Adapt Alayalite for annbenchmark

.github/workflows/benchmarks.yml

@@ -29,6 +29,7 @@ jobs:
       matrix:
         dataset: [random-xs-20-angular]
         library:
+          - alayalite
           - annoy
           - balltree
           - bruteforce

README.md

@@ -52,6 +52,7 @@ Evaluated
 * [kgn](https://github.com/Henry-yan/kgn)
 * [vsag](https://github.com/antgroup/vsag)
 * [PGVectorScale](https://github.com/timescale/pgvectorscale/tree/main)
+* [AlayaLite](https://github.com/AlayaDB-AI/AlayaLite.git) ![https://img.shields.io/github/stars/AlayaDB-AI/AlayaLite?style=social](https://img.shields.io/github/stars/AlayaDB-AI/AlayaLite?style=social)
 
 Data sets
 =========

ann_benchmarks/algorithms/alayalite/Dockerfile

@@ -0,0 +1,6 @@
+FROM ann-benchmarks
+
+RUN apt-get update && apt-get install -y wget
+RUN wget https://github.com/AlayaDB-AI/AlayaLite/releases/download/v0.1.0a2/alayalite-0.1.0a2-cp310-cp310-linux_x86_64.whl -O /home/app/alayalite-0.1.0a2-cp310-cp310-linux_x86_64.whl
+RUN python3 -m pip install /home/app/alayalite-0.1.0a2-cp310-cp310-linux_x86_64.whl
+RUN python3 -c 'import alayalite'

ann_benchmarks/algorithms/alayalite/config.yml

@@ -0,0 +1,20 @@
+float:
+  any:
+  - base_args: ['@metric', '@dimension']
+    constructor: AlayaLite
+    disabled: false
+    docker_tag: ann-benchmarks-alayalite
+    module: ann_benchmarks.algorithms.alayalite
+    name: AlayaLite
+    run_groups:
+      alayalite:
+        #### index_type HNSW FLAT
+        args: 
+          M: 32 
+          R: [32]
+          L: 200
+          index_type: NONE
+          quantization_type: RABITQ
+          fit_threads: 1 # only batch search
+          search_threads: 1 # only batch search
+        query_args: [[10, 20, 40, 60, 80, 120, 200, 400, 600, 800]]

ann_benchmarks/algorithms/alayalite/module.py

@@ -0,0 +1,62 @@
+import os
+from dataclasses import dataclass, field
+import numpy as np
+from alayalite import Client
+from alayalite import Index
+from ..base.module import BaseANN
+
+
+class AlayaLite(BaseANN):
+    def __init__(self, metric, dim, method_param):
+        self.index_save_dir = 'alaya_indices'
+        self.client = Client(self.index_save_dir)
+        self.index = None
+        self.ef = None
+        self.dim = dim
+        self.metric = metric
+
+        self.index_type = method_param['index_type']
+        self.quantization_type = method_param['quantization_type']
+        self.fit_threads = method_param['fit_threads']
+        self.search_threads = method_param['search_threads']
+        self.R = method_param['R']
+        self.L = method_param['L']
+        self.M = method_param['M']
+
+        self.save_index_name = f'alayalite_index_it_{self.index_type}_qt_{self.quantization_type}_dim_{self.dim}_metric_{self.metric}_M{self.M}.idx'
+        print("alaya init done")
+
+    def fit(self, X: np.array) -> None:
+
+        if os.path.exists(os.path.join(self.index_save_dir, self.save_index_name)):
+            self.index = Index.load(self.index_save_dir, self.save_index_name)
+            print('load index from cache')
+        else:
+            X = X.astype(np.float32)
+            self.index = self.client.create_index(name=self.save_index_name, metric=self.metric, quantization_type=self.quantization_type, capacity=X.shape[0])
+            self.index.fit(vectors=X, num_threads=self.fit_threads)
+            self.client.save_index(self.save_index_name)
+            print('save index to cache')
+
+    def set_query_arguments(self, ef):
+        self.ef = int(ef)
+
+    def prepare_query(self, q: np.array, n: int):
+        self.q = q
+        self.n = n
+
+    def run_prepared_query(self):
+        self.res = self.index.search(query=self.q, topk=self.n, ef_search=self.ef)
+
+    def batch_query(self, X: np.array, n: int) -> None:
+        self.res = self.index.batch_search(queries=X, topk=n, ef_search=self.ef)
+
+    def get_prepared_query_results(self):
+        return self.res
+
+    def get_batch_results(self) -> np.array:
+        return self.res
+
+    def __str__(self) -> str:
+        return "AlayaLite"
+