Skip to main content

Shimaenaga: Attentive Histogram GBDT with sample-level token attention

Project description

Shimaenaga

Attentive Histogram GBDT — サンプル単位のトークン注意機構を備えた勾配ブースティング

Shimaenaga は、LightGBM スタイルのヒストグラムベース GBDT に、サンプルごとの トークン注意機構(attention)を組み込んだ機械学習ライブラリです。特徴量を 「トークン」と呼ぶサブセットに分割して各トークンごとに木を成長させ、 サンプルごとに学習された注意重みでそれらを混合します。これにより、 純粋な GBDT では捉えにくい特徴グループ間の相互作用を表現できます。

  • コアは C++17(依存ライブラリなし、OpenMP 並列)
  • Python から scikit-learn 互換 API で利用可能(fit / predict / predict_proba / clone / GridSearchCV / cross_val_score 対応)
  • 回帰(L2 / Huber / Quantile / MAE)・二値分類・多クラス分類・ ランキング(LambdaMART)をサポート

アーキテクチャ概要

学習は 2 相構成です。Phase A で木構造を成長させ、Phase B で Newton 法によるパラメータの再適合(refit)を行います。

入力 X  ──→  [BinMapper]  ──→  ビン化された特徴

Phase A ─── 各トークン p ごとにツリー成長 (LightGBM leaf-wise)
             ゲートツリー(ルーティング用)も同時に成長

Phase B ─── Newton refit(反復最適化)
             B1: 葉値 v_{pℓk}
             B2: 読み出し QK 埋め込み (q, k, b)
             B3: 自己注意 QK(Tier-2 のみ)
             B4: ヘッド重み ρ_h

注意機構

トークン p: 特徴のサブセット S_p から学習した「部品」
ゲート:     α_{ihp} = softmax(q_{h,gate_leaf}^T k_{h,p,token_leaf} / √d_a)
混合:       β_{ip}  = Σ_h ρ_h · α_{ihp}   (サンプルごとの部品重み)
出力:       φ_{ik}  = Σ_p β_{ip} · v_{p,leaf_p(i),k}

各サンプルがどのトークン(特徴グループ)をどれだけ重視するかを、 ゲートツリーと QK 埋め込みが決定します。

Tier システム

Tier 機能 用途
0 純粋 GBDT(LightGBM 退化形) ベースライン、最速
1 + Attentive Readout 特徴グループ間の相互作用
2 + Self-Attention(Tier-1 の上) 複雑なトークン間依存関係

tier-1/2 の表現力は tier-0 を常に包含するよう設計されています (最初のトークンは全特徴を使用)。


ビルドとインストール

必要環境

  • C++17 コンパイラ(GCC ≥ 9, Clang ≥ 11, MSVC 2019 以降)
  • CMake ≥ 3.20
  • Python ≥ 3.8 + NumPy
  • scikit-learn(sklearn 互換 API と examples の実行に使用)
  • (オプション)pybind11 — ネイティブ拡張としてビルドする場合のみ

1. C++ コアのビルド(必須)

cmake -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build

これで共有ライブラリ build/libshimaenaga_core.dylib(macOS)/ .so(Linux)が生成されます。

2. Python パッケージの利用

Python パッケージ shimaenagapybind11 なしでも動作します。 pybind11 拡張が見つからない場合、ビルド済み共有ライブラリを ctypes 経由で 自動的に呼び出します(機能は同一)。

# 方法 A: そのまま使う(ビルド済み共有ライブラリを自動検出)
PYTHONPATH=python python3 -c "import shimaenaga; print(shimaenaga.__version__)"

# 方法 B: 開発インストール
pip install -e .

# 方法 C: pybind11 ネイティブ拡張としてビルド
pip install pybind11
pip install -e .

クイックスタート

from shimaenaga import ShimaenagaRegressor, ShimaenagaClassifier, ShimaenagaRanker

回帰

from sklearn.datasets import fetch_california_housing
from sklearn.model_selection import train_test_split
from shimaenaga import ShimaenagaRegressor

data = fetch_california_housing()
X_train, X_test, y_train, y_test = train_test_split(
    data.data, data.target, test_size=0.2, random_state=42
)

model = ShimaenagaRegressor(
    tier=1,             # Tier-1: Attentive Readout
    num_tokens=4,       # 特徴を 4 グループに分割
    num_heads=2,        # 注意ヘッド数
    d_attn=4,           # 注意埋め込み次元
    num_iterations=300,
    learning_rate=0.05,
)
model.fit(X_train, y_train)
y_pred = model.predict(X_test)

二値分類

from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from shimaenaga import ShimaenagaClassifier

data = load_breast_cancer()
X_train, X_test, y_train, y_test = train_test_split(
    data.data, data.target, test_size=0.2, random_state=42
)

clf = ShimaenagaClassifier(
    num_class=1,        # 1 = 二値(既定)
    tier=1,
    num_tokens=4,
    num_iterations=200,
)
clf.fit(X_train, y_train)
proba = clf.predict_proba(X_test)   # shape (n, 2)
labels = clf.predict(X_test)

多クラス分類

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from shimaenaga import ShimaenagaClassifier

data = load_iris()
X_train, X_test, y_train, y_test = train_test_split(
    data.data, data.target, test_size=0.2, random_state=42
)

clf = ShimaenagaClassifier(
    num_class=3,        # K クラス softmax
    tier=1,
    num_tokens=2,
    num_iterations=100,
)
clf.fit(X_train, y_train)
proba = clf.predict_proba(X_test)   # shape (n, 3)

ランキング(LambdaMART)

from shimaenaga import ShimaenagaRanker

# group: クエリごとのサンプル数(合計 = len(X_train))
model = ShimaenagaRanker(tier=1, num_iterations=200)
model.fit(X_train, y_train, group=group_train)
scores = model.predict(X_test)

# 検証セット付き(eval_group 必須 — valid の NDCG をクエリ単位で計算するため)
model.fit(X_train, y_train, group=group_train,
          eval_set=[(X_valid, y_valid)], eval_group=[group_valid],
          early_stopping_rounds=50)

注意重みの取得(解釈性)

diag = model.attention_diagnostics(X_test)
beta = diag["beta"]      # shape (n, num_tokens): 各サンプルのトークン重み
                         # (ブースティングブロック平均、行和 = 1)

Early Stopping

model = ShimaenagaRegressor(num_iterations=1000, early_stopping_rounds=50)
model.fit(X_train, y_train, eval_set=[(X_valid, y_valid)])
print(model.best_iteration_)

モデルの保存・ロード

model.save_model("model.sbb")       # バイナリ形式 (.sbb, 形式 v3)

# ロード: 同じ設定でインスタンスを作り、一度 fit で booster を初期化してから load
loaded = ShimaenagaRegressor(tier=1, num_tokens=4)
loaded.fit(X_tiny, y_tiny)          # booster の初期化(小さなデータで可)
loaded.load_model("model.sbb")
pred = loaded.predict(X_test)

パラメータ一覧

既定値はコンストラクタ(python/shimaenaga/sklearn_api.py)および C++ Config と一致しています。

基本設定

パラメータ 既定値 説明
objective "regression" "regression" / "huber" / "quantile" / "mae" / "binary" / "multiclass" / "lambdarank"
num_iterations 1000 ブースティング反復数
learning_rate 0.05 学習率 η
num_class 1 クラス数(Classifier のみ。1=二値, K>1=多クラス)
huber_alpha 0.9 Huber の適応的 δ = 残差絶対値の α 分位点(毎反復更新)
quantile_alpha 0.5 quantile 回帰の分位点(pinball 損失)
num_threads 0 スレッド数(0 = 自動)
seed 0 乱数シード(同一シードでビット単位に再現)

アーキテクチャ

パラメータ 既定値 説明
tier 2 0=純 GBDT, 1=注意読み出し, 2=+自己注意
num_tokens 8 トークン数 P(上限 16)
num_heads 2 注意ヘッド数 H(上限 4)
attention_mode "qk_leaf" "qk_leaf"(学習済 QK)/ "score_tree"(直接スコア)
d_attn 4 QK 埋め込み次元 d_a(上限 8)
eta_attn 0.5 自己注意の混合率 η(Tier-2)
attn_mask "full" "full" / "feature_local"(特徴を共有するトークン間のみ自己注意。Tier-2 の過学習・メモリ抑制)

ツリー

パラメータ 既定値 説明
token_num_leaves 31 トークンツリーの最大葉数
gate_num_leaves 31 ゲートツリーの最大葉数
max_depth -1 木の最大深さ(-1 = 無制限。leaf-wise 成長の深さ上限)
min_data_in_leaf 20 葉の最小サンプル数(別名 min_child_samples)
min_sum_hessian_in_leaf 1e-3 葉の最小ヘシアン量(別名 min_child_weight)
max_bin 255 ヒストグラムビン数(上限 256)

正則化

パラメータ 既定値 説明
lambda_l1 0.0 分割ゲインの L1 正則化(別名 reg_alpha)
lambda_v 1.0 葉値の L2 正則化(別名 reg_lambda)
lambda_q 0.1 クエリベクトルの L2 正則化
lambda_k 0.1 キーベクトルの L2 正則化
lambda_z 0.1 スコアツリーパラメータの正則化
lambda_ent 1e-3 注意分布のエントロピー正則化(崩壊防止)
lambda_div 1e-3 ヘッド多様性正則化(cos²(Ā_h, Ā_h') を抑制、Tier-2)
min_gain_to_split 0.0 分割に要求する最小ゲイン
bagging_fraction 1.0 行サブサンプリング率(別名 subsample)
bagging_freq 0 バギングマスクの再抽選間隔(反復数)
feature_fraction 1.0 木ごとの特徴サブサンプリング率(別名 colsample_bytree)
early_stopping_rounds 0 検証メトリクスが改善しない場合に打ち切る反復数(eval_set 必須)

早期打ち切りの監視メトリクスは回帰 = RMSE(huber は MAE、quantile は pinball)、 二値 = logloss、多クラス = logloss、ランキング = 1 − NDCG@T です。

Phase B 内部設定

通常は変更不要です。

パラメータ 既定値 説明
inner_refit_steps 2 Phase B の Newton 反復数
attn_warmup 10 α=1/P 固定で value のみ学習する初期反復数(実効値は min(attn_warmup, num_iterations/5))
attn_step_clip 1.0 ω / b / bA のスカラー Newton ステップ上限(0=無効)。softmax の 0/1 飽和による attention collapse を防止
beta_uniform_mix 0.1 Phase A のトークン重み w = (1−γ)β̂ + γ/P の γ。飢餓トークンの回復を保証

Tier ガイド 0 1 2

データの特性 推奨設定
特徴数 < 20 tier=0 または tier=1(num_tokens=2〜4)
特徴数 20〜100 tier=1(num_tokens=4〜8)
特徴数 > 100 tier=1(num_tokens=8〜16)または tier=2
サンプル数 < 500 tier=0 または tier=1(min_data_in_leaf を大きく)

使用ライブラリ・著作権表示

Shimaenaga は以下のサードパーティライブラリを使用しています。

ランタイム依存

ライブラリ ライセンス 著作権
NumPy BSD 3-Clause Copyright (c) 2005-2025, NumPy Developers
scikit-learn BSD 3-Clause Copyright (c) 2007-2026, The scikit-learn developers

ビルド時依存

ライブラリ ライセンス 著作権
pybind11 BSD 3-Clause Copyright (c) 2016 Wenzel Jakob and others

アルゴリズム参照・著作権表示

Shimaenaga は以下のプロジェクトで発表されたアルゴリズムや手法を参考に実装しています。

XGBoost

  • Project: https://github.com/dmlc/xgboost
  • License: Apache 2.0
  • Reference: Tianqi Chen and Carlos Guestrin. "XGBoost: A Scalable Tree Boosting System." KDD 2016. https://arxiv.org/abs/1603.02754
  • 参考にした手法: L1/L2 正則化付き葉の重み計算式、ヘシアンベースの分割ゲイン、最小子ノード制約

LightGBM

CatBoost

  • Project: https://github.com/catboost/catboost
  • License: Apache 2.0
  • Reference: Liudmila Prokhorenkova et al. "CatBoost: unbiased boosting with categorical features." NeurIPS 2018. https://arxiv.org/abs/1706.09516
  • 参考にした手法: Ordered Boosting(予測シフト低減)、Ordered Target Encoding(カテゴリカル特徴量)、対称木(Oblivious Tree)成長

Gradient Boosting Machine

  • Reference: Jerome H. Friedman. "Greedy Function Approximation: A Gradient Boosting Machine." Annals of Statistics, 2001. https://www.jstor.org/stable/2699986
  • 参考にした手法: 勾配ブースティング全体の理論的基盤(残差への逐次的な関数近似、加法モデル)

Attention Is All You Need

  • Reference: Ashish Vaswani et al. "Attention Is All You Need." NeurIPS 2017. https://arxiv.org/abs/1706.03762
  • 参考にした手法: スケーリング付き内積注意(scaled dot-product attention, softmax(QK^T / √d))、マルチヘッド注意 — Tier-2 自己注意およびゲート機構の設計に対応

LambdaMART


ライセンス

MIT License. 詳細は LICENSE を参照してください。 サードパーティ(pybind11)の著作権表示は THIRD_PARTY_NOTICES.md を参照してください。

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

shimaenaga-1.3.0.tar.gz (23.6 kB view details)

Uploaded Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

shimaenaga-1.3.0-cp313-cp313-win_amd64.whl (293.0 kB view details)

Uploaded CPython 3.13Windows x86-64

shimaenaga-1.3.0-cp313-cp313-manylinux_2_28_x86_64.whl (460.1 kB view details)

Uploaded CPython 3.13manylinux: glibc 2.28+ x86-64

shimaenaga-1.3.0-cp313-cp313-macosx_11_0_arm64.whl (279.3 kB view details)

Uploaded CPython 3.13macOS 11.0+ ARM64

shimaenaga-1.3.0-cp312-cp312-win_amd64.whl (293.0 kB view details)

Uploaded CPython 3.12Windows x86-64

shimaenaga-1.3.0-cp312-cp312-manylinux_2_28_x86_64.whl (460.1 kB view details)

Uploaded CPython 3.12manylinux: glibc 2.28+ x86-64

shimaenaga-1.3.0-cp312-cp312-macosx_11_0_arm64.whl (279.3 kB view details)

Uploaded CPython 3.12macOS 11.0+ ARM64

shimaenaga-1.3.0-cp311-cp311-win_amd64.whl (289.1 kB view details)

Uploaded CPython 3.11Windows x86-64

shimaenaga-1.3.0-cp311-cp311-manylinux_2_28_x86_64.whl (458.5 kB view details)

Uploaded CPython 3.11manylinux: glibc 2.28+ x86-64

shimaenaga-1.3.0-cp311-cp311-macosx_11_0_arm64.whl (277.2 kB view details)

Uploaded CPython 3.11macOS 11.0+ ARM64

shimaenaga-1.3.0-cp310-cp310-win_amd64.whl (287.8 kB view details)

Uploaded CPython 3.10Windows x86-64

shimaenaga-1.3.0-cp310-cp310-manylinux_2_28_x86_64.whl (457.1 kB view details)

Uploaded CPython 3.10manylinux: glibc 2.28+ x86-64

shimaenaga-1.3.0-cp310-cp310-macosx_11_0_arm64.whl (276.2 kB view details)

Uploaded CPython 3.10macOS 11.0+ ARM64

shimaenaga-1.3.0-cp39-cp39-win_amd64.whl (288.0 kB view details)

Uploaded CPython 3.9Windows x86-64

shimaenaga-1.3.0-cp39-cp39-manylinux_2_28_x86_64.whl (457.3 kB view details)

Uploaded CPython 3.9manylinux: glibc 2.28+ x86-64

shimaenaga-1.3.0-cp39-cp39-macosx_11_0_arm64.whl (276.3 kB view details)

Uploaded CPython 3.9macOS 11.0+ ARM64

File details

Details for the file shimaenaga-1.3.0.tar.gz.

File metadata

  • Download URL: shimaenaga-1.3.0.tar.gz
  • Upload date:
  • Size: 23.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for shimaenaga-1.3.0.tar.gz
Algorithm Hash digest
SHA256 b1470f9c7b86b09a27ddbf82fcb6710f2964ae3f0f7c5dc8eb7d381cbfe8e440
MD5 940b8f59dcaa4dbc1631e08fe3367f7f
BLAKE2b-256 447b42ed74f75bb0c6b90f050f1765dfbd3f5ad38ba741291ae263ad5b2ffefa

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0.tar.gz:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp313-cp313-win_amd64.whl.

File metadata

  • Download URL: shimaenaga-1.3.0-cp313-cp313-win_amd64.whl
  • Upload date:
  • Size: 293.0 kB
  • Tags: CPython 3.13, Windows x86-64
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for shimaenaga-1.3.0-cp313-cp313-win_amd64.whl
Algorithm Hash digest
SHA256 884ed8a7357e15f0a417d6a4e5af3e422e0a034b9387ec23ffd6e3ee6f530ff6
MD5 9092dbd62fbdc785858bb41915cf0558
BLAKE2b-256 67deb19ae55546fcaa634b3c539633873397322db608427ae7ada25ea4dd3243

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp313-cp313-win_amd64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp313-cp313-manylinux_2_28_x86_64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp313-cp313-manylinux_2_28_x86_64.whl
Algorithm Hash digest
SHA256 5ec04c8abcd67b451c74380d94f83215339e8d1e0ee4f236a1117ea8f3c611d1
MD5 3f68d1ea7bd417d14d94d19bc6c53bbe
BLAKE2b-256 3c40162476e8f4c5b786e27276c50b4a6768b21167b22f3054da2e5a812c5d11

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp313-cp313-manylinux_2_28_x86_64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp313-cp313-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp313-cp313-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 91e170a3b36c5a7c6aa33091411795f23b9817290becf37c4bb6a2fb95a28aaf
MD5 895bcd68ccae7c5be9e2ddad96baf379
BLAKE2b-256 eb489b78cc708b276b0b6489233c8fcf387692ef54f1e9a195c84d3b23b86c87

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp313-cp313-macosx_11_0_arm64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp312-cp312-win_amd64.whl.

File metadata

  • Download URL: shimaenaga-1.3.0-cp312-cp312-win_amd64.whl
  • Upload date:
  • Size: 293.0 kB
  • Tags: CPython 3.12, Windows x86-64
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for shimaenaga-1.3.0-cp312-cp312-win_amd64.whl
Algorithm Hash digest
SHA256 356b207e0cfc6d96501a4816b29f6e9060776982f10fc1d983036eacbcafca79
MD5 27e9a2ae981cc36d1f4fa7c8102d5aee
BLAKE2b-256 5718c5b8f1b8b29dcef8200b9b9104acdf000289cb6b182494e26d5469c1475b

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp312-cp312-win_amd64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp312-cp312-manylinux_2_28_x86_64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp312-cp312-manylinux_2_28_x86_64.whl
Algorithm Hash digest
SHA256 9084e0653c5f6ba676b99c0e118367daa149d8a601d82860104e8df9e523e177
MD5 96b58d966a750af964920375c013265c
BLAKE2b-256 ea1ab9af2329ef68933e2d393dbf3f46c7ecfccda83bc5f95ea0bc9ef843c251

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp312-cp312-manylinux_2_28_x86_64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp312-cp312-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp312-cp312-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 9c6dae1b59399b5b8d8404a2030b87b352c99d73cee05ba4be29939378972623
MD5 0ced06bc6fa85d40602156308c8f86ca
BLAKE2b-256 33ca5d0fb62aa1d22f68838a1d3d12f205e91b0656e9ba4f97bf886378cd093c

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp312-cp312-macosx_11_0_arm64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp311-cp311-win_amd64.whl.

File metadata

  • Download URL: shimaenaga-1.3.0-cp311-cp311-win_amd64.whl
  • Upload date:
  • Size: 289.1 kB
  • Tags: CPython 3.11, Windows x86-64
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for shimaenaga-1.3.0-cp311-cp311-win_amd64.whl
Algorithm Hash digest
SHA256 6c50c366ff529540f027c47d57d0f3c974d41c3709d442b3b4a25ce885f22c97
MD5 b257915655f41cf097f12875d1872e1c
BLAKE2b-256 9b31bd4e330406dc45f6779eb3e1d990114c79e567d3d8bba84c48dd02106933

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp311-cp311-win_amd64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp311-cp311-manylinux_2_28_x86_64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp311-cp311-manylinux_2_28_x86_64.whl
Algorithm Hash digest
SHA256 4b09fa3f591df712158c500c7922202b9603d2e9249847bd2cbdead891fe86d4
MD5 87505035d9993c5c0f0e3b37362f9f38
BLAKE2b-256 8ca45dc3a27e1f5d3a9157e525394e5370db80c26428c0fb42e05fc10d5f0116

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp311-cp311-manylinux_2_28_x86_64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp311-cp311-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp311-cp311-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 eff6739de372d8304bf7d3bd522cab4a21db907586a8ddb8d808ad674e9cffd7
MD5 aae3fa79fb43c31badd1bda9f817cb9a
BLAKE2b-256 0303bb3fd422ceaa84cf4b519e3005120e2604013e1afffbf2ce5e48c5ce0458

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp311-cp311-macosx_11_0_arm64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp310-cp310-win_amd64.whl.

File metadata

  • Download URL: shimaenaga-1.3.0-cp310-cp310-win_amd64.whl
  • Upload date:
  • Size: 287.8 kB
  • Tags: CPython 3.10, Windows x86-64
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for shimaenaga-1.3.0-cp310-cp310-win_amd64.whl
Algorithm Hash digest
SHA256 765281787815ebcdc49f251ce138b28b0ca9b58f42e178a80b8aad1eab271b13
MD5 5d667f4974f85a1b41dcce898827f72c
BLAKE2b-256 4dfcb80b3d69ebc1d063cc0c1cbd06b475af5b95ed230088a3fdb6130c463f7d

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp310-cp310-win_amd64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp310-cp310-manylinux_2_28_x86_64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp310-cp310-manylinux_2_28_x86_64.whl
Algorithm Hash digest
SHA256 e75c77351d299d98386fe47f0a675dddafc9b5a2062b9ffe6fbcecc85503d81b
MD5 106ab0e56a6ca081821f34fe0d5b93d6
BLAKE2b-256 ec3125f9648dca4fc87ddfc4e0a0066334276ad2549399dd6456c77a2c6d08bf

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp310-cp310-manylinux_2_28_x86_64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp310-cp310-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp310-cp310-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 f5290f41b53923e2591aec5be871df5c351e93de4106967c53eb1257649a02ab
MD5 039ad656c3a33af2447e9fa352c9a34b
BLAKE2b-256 762c1c9e016a9667b560686b40c762cb548abe87d667419a0d85300480973fd4

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp310-cp310-macosx_11_0_arm64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp39-cp39-win_amd64.whl.

File metadata

  • Download URL: shimaenaga-1.3.0-cp39-cp39-win_amd64.whl
  • Upload date:
  • Size: 288.0 kB
  • Tags: CPython 3.9, Windows x86-64
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for shimaenaga-1.3.0-cp39-cp39-win_amd64.whl
Algorithm Hash digest
SHA256 9d8593847d191d9b8f26ae7207f02033afa2a631e16256d5826f84c72878a4d0
MD5 a11bac471babcdd310a22ca91f577752
BLAKE2b-256 f33a639fe3b4e1d4d880112fbedd6d9758d78630c3876172f0a39bd9e12b3eb3

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp39-cp39-win_amd64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp39-cp39-manylinux_2_28_x86_64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp39-cp39-manylinux_2_28_x86_64.whl
Algorithm Hash digest
SHA256 81b3b85dd1e0c3a54ed8f8dea11413ba7959c6e17c4557ad0bc5cc54efb031f0
MD5 e84caa59c3a4f4bce13da320282e3de5
BLAKE2b-256 62552d53ac9dca21d20204c64ef51b13726a3f945f8bada54eea86f218f6b6ae

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp39-cp39-manylinux_2_28_x86_64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file shimaenaga-1.3.0-cp39-cp39-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for shimaenaga-1.3.0-cp39-cp39-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 0b25035732df43959c5123fb43fffc7cd02218dbf2c1609dddc67c62185451f0
MD5 a9ed8fc19881805556e0f4bffaf02fb1
BLAKE2b-256 9b662cca62a9aa54d6f0a1f747bf99867d7497d6135e3243425ee952b69e9ca8

See more details on using hashes here.

Provenance

The following attestation bundles were made for shimaenaga-1.3.0-cp39-cp39-macosx_11_0_arm64.whl:

Publisher: pypi-publish.yml on shota-iwamoto/shimaenaga

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page