@metric", "Float"]
      run-groups :
        panng:
           args : [{'edge': 20, 'pathadj': 40, 'searchedge': 60}]
           query-args : [[0.6, 0.8, 0.9, 1.0, 1.02, 1.05, 1.1, 1.2]]
    NGT-onng:
      docker-tag: ann-benchmarks-ngt
      module: ann_benchmarks.algorithms.onng_ngt
      constructor : ONNG
      base-args : ["@metric", "Float", 0.1]
      run-groups :
        onng:
           args : [{'edge': 100, 'outdegree': 10, 'indegree': 120}]
           query-args : [[0.6, 0.9, 1.0, 1.02, 1.03, 1.05, 1.07, 1.1, 1.2]]
        onng-80-60-graph:
           args : [{'edge': 100, 'outdegree': 10, 'indegree': 80, 'search_edge': 60, 'tree': False}]
           query-args : [[0.6, 0.9, 0.995, 0.998, 1.0, 1.01, 1.02, 1.05, 1.07]]
    mrpt:
      docker-tag: ann-benchmarks-mrpt
      module: ann_benchmarks.algorithms.mrpt
      constructor: MRPT
      base-args: ["@metric"]
      run-groups:
        mrpt:
          args: ["@count"]
          query-args: [[0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.85, 0.9,
                        0.925, 0.95, 0.97, 0.98, 0.99, 0.995]]
    elastiknn-exact:
      # Disabled because they are < 10 q/s, which will timeout on default settings.
      disabled: true
      docker-tag: ann-benchmarks-elastiknn
      module: ann_benchmarks.algorithms.elastiknn
      constructor: Exact
      base-args: [ "@metric", "@dimension" ]
      run-groups:
        exact:
          args: []

  euclidean:
    kgraph:
      docker-tag: ann-benchmarks-kgraph
      module: ann_benchmarks.algorithms.kgraph
      constructor: KGraph
      base-args: ["@metric"]
      run-groups:
        kgraph:
          args: [ {'reverse': -1}, True] # XXX: hard-codes save_index as True
          query-args: [[1, 2, 3, 4, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100]]
    hnsw(nmslib):
      docker-tag: ann-benchmarks-nmslib
      module: ann_benchmarks.algorithms.nmslib
      constructor: NmslibReuseIndex
      base-args: ["@metric", "hnsw"]
      run-groups:
        M-32:
          # If a run group has an array called @arg-groups instead of one
          # called @args, then every element in that array will be separately
          # expanded before then taking the Cartesian product of all of those
          # expansions.
          #
          # Yes, this is a bit of a hack, but some constructors are weird.
          # (This one used to require that dictionaries be encoded as lists
          # of environment variable-style strings -- ["M=32", "post=2",
          # "efConstruction=400"] -- which didn't work with this at all...)
          arg-groups:
            - {"M": 32, "post": 2, "efConstruction": 400}
            - False
          query-args: [[20, 30, 40, 50, 60, 70, 80, 90, 100, 120, 140, 160, 200,
              300, 400]]
        M-20:
          arg-groups:
            - {"M": 20, "post": 2, "efConstruction": 400}
            - False
          query-args: [[2, 5, 10, 15, 20, 30, 40, 50, 70, 80, 120, 200, 400]]
        M-12:
          arg-groups:
            - {"M": 12, "post": 0, "efConstruction": 400}
            - False
          query-args: [[1, 2, 5, 10, 15, 20, 30, 40, 50, 70, 80, 120]]
        M-4:
          arg-groups:
            - {"M": 4, "post": 0, "efConstruction": 400}
            - False
          query-args: [[1, 2, 5, 10, 20, 30, 50, 70, 90, 120]]
        M-8:
          arg-groups:
            - {"M": 8, "post": 0, "efConstruction": 400}
            - False
          query-args: [[1, 2, 5, 10, 20, 30, 50, 70, 90, 120, 160]]
    SW-graph(nmslib):
      docker-tag: ann-benchmarks-nmslib
      module: ann_benchmarks.algorithms.nmslib
      constructor: NmslibReuseIndex
      base-args: ["@metric", "sw-graph"]
      run-groups:
        NN-24:
          arg-groups:
            - {"NN": 24}
            - False
          query-args: [[800, 400, 200, 100, 50, 30, 20, 15, 10, 5, 1]]
        NN-16:
          arg-groups:
            - {"NN": 16}
            - False
          query-args: [[800, 400, 200, 100, 50, 30, 20, 15, 10, 5, 1]]
        NN-10:
          arg-groups:
            - {"NN": 10}
            - False
          query-args: [[800, 400, 200, 100, 50, 30, 20, 15, 10, 5, 1]]
        NN-5:
          arg-groups:
            - {"NN": 5}
            - False
          query-args: [[30, 25, 20, 15, 10, 5, 4, 3, 2, 1]]
    pynndescent:
      docker-tag: ann-benchmarks-pynndescent
      module: ann_benchmarks.algorithms.pynndescent
      constructor: PyNNDescent
      base-args: ["@metric"]
      run-groups:
        NN-10:
          arg-groups:
            - {"n_neighbors": 10, "diversify_prob": 1.0,
               "pruning_degree_multiplier":[0.5, 1.0], "leaf_size": 24}
          query-args: [[0.0, 0.01, 0.02, 0.03, 0.04, 0.06, 0.08, 0.10, 0.12]]
        NN-20:
          arg-groups:
            - {"n_neighbors": 20, "diversify_prob": 1.0,
               "pruning_degree_multiplier":[0.75, 1.5], "leaf_size": 24}
          query-args: [[0.0, 0.02, 0.04, 0.06, 0.08, 0.10, 0.12, 0.14, 0.16, 0.20]]
        NN-40:
          arg-groups:
            - {"n_neighbors": 40, "diversify_prob": [0.0, 1.0],
               "pruning_degree_multiplier":[1.0, 2.0], "leaf_size": 36}
          query-args: [[0.0, 0.02, 0.04, 0.08, 0.12, 0.16, 0.20, 0.24, 0.28, 0.32]]
        NN-60:
          arg-groups:
            - {"n_neighbors": 60, "diversify_prob": 0.0,
               "pruning_degree_multiplier":[2.0, 3.0], "leaf_size": 48}
          query-args: [[0.0, 0.04, 0.08, 0.12, 0.16, 0.20, 0.24, 0.28, 0.32, 0.36]]
    elasticsearch:
      docker-tag: ann-benchmarks-elasticsearch
      module: ann_benchmarks.algorithms.elasticsearch
      constructor: ElasticsearchScriptScoreQuery
      base-args: [ "@metric", "@dimension" ]
      run-groups:
        empty:
          args: []
    elastiknn-l2lsh:
      docker-tag: ann-benchmarks-elastiknn
      module: ann_benchmarks.algorithms.elastiknn
      constructor: L2Lsh
      base-args: []
      run-groups:
        elastiknn-lsh:
          args:
          - [50, 75, 100]
          - [3, 4]
          - [1, 3, 7]
          query-args:
          - [1000, 10000]
          - [0, 6]
    opendistroknn:
      docker-tag: ann-benchmarks-opendistroknn
      module: ann_benchmarks.algorithms.opendistroknn
      constructor: OpenDistroKNN
      base-args: ["@metric", "@dimension"]
      run-groups:
        M-4:
          arg-groups:
            - {"M": 4,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-8:
          arg-groups:
            - {"M": 8,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-12:
          arg-groups:
            - {"M": 12,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-16:
          arg-groups:
            - {"M": 16,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-24:
          arg-groups:
            - {"M": 24,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-36:
          arg-groups:
            - {"M": 36,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-48:
          arg-groups:
            - {"M": 48,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]

  angular:
    puffinn:
      docker-tag: ann-benchmarks-puffinn
      module: ann_benchmarks.algorithms.puffinn
      constructor: Puffinn
      base-args: ["@metric"]
      run-groups:
        base:
            args: [
              [ 268435456, 536870912, 1073741824, 2147483648, 4294967296],
              ['fht_crosspolytope'],
            ]
            query-args: [[0.1, 0.2, 0.5, 0.7, 0.9, 0.95, 0.99]]
    kgraph:
      docker-tag: ann-benchmarks-kgraph
      module: ann_benchmarks.algorithms.kgraph
      constructor: KGraph
      base-args: ["@metric"]
      run-groups:
        kgraph:
          args: [{'reverse': -1, "K": 200, "L": 300, "S": 20}, False]
          query-args: [[1, 2, 3, 4, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100]]
    hnsw(nmslib):
      docker-tag: ann-benchmarks-nmslib
      module: ann_benchmarks.algorithms.nmslib
      constructor: NmslibReuseIndex
      base-args: ["@metric", "hnsw"]
      run-groups:
        M-48:
          arg-groups:
            - {"M": 48, "post": 2, "efConstruction": 800}
            - False
          query-args: [[50, 70, 90, 120, 160, 200, 400, 600, 700, 800, 1000,
              1400, 1600, 2000]]
        M-32:
          arg-groups:
            - {"M": 32, "post": 2, "efConstruction": 800}
            - False
          query-args: [[10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 120, 140, 160,
              200, 300, 400, 600, 700, 800, 1000, 1200, 1400, 1600, 2000]]
        M-20:
          arg-groups:
            - {"M": 20, "post": 0, "efConstruction": 800}
            - False
          query-args: [[2, 5, 10, 15, 20, 30, 40, 50, 70, 80]]
        M-12:
          arg-groups:
            - {"M": 12, "post": 0, "efConstruction": 800}
            - False
          query-args: [[1, 2, 5, 10, 15, 20, 30, 40, 50, 70, 80]]
    SW-graph(nmslib):
      docker-tag: ann-benchmarks-nmslib
      module: ann_benchmarks.algorithms.nmslib
      constructor: NmslibReuseIndex
      base-args: ["@metric", "sw-graph"]
      run-groups:
        NN-30:
          arg-groups:
            - {"NN": 30}
            - False
          query-args: [[700, 650, 550, 450, 350, 275, 200, 150, 120, 80,
              50, 30]]
        NN-15:
          arg-groups:
            - {"NN": 15}
            - False
          query-args: [[80, 50, 30, 20]]
        NN-3:
          arg-groups:
            - {"NN": 3}
            - False
          query-args: [[120, 80, 60, 40, 20, 10, 8, 4, 2]]
    rpforest:
      docker-tag: ann-benchmarks-rpforest
      module: ann_benchmarks.algorithms.rpforest
      constructor: RPForest
      run-groups:
        base:
          args: [[3, 10, 40, 100, 400],
                [3, 10, 40, 100, 400]]
    pynndescent:
      docker-tag: ann-benchmarks-pynndescent
      module: ann_benchmarks.algorithms.pynndescent
      constructor: PyNNDescent
      base-args: ["@metric"]
      run-groups:
        NN-20:
          arg-groups:
            - {"n_neighbors": [20], "diversify_prob": [1.0],
               "pruning_degree_multiplier":[0.5, 1.0], leaf_size: 20}
          query-args: [[0.0, 0.02, 0.04, 0.06, 0.08, 0.10, 0.12]]
        NN-40:
          arg-groups:
            - {"n_neighbors": [40], "diversify_prob": [0.5, 1.0],
               "pruning_degree_multiplier":[1.5], leaf_size: 25}
          query-args: [[0.0, 0.02, 0.04, 0.06, 0.08, 0.10, 0.12, 0.14, 0.16]]
        NN-80-fast:
          arg-groups:
            - {"n_neighbors": 80, diversify_prob: 1.0,
               "pruning_degree_multiplier": 2.0, leaf_size: 20}
          query-args: [[0.0, 0.02, 0.04, 0.08, 0.12, 0.16, 0.20, 0.24]]
        NN-80-accurate:
          arg-groups:
            - {"n_neighbors": 80, diversify_prob: 0.25,
                "pruning_degree_multiplier": 2.0, leaf_size: 30}
          query-args: [[0.08, 0.12, 0.16, 0.20, 0.24, 0.28, 0.32, 0.36]]
        NN-120-fast:
          arg-groups:
            - {"n_neighbors": 120, "diversify_prob": 1.0,
               "pruning_degree_multiplier":2.5, leaf_size: 20}
          query-args: [[0.0, 0.04, 0.08, 0.16, 0.20, 0.24, 0.28, 0.32]]
        NN-120-accurate:
          arg-groups:
            - {"n_neighbors": 120, "diversify_prob": 0.125,
               "pruning_degree_multiplier":2.5, leaf_size: 35}
          query-args: [[0.16, 0.20, 0.24, 0.28, 0.32, 0.36]]
    scann:
      docker-tag: ann-benchmarks-scann
      module: ann_benchmarks.algorithms.scann
      constructor: Scann
      run-groups:
        scann1:
          args: [[2000], [0.2], [2]]
          query-args: [[[30, 120], [35, 100], [40, 80], [45, 80], [50, 80], [55, 95], [60, 110], [65, 110], [75, 110], [90, 110], [110, 120], [130, 150], [150, 200], [170, 200], [200, 300], [220, 500], [250, 500], [310, 300], [400, 300], [500, 500], [800, 1000]]]
        scann2:
          args: [[1500], [0.55], [1]]
          query-args: [[[8, 25], [10, 25], [12, 25], [13, 25], [14, 27], [15, 30], [17, 30], [18, 40], [20, 40], [22, 40], [25, 50], [30, 50], [35, 55], [50, 60], [60, 60], [80, 80], [100, 100]]]
        scann3:
          args: [[1000], [.2], [1]]
          query-args: [[[9, 25], [11, 35], [12, 35], [13, 35], [14, 40], [15, 40], [16, 40], [17, 45], [20, 45], [20, 55], [25, 55], [25, 70], [30, 70], [40, 90], [50, 100], [60, 120], [70, 140]]]
    elasticsearch:
      docker-tag: ann-benchmarks-elasticsearch
      module: ann_benchmarks.algorithms.elasticsearch
      constructor: ElasticsearchScriptScoreQuery
      base-args: [ "@metric", "@dimension" ]
      run-groups:
        empty:
          args: []
    opendistroknn:
      docker-tag: ann-benchmarks-opendistroknn
      module: ann_benchmarks.algorithms.opendistroknn
      constructor: OpenDistroKNN
      base-args: ["@metric", "@dimension"]
      run-groups:
        M-4:
          arg-groups:
            - {"M": 4,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-8:
          arg-groups:
            - {"M": 8,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-12:
          arg-groups:
            - {"M": 12,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-16:
          arg-groups:
            - {"M": 16,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-24:
          arg-groups:
            - {"M": 24,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-36:
          arg-groups:
            - {"M": 36,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]
        M-48:
          arg-groups:
            - {"M": 48,  "efConstruction": 500}
          query-args: [[10, 20, 40, 80, 120, 200, 400, 600, 800]]

bit:
  hamming:
    mih:
      docker-tag: ann-benchmarks-mih
      module: ann_benchmarks.algorithms.subprocess
      constructor: BitSubprocessPrepared
      base-args: [["mih/bin/mihwrapper"]]
      run-groups:
        base:
          # Note that MIH assumes that 5 <= dimension / chunks <= 37 and segfaults otherwise.
          # This parameter space is explored below using the `chunk-factor`.
          # A factor of 0 means that dimension / chunks = 5 which proceeds linearly to
          # dimension / chunks = 37 for a factor of 1.
          # The r parameter denotes a reordering with n / r steps. If r is 0, no reordering
          # takes places. Otherwise, n / r points from the dataset are used for a
          # re-ording process. MIH requires either r = 0 or r >= 2.
          args: { "d" : "@dimension", "r" : [0, 2, 3],
            "chunk-factor" : [0.2, 0.5, 0.8] }
    kgraph:
      docker-tag: ann-benchmarks-kgraph
      module: ann_benchmarks.algorithms.kgraph
      constructor: KGraph
      base-args: ["euclidean"]
      run-groups:
        kgraph:
          args: [[1, 2, 3, 4, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100],
                 {'reverse': -1, "K": 200, "L": 300, "S": 20}, False]
    hnsw(nmslib):
      docker-tag: ann-benchmarks-nmslib
      module: ann_benchmarks.algorithms.nmslib
      constructor: NmslibReuseIndex
      base-args: ["euclidean", "hnsw"]
      run-groups:
        M-48:
          arg-groups:
            - {"M": 48, "post": 2, "efConstruction": 800}
            - False
          query-args: [[50, 70, 90, 120, 160, 200, 400, 600, 700, 800, 1000,
              1400, 1600, 2000]]
        M-32:
          arg-groups:
            - {"M": 32, "post": 2, "efConstruction": 800}
            - False
          query-args: [[10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 120, 140, 160,
              200, 300, 400, 600, 700, 800, 1000, 1200, 1400, 1600, 2000]]
        M-20:
          arg-groups:
            - {"M": 20, "post": 0, "efConstruction": 800}
            - False
          query-args: [[2, 5, 10, 15, 20, 30, 40, 50, 70, 80]]
        M-12:
          arg-groups:
            - {"M": 12, "post": 0, "efConstruction": 800}
            - False
          query-args: [[1, 2, 5, 10, 15, 20, 30, 40, 50, 70, 80]]
    pynndescent:
      docker-tag: ann-benchmarks-pynndescent
      module: ann_benchmarks.algorithms.pynndescent
      constructor: PyNNDescent
      base-args: ["@metric"]
      run-groups:
        NN-20:
          arg-groups:
            - {"n_neighbors": [20], "diversify_prob": [0.75, 1.0],
               "pruning_degree_multiplier":[1.0, 1.5], leaf_size: 32}
          query-args: [[0.0, 0.01, 0.02, 0.04, 0.08, 0.12, 0.16]]
        NN-40:
          arg-groups:
            - {"n_neighbors": [40], "diversify_prob": [0.5, 1.0],
               "pruning_degree_multiplier":[1.5, 2.0], leaf_size: 48}
          query-args: [[0.0, 0.04, 0.08, 0.12, 0.16, 0.20, 0.24]]
        NN-80:
          arg-groups:
            - {"n_neighbors": [80], "diversify_prob": [0.25, 1.0],
               "pruning_degree_multiplier":[1.75, 2.25], leaf_size: 64}
          query-args: [[0.0, 0.08, 0.12, 0.16, 0.20, 0.24, 0.28, 0.32]]
        NN-120:
          arg-groups:
            - {"n_neighbors": [120], "diversify_prob": [0.0, 1.0],
               "pruning_degree_multiplier":[2.0, 2.5], leaf_size: 80}
          query-args: [[0.08, 0.16, 0.20, 0.24, 0.28, 0.32, 0.36]]
    annoy:
      docker-tag: ann-benchmarks-annoy
      module: ann_benchmarks.algorithms.annoy
      constructor: Annoy
      base-args: ["@metric"]
      run-groups:
        annoy:
          args: [[100, 200, 400]]
          query-args: [[100, 200, 400, 1000, 2000, 4000, 10000, 20000, 40000,
                      100000, 200000, 400000]]
          # This run group produces 3 algorithm instances -- Annoy("angular",
          # 100), Annoy("angular", 200), and Annoy("angular", 400) -- each of
          # which will be used to run 12 different queries.
    faiss-ivf:
      docker-tag: ann-benchmarks-faiss
      module: ann_benchmarks.algorithms.faiss
      constructor: FaissIVF
      base-args: ["euclidean"]
      run-groups:
        base:
          args: [[32,64,128,256,512,1024,2048,4096,8192]]
          query-args: [[1, 5, 10, 50, 100, 200]]
  jaccard:
    bf:
      docker-tag: ann-benchmarks-sklearn
      module: ann_benchmarks.algorithms.bruteforce
      constructor: BruteForceBLAS
      base-args: ["@metric"]
      run-groups:
        base:
          args: {}
    datasketch:
      docker-tag: ann-benchmarks-datasketch
      module: ann_benchmarks.algorithms.datasketch
      constructor: DataSketch
      base-args: ["@metric"]
      run-groups:
        base:
          args:  [[128, 256, 512],[8, 16, 32, 64, 128]]
    puffinn:
      docker-tag: ann-benchmarks-puffinn
      module: ann_benchmarks.algorithms.puffinn
      constructor: Puffinn
      base-args: ["@metric"]
      run-groups:
        base:
            args: [
              [ 16000000, 512000000],
              ['1bit_minhash'],
            ]
            query-args: [[0.1, 0.2, 0.5, 0.7, 0.9, 0.95, 0.99]]
    pynndescent:
      docker-tag: ann-benchmarks-pynndescent
      module: ann_benchmarks.algorithms.pynndescent
      constructor: PyNNDescent
      base-args: ["@metric"]
      run-groups:
        NN-20:
          arg-groups:
            - {"n_neighbors": [20], "diversify_prob": [1.0],
               "pruning_degree_multiplier":[0.5, 1.0], leaf_size: 20}
          query-args: [[0.0, 0.02, 0.04, 0.06, 0.08, 0.10, 0.12]]
        NN-40:
          arg-groups:
            - {"n_neighbors": [40], "diversify_prob": [0.5, 1.0],
               "pruning_degree_multiplier":[1.5], leaf_size: 25}
          query-args: [[0.0, 0.02, 0.04, 0.06, 0.08, 0.10, 0.12, 0.14, 0.16]]
        NN-80-fast:
          arg-groups:
            - {"n_neighbors": 80, diversify_prob: 1.0,
               "pruning_degree_multiplier": 2.0, leaf_size: 20}
          query-args: [[0.0, 0.02, 0.04, 0.08, 0.12, 0.16, 0.20, 0.24]]
        NN-80-accurate:
          arg-groups:
            - {"n_neighbors": 80, diversify_prob: 0.25,
                "pruning_degree_multiplier": 2.0, leaf_size: 30}
          query-args: [[0.08, 0.12, 0.16, 0.20, 0.24, 0.28, 0.32, 0.36]]
        NN-120-fast:
          arg-groups:
            - {"n_neighbors": 120, "diversify_prob": 1.0,
               "pruning_degree_multiplier":2.5, leaf_size: 20}
          query-args: [[0.0, 0.04, 0.08, 0.16, 0.20, 0.24, 0.28, 0.32]]
        NN-120-accurate:
          arg-groups:
            - {"n_neighbors": 120, "diversify_prob": 0.125,
               "pruning_degree_multiplier":2.5, leaf_size: 35}
          query-args: [[0.16, 0.20, 0.24, 0.28, 0.32, 0.36]]