usearch 0.8.0 → 0.9.0

package/.github/workflows/release.yml

@@ -160,11 +160,51 @@ jobs:
     - name: Run tests
       run: swift test
 
+  # Publishes to both Docker Hub and GitHub Container Registry
+  # https://docs.github.com/en/actions/publishing-packages/publishing-docker-images#publishing-images-to-docker-hub-and-github-packages
+  publish_docker:
+    name: Publish Docker Image
+    needs: versioning
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v3
+      with:
+        ref: 'main'  
+    - run: git submodule update --init --recursive
+    - name: Login to Docker Hub
+      uses: docker/login-action@v2
+      with:
+        username: ${{ secrets.DOCKERHUB_USERNAME }}
+        password: ${{ secrets.DOCKERHUB_TOKEN }}
+    - name: Log in to GitHub Registry
+      uses: docker/login-action@2
+      with:
+        registry: ghcr.io
+        username: ${{ github.actor }}
+        password: ${{ secrets.GITHUB_TOKEN }}
+    - name: Set up Docker Buildx
+      uses: docker/setup-buildx-action@v2
+    - name: Extract tags & labels for Docker
+      id: meta
+      uses: docker/metadata-action@4
+      with:
+        images: |
+          unum/usearch
+          ghcr.io/${{ github.repository }}
+    - name: Build and push
+      uses: docker/build-push-action@v4
+      with:
+        context: .
+        push: true
+        tags: ${{ steps.meta.outputs.tags }}
+        labels: ${{ steps.meta.outputs.labels }}
+
+
   build_docs:
     name: Build Docs
     runs-on: ubuntu-22.04
     if: ${{ always() }}
-    needs: [publish_python, publish_javascript, publish_rust, publish_java, publish_swift]
+    needs: [publish_python, publish_javascript, publish_rust, publish_java, publish_swift, publish_docker]
     permissions:
       contents: write
     steps:

package/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "usearch"
-version = "0.8.0"
+version = "0.9.0"
 authors = ["Ashot Vardanian <1983160+ashvardanian@users.noreply.github.com>"]
 description = "Smaller & Faster Single-File Vector Search Engine from Unum"
 edition = "2021"

package/Dockerfile

@@ -1,6 +1,8 @@
 # syntax=docker/dockerfile:1
 
 FROM ubuntu:23.04
+ENV DEBIAN_FRONTEND=noninteractive
+
 RUN apt-get update && \
     apt-get install -y --no-install-recommends python3.11 python3-pip && \
     apt-get clean && \
@@ -10,5 +12,5 @@ RUN pip3 install --no-cache-dir --break-system-packages ucall usearch
 WORKDIR /usearch
 COPY python/usearch/server.py server.py
 
-CMD ["python3", "./server.py"]
+ENTRYPOINT ["python3", "./server.py"]
 EXPOSE 8545

package/README.md

@@ -417,7 +417,7 @@ assert!(new_haversine(&quant, 0, 0, 0).is_ok());
 
 ```xml
 <dependency>
-  <groupId>cloud.unum.usearch</groupId>
+  <groupId>cloud.unum</groupId>
   <artifactId>usearch</artifactId>
   <version>0.2.3</version>
 </dependency>

package/VERSION

@@ -1 +1 @@
-0.8.0
+0.9.0

package/build.gradle

@@ -6,7 +6,7 @@ plugins {
     id 'maven-publish'
 }
 
-group "cloud.unum.usearch"
+group "cloud.unum"
 version = file("VERSION").text.trim()
 
 sourceCompatibility = 1.8

package/conanfile.py

@@ -4,7 +4,7 @@ from conans import ConanFile
 class USearchConan(ConanFile):
 
     name = 'USearch'
-    version = '0.8.0'
+    version = '0.9.0'
     license = 'Apache 2.0'
     url = 'https://github.com/unum-cloud/usearch'
     description = 'Smaller & Faster Single-File Vector Search Engine from Unum'

package/docs/compilation.md

@@ -136,4 +136,19 @@ brew install --cask wolfram-engine
 
 ```sh
 docker build -t unum/usearch . && docker run unum/usearch
+```
+
+For multi-architecture builds and publications:
+
+```sh
+version=$(cat VERSION)
+docker buildx create --use &&
+    docker login &&
+    docker buildx build \
+        --platform "linux/amd64,linux/arm64" \
+        --build-arg version=$version \
+        --file Dockerfile \
+        --tag unum/usearch:$version \
+        --tag unum/usearch:latest \
+        --push .
 ```

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "usearch",
-    "version": "0.8.0",
+    "version": "0.9.0",
     "description": "Smaller & Faster Single-File Vector Search Engine from Unum",
     "author": "Ashot Vardanian",
     "license": "Apache 2.0",

package/pyproject.toml

@@ -10,7 +10,7 @@ filterwarnings = ["error"]
 
 [tool.cibuildwheel]
 test-requires = ["pytest", "numpy"]
-test-command = "pytest {project}/python/test.py"
+test-command = "pytest {project}/python/scripts/test.py"
 build-verbosity = 0
 skip = ["*musllinux*", "*i686*", "pp*"]
 

package/python/scripts/bench.py

@@ -1,9 +1,8 @@
 import time
 import numpy as np
-import fire
+import math
 
-import numpy as np
-from numba import cfunc, types, carray
+import fire
 
 from usearch.index import Index
 from usearch.io import load_matrix
@@ -51,25 +50,6 @@ def bench_usearch(index, vectors: np.array, queries: np.array, neighbors: np.arr
     recall_at_one /= queries.shape[0]
     print(f'- Recall@1: {recall_at_one * 100:.2f} %')
 
-# Showcases how to use Numba to JIT-compile similarity measures for USearch.
-# https://numba.readthedocs.io/en/stable/reference/jit-compilation.html#c-callbacks
-
-
-signature_f32 = types.float32(
-    types.CPointer(types.float32),
-    types.CPointer(types.float32),
-    types.size_t, types.size_t)
-
-
-@cfunc(signature_f32)
-def inner_product_f32(a, b, n, m):
-    a_array = carray(a, n)
-    b_array = carray(b, n)
-    c = types.float32(0)
-    for i in range(n):
-        c += a_array[i] * b_array[i]
-    return types.float32(1 - c)
-
 
 def main(
     vectors: str,
@@ -94,16 +74,6 @@ def main(
     )
     bench_usearch(index, vectors_mat, queries_mat, neighbors_mat)
 
-    print('USearch: HNSW + Numba JIT')
-    index = Index(
-        ndim=dim,
-        expansion_add=expansion_add,
-        expansion_search=expansion_search,
-        connectivity=connectivity,
-        metric_pointer=inner_product_f32.address,
-    )
-    bench_usearch(index, vectors_mat, queries_mat, neighbors_mat)
-
     print('USearch: HNSW + Half Precision')
     index = Index(
         ndim=dim,
@@ -124,27 +94,94 @@ def main(
     )
     bench_usearch(index, vectors_mat, queries_mat, neighbors_mat)
 
+    # Showcases how to use Numba to JIT-compile similarity measures for USearch.
+    # https://numba.readthedocs.io/en/stable/reference/jit-compilation.html#c-callbacks
+    try:
+        from numba import cfunc, types, carray
+
+        signature_f32 = types.float32(
+            types.CPointer(types.float32),
+            types.CPointer(types.float32),
+            types.size_t, types.size_t)
+
+        @cfunc(signature_f32)
+        def inner_product_f32(a, b, n, m):
+            a_array = carray(a, n)
+            b_array = carray(b, n)
+            c = types.float32(0)
+            for i in range(n):
+                c += a_array[i] * b_array[i]
+            return types.float32(1 - c)
+
+        signature_f8 = types.float32(
+            types.CPointer(types.int8),
+            types.CPointer(types.int8),
+            types.size_t, types.size_t)
+
+        @cfunc(signature_f8)
+        def cos_f8(a, b, n, m):
+            a_array = carray(a, n)
+            b_array = carray(b, n)
+            c = types.int32(0)
+            a_sq = types.int32(0)
+            b_sq = types.int32(0)
+            for i in range(n):
+                ai = types.int16(a_array[i])
+                bi = types.int16(b_array[i])
+                c += ai * bi
+                a_sq += ai * ai
+                b_sq += bi * bi
+            a_norm = math.sqrt(types.float32(a_sq))
+            b_norm = math.sqrt(types.float32(b_sq))
+            c_norm = types.float32(c) / (a_norm * b_norm)
+            return types.float32(1 - c_norm)
+
+        print('USearch: HNSW + Numba JIT')
+        index = Index(
+            ndim=dim,
+            expansion_add=expansion_add,
+            expansion_search=expansion_search,
+            connectivity=connectivity,
+            metric_pointer=inner_product_f32.address,
+        )
+        bench_usearch(index, vectors_mat, queries_mat, neighbors_mat)
+
+        print('USearch: HNSW + Quarter Precision + Numba JIT')
+        index = Index(
+            ndim=dim,
+            dtype='f8',
+            expansion_add=expansion_add,
+            expansion_search=expansion_search,
+            connectivity=connectivity,
+            metric_pointer=cos_f8.address,
+        )
+        bench_usearch(index, vectors_mat, queries_mat, neighbors_mat)
+    except ImportError:
+        print('Skipping Numba benchmarks')
+
     # Don't depend on the FAISS installation for benchmarks
     try:
         from faiss import IndexHNSWFlat, IndexIVFPQ
+
+        print('FAISS: HNSW')
+        index = IndexHNSWFlat(dim, connectivity)
+        index.hnsw.efSearch = expansion_search
+        index.hnsw.efConstruction = expansion_add
+        bench_faiss(index, vectors_mat, queries_mat, neighbors_mat)
+
+        print('FAISS: HNSW + IVFPQ')
+        # Number of inverted lists (number of partitions or cells).
+        nlist = 10000
+        nsegment = 16  # Number of segments for PQ (number of subquantizers).
+        nbit = 8       # Number of bits to encode each segment.
+        coarse_quantizer = IndexHNSWFlat(dim, connectivity)
+        index = IndexIVFPQ(coarse_quantizer, dim, nlist, nsegment, nbit)
+        index.train(vectors_mat)
+        index.nprobe = 10
+        bench_faiss(index, vectors_mat, queries_mat, neighbors_mat)
+
     except ImportError:
-        return
-
-    print('FAISS: HNSW')
-    index = IndexHNSWFlat(dim, connectivity)
-    index.hnsw.efSearch = expansion_search
-    index.hnsw.efConstruction = expansion_add
-    bench_faiss(index, vectors_mat, queries_mat, neighbors_mat)
-
-    print('FAISS: HNSW + IVFPQ')
-    nlist = 10000  # Number of inverted lists (number of partitions or cells).
-    nsegment = 16  # Number of segments for PQ (number of subquantizers).
-    nbit = 8       # Number of bits to encode each segment.
-    coarse_quantizer = IndexHNSWFlat(dim, connectivity)
-    index = IndexIVFPQ(coarse_quantizer, dim, nlist, nsegment, nbit)
-    index.train(vectors_mat)
-    index.nprobe = 10
-    bench_faiss(index, vectors_mat, queries_mat, neighbors_mat)
+        print('Skipping FAISS benchmarks')
 
 
 if __name__ == '__main__':

package/python/scripts/bench_params.py

@@ -1,6 +1,7 @@
 import os
 import time
 import numpy as np
+
 import fire
 import plotly.express as px
 import numpy as np