Shinrin

shinrin

Shinrin (森林, "forest" in Japanese) is a scikit-learn-compatible library for decision tree and forest models and tabular neural networks, with Rust and Mojo bindings for performance and ONNX export support.

Since skope-rules and scikit-garden are no longer actively maintained, this project aims to bring them together with extensions for tree models — including SHAP explanations, ONNX export, and benchmarking utilities.

Shinrin also includes TabM — a parameter-efficient ensemble MLP for tabular data (ICLR 2025) that trains an ensemble of k members jointly through BatchEnsemble-style multiplicative adapters, matching ensemble accuracy at a fraction of the training cost. Training runs entirely on NumPy or Mojo kernels — PyTorch is not required.

Features

Mondrian Trees & Forests — Full scikit-learn API compatibility
CORELS Optimal Rule Lists — Certifiably optimal rule lists for binary data (CorelsClassifier), vendored from pycorels with bundled mini-GMP (no system dependency)
GOSDT Optimal Sparse Decision Trees — Globally optimized sparse decision trees with reference-ensemble guesses (GOSDTClassifier, ThresholdGuessBinarizer), vendored from gosdt-guesses; runs single-threaded with no TBB/GMP system dependencies
Tabular Neural Networks — scikit-learn compatible MLPClassifier/MLPRegressor and TabMClassifier/TabMRegressor with optional PLE embeddings and Mojo-accelerated training
TabM Neural Networks — Parameter-efficient ensemble MLPs for tabular data with BatchEnsemble-style multiplicative adapters (ICLR 2025)
TabICL — Tabular in-context learning foundation model (torch/NumPy/Mojo backends)
TreeSHAP Explanations — TreeExplainer for single trees and forests with explanation() visualization helper
ONNX Export — Export trained trees, forests, and TabM models to ONNX format for deployment
Benchmarking — Built-in utilities for training speed, prediction speed, and model size
Rust & Mojo Bindings — Performance-critical code in Rust via PyO3 and Mojo kernels

Quick Start

Tree Models

from shinrin import MondrianTreeRegressor, MondrianForestClassifier
from shinrin import TreeExplainer, explanation

# Train a model
X, y = ...
tree = MondrianTreeRegressor(max_depth=8, random_state=0)
tree.fit(X, y)
predictions = tree.predict(X)

# Get SHAP explanations
explainer = TreeExplainer(tree)
shap_values = explainer.shap_values(X)
# Or use the convenience function:
# explanation(tree, X)  # opens matplotlib visualization

TabM Neural Networks

from shinrin import TabMClassifier, TabMRegressor

# Train a TabM model
model = TabMRegressor(
    hidden_layer_sizes=(256,),
    k=32,               # ensemble size
    random_state=0,
)
model.fit(X, y)
predictions = model.predict(X)

# Classification
clf = TabMClassifier(k=32, max_iter=200)
clf.fit(X, y)

CORELS Optimal Rule Lists

from shinrin import CorelsClassifier

# Binary features, binary classification — provably optimal rule list
clf = CorelsClassifier(c=0.01, verbosity=["rulelist"])
clf.fit(X, y, features=["feature1", "feature2"])
print(clf.rl())          # human-readable optimal rule list
predictions = clf.predict(X)

GOSDT Optimal Sparse Decision Trees

from shinrin import GOSDTClassifier, ThresholdGuessBinarizer

# Binarize continuous features via gradient-boosting threshold guesses
X_bin = ThresholdGuessBinarizer().fit_transform(X, y)

# Optionally guide the search with a blackbox reference ensemble
clf = GOSDTClassifier(regularization=0.05, depth_budget=4)
clf.fit(X_bin, y)                      # or: clf.fit(X_bin, y, y_ref=y_ref)
print(str(clf.trees_[0]))              # globally optimal tree
accuracy = clf.score(X_bin, y)

Native Backends

Both Mondrian trees and TabM support interchangeable native backends:

Rust (default) — PyO3/maturin extension for tree models
Mojo — Experimental Mojo port for TabM training kernels

Select the backend with environment variables:

SHINRIN_BACKEND=mojo python your_script.py          # TabM Mojo backend
SHINRIN_TABM_BACKEND=mojo python your_script.py     # TabM-specific backend

Benchmarks

See scripts/benchmarks/BENCHMARK.md for detailed benchmark results comparing Shinrin against LightGBM and scikit-learn SGD.

See scripts/benchmarks/TABM_BENCHMARK.md for TabM backend comparisons (NumPy vs Mojo vs PyTorch).

See scripts/benchmarks/GOSDT_BENCHMARK.md for GOSDT vs scikit-learn CART comparisons (just bench-gosdt).

To run benchmarks yourself: python scripts/benchmarks/bench_baselines.py (or just bench-backends for Rust vs Mojo backend comparisons, just bench-tabm for TabM backends).

Installation

pip install shinrin

Optional dependencies:

pip install shinrin[sklearn]   # scikit-learn for benchmarks and SHAP plotting
pip install shinrin[onnx]      # ONNX export
pip install shinrin[mojo]      # TabM Mojo kernels (`just build-tabm-mojo`)
pip install shinrin[full]      # All optional dependencies

Native backends

The tree/forest internals ship with two interchangeable native backends:

rust (default) – the original pyo3/maturin extension (shinrin._native)
mojo – an experimental Mojo port (shinrin._native_mojo)

Select the backend with the SHINRIN_BACKEND environment variable:

SHINRIN_BACKEND=mojo python your_script.py

The default remains rust; the Mojo backend is opt-in while Mojo is in alpha. Both backends produce identical trees for identical random states (verified by tests/test_mojo_parity.py). Build the Mojo extension with just build-mojo (requires the mojo package, e.g. pip install 'shinrin[mojo]').

API Overview

Models

Tree Models

Model	Description
`MondrianTreeRegressor`	Single Mondrian tree for regression
`MondrianTreeClassifier`	Single Mondrian tree for classification
`MondrianForestRegressor`	Ensemble of Mondrian trees for regression
`MondrianForestClassifier`	Ensemble of Mondrian trees for classification
`TabMClassifier` / `TabMRegressor`	Ensemble MLP trainers (NumPy / Mojo backends)
`TabICLClassifier` / `TabICLRegressor`	Tabular in-context learning estimators (TabICLv2)

TabM Neural Networks

Model	Description
`TabMRegressor`	TabM ensemble regressor (drop-in for `MLPRegressor`)
`TabMClassifier`	TabM ensemble classifier (drop-in for `MLPClassifier`)

TabM parameters:

Parameter	Default	Description
`hidden_layer_sizes`	`(256,)`	Backbone block widths
`k`	`32`	Number of ensemble members
`solver`	`'adam'`	`'adam'`, `'sgd'`, or `'lbfgs'`
`arch_type`	`'tabm'`	`'tabm'`, `'tabm-mini'`, or `'tabm-packed'`
`dropout`	`0.1`	Backbone dropout rate
`use_embeddings`	`True`	Piecewise-linear + linear embeddings for numeric features
`n_bins`	`64`	Quantile bins per numeric feature
`d_embedding`	`8`	Embedding width per numeric feature
`categorical_indices`	`None`	Columns to treat as categorical
`categorical_cardinality_threshold`	`32`	Max unique values for auto-detecting categoricals

Explanations (Tree Models)

from shinrin import TreeExplainer, explanation

explainer = TreeExplainer(model)
shap_values = explainer.shap_values(X)
expected_value = explainer.expected_value

# Quick visualization (requires matplotlib)
explanation(model, X)

ONNX Export

from shinrin.onnx import to_onnx, save_onnx

# Export to ONNX protobuf
onnx_model = to_onnx(model, X)

# Save to file
save_onnx(model, "model.onnx", X)

Works for tree/forest models and TabM. TabM exports a self-contained graph (preprocessing, embeddings, ensemble backbone, averaged head) that runs on raw feature vectors with any batch size:

import onnxruntime as ort
from shinrin import TabMRegressor
from shinrin.onnx import save_onnx

model = TabMRegressor(hidden_layer_sizes=(256,), k=32, random_state=0)
model.fit(X_train, y_train)
save_onnx(model, "tabm.onnx", X_train)

session = ort.InferenceSession("tabm.onnx")
predictions = session.run(None, {"X": X_test.astype(np.float32)})[0]

Benchmarking

from shinrin.benchmark import (
    benchmark_training,
    benchmark_prediction,
    benchmark_model_size,
    full_benchmark,
    print_benchmark_report,
)

models = {
    "shinrin_tree": MondrianTreeRegressor(max_depth=8),
    "shinrin_forest": MondrianForestRegressor(n_estimators=10, max_depth=8),
    "shinrin_tabm": TabMRegressor(hidden_layer_sizes=(256,), k=32),
}

results = full_benchmark(models, X_train, y_train, X_test)
print_benchmark_report(results)

Test Coverage

All vendored tests are included and passing — these are ported from scikit-garden and skope-rules to verify compatibility. Run pytest --cov=src/shinrin tests/ for a full coverage report.

TabM parity tests (tests/test_tabm_parity.py) verify that the Mojo kernels produce identical results to the NumPy reference implementation. TabM functional tests (tests/test_tabm.py) cover training, prediction, and determinism. TabM ONNX export tests (tests/test_tabm_onnx.py) verify onnxruntime inference parity for all architectures, tasks, and preprocessing configurations.

License

MIT

Download files

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

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distributions

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

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

shinrin-0.2.0-pp311-pypy311_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 PyPymanylinux: glibc 2.17+ x86-64

shinrin-0.2.0-pp311-pypy311_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 PyPymanylinux: glibc 2.17+ ARM64

shinrin-0.2.0-cp314-cp314t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.14tmanylinux: glibc 2.17+ x86-64

shinrin-0.2.0-cp314-cp314t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.14tmanylinux: glibc 2.17+ ARM64

shinrin-0.2.0-cp314-cp314-win_amd64.whl (960.1 kB view details)

Uploaded Aug 23, 2026 CPython 3.14Windows x86-64

shinrin-0.2.0-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.14manylinux: glibc 2.17+ x86-64

shinrin-0.2.0-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.14manylinux: glibc 2.17+ ARM64

shinrin-0.2.0-cp314-cp314-macosx_11_0_arm64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.14macOS 11.0+ ARM64

shinrin-0.2.0-cp314-cp314-macosx_10_12_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.14macOS 10.12+ x86-64

shinrin-0.2.0-cp313-cp313-win_amd64.whl (960.1 kB view details)

Uploaded Aug 23, 2026 CPython 3.13Windows x86-64

shinrin-0.2.0-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.13manylinux: glibc 2.17+ x86-64

shinrin-0.2.0-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.13manylinux: glibc 2.17+ ARM64

shinrin-0.2.0-cp313-cp313-macosx_11_0_arm64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.13macOS 11.0+ ARM64

shinrin-0.2.0-cp313-cp313-macosx_10_12_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.13macOS 10.12+ x86-64

shinrin-0.2.0-cp312-cp312-win_amd64.whl (960.6 kB view details)

Uploaded Aug 23, 2026 CPython 3.12Windows x86-64

shinrin-0.2.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.12manylinux: glibc 2.17+ x86-64

shinrin-0.2.0-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.12manylinux: glibc 2.17+ ARM64

shinrin-0.2.0-cp312-cp312-macosx_11_0_arm64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.12macOS 11.0+ ARM64

shinrin-0.2.0-cp312-cp312-macosx_10_12_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.12macOS 10.12+ x86-64

shinrin-0.2.0-cp311-cp311-win_amd64.whl (959.7 kB view details)

Uploaded Aug 23, 2026 CPython 3.11Windows x86-64

shinrin-0.2.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.11manylinux: glibc 2.17+ x86-64

shinrin-0.2.0-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.11manylinux: glibc 2.17+ ARM64

shinrin-0.2.0-cp311-cp311-macosx_11_0_arm64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.11macOS 11.0+ ARM64

shinrin-0.2.0-cp311-cp311-macosx_10_12_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.11macOS 10.12+ x86-64

shinrin-0.2.0-cp310-cp310-win_amd64.whl (959.6 kB view details)

Uploaded Aug 23, 2026 CPython 3.10Windows x86-64

shinrin-0.2.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.10manylinux: glibc 2.17+ x86-64

shinrin-0.2.0-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (1.1 MB view details)

Uploaded Aug 23, 2026 CPython 3.10manylinux: glibc 2.17+ ARM64

File details

Details for the file shinrin-0.2.0-pp311-pypy311_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.

File metadata

Download URL: shinrin-0.2.0-pp311-pypy311_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
Upload date: Aug 23, 2026
Size: 1.1 MB
Tags: PyPy, manylinux: glibc 2.17+ x86-64
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for shinrin-0.2.0-pp311-pypy311_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
Algorithm	Hash digest
SHA256	`4311451f50099739056beddfc88bafcdcb3f13eba63f159d37c776258717ab71`
MD5	`187a647b6afc5ca88321cadf3ed670d6`
BLAKE2b-256	`1076a00661600a2286a71f2066b3cf41ccff213ddfbfe163438b654934cef576`

Algorithm	Hash digest
SHA256	`2f55010fc032286e84cbaa38b95f075103080b06f7e952ee2e311c2b73786e09`
MD5	`3e3867b79e305227e3518ffd67248bc4`
BLAKE2b-256	`1275c2ab7bf7cdec63152110f29b8c75f12c36c3fe93887d9267fcb854eae65f`

Algorithm	Hash digest
SHA256	`6974c75d44327bb41d60f37e3b39b474d62e6e8e04bcc45b7131abd3fb2fa350`
MD5	`43f4168952344210ff6a1d9770b70aa4`
BLAKE2b-256	`8dc6ad3124d7145606b0f5631727e346a932e95f676fa25e98d5b8100740312a`

Algorithm	Hash digest
SHA256	`73dc8f60789abae77221bed5d6926f2d42f72a51cb3868466fa438d398b6482a`
MD5	`6c644689bbc62f0f9fdc9f15dcf94da6`
BLAKE2b-256	`068bbc7524eb68ab2ba9139e21b91f90155e9293b4451c5c076590f23fa89d58`

Algorithm	Hash digest
SHA256	`f9d5a4ffefa098420c25746dab06effbae94a4df957f881dc4a48a203cd49481`
MD5	`878c5b2fb5b466911fd67e136a370f0d`
BLAKE2b-256	`a8af617cdddb274bbb7a90e3c7c7894653f4d6a021da6c47f29cd01c0b0e6250`

Algorithm	Hash digest
SHA256	`b23925cb92bfa1aca1f200f5b59af700ad3ed572e39eb06f88c2f5fbf95f0398`
MD5	`841dd7d4f2fbc9d4827c50056358e118`
BLAKE2b-256	`3d1460e0426b98067d08cae9cd4f62fc663cad04458f7f1f4c48f0f26b69aa5f`

Algorithm	Hash digest
SHA256	`f2bd3284918c92973069ac9ab5550d5274cd97e2ba92b7066c42c8966506b5f9`
MD5	`74c482bf5ac17fe86f0c0165196f4783`
BLAKE2b-256	`f5b60b28e980c9e2b6d42fe739b4f3aa4d807b7781c2ddd4771d110ad4ff4145`

Algorithm	Hash digest
SHA256	`c17178fa2d6308c46c92d250b6d914367b0d9d0e66207691323e299568ea3f2a`
MD5	`5341bcdf859965eaef18cbf37c0044c0`
BLAKE2b-256	`331ac07a52452d90f57a2c15dba287443948a70455421dd09490a4ef887e5401`

Algorithm	Hash digest
SHA256	`8df078c5f6167ac3f50f1cb657080dd4e5946c6f25410912d0b72999e87f25f7`
MD5	`9a0a7a9b5487bc39d640f38911a1fad7`
BLAKE2b-256	`656b2f29315c16e8a9b08e762464658277f041eafc65158db37e8cd9cba3105d`

Algorithm	Hash digest
SHA256	`5ec0b3cbef8ca5e7bb06d68b71dde9d74348bd06063ea916590ba47e96150b30`
MD5	`e269bcf2409ef1a54f81d06f51f5b6b4`
BLAKE2b-256	`9d0a22277a95a811372ce7db1326af12acc0705ceb90f1f856ae66830ca6cd7f`

Algorithm	Hash digest
SHA256	`030d1474a1c56ced8e4870a94b8197e0e16c8c68b6d1e41bdfbfec9d155a972a`
MD5	`88e103d9cfe03d68076bd18a3c634412`
BLAKE2b-256	`fb4e2b12616b86560832cd74b4e2a794825510d41ed66c044810a4b7bd506f47`

Algorithm	Hash digest
SHA256	`f37f4588e4982fed4823118e37a4d5aede4b9f53a6603e110137319326fb8e3a`
MD5	`9a94b77f855c83b441398c1b74fecfe9`
BLAKE2b-256	`494a7c1002e9cbe7e118684dd5935066dd08c9cee6cdb81f909c3bb524a0049d`

Algorithm	Hash digest
SHA256	`ab536c60cf177d31dcce407bfc67cf4e988afe6e125fd1ef5e4e17c9c236a704`
MD5	`eca6dcdb53a9b92523bc1b37cf832f31`
BLAKE2b-256	`eb33d2d69fc92ef97a66ad2c3dc9b63442e6dc46a6c101d9f7206926beeb750c`

Algorithm	Hash digest
SHA256	`ce47232954ee204202b89bc297366ce0d8f7effab92a265f976acf886de7a0c7`
MD5	`685f554cc2368ded87e1f7c5dfe3c9fc`
BLAKE2b-256	`f24853ca10cd571faa89cf7416f0e9a799c0142cb9880647a89bdefc91e61a3b`

Algorithm	Hash digest
SHA256	`608634419e4a978dcbc7630d2d7458348bc6fa6623745053689efb7b0620e72e`
MD5	`7c75a0f08817b424c9e1559b8928fee2`
BLAKE2b-256	`6c9311851c60dc97be6b9a988a5aaabbb689c0d7a1817459c28fa25bfa1c665c`

Algorithm	Hash digest
SHA256	`fdd1c89a0ee5cffde8c9ff8a491d5bfda9adf3ba3762677ad1268947d40bf1ca`
MD5	`90fd565bec2c4cb1f20a09f5f0d86189`
BLAKE2b-256	`272a1ae9bb0329eee91a1207434bceb9c4df38f089b2cc0011711edfa58d4cef`

Algorithm	Hash digest
SHA256	`d21fa3107ff5e89aac402204faeaf6ed63e86183a272cf62fb75fa9285393f2e`
MD5	`86c4cd9b035b55c379dd487d6a63281d`
BLAKE2b-256	`5b26261296aa074c8971404ec06d3b434edc35635829ffff626bee937c0d730f`

Algorithm	Hash digest
SHA256	`40ea91f1da3f398166f559508f694e5088392d14072c512534c9ee194a46982f`
MD5	`43a8c54aa00f0397b53cad196af04afc`
BLAKE2b-256	`83fe67ca5c631fd057886acdd62ecd100b9d2c8b2c8047d9a0b66a4d42b439a5`

Algorithm	Hash digest
SHA256	`ce0409cfd4fb6a9a2c465f61b323367bee6384f944617488eee0a50040d86ad3`
MD5	`efffb02da0c46f0623e916fe0cbd7336`
BLAKE2b-256	`3d70fa4368e441b529490d14b5667a538810fdc5e7216127a709deaf77e0bbd8`

Algorithm	Hash digest
SHA256	`bf7536d623acbabb616b94215e482c5a34a1f74f771d46804a4f09049ed6fa92`
MD5	`8558b005f9b837b5df3f36f805bc9bb3`
BLAKE2b-256	`4376126044af97c8bebd0ce79555c2316e7c615bf76cadf55a9d8d71e0bf999b`

Algorithm	Hash digest
SHA256	`1c895f93e663c5efdec8d52528a28a49273fa536c762c53e6c91d59dfe973e81`
MD5	`5373c306fc022d6e66060d34b3f28d6a`
BLAKE2b-256	`2b5a10a0cdc8f7b73c018c6b1fe9e53c4e98da7b14b5860b607a1824c1b13063`

shinrin 0.2.0

Shinrin

Features

Quick Start

Tree Models

TabM Neural Networks

CORELS Optimal Rule Lists

GOSDT Optimal Sparse Decision Trees

Native Backends

Benchmarks

Installation

Native backends

API Overview

Models

Tree Models

TabM Neural Networks

Explanations (Tree Models)

ONNX Export

Benchmarking

Test Coverage

License

Key dates

1 maintainer

Credits

Requires

Provides Extra

Download files

Source Distributions

Built Distributions

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Algorithm	Hash digest
SHA256	`dfeef00f9bd2ab68dd2f6d9711bd193beee6dfedf139526df40116899d052068`
MD5	`ed8142ff373e8c3510e0df640e51ea1d`
BLAKE2b-256	`431d796cd00fbdb258bb8bc304f6da95d03d7558b38aa035e788ae8c7ba5086f`

Algorithm	Hash digest
SHA256	`c2e06a15cca3ad4e493b7b41f27c32bac2d439c6cfe89b6fabe06099e1c18dc7`
MD5	`6e23956a59332185d0cfd2216232ff47`
BLAKE2b-256	`f63b27baf78cc0d8ce3f162bc3433f426f5506c352ac468b15177b675d4ee61b`

Algorithm	Hash digest
SHA256	`d05c175bcd8db26389e891b2ad62b104b77f172af87db71ae17349201a03aee4`
MD5	`a9911bcd9e3770c474c15796de66288e`
BLAKE2b-256	`590a8c54d9c57f03afa602422af608595954af9c836ac8eea1242d07089ef781`

Algorithm	Hash digest
SHA256	`a6be5ae33daf6b6914881860a4ae5808a903e7b6d29462f63c5b2c235ae0970c`
MD5	`b78dc3227954d58df91c992deca84329`
BLAKE2b-256	`4603563a152aab2a019988a47b6768b7955698183e6b375991f966cccbc436b7`

Algorithm	Hash digest
SHA256	`9fe23b9b3087a296e0b4728f388ba9792b81ce25ea058e68864004ced9da4506`
MD5	`73746c821ae4ca66afd3de640c4f7c7e`
BLAKE2b-256	`a200e46b279d6b53bb6a335bb64fe34bdd59b425a8877589afdc68a663d27ee5`

Algorithm	Hash digest
SHA256	`71ccfe8af7f6bbedf9d1736a0f1858da291d8338fadb9660c8643d03e48aa104`
MD5	`af9a169fa1fed3f43d248f1abb1ab124`
BLAKE2b-256	`c916274f4d83917dd096ac56c9f9dfcdeded3e1c991885de3f14c328dbb2b263`