Upload folder using huggingface_hub

.gitattributes

@@ -3,3 +3,4 @@
 *.pkl filter=lfs diff=lfs merge=lfs -text
 *.pt filter=lfs diff=lfs merge=lfs -text
 *.bin filter=lfs diff=lfs merge=lfs -text
+models/model.onnx filter=lfs diff=lfs merge=lfs -text

models/model.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7c7423d4f48ddb07090b4db04b566ecf778384e0f9d5278c38f528a42a04621
+size 16320803

pyproject.toml

@@ -32,6 +32,8 @@ dev = [
     "evidently>=0.7",
     "httpx>=0.28.1",
     "line-profiler>=5.0.2",
+    "onnxmltools>=1.16.0",
+    "onnxruntime>=1.26.0",
     "pytest>=9.0.3",
     "pytest-cov>=7.1.0",
     "ruff>=0.15.12",

scripts/benchmark_onnx.py

@@ -0,0 +1,203 @@
+"""Benchmark LightGBM joblib vs ONNX Runtime inference.
+
+Loads N real feature rows from `data/reference_dataset.parquet`, runs each
+sample one at a time through both backends (simulating the per-request path),
+and reports p50/p95/p99 latency plus numerical equivalence.
+
+Run:
+    uv run python scripts/benchmark_onnx.py --n 1000
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import time
+from pathlib import Path
+from statistics import median
+
+import joblib
+import numpy as np
+import onnxruntime as ort
+import pandas as pd
+
+ROOT = Path(__file__).resolve().parents[1]
+DEFAULT_MODEL_PATH = ROOT / "models" / "model.joblib"
+DEFAULT_ONNX_PATH = ROOT / "models" / "model.onnx"
+DEFAULT_FEATURE_NAMES_PATH = ROOT / "models" / "feature_names.json"
+DEFAULT_REFERENCE_PATH = ROOT / "data" / "reference_dataset.parquet"
+
+
+def percentile(samples: list[float], pct: float) -> float:
+    """Linear-interpolated percentile (pct in 0..100)."""
+    return float(np.percentile(samples, pct))
+
+
+def fmt_ms(seconds: float) -> str:
+    return f"{seconds * 1000:.3f} ms"
+
+
+def bench_lightgbm(model, samples_df: pd.DataFrame) -> tuple[list[float], np.ndarray]:
+    """Run predict_proba one row at a time. Returns (latencies_s, probas)."""
+    latencies: list[float] = []
+    probas = np.empty(len(samples_df), dtype=np.float64)
+    for i in range(len(samples_df)):
+        row = samples_df.iloc[[i]]
+        t0 = time.perf_counter()
+        p = model.predict_proba(row)[0, 1]
+        latencies.append(time.perf_counter() - t0)
+        probas[i] = p
+    return latencies, probas
+
+
+def bench_onnx(
+    session: ort.InferenceSession,
+    input_name: str,
+    output_name: str,
+    samples_array: np.ndarray,
+) -> tuple[list[float], np.ndarray]:
+    """Run ONNX inference one row at a time. Returns (latencies_s, probas)."""
+    latencies: list[float] = []
+    probas = np.empty(len(samples_array), dtype=np.float64)
+    for i in range(len(samples_array)):
+        # Reshape to (1, n_features); float32 is required by the ONNX graph.
+        row = samples_array[i : i + 1]
+        t0 = time.perf_counter()
+        out = session.run([output_name], {input_name: row})[0]
+        latencies.append(time.perf_counter() - t0)
+        # zipmap=False → out shape is (1, 2): [prob_class_0, prob_class_1].
+        probas[i] = out[0, 1]
+    return latencies, probas
+
+
+def summarise(name: str, latencies: list[float]) -> dict[str, float]:
+    return {
+        "backend": name,
+        "n": len(latencies),
+        "p50_ms": percentile(latencies, 50) * 1000,
+        "p95_ms": percentile(latencies, 95) * 1000,
+        "p99_ms": percentile(latencies, 99) * 1000,
+        "mean_ms": float(np.mean(latencies)) * 1000,
+        "total_s": sum(latencies),
+    }
+
+
+def print_table(rows: list[dict[str, float]]) -> None:
+    print(
+        f"{'Backend':<14}{'n':>6}{'p50':>12}{'p95':>12}{'p99':>12}{'mean':>12}"
+    )
+    for r in rows:
+        print(
+            f"{r['backend']:<14}{r['n']:>6}"
+            f"{r['p50_ms']:>10.3f}ms"
+            f"{r['p95_ms']:>10.3f}ms"
+            f"{r['p99_ms']:>10.3f}ms"
+            f"{r['mean_ms']:>10.3f}ms"
+        )
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--n", type=int, default=1000, help="Number of inferences")
+    parser.add_argument("--warmup", type=int, default=20)
+    parser.add_argument("--model", type=Path, default=DEFAULT_MODEL_PATH)
+    parser.add_argument("--onnx", type=Path, default=DEFAULT_ONNX_PATH)
+    parser.add_argument(
+        "--feature-names", type=Path, default=DEFAULT_FEATURE_NAMES_PATH
+    )
+    parser.add_argument("--reference", type=Path, default=DEFAULT_REFERENCE_PATH)
+    parser.add_argument("--seed", type=int, default=42)
+    parser.add_argument(
+        "--out",
+        type=Path,
+        default=None,
+        help="Optional path to write a JSON report (e.g. profiling/benchmark_onnx.json).",
+    )
+    args = parser.parse_args()
+
+    feature_names = json.loads(args.feature_names.read_text())
+
+    # Sample N rows from the reference dataset, aligned to model feature order.
+    reference = pd.read_parquet(args.reference)
+    rng = np.random.default_rng(args.seed)
+    idx = rng.choice(len(reference), size=args.n + args.warmup, replace=True)
+    samples_df = reference.iloc[idx][feature_names].reset_index(drop=True)
+    samples_array = samples_df.to_numpy(dtype=np.float32)
+
+    # --- Load both backends ---
+    print("Loading LightGBM model...")
+    model = joblib.load(args.model)
+    raw_model = model.get_raw_model() if hasattr(model, "get_raw_model") else model
+
+    print("Loading ONNX session...")
+    session = ort.InferenceSession(
+        str(args.onnx), providers=["CPUExecutionProvider"]
+    )
+    input_name = session.get_inputs()[0].name
+    output_name = session.get_outputs()[1].name  # probabilities output
+
+    # --- Warmup (excluded from stats) ---
+    print(f"Warmup x{args.warmup}...")
+    bench_lightgbm(raw_model, samples_df.iloc[: args.warmup])
+    bench_onnx(session, input_name, output_name, samples_array[: args.warmup])
+
+    # --- Measured runs ---
+    print(f"Benchmarking LightGBM (n={args.n})...")
+    lgbm_lat, lgbm_proba = bench_lightgbm(raw_model, samples_df.iloc[args.warmup :])
+
+    print(f"Benchmarking ONNX     (n={args.n})...")
+    onnx_lat, onnx_proba = bench_onnx(
+        session, input_name, output_name, samples_array[args.warmup :]
+    )
+
+    # --- Numerical equivalence ---
+    diff = np.abs(lgbm_proba - onnx_proba)
+    print()
+    print("=== Numerical equivalence ===")
+    print(f"max  |delta_proba| = {diff.max():.2e}")
+    print(f"mean |delta_proba| = {diff.mean():.2e}")
+    print(f"# rows with |delta| > 1e-5: {(diff > 1e-5).sum()} / {len(diff)}")
+
+    # --- Latency summary ---
+    print()
+    print("=== Latency (single-row predict_proba) ===")
+    rows = [
+        summarise("LightGBM", lgbm_lat),
+        summarise("ONNX-Runtime", onnx_lat),
+    ]
+    print_table(rows)
+
+    speedup = median(lgbm_lat) / median(onnx_lat) if median(onnx_lat) > 0 else float("inf")
+    gain_pct = (1 - median(onnx_lat) / median(lgbm_lat)) * 100
+    print()
+    print(f"Speed-up (p50): x{speedup:.2f}")
+    print(f"Gain (p50): {gain_pct:.1f}%")
+
+    if args.out is not None:
+        report = {
+            "config": {
+                "n": args.n,
+                "warmup": args.warmup,
+                "seed": args.seed,
+                "model_joblib": str(args.model),
+                "model_onnx": str(args.onnx),
+                "reference": str(args.reference),
+            },
+            "equivalence": {
+                "max_abs_delta": float(diff.max()),
+                "mean_abs_delta": float(diff.mean()),
+                "rows_above_1e-5": int((diff > 1e-5).sum()),
+                "n_compared": int(len(diff)),
+            },
+            "latency": rows,
+            "speedup_p50": speedup,
+            "gain_p50_pct": gain_pct,
+        }
+        args.out.parent.mkdir(parents=True, exist_ok=True)
+        args.out.write_text(json.dumps(report, indent=2))
+        print()
+        print(f"Wrote report → {args.out}")
+
+
+if __name__ == "__main__":
+    main()

scripts/export_to_onnx.py

@@ -0,0 +1,75 @@
+"""Convert the LightGBM credit-scoring model to ONNX format.
+
+Reads `models/model.joblib`, unwraps the MLflow PyFunc layer if present,
+converts the underlying LightGBM model to ONNX via `onnxmltools`, and writes
+`models/model.onnx`.
+
+Run:
+    uv run python scripts/export_to_onnx.py
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+
+import joblib
+from onnxmltools import convert_lightgbm
+from onnxmltools.convert.common.data_types import FloatTensorType
+
+ROOT = Path(__file__).resolve().parents[1]
+DEFAULT_MODEL_PATH = ROOT / "models" / "model.joblib"
+DEFAULT_FEATURE_NAMES_PATH = ROOT / "models" / "feature_names.json"
+DEFAULT_OUT_PATH = ROOT / "models" / "model.onnx"
+
+
+def export(
+    model_path: Path,
+    feature_names_path: Path,
+    out_path: Path,
+    target_opset: int = 13,
+) -> Path:
+    """Convert LightGBM joblib → ONNX file. Returns the output path."""
+    model = joblib.load(model_path)
+    raw_model = model.get_raw_model() if hasattr(model, "get_raw_model") else model
+
+    feature_names = json.loads(feature_names_path.read_text())
+    n_features = len(feature_names)
+
+    # ONNX requires a fixed-shape input declaration. None = dynamic batch dim.
+    initial_types = [("input", FloatTensorType([None, n_features]))]
+
+    onnx_model = convert_lightgbm(
+        raw_model,
+        initial_types=initial_types,
+        target_opset=target_opset,
+        zipmap=False,  # Output raw probability array instead of list-of-dicts.
+    )
+
+    out_path.write_bytes(onnx_model.SerializeToString())
+    return out_path
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--model", type=Path, default=DEFAULT_MODEL_PATH)
+    parser.add_argument(
+        "--feature-names", type=Path, default=DEFAULT_FEATURE_NAMES_PATH
+    )
+    parser.add_argument("--out", type=Path, default=DEFAULT_OUT_PATH)
+    parser.add_argument("--opset", type=int, default=13)
+    args = parser.parse_args()
+
+    out = export(
+        model_path=args.model,
+        feature_names_path=args.feature_names,
+        out_path=args.out,
+        target_opset=args.opset,
+    )
+    size_mb = out.stat().st_size / (1024 * 1024)
+    print(f"Wrote {out} ({size_mb:.2f} MB)")
+
+
+if __name__ == "__main__":
+    main()

uv.lock

@@ -1,6 +1,10 @@
 version = 1
 revision = 3
 requires-python = "==3.12.*"
+resolution-markers = [
+    "platform_machine != 's390x'",
+    "platform_machine == 's390x'",
+]
 
 [[package]]
 name = "aiohappyeyeballs"
@@ -680,6 +684,14 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/4f/af/72ad54402e599152de6d067324c46fe6a4f531c7c65baf7e96c63db55eaf/flask_cors-6.0.2-py3-none-any.whl", hash = "sha256:e57544d415dfd7da89a9564e1e3a9e515042df76e12130641ca6f3f2f03b699a", size = 13257, upload-time = "2025-12-12T20:31:41.3Z" },
 ]
 
+[[package]]
+name = "flatbuffers"
+version = "25.12.19"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e8/2d/d2a548598be01649e2d46231d151a6c56d10b964d94043a335ae56ea2d92/flatbuffers-25.12.19-py2.py3-none-any.whl", hash = "sha256:7634f50c427838bb021c2d66a3d1168e9d199b0607e6329399f04846d42e20b4", size = 26661, upload-time = "2025-12-19T23:16:13.622Z" },
+]
+
 [[package]]
 name = "fonttools"
 version = "4.62.1"
@@ -1585,6 +1597,22 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/9b/f7/4a5e785ec9fbd65146a27b6b70b6cdc161a66f2024e4b04ac06a67f5578b/mistune-3.2.0-py3-none-any.whl", hash = "sha256:febdc629a3c78616b94393c6580551e0e34cc289987ec6c35ed3f4be42d0eee1", size = 53598, upload-time = "2025-12-23T11:36:33.211Z" },
 ]
 
+[[package]]
+name = "ml-dtypes"
+version = "0.5.4"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "numpy" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/0e/4a/c27b42ed9b1c7d13d9ba8b6905dece787d6259152f2309338aed29b2447b/ml_dtypes-0.5.4.tar.gz", hash = "sha256:8ab06a50fb9bf9666dd0fe5dfb4676fa2b0ac0f31ecff72a6c3af8e22c063453", size = 692314, upload-time = "2025-11-17T22:32:31.031Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a8/b8/3c70881695e056f8a32f8b941126cf78775d9a4d7feba8abcb52cb7b04f2/ml_dtypes-0.5.4-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:a174837a64f5b16cab6f368171a1a03a27936b31699d167684073ff1c4237dac", size = 676927, upload-time = "2025-11-17T22:31:48.182Z" },
+    { url = "https://files.pythonhosted.org/packages/54/0f/428ef6881782e5ebb7eca459689448c0394fa0a80bea3aa9262cba5445ea/ml_dtypes-0.5.4-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:a7f7c643e8b1320fd958bf098aa7ecf70623a42ec5154e3be3be673f4c34d900", size = 5028464, upload-time = "2025-11-17T22:31:50.135Z" },
+    { url = "https://files.pythonhosted.org/packages/3a/cb/28ce52eb94390dda42599c98ea0204d74799e4d8047a0eb559b6fd648056/ml_dtypes-0.5.4-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9ad459e99793fa6e13bd5b7e6792c8f9190b4e5a1b45c63aba14a4d0a7f1d5ff", size = 5009002, upload-time = "2025-11-17T22:31:52.001Z" },
+    { url = "https://files.pythonhosted.org/packages/f5/f0/0cfadd537c5470378b1b32bd859cf2824972174b51b873c9d95cfd7475a5/ml_dtypes-0.5.4-cp312-cp312-win_amd64.whl", hash = "sha256:c1a953995cccb9e25a4ae19e34316671e4e2edaebe4cf538229b1fc7109087b7", size = 212222, upload-time = "2025-11-17T22:31:53.742Z" },
+    { url = "https://files.pythonhosted.org/packages/16/2e/9acc86985bfad8f2c2d30291b27cd2bb4c74cea08695bd540906ed744249/ml_dtypes-0.5.4-cp312-cp312-win_arm64.whl", hash = "sha256:9bad06436568442575beb2d03389aa7456c690a5b05892c471215bfd8cf39460", size = 160793, upload-time = "2025-11-17T22:31:55.358Z" },
+]
+
 [[package]]
 name = "mlflow"
 version = "3.11.1"
@@ -1919,6 +1947,8 @@ dev = [
     { name = "evidently" },
     { name = "httpx" },
     { name = "line-profiler" },
+    { name = "onnxmltools" },
+    { name = "onnxruntime" },
     { name = "pytest" },
     { name = "pytest-cov" },
     { name = "ruff" },
@@ -1956,6 +1986,8 @@ dev = [
     { name = "evidently", specifier = ">=0.7" },
     { name = "httpx", specifier = ">=0.28.1" },
     { name = "line-profiler", specifier = ">=5.0.2" },
+    { name = "onnxmltools", specifier = ">=1.16.0" },
+    { name = "onnxruntime", specifier = ">=1.26.0" },
     { name = "pytest", specifier = ">=9.0.3" },
     { name = "pytest-cov", specifier = ">=7.1.0" },
     { name = "ruff", specifier = ">=0.15.12" },
@@ -1963,6 +1995,59 @@ dev = [
     { name = "streamlit", specifier = ">=1.57.0" },
 ]
 
+[[package]]
+name = "onnx"
+version = "1.21.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "ml-dtypes" },
+    { name = "numpy" },
+    { name = "protobuf" },
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/c5/93/942d2a0f6a70538eea042ce0445c8aefd46559ad153469986f29a743c01c/onnx-1.21.0.tar.gz", hash = "sha256:4d8b67d0aaec5864c87633188b91cc520877477ec0254eda122bef8be43cd764", size = 12074608, upload-time = "2026-03-27T21:33:36.118Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7d/ae/cb644ec84c25e63575d9d8790fdcc5d1a11d67d3f62f872edb35fa38d158/onnx-1.21.0-cp312-abi3-macosx_12_0_universal2.whl", hash = "sha256:fc2635400fe39ff37ebc4e75342cc54450eadadf39c540ff132c319bf4960095", size = 17965930, upload-time = "2026-03-27T21:32:48.089Z" },
+    { url = "https://files.pythonhosted.org/packages/6f/b6/eeb5903586645ef8a49b4b7892580438741acc3df91d7a5bd0f3a59ea9cb/onnx-1.21.0-cp312-abi3-manylinux_2_26_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:9003d5206c01fa2ff4b46311566865d8e493e1a6998d4009ec6de39843f1b59b", size = 17531344, upload-time = "2026-03-27T21:32:50.837Z" },
+    { url = "https://files.pythonhosted.org/packages/a7/00/4823f06357892d1e60d6f34e7299d2ba4ed2108c487cc394f7ce85a3ff14/onnx-1.21.0-cp312-abi3-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:a9261bd580fb8548c9c37b3c6750387eb8f21ea43c63880d37b2c622e1684285", size = 17613697, upload-time = "2026-03-27T21:32:54.222Z" },
+    { url = "https://files.pythonhosted.org/packages/23/1d/391f3c567ae068c8ac4f1d1316bae97c9eb45e702f05975fe0e17ad441f0/onnx-1.21.0-cp312-abi3-win32.whl", hash = "sha256:9ea4e824964082811938a9250451d89c4ec474fe42dd36c038bfa5df31993d1e", size = 16287200, upload-time = "2026-03-27T21:32:57.277Z" },
+    { url = "https://files.pythonhosted.org/packages/9c/a6/5eefbe5b40ea96de95a766bd2e0e751f35bdea2d4b951991ec9afaa69531/onnx-1.21.0-cp312-abi3-win_amd64.whl", hash = "sha256:458d91948ad9a7729a347550553b49ab6939f9af2cddf334e2116e45467dc61f", size = 16441045, upload-time = "2026-03-27T21:33:00.081Z" },
+    { url = "https://files.pythonhosted.org/packages/63/c4/0ed8dc037a39113d2a4d66e0005e07751c299c46b993f1ad5c2c35664c20/onnx-1.21.0-cp312-abi3-win_arm64.whl", hash = "sha256:ca14bc4842fccc3187eb538f07eabeb25a779b39388b006db4356c07403a7bbb", size = 16403134, upload-time = "2026-03-27T21:33:03.987Z" },
+]
+
+[[package]]
+name = "onnxmltools"
+version = "1.16.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "numpy" },
+    { name = "onnx" },
+    { name = "protobuf" },
+    { name = "skl2onnx" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/41/3e/85a40b6e56a8aaa45bffc9eb00f93182b87841b4dc5a4198ea609993e17c/onnxmltools-1.16.0.tar.gz", hash = "sha256:cd76e0a7ba6a3c4ca4acf3b4c7973cda6a70f2edc146ab11d4efc3dfbee6805a", size = 208397, upload-time = "2026-01-30T12:45:06.1Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/fb/e6/6713d9a089a6861b4bf748f02a6238cb2759968aadf672dccef3e960376b/onnxmltools-1.16.0-py3-none-any.whl", hash = "sha256:7b27196e7dcc0d9de29110f211e7941ad1c71dd97606baa729144d9acd105d3c", size = 303991, upload-time = "2026-01-30T12:45:04.809Z" },
+]
+
+[[package]]
+name = "onnxruntime"
+version = "1.26.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "flatbuffers" },
+    { name = "numpy" },
+    { name = "packaging" },
+    { name = "protobuf" },
+]
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/81/b1/d111b1df656761f980d9e298a60039a9cb66036b1d039e777537743d0ac3/onnxruntime-1.26.0-cp312-cp312-macosx_14_0_arm64.whl", hash = "sha256:05b028781b322ad74b57ce5b50aa5280bb1fe96ceec334628ade681e0b24c1ac", size = 18016624, upload-time = "2026-05-12T00:41:01.735Z" },
+    { url = "https://files.pythonhosted.org/packages/f6/a0/3f9d896a0385a36bd04345d6d0b802821a5782adde562e7e135f6bb71c73/onnxruntime-1.26.0-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:91f2bb870a4b9224eba0a6728c1fa7a9e552b8e59e1083c51fbbc3d013f2b5c0", size = 16052692, upload-time = "2026-05-08T19:07:13.829Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/43/2a4e04f8dbeffad19bbcced4bcd4289bf478921518437404d6b92bdf213b/onnxruntime-1.26.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:9b6dd70599005bd1bf29779f04a91978b92b5e719c11a20068a8f8e535f725b6", size = 18185439, upload-time = "2026-05-08T19:07:36.299Z" },
+    { url = "https://files.pythonhosted.org/packages/44/fc/026d0a7162b9c2153dac292baea9e027c42304dc1d9dc6f8ff5b4cfbaedd/onnxruntime-1.26.0-cp312-cp312-win_amd64.whl", hash = "sha256:a26374dc7fbcaae593601086b242120e13f2310558df0991da6dd8b8fac00414", size = 13026427, upload-time = "2026-05-08T19:08:03.503Z" },
+    { url = "https://files.pythonhosted.org/packages/3e/27/1dcf88e45e4c69db5f7b106f2dacc3801ba98994e082ca03e1dfdf7bfe57/onnxruntime-1.26.0-cp312-cp312-win_arm64.whl", hash = "sha256:54a8053410fd31fd66469bd754fcfe8a4df9f7eb44756b4b5479bf50c842d948", size = 12796647, upload-time = "2026-05-08T19:07:52.108Z" },
+]
+
 [[package]]
 name = "opentelemetry-api"
 version = "1.41.0"
@@ -2835,6 +2920,19 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/b7/ce/149a00dd41f10bc29e5921b496af8b574d8413afcd5e30dfa0ed46c2cc5e/six-1.17.0-py2.py3-none-any.whl", hash = "sha256:4721f391ed90541fddacab5acf947aa0d3dc7d27b2e1e8eda2be8970586c3274", size = 11050, upload-time = "2024-12-04T17:35:26.475Z" },
 ]
 
+[[package]]
+name = "skl2onnx"
+version = "1.20.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "onnx" },
+    { name = "scikit-learn" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/cb/39/a5015fefb613d5172541740540851a301c53392b57051cf4d313cb6d5718/skl2onnx-1.20.0.tar.gz", hash = "sha256:c74ea827d92ba186fe659695e8fc989cd97bfc320edce3d32b9936a5878da10a", size = 956369, upload-time = "2026-01-30T10:52:07.694Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/24/d3/b0db77025a4683ec1b9aafc301b78c7e2e2059a1e2543e918435f3d03582/skl2onnx-1.20.0-py3-none-any.whl", hash = "sha256:30cac34803d1776c14b336ae945e48ef28debfc339215acde1cc04b963ed3f7b", size = 317169, upload-time = "2026-01-30T10:52:05.824Z" },
+]
+
 [[package]]
 name = "skops"
 version = "0.14.0"