Submission: owenqueen

.github/workflows/submit_owenqueen.yaml

@@ -0,0 +1,38 @@
+---
+# This workflow will install Python dependencies and run tests
+
+name: Unit test and code coverage
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+jobs:
+  build:
+
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ['3.13']
+        submission: ['owenqueen']
+
+    steps:
+      - uses: actions/checkout@v3
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v4
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Install dependencies
+        run: |
+          sudo apt-get update
+          sudo apt install libbz2-dev
+          python -m pip install --upgrade pip
+          pip install wheel
+          pip install .
+          pip install .[dev]
+          if [ -f requirements_${{ matrix.submission }}.txt ]; then pip install -r requirements_${{ matrix.submission }}.txt; fi
+      - name: Run unit tests with pytest
+        run: |
+          python -m pytest tests/test_${{ matrix.submission }}.py

requirements_owenqueen.txt

@@ -0,0 +1,3 @@
+scikit-learn
+numpy
+pandas

tests/test_owenqueen.py

@@ -0,0 +1,221 @@
+import os
+import pickle
+from pathlib import Path
+
+import numpy as np
+import qp
+import tables_io
+import pytest
+from sklearn.ensemble import HistGradientBoostingRegressor
+
+# These are used by test scripts
+from pz_data_challenge.taskset_1 import run_taskset_1
+
+from pz_data_challenge import submit_utils
+
+# Change these to match the name of the submission
+# and a URL to download the submission data files
+# and needed model files
+SUBMISSION_NAME: str = "owenqueen"
+SUBMISSION_URL: str = "https://github.com/owencqueen/pz_data_challenge/releases/download/submit-owenqueen/submit_owenqueen.tgz"
+
+# don't change these
+SUBMIT_DIR: str = f"submissions/{SUBMISSION_NAME}"
+PUBLIC_AREA: str = "tests/public"
+
+# ---------------------------------------------------------------------------
+# Photo-z model: quantile regression on magnitudes + colors.
+#
+# HistGradientBoostingRegressor handles NaN (non-detections) natively, so we
+# can feed magnitudes/colors directly without imputation. We fit one regressor
+# per target quantile of redshift; the per-object predicted quantiles define a
+# full p(z) via qp's quantile parameterization. The median quantile is used as
+# the `zmode` point estimate.
+# ---------------------------------------------------------------------------
+
+# Bands ordered by effective wavelength (blue -> red): LSST ugrizy + Roman YJH.
+_BAND_ORDER = ["u", "g", "r", "i", "z", "y", "Y", "J", "H"]
+_LSST_BANDS = {"u", "g", "r", "i", "z", "y"}
+
+# CDF levels at which we predict redshift quantiles.
+_QUANTILES = np.array(
+    [0.02, 0.05, 0.10, 0.16, 0.25, 0.40, 0.50, 0.60, 0.75, 0.84, 0.90, 0.95, 0.98]
+)
+
+
+def _band_col(b: str) -> str:
+    return f"mag_{b}_lsst" if b in _LSST_BANDS else f"mag_{b}_roman"
+
+
+def _load_df(path: str | Path):
+    return tables_io.convert(tables_io.read(str(path)), tables_io.types.PD_DATAFRAME)
+
+
+def _build_features(df) -> tuple[np.ndarray, list[str]]:
+    """Magnitudes + adjacent colors for whatever bands are present."""
+    present = [b for b in _BAND_ORDER if _band_col(b) in df.columns]
+    mags = {b: df[_band_col(b)].to_numpy(dtype=float) for b in present}
+
+    feats: list[np.ndarray] = []
+    names: list[str] = []
+    for b in present:
+        feats.append(mags[b])
+        names.append(f"mag_{b}")
+    for b1, b2 in zip(present[:-1], present[1:]):
+        feats.append(mags[b1] - mags[b2])
+        names.append(f"col_{b1}_{b2}")
+
+    return np.column_stack(feats), names
+
+
+def _train_model(train_file: str | Path) -> dict:
+    """Fit one quantile regressor per CDF level; return a picklable dict."""
+    df = _load_df(train_file)
+    X, feat_names = _build_features(df)
+    y = df["redshift"].to_numpy(dtype=float)
+
+    regressors = []
+    for q in _QUANTILES:
+        reg = HistGradientBoostingRegressor(
+            loss="quantile",
+            quantile=float(q),
+            max_iter=300,
+            learning_rate=0.08,
+            max_leaf_nodes=63,
+            min_samples_leaf=60,
+            l2_regularization=1.0,
+            early_stopping=False,
+            random_state=42,
+        )
+        reg.fit(X, y)
+        regressors.append(reg)
+
+    return {
+        "quantiles": _QUANTILES,
+        "regressors": regressors,
+        "feat_names": feat_names,
+        "z_max": float(y.max()),
+    }
+
+
+def _estimate(model: dict, test_file: str | Path, output_file: str | Path) -> None:
+    """Predict per-object redshift quantiles and write a qp.quant ensemble."""
+    df = _load_df(test_file)
+    X, _ = _build_features(df)
+
+    quants = model["quantiles"]
+    # (n_obj, n_quant) matrix of predicted z at each CDF level.
+    preds = np.column_stack([reg.predict(X) for reg in model["regressors"]])
+    # Enforce monotonic, non-negative quantiles (fix any quantile crossing).
+    preds = np.clip(preds, 0.0, None)
+    preds = np.maximum.accumulate(preds, axis=1)
+
+    median_idx = int(np.argmin(np.abs(quants - 0.5)))
+    zmode = preds[:, median_idx]
+
+    ens = qp.Ensemble(qp.quant, data=dict(quants=quants, locs=preds))
+    ens.set_ancil(
+        {
+            "object_id": df["object_id"].astype("int64").to_numpy(),
+            "zmode": zmode,
+        }
+    )
+
+    os.makedirs(os.path.dirname(os.path.abspath(str(output_file))), exist_ok=True)
+    ens.write_to(str(output_file))
+
+
+@pytest.fixture(name="setup_submit_area", scope="module")
+def setup_submit_area(request: pytest.FixtureRequest) -> int:
+    """
+    A pytest fixture to download the submission data
+
+    If all the submission data are in a tar file with the
+    proper structure you should not need to change this function.
+    """
+
+    if not os.path.exists(SUBMIT_DIR):
+        if not SUBMISSION_URL:
+            raise ValueError(f"SUBMISSION_URL in tests/test_{SUBMISSION_NAME}.py has not been set")
+        submit_utils.download_and_extract_tar(SUBMISSION_URL, SUBMIT_DIR)
+
+    def teardown_submit_area() -> None:
+        if not os.environ.get("NO_TEARDOWN"):
+            os.system(f"\\rm -rf {SUBMIT_DIR}")
+
+    try:
+        os.makedirs(os.path.join(SUBMIT_DIR, "outputs_2"))
+    except Exception:
+        pass
+
+    try:
+        os.makedirs(os.path.join(SUBMIT_DIR, "outputs_3"))
+    except Exception:
+        pass
+
+    request.addfinalizer(teardown_submit_area)
+
+    return 0
+
+
+# ---------------------------------------------------------------------------
+# Subtask 2: estimation only -- load a pre-trained model and estimate p(z).
+# ---------------------------------------------------------------------------
+def run_taskset_1_estimation_only(
+    model_file: str | Path,
+    test_file: str | Path,
+    output_file: str | Path,
+) -> None:
+    """
+    User supplied function to run estimation for task set 1
+
+    This function should use a model stored in model_file, which
+    is downloaded as part of the submission tar file.
+
+    This function should write output data to output_file in qp
+    format.
+    """
+    with open(model_file, "rb") as fin:
+        model = pickle.load(fin)
+    _estimate(model, test_file, output_file)
+
+
+# ---------------------------------------------------------------------------
+# Subtask 3: train a model on train_file, then estimate p(z) on test_file.
+# ---------------------------------------------------------------------------
+def run_taskset_1_training_and_estimation(
+    train_file: str | Path,
+    test_file: str | Path,
+    output_file: str | Path,
+) -> None:
+    """
+    User supplied function to run training and estimation for task set 1
+
+    This function should train a model and use it.
+
+    This function should write output data to output_file in qp
+    format.
+    """
+    model = _train_model(train_file)
+    _estimate(model, test_file, output_file)
+
+
+def test_example_taskset_1(
+    setup_public_area: int,
+    setup_submit_area: int,
+) -> None:
+    """
+    Test function to validate a submission for Taskset 1
+
+    You should not need to change this function
+    """
+
+    assert setup_public_area == 0
+    assert setup_submit_area == 0
+
+    run_taskset_1(
+        PUBLIC_AREA,
+        SUBMISSION_NAME,
+        run_taskset_1_estimation_only,
+        run_taskset_1_training_and_estimation,
+    )