feat: init repo with pgvecto.rs support (#1)

* feat: init with pgvecto.rs Signed-off-by: Keming <[email protected]> * ignore py Signed-off-by: Keming <[email protected]> * re-arrange the structure, add pseudo reader Signed-off-by: Keming <[email protected]> * use vector type Signed-off-by: Keming <[email protected]> * re struct Signed-off-by: Keming <[email protected]> * basic func Signed-off-by: Keming <[email protected]> * fix cli Signed-off-by: Keming <[email protected]> * add tqdm Signed-off-by: Keming <[email protected]> * feat: add dataset downloader (#1) * feat: add dataset downloader Signed-off-by: Keming <[email protected]> * fix insert memory usage Signed-off-by: Keming <[email protected]> --------- Signed-off-by: Keming <[email protected]> * add links in readme Signed-off-by: Keming <[email protected]> --------- Signed-off-by: Keming <[email protected]>
tensorchord · Jan 23, 2024 · 2cb974a · 2cb974a
1 parent e527067
commit 2cb974a
Show file tree

Hide file tree

Showing 18 changed files with 926 additions and 0 deletions.
diff --git a/.gitignore b/.gitignore
@@ -0,0 +1,166 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+.ruff_cache/
+.csv
+.json
+.hdf5
+datasets/
diff --git a/Makefile b/Makefile
@@ -0,0 +1,20 @@
+PY_SOURCE=vector_bench
+
+dev:
+	@pip install -e .[dev]
+
+lint:
+	@ruff check ${PY_SOURCE}
+
+format:
+	@ruff check --fix ${PY_SOURCE}
+	@ruff format ${PY_SOURCE}
+
+clean:
+	@-rm -rf dist build __pycache__ *.egg-info argstruct/__version__.py
+
+build:
+	@python -m build
+
+test:
+	@pytest -v tests
diff --git a/README.md b/README.md
@@ -1 +1,32 @@
 # Vector DB Benchmark
+
+Supported databases/extensions:
+
+- [x] [`pgvecto.rs`](https://github.com/tensorchord/pgvecto.rs)
+- [ ] [`pgvector`](https://github.com/pgvector/pgvector)
+- [ ] [`qdrant`](https://github.com/qdrant/qdrant/)
+
+Supported datasets:
+
+- [x] random generated
+- [x] GIST 960
+
+
+## Installation
+
+```bash
+pip install vector_bench[pgvectors]
+```
+
+## Run
+
+```bash
+# help
+vector_bench --help
+# only insert the data
+vector_bench --insert --url postgresql://postgres:[email protected]:5432/postgres -s gist_960_l2
+# only query the data (make sure the data is already inserted)
+vector_bench --query --url postgresql://postgres:password@localhost:5432/postgres -s gist_960_l2
+# insert and query the data
+vector_bench --insert --query --url postgresql://postgres:password@localhost:5432/postgres -s gist_960_l2
+```
diff --git a/pyproject.toml b/pyproject.toml
@@ -0,0 +1,41 @@
+[project]
+name = "vectordb-benchmark"
+dynamic = ["version"]
+requires-python = ">=3.8"
+readme = "README.md"
+license = {text = "Apache-2.0"}
+dependencies = [
+    "msgspec~=0.18.5",
+    "h5py~=3.10.0",
+    "numpy~=1.26.3",
+    "tqdm~=4.66",
+    "httpx",
+]
+[project.optional-dependencies]
+dev = [
+    "ruff~=0.1.3",
+    "pytest~=7.4",
+]
+pgvectors = [
+    "psycopg[binary]",
+]
+[project.urls]
+"Homepage" = "https://github.com/tensorchord/vectordb-benchmark"
+[build-system]
+requires = ["setuptools", "setuptools_scm>=7.0"]
+build-backend = "setuptools.build_meta"
+[project.scripts]
+"vector_bench" = "vector_bench.main:main"
+
+[tool.setuptools_scm]
+fallback_version = "0.0.0"
+
+[tool.ruff]
+target-version = "py38"
+[tool.ruff.lint]
+select = ["E", "F", "G", "B", "I", "SIM", "TID", "PL", "RUF"]
+ignore = ["E501"]
+[tool.ruff.lint.isort]
+known-first-party = ["argstruct"]
+[tool.ruff.pylint]
+max-args = 7
diff --git a/vector_bench/args.py b/vector_bench/args.py
@@ -0,0 +1,33 @@
+from argparse import ArgumentParser
+
+from vector_bench.dataset.source import DataSource
+
+
+def build_arg_parser():
+    parser = ArgumentParser()
+    parser.add_argument(
+        "--client",
+        "-c",
+        choices=["pgvecto_rs", "pgvector", "qdrant"],
+        default="pgvecto_rs",
+        help="client type",
+    )
+    parser.add_argument(
+        "--worker-num",
+        "-w",
+        type=int,
+        help="number of workers, if not set, use min(32, cpu_thread + 4)",
+    )
+    parser.add_argument("--url", "-u", help="database url")
+    parser.add_argument(
+        "--source", "-s", choices=DataSource.list(), help="dataset source"
+    )
+    parser.add_argument("--query", action="store_true", help="query benchmark")
+    parser.add_argument("--insert", action="store_true", help="insert data")
+    return parser
+
+
+if __name__ == "__main__":
+    parser = build_arg_parser()
+    args = parser.parse_args()
+    print(args)
diff --git a/vector_bench/bench.py b/vector_bench/bench.py
@@ -0,0 +1,93 @@
+from concurrent.futures import CancelledError, ThreadPoolExecutor, as_completed
+from itertools import islice
+from time import perf_counter
+from typing import Iterable, Optional
+
+from vector_bench.client import DataBaseClient
+from vector_bench.client.base import BaseClient
+from vector_bench.dataset import DatasetReader
+from vector_bench.dataset.base import BaseReader
+from vector_bench.log import logger
+from vector_bench.spec import BenchmarkResult, DatabaseConfig, DatasetConfig, Query
+
+
+def batched(iterable: Iterable, n: int):
+    it = iter(iterable)
+    while batch := tuple(islice(it, n)):
+        yield batch
+
+
+class Benchmark:
+    def __init__(
+        self,
+        db_config: DatabaseConfig,
+        dataset_config: DatasetConfig,
+        worker_num: Optional[int] = None,
+    ) -> None:
+        self.client: BaseClient = DataBaseClient.select(db_config.name).from_config(
+            db_config
+        )
+        self.reader: BaseReader = DatasetReader.select(
+            dataset_config.type.value
+        ).from_config(dataset_config)
+        self.worker_num: Optional[int] = worker_num
+        self.query_result: BenchmarkResult = BenchmarkResult()
+
+    def insert(self):
+        logger.info("inserting records...")
+        epoch_size, batch_size = 10000, 20
+        with ThreadPoolExecutor(self.worker_num) as executor:
+            logger.info("using %s executors", executor._max_workers)
+            for i, epoch in enumerate(batched(self.reader.read_record(), epoch_size)):
+                epoch_start = perf_counter()
+                for future in as_completed(
+                    executor.submit(self.client.insert_batch, records)
+                    for records in batched(epoch, batch_size)
+                ):
+                    try:
+                        future.result()
+                    except (CancelledError, TimeoutError) as err:
+                        logger.exception("failed to insert records", exc_info=err)
+                logger.info(
+                    "finished %s records with RPS(%.3f)",
+                    (i + 1) * epoch_size,
+                    epoch_size / (perf_counter() - epoch_start),
+                )
+
+    def _query_helper(self, query: Query):
+        start_time = perf_counter()
+        records = self.client.query(query.vector, len(query.expect_ids))
+        elapsed = perf_counter() - start_time
+        self.query_result.query += 1
+        self.query_result.latency.append(elapsed)
+        self.query_result.precision.append(
+            len(
+                set(query.expect_ids).intersection(set(record.id for record in records))
+            )
+            / len(query.expect_ids)
+        )
+
+    def query(self) -> BenchmarkResult:
+        logger.info("querying...")
+        epoch_size = 100
+        with ThreadPoolExecutor(self.worker_num) as executor:
+            logger.info("using %s executors", executor._max_workers)
+            start = perf_counter()
+            for i, epoch in enumerate(batched(self.reader.read_query(), epoch_size)):
+                epoch_start = perf_counter()
+                for future in as_completed(
+                    executor.submit(self._query_helper, query) for query in epoch
+                ):
+                    try:
+                        future.result()
+                    except (CancelledError, TimeoutError) as err:
+                        logger.exception("failed to query", exc_info=err)
+                logger.info(
+                    "finished %s queries with RPS (%.3f)",
+                    (i + 1) * epoch_size,
+                    epoch_size / (perf_counter() - epoch_start),
+                )
+            self.query_result.total_second = perf_counter() - start
+            self.query_result.worker_num = executor._max_workers
+
+        return self.query_result
diff --git a/vector_bench/client/__init__.py b/vector_bench/client/__init__.py
@@ -0,0 +1,6 @@
+from vector_bench.client.pgvecto_rs import PgVectorsClient
+from vector_bench.spec import EnumSelector
+
+
+class DataBaseClient(EnumSelector):
+    PGVECTO_RS = PgVectorsClient