edmkit 0.0.1a2__tar.gz → 0.0.3__tar.gz

{edmkit-0.0.1a2 → edmkit-0.0.3}/PKG-INFO

@@ -1,13 +1,14 @@
-Metadata-Version: 2.4
+Metadata-Version: 2.3
 Name: edmkit
-Version: 0.0.1a2
+Version: 0.0.3
 Summary: Simple EDM (Empirical Dynamic Modeling) library
+Author: FUJISHIGE TEMMA
 Author-email: FUJISHIGE TEMMA <tenma.x0@gmail.com>
-License-File: LICENSE
-Requires-Python: >=3.13
 Requires-Dist: numpy>=2.4.3
 Requires-Dist: scipy>=1.17.1
 Requires-Dist: tinygrad>=0.11.0
+Requires-Dist: usearch>=2.23.0
+Requires-Python: >=3.13
 Description-Content-Type: text/markdown
 
 # edmkit

{edmkit-0.0.1a2 → edmkit-0.0.3}/pyproject.toml

@@ -1,19 +1,24 @@
 [project]
 name = "edmkit"
-version = "0.0.1-alpha.2"
+version = "0.0.3"
 description = "Simple EDM (Empirical Dynamic Modeling) library"
 authors = [{ name = "FUJISHIGE TEMMA", email = "tenma.x0@gmail.com" }]
 readme = "README.md"
 requires-python = ">= 3.13"
-dependencies = ["numpy>=2.4.3", "scipy>=1.17.1", "tinygrad>=0.11.0"]
+dependencies = [
+    "numpy>=2.4.3",
+    "scipy>=1.17.1",
+    "tinygrad>=0.11.0",
+    "usearch>=2.23.0",
+]
 
 [dependency-groups]
 dev = [
     "hypothesis>=6.151.9",
     "pytest>=9.0.2",
-    "ruff>=0.15.5",
+    "ruff>=0.15.6",
     "tabulate>=0.10.0",
-    "ty>=0.0.21",
+    "ty>=0.0.23",
 ]
 
 [tool.pytest.ini_options]
@@ -23,6 +28,11 @@ markers = [
     "gpu: marks tests requiring tinygrad tensor backend (deselect with '-m \"not gpu\"')",
 ]
 
+
 [build-system]
-requires = ["hatchling"]
-build-backend = "hatchling.build"
+requires = ["uv_build>=0.10.9,<0.11.0"]
+build-backend = "uv_build"
+
+[tool.uv.build-backend]
+module-name = "edmkit"
+namespace = true

{edmkit-0.0.1a2 → edmkit-0.0.3}/src/edmkit/simplex_projection.py

@@ -1,6 +1,7 @@
 import numpy as np
 from scipy.spatial import KDTree
 from tinygrad import Tensor, dtypes
+from usearch.index import Index
 
 from edmkit.util import pairwise_distance
 
@@ -73,6 +74,20 @@ def simplex_projection(
     return _numpy(X, Y, query_points) if not use_tensor else _tensor(X, Y, query_points)
 
 
+def knn(X: np.ndarray, query_points: np.ndarray, k: int) -> tuple[np.ndarray, np.ndarray]:
+    N, E = X.shape
+    if E >= 15 and N >= 10_000:
+        index = Index(ndim=E, metric="l2sq")
+        index.add(np.arange(len(X)), np.ascontiguousarray(X, dtype=np.float32))
+        matches = index.search(np.ascontiguousarray(query_points, dtype=np.float32), k)
+        distances = np.atleast_2d(np.sqrt(np.asarray(matches.distances)))
+        indices = np.atleast_2d(np.asarray(matches.keys).astype(np.intp))
+        return distances, indices
+    else:
+        tree = KDTree(X)
+        return tree.query(query_points, k=k)
+
+
 def _numpy(
     X: np.ndarray,
     Y: np.ndarray,
@@ -117,8 +132,7 @@ def _numpy(
 
         k: int = X.shape[1] + 1
 
-        tree = KDTree(X)
-        distances, indices = tree.query(query_points, k=k)
+        distances, indices = knn(X, query_points, k)
 
         Y_neighbors = Y[indices]  # (M, k, E')
 
@@ -146,8 +160,7 @@ def _numpy(
         distances = np.empty((B, M, k))
         indices = np.empty((B, M, k), dtype=np.intp)
         for b in range(B):
-            tree = KDTree(X[b])
-            distances[b], indices[b] = tree.query(query_points[b], k=k)
+            distances[b], indices[b] = knn(X[b], query_points[b], k)
 
         batch_idx = np.arange(B)[:, None, None]  # (B, 1, 1)
         Y_neighbors = Y[batch_idx, indices]  # (B, M, k, E')
@@ -219,10 +232,10 @@ def _tensor(
         Y_neighbors = Y_tensor[indices]  # (M, k, E')
 
         d_min = distances[:, :1].clip(min_=1e-6)  # (M, 1)
-        weights: Tensor = (-distances / d_min).exp()  # (M, k) # type: ignore
+        weights: Tensor = (-distances / d_min).exp()  # (M, k)
 
-        weighted_sum: Tensor = (weights.unsqueeze(-1) * Y_neighbors).sum(axis=1)  # (M, E') # type: ignore
-        predictions: Tensor = weighted_sum / weights.sum(axis=1, keepdim=True)  # (M, E') # type: ignore
+        weighted_sum: Tensor = (weights.unsqueeze(-1) * Y_neighbors).sum(axis=1)  # (M, E')
+        predictions: Tensor = weighted_sum / weights.sum(axis=1, keepdim=True)  # (M, E')
 
         return predictions.numpy().squeeze()
     # X (B, N, E), Y (B, N, E'), query_points (B, M, E)
@@ -268,10 +281,10 @@ def _tensor(
         # ---------------------------------------------------------------------------
 
         d_min = distances[:, :, :1].clip(min_=1e-6)  # (B, M, 1)
-        weights: Tensor = (-distances / d_min).exp()  # (B, M, k) # type: ignore
+        weights: Tensor = (-distances / d_min).exp()  # (B, M, k)
 
-        weighted_sum: Tensor = (weights.unsqueeze(-1) * Y_neighbors).sum(axis=2)  # (B, M, E') # type: ignore
-        predictions: Tensor = weighted_sum / weights.sum(axis=2, keepdim=True)  # (B, M, E') # type: ignore
+        weighted_sum: Tensor = (weights.unsqueeze(-1) * Y_neighbors).sum(axis=2)  # (B, M, E')
+        predictions: Tensor = weighted_sum / weights.sum(axis=2, keepdim=True)  # (B, M, E')
 
         return predictions.numpy()
     else:

{edmkit-0.0.1a2 → edmkit-0.0.3}/src/edmkit/util.py

@@ -27,7 +27,7 @@ def pad(As: list[np.ndarray]):
 
     B = len(As)
     L = As[0].shape[0]
-    max_D = max(t.shape[-1] for t in As)  # type: ignore
+    max_D = max(t.shape[-1] for t in As)
 
     A = np.zeros((B, L, max_D), dtype=As[0].dtype)
     for i, x in enumerate(As):
@@ -74,7 +74,7 @@ def pairwise_distance(A: Tensor, B: Tensor | None = None) -> Tensor:
     A_sq = A.pow(2).sum(-1, keepdim=True)
     B_sq = B.pow(2).sum(-1, keepdim=True).transpose(-1, -2)
 
-    D: Tensor = A_sq + B_sq - 2 * A.matmul(B.transpose(-1, -2))  # type: ignore
+    D: Tensor = A_sq + B_sq - 2 * A.matmul(B.transpose(-1, -2))
 
     return D.clamp(min_=0)
 

edmkit-0.0.1a2/.claude/settings.json

@@ -1,13 +0,0 @@
-{
-  "permissions": {
-    "allow": [
-      "Bash(grep:*)",
-      "Bash(uv run ruff:*)",
-      "Bash(uv run ty:*)",
-      "Bash(uv run pytest:*)",
-      "Bash(uv run python:*)"
-    ],
-    "deny": [],
-    "ask": []
-  }
-}

edmkit-0.0.1a2/.github/workflows/ci.yaml

@@ -1,41 +0,0 @@
-name: Test
-
-on:
-  pull_request:
-  push:
-    branches:
-      - main
-
-jobs:
-  lint:
-    name: Lint
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v6
-      - uses: astral-sh/setup-uv@v7
-      - name: Ruff lint
-        run: uv run ruff check .
-      - name: Ruff format
-        run: uv run ruff format --diff .
-
-  type-check:
-    name: Type check
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v6
-      - uses: astral-sh/setup-uv@v7
-      - name: Ty type check
-        run: uv run ty check
-
-  test:
-    name: Test (${{ matrix.os }})
-    strategy:
-      matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
-    runs-on: ${{ matrix.os }}
-    env:
-      HYPOTHESIS_PROFILE: ci
-    steps:
-      - uses: actions/checkout@v6
-      - uses: astral-sh/setup-uv@v7
-      - run: uv run pytest

edmkit-0.0.1a2/.github/workflows/release.yaml

@@ -1,61 +0,0 @@
-name: Release
-
-on:
-  push:
-    branches:
-      - main
-
-jobs:
-  check-version:
-    name: Check if version is new
-    runs-on: ubuntu-latest
-    outputs:
-      should-release: ${{ steps.check.outputs.should-release }}
-      version: ${{ steps.check.outputs.version }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v6
-      - name: Install uv
-        uses: astral-sh/setup-uv@v7
-      - name: Check PyPI for existing version
-        id: check
-        run: |
-          VERSION=$(uv run python -c "import tomllib; print(tomllib.load(open('pyproject.toml','rb'))['project']['version'])")
-          echo "version=$VERSION" >> "$GITHUB_OUTPUT"
-          STATUS=$(curl -s -o /dev/null -w '%{http_code}' "https://pypi.org/pypi/edmkit/$VERSION/json")
-          if [ "$STATUS" = "404" ]; then
-            echo "should-release=true" >> "$GITHUB_OUTPUT"
-          else
-            echo "should-release=false" >> "$GITHUB_OUTPUT"
-          fi
-
-  pypi:
-    name: Publish to PyPI
-    needs: check-version
-    if: needs.check-version.outputs.should-release == 'true'
-    runs-on: ubuntu-latest
-    environment:
-      name: pypi
-    permissions:
-      id-token: write
-      contents: read
-      attestations: write
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v6
-      - name: Install uv
-        uses: astral-sh/setup-uv@v7
-      - name: Install Python 3.13
-        run: uv python install 3.13
-      - name: Build
-        run: uv build
-      - name: Smoke test (wheel)
-        run: uv run --isolated --no-project --with dist/*.whl tests/smoke_test.py
-      - name: Smoke test (source distribution)
-        run: uv run --isolated --no-project --with dist/*.tar.gz tests/smoke_test.py
-      - name: Generate attestations
-        uses: actions/attest@v4
-        with:
-          subject-path: "dist/*"
-      - name: Publish
-        run: uv publish --check-url https://pypi.org/simple/

edmkit-0.0.1a2/.gitignore

@@ -1,12 +0,0 @@
-# python generated files
-__pycache__/
-*.py[oc]
-build/
-dist/
-wheels/
-*.egg-info
-
-# venv
-.venv
-.hypothesis/*
-!.hypothesis/examples/

edmkit-0.0.1a2/.hypothesis/examples/04e6b3400353b141/1626189b6bf56f80

@@ -1,3 +0,0 @@
-���U������

-Ẽh���,z�#So�
-P$�p��lj�S

edmkit-0.0.1a2/.hypothesis/examples/04e6b3400353b141/d6ffe0a2e4161c9f

@@ -1,3 +0,0 @@
-���U������

-Ẽh���,z�#So�
-P$�p��lj�S.secondary

edmkit-0.0.1a2/.python-version

@@ -1 +0,0 @@
-3.13.0

edmkit-0.0.1a2/CLAUDE.md

@@ -1 +0,0 @@
-@AGENTS.md

edmkit-0.0.1a2/LICENSE

@@ -1,21 +0,0 @@
-MIT License
-
-Copyright (c) 2025 temma
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.

edmkit-0.0.1a2/benchmarks/ccm.py

@@ -1,124 +0,0 @@
-from __future__ import annotations
-
-import time
-from dataclasses import dataclass
-from functools import partial
-from itertools import product
-
-import numpy as np
-from tabulate import tabulate
-
-from edmkit.ccm import ccm
-from edmkit.simplex_projection import simplex_projection
-
-rng = np.random.default_rng(42)
-
-N_POINTS = (100, 1000, 10000)
-X_DIMS = (1, 10, 100)
-Y_DIMS = (1, 10, 100)
-N_SAMPLES = (10, 20, 30)
-LIBRARY_RATIO = 0.7
-LIB_SIZE_STEPS = 4
-BATCH_LIMIT = 8
-
-
-@dataclass(frozen=True)
-class BenchmarkCase:
-    n_points: int
-    x_dim: int
-    y_dim: int
-    n_samples: int
-
-    @property
-    def name(self) -> str:
-        return f"n{self.n_points}_x{self.x_dim}_y{self.y_dim}_s{self.n_samples}"
-
-    @property
-    def batch_size(self) -> int:
-        return min(BATCH_LIMIT, self.n_samples)
-
-
-def iter_cases() -> list[BenchmarkCase]:
-    return [BenchmarkCase(n, xd, yd, ns) for n, xd, yd, ns in product(N_POINTS, X_DIMS, Y_DIMS, N_SAMPLES)]
-
-
-def synthesize_series(n_points: int, x_dim: int, y_dim: int) -> tuple[np.ndarray, np.ndarray]:
-    """Generate coupled synthetic data with controllable dimensionality."""
-
-    drivers = rng.standard_normal((n_points, x_dim))
-    weights = rng.standard_normal((x_dim, max(y_dim, 1)))
-    responses = drivers @ weights + 0.1 * rng.standard_normal((n_points, max(y_dim, 1)))
-
-    X = drivers if x_dim > 1 else drivers.reshape(n_points, 1)
-    Y = responses[:, :y_dim] if y_dim > 1 else responses[:, 0]
-
-    return X, Y
-
-
-def compute_lib_sizes(min_size: int, pool_length: int, steps: int) -> np.ndarray:
-    base = max(min_size, 2)
-    max_candidate = max(pool_length, base)
-    if max_candidate == base:
-        max_candidate = base + max(2, base // 2)
-
-    values = np.linspace(base, max_candidate, num=steps)
-    lib_sizes = np.unique(np.maximum(values.astype(int), base))
-    if lib_sizes.size == 0:
-        lib_sizes = np.array([base])
-    return lib_sizes
-
-
-def main() -> None:
-    predict_func = partial(simplex_projection, use_tensor=False)
-    rows: list[list[str | int]] = []
-
-    for case in iter_cases():
-        X, Y = synthesize_series(case.n_points, case.x_dim, case.y_dim)
-        split = max(2, min(case.n_points - 1, int(case.n_points * LIBRARY_RATIO)))
-        library_pool = np.arange(split)
-        prediction_pool = np.arange(split, case.n_points)
-        lib_sizes = compute_lib_sizes(case.x_dim + 1, library_pool.size, LIB_SIZE_STEPS)
-
-        start = time.perf_counter()
-        ccm(
-            X=X,
-            Y=Y,
-            lib_sizes=lib_sizes,
-            predict_func=predict_func,
-            n_samples=case.n_samples,
-            library_pool=library_pool,
-            prediction_pool=prediction_pool,
-            batch_size=case.batch_size,
-        )
-        elapsed = time.perf_counter() - start
-
-        rows.append(
-            [
-                case.name,
-                case.n_points,
-                case.x_dim,
-                case.y_dim,
-                case.n_samples,
-                library_pool.size,
-                f"{lib_sizes.tolist()}",
-                case.batch_size,
-                f"{elapsed:.3f}",
-            ]
-        )
-
-    headers = [
-        "scenario",
-        "n_points",
-        "x_dim",
-        "y_dim",
-        "n_samples",
-        "pool",
-        "lib sizes",
-        "batch",
-        "time (s)",
-    ]
-    print(tabulate(rows, headers=headers, tablefmt="github"))
-
-
-if __name__ == "__main__":
-    main()

edmkit-0.0.1a2/ruff.toml

@@ -1 +0,0 @@
-line-length = 150

edmkit-0.0.1a2/src/edmkit/__init__.py

@@ -1,5 +0,0 @@
-# ruff: noqa: F401
-from .ccm import bootstrap
-from .embedding import lagged_embed
-from .simplex_projection import simplex_projection
-from .smap import smap