vectordb-bench 0.0.16__py3-none-any.whl → 0.0.18__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- vectordb_bench/__init__.py +1 -1
- vectordb_bench/backend/clients/__init__.py +13 -0
- vectordb_bench/backend/clients/alloydb/alloydb.py +372 -0
- vectordb_bench/backend/clients/alloydb/cli.py +147 -0
- vectordb_bench/backend/clients/alloydb/config.py +168 -0
- vectordb_bench/backend/clients/api.py +2 -0
- vectordb_bench/backend/clients/milvus/milvus.py +2 -1
- vectordb_bench/backend/dataset.py +8 -5
- vectordb_bench/backend/runner/mp_runner.py +94 -4
- vectordb_bench/backend/runner/rate_runner.py +79 -0
- vectordb_bench/backend/runner/read_write_runner.py +112 -0
- vectordb_bench/backend/runner/util.py +32 -0
- vectordb_bench/backend/task_runner.py +5 -5
- vectordb_bench/cli/vectordbbench.py +2 -0
- vectordb_bench/frontend/components/concurrent/charts.py +25 -9
- vectordb_bench/frontend/config/dbCaseConfigs.py +166 -0
- vectordb_bench/frontend/pages/concurrent.py +5 -1
- vectordb_bench/metric.py +1 -0
- vectordb_bench/models.py +9 -0
- {vectordb_bench-0.0.16.dist-info → vectordb_bench-0.0.18.dist-info}/METADATA +31 -26
- {vectordb_bench-0.0.16.dist-info → vectordb_bench-0.0.18.dist-info}/RECORD +25 -19
- {vectordb_bench-0.0.16.dist-info → vectordb_bench-0.0.18.dist-info}/WHEEL +1 -1
- {vectordb_bench-0.0.16.dist-info → vectordb_bench-0.0.18.dist-info}/LICENSE +0 -0
- {vectordb_bench-0.0.16.dist-info → vectordb_bench-0.0.18.dist-info}/entry_points.txt +0 -0
- {vectordb_bench-0.0.16.dist-info → vectordb_bench-0.0.18.dist-info}/top_level.txt +0 -0
vectordb_bench/__init__.py
CHANGED
@@ -17,7 +17,7 @@ class config:
|
|
17
17
|
|
18
18
|
DEFAULT_DATASET_URL = env.str("DEFAULT_DATASET_URL", AWS_S3_URL)
|
19
19
|
DATASET_LOCAL_DIR = env.path("DATASET_LOCAL_DIR", "/tmp/vectordb_bench/dataset")
|
20
|
-
NUM_PER_BATCH = env.int("NUM_PER_BATCH",
|
20
|
+
NUM_PER_BATCH = env.int("NUM_PER_BATCH", 100)
|
21
21
|
|
22
22
|
DROP_OLD = env.bool("DROP_OLD", True)
|
23
23
|
USE_SHUFFLED_DATA = env.bool("USE_SHUFFLED_DATA", True)
|
@@ -32,6 +32,7 @@ class DB(Enum):
|
|
32
32
|
PgVectoRS = "PgVectoRS"
|
33
33
|
PgVectorScale = "PgVectorScale"
|
34
34
|
PgDiskANN = "PgDiskANN"
|
35
|
+
AlloyDB = "AlloyDB"
|
35
36
|
Redis = "Redis"
|
36
37
|
MemoryDB = "MemoryDB"
|
37
38
|
Chroma = "Chroma"
|
@@ -97,6 +98,10 @@ class DB(Enum):
|
|
97
98
|
if self == DB.AWSOpenSearch:
|
98
99
|
from .aws_opensearch.aws_opensearch import AWSOpenSearch
|
99
100
|
return AWSOpenSearch
|
101
|
+
|
102
|
+
if self == DB.AlloyDB:
|
103
|
+
from .alloydb.alloydb import AlloyDB
|
104
|
+
return AlloyDB
|
100
105
|
|
101
106
|
@property
|
102
107
|
def config_cls(self) -> Type[DBConfig]:
|
@@ -156,6 +161,10 @@ class DB(Enum):
|
|
156
161
|
if self == DB.AWSOpenSearch:
|
157
162
|
from .aws_opensearch.config import AWSOpenSearchConfig
|
158
163
|
return AWSOpenSearchConfig
|
164
|
+
|
165
|
+
if self == DB.AlloyDB:
|
166
|
+
from .alloydb.config import AlloyDBConfig
|
167
|
+
return AlloyDBConfig
|
159
168
|
|
160
169
|
def case_config_cls(self, index_type: IndexType | None = None) -> Type[DBCaseConfig]:
|
161
170
|
if self == DB.Milvus:
|
@@ -197,6 +206,10 @@ class DB(Enum):
|
|
197
206
|
if self == DB.PgDiskANN:
|
198
207
|
from .pgdiskann.config import _pgdiskann_case_config
|
199
208
|
return _pgdiskann_case_config.get(index_type)
|
209
|
+
|
210
|
+
if self == DB.AlloyDB:
|
211
|
+
from .alloydb.config import _alloydb_case_config
|
212
|
+
return _alloydb_case_config.get(index_type)
|
200
213
|
|
201
214
|
# DB.Pinecone, DB.Chroma, DB.Redis
|
202
215
|
return EmptyDBCaseConfig
|
@@ -0,0 +1,372 @@
|
|
1
|
+
"""Wrapper around the alloydb vector database over VectorDB"""
|
2
|
+
|
3
|
+
import logging
|
4
|
+
import pprint
|
5
|
+
from contextlib import contextmanager
|
6
|
+
from typing import Any, Generator, Optional, Tuple, Sequence
|
7
|
+
|
8
|
+
import numpy as np
|
9
|
+
import psycopg
|
10
|
+
from pgvector.psycopg import register_vector
|
11
|
+
from psycopg import Connection, Cursor, sql
|
12
|
+
|
13
|
+
from ..api import VectorDB
|
14
|
+
from .config import AlloyDBConfigDict, AlloyDBIndexConfig, AlloyDBScaNNConfig
|
15
|
+
|
16
|
+
log = logging.getLogger(__name__)
|
17
|
+
|
18
|
+
|
19
|
+
class AlloyDB(VectorDB):
|
20
|
+
"""Use psycopg instructions"""
|
21
|
+
|
22
|
+
conn: psycopg.Connection[Any] | None = None
|
23
|
+
cursor: psycopg.Cursor[Any] | None = None
|
24
|
+
|
25
|
+
_filtered_search: sql.Composed
|
26
|
+
_unfiltered_search: sql.Composed
|
27
|
+
|
28
|
+
def __init__(
|
29
|
+
self,
|
30
|
+
dim: int,
|
31
|
+
db_config: AlloyDBConfigDict,
|
32
|
+
db_case_config: AlloyDBIndexConfig,
|
33
|
+
collection_name: str = "alloydb_collection",
|
34
|
+
drop_old: bool = False,
|
35
|
+
**kwargs,
|
36
|
+
):
|
37
|
+
self.name = "AlloyDB"
|
38
|
+
self.db_config = db_config
|
39
|
+
self.case_config = db_case_config
|
40
|
+
self.table_name = collection_name
|
41
|
+
self.dim = dim
|
42
|
+
|
43
|
+
self._index_name = "alloydb_index"
|
44
|
+
self._primary_field = "id"
|
45
|
+
self._vector_field = "embedding"
|
46
|
+
|
47
|
+
# construct basic units
|
48
|
+
self.conn, self.cursor = self._create_connection(**self.db_config)
|
49
|
+
|
50
|
+
# create vector extension
|
51
|
+
self.cursor.execute("CREATE EXTENSION IF NOT EXISTS alloydb_scann CASCADE")
|
52
|
+
self.conn.commit()
|
53
|
+
|
54
|
+
log.info(f"{self.name} config values: {self.db_config}\n{self.case_config}")
|
55
|
+
if not any(
|
56
|
+
(
|
57
|
+
self.case_config.create_index_before_load,
|
58
|
+
self.case_config.create_index_after_load,
|
59
|
+
)
|
60
|
+
):
|
61
|
+
err = f"{self.name} config must create an index using create_index_before_load or create_index_after_load"
|
62
|
+
log.error(err)
|
63
|
+
raise RuntimeError(
|
64
|
+
f"{err}\n{pprint.pformat(self.db_config)}\n{pprint.pformat(self.case_config)}"
|
65
|
+
)
|
66
|
+
|
67
|
+
if drop_old:
|
68
|
+
self._drop_index()
|
69
|
+
self._drop_table()
|
70
|
+
self._create_table(dim)
|
71
|
+
if self.case_config.create_index_before_load:
|
72
|
+
self._create_index()
|
73
|
+
|
74
|
+
self.cursor.close()
|
75
|
+
self.conn.close()
|
76
|
+
self.cursor = None
|
77
|
+
self.conn = None
|
78
|
+
|
79
|
+
@staticmethod
|
80
|
+
def _create_connection(**kwargs) -> Tuple[Connection, Cursor]:
|
81
|
+
conn = psycopg.connect(**kwargs)
|
82
|
+
register_vector(conn)
|
83
|
+
conn.autocommit = False
|
84
|
+
cursor = conn.cursor()
|
85
|
+
|
86
|
+
assert conn is not None, "Connection is not initialized"
|
87
|
+
assert cursor is not None, "Cursor is not initialized"
|
88
|
+
return conn, cursor
|
89
|
+
|
90
|
+
def _generate_search_query(self, filtered: bool=False) -> sql.Composed:
|
91
|
+
search_query = sql.Composed(
|
92
|
+
[
|
93
|
+
sql.SQL(
|
94
|
+
"SELECT id FROM public.{table_name} {where_clause} ORDER BY embedding "
|
95
|
+
).format(
|
96
|
+
table_name=sql.Identifier(self.table_name),
|
97
|
+
where_clause=sql.SQL("WHERE id >= %s") if filtered else sql.SQL(""),
|
98
|
+
),
|
99
|
+
sql.SQL(self.case_config.search_param()["metric_fun_op"]),
|
100
|
+
sql.SQL(" %s::vector LIMIT %s::int"),
|
101
|
+
]
|
102
|
+
)
|
103
|
+
return search_query
|
104
|
+
|
105
|
+
@contextmanager
|
106
|
+
def init(self) -> Generator[None, None, None]:
|
107
|
+
"""
|
108
|
+
Examples:
|
109
|
+
>>> with self.init():
|
110
|
+
>>> self.insert_embeddings()
|
111
|
+
>>> self.search_embedding()
|
112
|
+
"""
|
113
|
+
|
114
|
+
self.conn, self.cursor = self._create_connection(**self.db_config)
|
115
|
+
|
116
|
+
# index configuration may have commands defined that we should set during each client session
|
117
|
+
session_options: Sequence[dict[str, Any]] = self.case_config.session_param()["session_options"]
|
118
|
+
|
119
|
+
if len(session_options) > 0:
|
120
|
+
for setting in session_options:
|
121
|
+
command = sql.SQL("SET {setting_name} " + "= {val};").format(
|
122
|
+
setting_name=sql.Identifier(setting['parameter']['setting_name']),
|
123
|
+
val=sql.Identifier(str(setting['parameter']['val'])),
|
124
|
+
)
|
125
|
+
log.debug(command.as_string(self.cursor))
|
126
|
+
self.cursor.execute(command)
|
127
|
+
self.conn.commit()
|
128
|
+
|
129
|
+
self._filtered_search = self._generate_search_query(filtered=True)
|
130
|
+
self._unfiltered_search = self._generate_search_query()
|
131
|
+
|
132
|
+
try:
|
133
|
+
yield
|
134
|
+
finally:
|
135
|
+
self.cursor.close()
|
136
|
+
self.conn.close()
|
137
|
+
self.cursor = None
|
138
|
+
self.conn = None
|
139
|
+
|
140
|
+
def _drop_table(self):
|
141
|
+
assert self.conn is not None, "Connection is not initialized"
|
142
|
+
assert self.cursor is not None, "Cursor is not initialized"
|
143
|
+
log.info(f"{self.name} client drop table : {self.table_name}")
|
144
|
+
|
145
|
+
self.cursor.execute(
|
146
|
+
sql.SQL("DROP TABLE IF EXISTS public.{table_name}").format(
|
147
|
+
table_name=sql.Identifier(self.table_name)
|
148
|
+
)
|
149
|
+
)
|
150
|
+
self.conn.commit()
|
151
|
+
|
152
|
+
def ready_to_load(self):
|
153
|
+
pass
|
154
|
+
|
155
|
+
def optimize(self):
|
156
|
+
self._post_insert()
|
157
|
+
|
158
|
+
def _post_insert(self):
|
159
|
+
log.info(f"{self.name} post insert before optimize")
|
160
|
+
if self.case_config.create_index_after_load:
|
161
|
+
self._drop_index()
|
162
|
+
self._create_index()
|
163
|
+
|
164
|
+
def _drop_index(self):
|
165
|
+
assert self.conn is not None, "Connection is not initialized"
|
166
|
+
assert self.cursor is not None, "Cursor is not initialized"
|
167
|
+
log.info(f"{self.name} client drop index : {self._index_name}")
|
168
|
+
|
169
|
+
drop_index_sql = sql.SQL("DROP INDEX IF EXISTS {index_name}").format(
|
170
|
+
index_name=sql.Identifier(self._index_name)
|
171
|
+
)
|
172
|
+
log.debug(drop_index_sql.as_string(self.cursor))
|
173
|
+
self.cursor.execute(drop_index_sql)
|
174
|
+
self.conn.commit()
|
175
|
+
|
176
|
+
def _set_parallel_index_build_param(self):
|
177
|
+
assert self.conn is not None, "Connection is not initialized"
|
178
|
+
assert self.cursor is not None, "Cursor is not initialized"
|
179
|
+
|
180
|
+
index_param = self.case_config.index_param()
|
181
|
+
|
182
|
+
if index_param["enable_pca"] is not None:
|
183
|
+
self.cursor.execute(
|
184
|
+
sql.SQL("SET scann.enable_pca TO {};").format(
|
185
|
+
index_param["enable_pca"]
|
186
|
+
)
|
187
|
+
)
|
188
|
+
self.cursor.execute(
|
189
|
+
sql.SQL("ALTER USER {} SET scann.enable_pca TO {};").format(
|
190
|
+
sql.Identifier(self.db_config["user"]),
|
191
|
+
index_param["enable_pca"],
|
192
|
+
)
|
193
|
+
)
|
194
|
+
self.conn.commit()
|
195
|
+
|
196
|
+
if index_param["maintenance_work_mem"] is not None:
|
197
|
+
self.cursor.execute(
|
198
|
+
sql.SQL("SET maintenance_work_mem TO {};").format(
|
199
|
+
index_param["maintenance_work_mem"]
|
200
|
+
)
|
201
|
+
)
|
202
|
+
self.cursor.execute(
|
203
|
+
sql.SQL("ALTER USER {} SET maintenance_work_mem TO {};").format(
|
204
|
+
sql.Identifier(self.db_config["user"]),
|
205
|
+
index_param["maintenance_work_mem"],
|
206
|
+
)
|
207
|
+
)
|
208
|
+
self.conn.commit()
|
209
|
+
|
210
|
+
if index_param["max_parallel_workers"] is not None:
|
211
|
+
self.cursor.execute(
|
212
|
+
sql.SQL("SET max_parallel_maintenance_workers TO '{}';").format(
|
213
|
+
index_param["max_parallel_workers"]
|
214
|
+
)
|
215
|
+
)
|
216
|
+
self.cursor.execute(
|
217
|
+
sql.SQL(
|
218
|
+
"ALTER USER {} SET max_parallel_maintenance_workers TO '{}';"
|
219
|
+
).format(
|
220
|
+
sql.Identifier(self.db_config["user"]),
|
221
|
+
index_param["max_parallel_workers"],
|
222
|
+
)
|
223
|
+
)
|
224
|
+
self.cursor.execute(
|
225
|
+
sql.SQL("SET max_parallel_workers TO '{}';").format(
|
226
|
+
index_param["max_parallel_workers"]
|
227
|
+
)
|
228
|
+
)
|
229
|
+
self.cursor.execute(
|
230
|
+
sql.SQL(
|
231
|
+
"ALTER USER {} SET max_parallel_workers TO '{}';"
|
232
|
+
).format(
|
233
|
+
sql.Identifier(self.db_config["user"]),
|
234
|
+
index_param["max_parallel_workers"],
|
235
|
+
)
|
236
|
+
)
|
237
|
+
self.cursor.execute(
|
238
|
+
sql.SQL(
|
239
|
+
"ALTER TABLE {} SET (parallel_workers = {});"
|
240
|
+
).format(
|
241
|
+
sql.Identifier(self.table_name),
|
242
|
+
index_param["max_parallel_workers"],
|
243
|
+
)
|
244
|
+
)
|
245
|
+
self.conn.commit()
|
246
|
+
|
247
|
+
results = self.cursor.execute(
|
248
|
+
sql.SQL("SHOW max_parallel_maintenance_workers;")
|
249
|
+
).fetchall()
|
250
|
+
results.extend(
|
251
|
+
self.cursor.execute(sql.SQL("SHOW max_parallel_workers;")).fetchall()
|
252
|
+
)
|
253
|
+
results.extend(
|
254
|
+
self.cursor.execute(sql.SQL("SHOW maintenance_work_mem;")).fetchall()
|
255
|
+
)
|
256
|
+
log.info(f"{self.name} parallel index creation parameters: {results}")
|
257
|
+
|
258
|
+
def _create_index(self):
|
259
|
+
assert self.conn is not None, "Connection is not initialized"
|
260
|
+
assert self.cursor is not None, "Cursor is not initialized"
|
261
|
+
log.info(f"{self.name} client create index : {self._index_name}")
|
262
|
+
|
263
|
+
index_param = self.case_config.index_param()
|
264
|
+
self._set_parallel_index_build_param()
|
265
|
+
options = []
|
266
|
+
for option in index_param["index_creation_with_options"]:
|
267
|
+
if option['val'] is not None:
|
268
|
+
options.append(
|
269
|
+
sql.SQL("{option_name} = {val}").format(
|
270
|
+
option_name=sql.Identifier(option['option_name']),
|
271
|
+
val=sql.Identifier(str(option['val'])),
|
272
|
+
)
|
273
|
+
)
|
274
|
+
if any(options):
|
275
|
+
with_clause = sql.SQL("WITH ({});").format(sql.SQL(", ").join(options))
|
276
|
+
else:
|
277
|
+
with_clause = sql.Composed(())
|
278
|
+
|
279
|
+
index_create_sql = sql.SQL(
|
280
|
+
"""
|
281
|
+
CREATE INDEX IF NOT EXISTS {index_name} ON public.{table_name}
|
282
|
+
USING {index_type} (embedding {embedding_metric})
|
283
|
+
"""
|
284
|
+
).format(
|
285
|
+
index_name=sql.Identifier(self._index_name),
|
286
|
+
table_name=sql.Identifier(self.table_name),
|
287
|
+
index_type=sql.Identifier(index_param["index_type"]),
|
288
|
+
embedding_metric=sql.Identifier(index_param["metric"]),
|
289
|
+
)
|
290
|
+
|
291
|
+
index_create_sql_with_with_clause = (
|
292
|
+
index_create_sql + with_clause
|
293
|
+
).join(" ")
|
294
|
+
log.debug(index_create_sql_with_with_clause.as_string(self.cursor))
|
295
|
+
self.cursor.execute(index_create_sql_with_with_clause)
|
296
|
+
self.conn.commit()
|
297
|
+
|
298
|
+
def _create_table(self, dim: int):
|
299
|
+
assert self.conn is not None, "Connection is not initialized"
|
300
|
+
assert self.cursor is not None, "Cursor is not initialized"
|
301
|
+
|
302
|
+
try:
|
303
|
+
log.info(f"{self.name} client create table : {self.table_name}")
|
304
|
+
|
305
|
+
# create table
|
306
|
+
self.cursor.execute(
|
307
|
+
sql.SQL(
|
308
|
+
"CREATE TABLE IF NOT EXISTS public.{table_name} (id BIGINT PRIMARY KEY, embedding vector({dim}));"
|
309
|
+
).format(table_name=sql.Identifier(self.table_name), dim=dim)
|
310
|
+
)
|
311
|
+
self.conn.commit()
|
312
|
+
except Exception as e:
|
313
|
+
log.warning(
|
314
|
+
f"Failed to create alloydb table: {self.table_name} error: {e}"
|
315
|
+
)
|
316
|
+
raise e from None
|
317
|
+
|
318
|
+
def insert_embeddings(
|
319
|
+
self,
|
320
|
+
embeddings: list[list[float]],
|
321
|
+
metadata: list[int],
|
322
|
+
**kwargs: Any,
|
323
|
+
) -> Tuple[int, Optional[Exception]]:
|
324
|
+
assert self.conn is not None, "Connection is not initialized"
|
325
|
+
assert self.cursor is not None, "Cursor is not initialized"
|
326
|
+
|
327
|
+
try:
|
328
|
+
metadata_arr = np.array(metadata)
|
329
|
+
embeddings_arr = np.array(embeddings)
|
330
|
+
|
331
|
+
with self.cursor.copy(
|
332
|
+
sql.SQL("COPY public.{table_name} FROM STDIN (FORMAT BINARY)").format(
|
333
|
+
table_name=sql.Identifier(self.table_name)
|
334
|
+
)
|
335
|
+
) as copy:
|
336
|
+
copy.set_types(["bigint", "vector"])
|
337
|
+
for i, row in enumerate(metadata_arr):
|
338
|
+
copy.write_row((row, embeddings_arr[i]))
|
339
|
+
self.conn.commit()
|
340
|
+
|
341
|
+
if kwargs.get("last_batch"):
|
342
|
+
self._post_insert()
|
343
|
+
|
344
|
+
return len(metadata), None
|
345
|
+
except Exception as e:
|
346
|
+
log.warning(
|
347
|
+
f"Failed to insert data into alloydb table ({self.table_name}), error: {e}"
|
348
|
+
)
|
349
|
+
return 0, e
|
350
|
+
|
351
|
+
def search_embedding(
|
352
|
+
self,
|
353
|
+
query: list[float],
|
354
|
+
k: int = 100,
|
355
|
+
filters: dict | None = None,
|
356
|
+
timeout: int | None = None,
|
357
|
+
) -> list[int]:
|
358
|
+
assert self.conn is not None, "Connection is not initialized"
|
359
|
+
assert self.cursor is not None, "Cursor is not initialized"
|
360
|
+
|
361
|
+
q = np.asarray(query)
|
362
|
+
if filters:
|
363
|
+
gt = filters.get("id")
|
364
|
+
result = self.cursor.execute(
|
365
|
+
self._filtered_search, (gt, q, k), prepare=True, binary=True
|
366
|
+
)
|
367
|
+
else:
|
368
|
+
result = self.cursor.execute(
|
369
|
+
self._unfiltered_search, (q, k), prepare=True, binary=True
|
370
|
+
)
|
371
|
+
|
372
|
+
return [int(i[0]) for i in result.fetchall()]
|
@@ -0,0 +1,147 @@
|
|
1
|
+
from typing import Annotated, Optional, TypedDict, Unpack
|
2
|
+
|
3
|
+
import click
|
4
|
+
import os
|
5
|
+
from pydantic import SecretStr
|
6
|
+
|
7
|
+
from vectordb_bench.backend.clients.api import MetricType
|
8
|
+
|
9
|
+
from ....cli.cli import (
|
10
|
+
CommonTypedDict,
|
11
|
+
cli,
|
12
|
+
click_parameter_decorators_from_typed_dict,
|
13
|
+
get_custom_case_config,
|
14
|
+
run,
|
15
|
+
)
|
16
|
+
from vectordb_bench.backend.clients import DB
|
17
|
+
|
18
|
+
|
19
|
+
class AlloyDBTypedDict(CommonTypedDict):
|
20
|
+
user_name: Annotated[
|
21
|
+
str, click.option("--user-name", type=str, help="Db username", required=True)
|
22
|
+
]
|
23
|
+
password: Annotated[
|
24
|
+
str,
|
25
|
+
click.option("--password",
|
26
|
+
type=str,
|
27
|
+
help="Postgres database password",
|
28
|
+
default=lambda: os.environ.get("POSTGRES_PASSWORD", ""),
|
29
|
+
show_default="$POSTGRES_PASSWORD",
|
30
|
+
),
|
31
|
+
]
|
32
|
+
|
33
|
+
host: Annotated[
|
34
|
+
str, click.option("--host", type=str, help="Db host", required=True)
|
35
|
+
]
|
36
|
+
db_name: Annotated[
|
37
|
+
str, click.option("--db-name", type=str, help="Db name", required=True)
|
38
|
+
]
|
39
|
+
maintenance_work_mem: Annotated[
|
40
|
+
Optional[str],
|
41
|
+
click.option(
|
42
|
+
"--maintenance-work-mem",
|
43
|
+
type=str,
|
44
|
+
help="Sets the maximum memory to be used for maintenance operations (index creation). "
|
45
|
+
"Can be entered as string with unit like '64GB' or as an integer number of KB."
|
46
|
+
"This will set the parameters: max_parallel_maintenance_workers,"
|
47
|
+
" max_parallel_workers & table(parallel_workers)",
|
48
|
+
required=False,
|
49
|
+
),
|
50
|
+
]
|
51
|
+
max_parallel_workers: Annotated[
|
52
|
+
Optional[int],
|
53
|
+
click.option(
|
54
|
+
"--max-parallel-workers",
|
55
|
+
type=int,
|
56
|
+
help="Sets the maximum number of parallel processes per maintenance operation (index creation)",
|
57
|
+
required=False,
|
58
|
+
),
|
59
|
+
]
|
60
|
+
|
61
|
+
|
62
|
+
|
63
|
+
class AlloyDBScaNNTypedDict(AlloyDBTypedDict):
|
64
|
+
num_leaves: Annotated[
|
65
|
+
int,
|
66
|
+
click.option("--num-leaves", type=int, help="Number of leaves", required=True)
|
67
|
+
]
|
68
|
+
num_leaves_to_search: Annotated[
|
69
|
+
int,
|
70
|
+
click.option("--num-leaves-to-search", type=int, help="Number of leaves to search", required=True)
|
71
|
+
]
|
72
|
+
pre_reordering_num_neighbors: Annotated[
|
73
|
+
int,
|
74
|
+
click.option("--pre-reordering-num-neighbors", type=int, help="Pre-reordering number of neighbors", default=200)
|
75
|
+
]
|
76
|
+
max_top_neighbors_buffer_size: Annotated[
|
77
|
+
int,
|
78
|
+
click.option("--max-top-neighbors-buffer-size", type=int, help="Maximum top neighbors buffer size", default=20_000)
|
79
|
+
]
|
80
|
+
num_search_threads: Annotated[
|
81
|
+
int,
|
82
|
+
click.option("--num-search-threads", type=int, help="Number of search threads", default=2)
|
83
|
+
]
|
84
|
+
max_num_prefetch_datasets: Annotated[
|
85
|
+
int,
|
86
|
+
click.option("--max-num-prefetch-datasets", type=int, help="Maximum number of prefetch datasets", default=100)
|
87
|
+
]
|
88
|
+
quantizer: Annotated[
|
89
|
+
str,
|
90
|
+
click.option(
|
91
|
+
"--quantizer",
|
92
|
+
type=click.Choice(["SQ8", "FLAT"]),
|
93
|
+
help="Quantizer type",
|
94
|
+
default="SQ8"
|
95
|
+
)
|
96
|
+
]
|
97
|
+
enable_pca: Annotated[
|
98
|
+
bool, click.option(
|
99
|
+
"--enable-pca",
|
100
|
+
type=click.Choice(["on", "off"]),
|
101
|
+
help="Enable PCA",
|
102
|
+
default="on"
|
103
|
+
)
|
104
|
+
]
|
105
|
+
max_num_levels: Annotated[
|
106
|
+
int,
|
107
|
+
click.option(
|
108
|
+
"--max-num-levels",
|
109
|
+
type=click.Choice([1, 2]),
|
110
|
+
help="Maximum number of levels",
|
111
|
+
default=1
|
112
|
+
)
|
113
|
+
]
|
114
|
+
|
115
|
+
|
116
|
+
@cli.command()
|
117
|
+
@click_parameter_decorators_from_typed_dict(AlloyDBScaNNTypedDict)
|
118
|
+
def AlloyDBScaNN(
|
119
|
+
**parameters: Unpack[AlloyDBScaNNTypedDict],
|
120
|
+
):
|
121
|
+
from .config import AlloyDBConfig, AlloyDBScaNNConfig
|
122
|
+
|
123
|
+
parameters["custom_case"] = get_custom_case_config(parameters)
|
124
|
+
run(
|
125
|
+
db=DB.AlloyDB,
|
126
|
+
db_config=AlloyDBConfig(
|
127
|
+
db_label=parameters["db_label"],
|
128
|
+
user_name=SecretStr(parameters["user_name"]),
|
129
|
+
password=SecretStr(parameters["password"]),
|
130
|
+
host=parameters["host"],
|
131
|
+
db_name=parameters["db_name"],
|
132
|
+
),
|
133
|
+
db_case_config=AlloyDBScaNNConfig(
|
134
|
+
num_leaves=parameters["num_leaves"],
|
135
|
+
quantizer=parameters["quantizer"],
|
136
|
+
enable_pca=parameters["enable_pca"],
|
137
|
+
max_num_levels=parameters["max_num_levels"],
|
138
|
+
num_leaves_to_search=parameters["num_leaves_to_search"],
|
139
|
+
max_top_neighbors_buffer_size=parameters["max_top_neighbors_buffer_size"],
|
140
|
+
pre_reordering_num_neighbors=parameters["pre_reordering_num_neighbors"],
|
141
|
+
num_search_threads=parameters["num_search_threads"],
|
142
|
+
max_num_prefetch_datasets=parameters["max_num_prefetch_datasets"],
|
143
|
+
max_parallel_workers=parameters["max_parallel_workers"],
|
144
|
+
maintenance_work_mem=parameters["maintenance_work_mem"],
|
145
|
+
),
|
146
|
+
**parameters,
|
147
|
+
)
|