ci: efficiency tests (#942)

.github/workflows/ci.yaml

@@ -59,7 +59,7 @@ jobs:
           CIBW_TEST_COMMAND: >
             uv pip install -r {project}/setup.py --extra all &&
             nbdev_test --path {project}/nbs/ --skip_file_re "(distributed|prophet).*.ipynb" --pause 1.0 --do_print --timing &&
-            pytest --durations=0 {project}/action_files
+            pytest --durations=0 {project}/action_files -k "not efficiency"
 
       - name: Build wheels and run local tests
         if: matrix.os-platform[0] != 'ubuntu-latest'
@@ -126,3 +126,23 @@ jobs:
           python -m src.experiment
           python -m src.evaluation --test
         working-directory: ./experiments/m3
+
+  efficiency-tests:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          submodules: "true"
+
+      - uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+        with:
+          python-version: "3.12"
+
+      - name: Install dependencies
+        run: pip install uv && uv pip install --system . pytest-codspeed pytest-xdist
+
+      - name: Run benchmarks
+        uses: CodSpeedHQ/action@fa1dcde8d58f2ab0b407a6a24d6cc5a8c1444a8c # 3.1.0
+        with:
+          token: ${{ secrets.CODESPEED_TOKEN }}
+          run: pytest action_files/test_efficiency.py --codspeed -n 2

action_files/test_efficiency.py

@@ -0,0 +1,33 @@
+import inspect
+
+import numpy as np
+import pytest
+
+import statsforecast.models
+
+
+models = [
+    statsforecast.models.AutoARIMA,
+    statsforecast.models.AutoCES,
+    statsforecast.models.AutoETS,
+    statsforecast.models.AutoTheta,
+    statsforecast.models.TBATS,
+    statsforecast.models.GARCH,
+]
+
+
+@pytest.fixture(scope="module")
+def y():
+    return np.arange(24)[np.arange(200) % 24]
+
+
+@pytest.mark.parametrize("model_cls", models)
+def test_efficiency(benchmark, y, model_cls):
+    if model_cls is statsforecast.models.AutoCES:
+        # fails to fit without model="N"
+        model = model_cls(season_length=24, model="N")
+    elif "season_length" in inspect.signature(model_cls).parameters:
+        model = model_cls(season_length=24)
+    else:
+        model = model_cls()
+    benchmark(model.forecast, y=y, h=48)