drop py38

.github/workflows/ci.yaml

@@ -18,7 +18,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.8", "3.9", "3.10", "3.11"]
+        python-version: ["3.9", "3.10", "3.11", "3.12"]
     env:
       AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID_NIXTLA_TMP }}
       AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY_NIXTLA_TMP }}
@@ -44,7 +44,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [macos-13, macos-14, windows-latest]
-        python-version: ["3.8", "3.9", "3.10", "3.11"]
+        python-version: ["3.9", "3.10", "3.11", "3.12"]
     steps:
       - name: Clone repo
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

README.md

@@ -57,6 +57,7 @@ for best practices.**
 - [m4](https://www.kaggle.com/code/lemuz90/m4-competition)
 - [m4-cv](https://www.kaggle.com/code/lemuz90/m4-competition-cv)
 - [favorita](https://www.kaggle.com/code/lemuz90/mlforecast-favorita)
+- [VN1](https://colab.research.google.com/drive/1UdhCAk49k6HgMezG-U_1ETnAB5pYvZk9)
 
 ## Why?
 

mlforecast/target_transforms.py

@@ -125,14 +125,14 @@ def inverse_transform(self, ga: GroupedArray) -> GroupedArray:
     def inverse_transform_fitted(self, ga: GroupedArray) -> GroupedArray:
         if self.fitted_[0].size < ga.data.size:
             raise ValueError("fitted differences are smaller than provided target.")
+        transformed = ga.data
         for d, fitted in zip(reversed(self.differences), reversed(self.fitted_)):
             fitted_ga = CoreGroupedArray(fitted, self.fitted_indptr_)
             adds = fitted_ga._lag(d)
             if adds.size > ga.data.size:
                 adds = CoreGroupedArray(adds, self.fitted_indptr_)._tails(ga.indptr)
-            transformed = ga.data + adds
-            ga = GroupedArray(transformed, ga.indptr)
-        return ga
+            transformed = transformed + adds
+        return GroupedArray(transformed, ga.indptr)
 
     def take(self, idxs: np.ndarray) -> "Differences":
         out = Differences(self.differences)
@@ -161,9 +161,11 @@ def stack(scalers: Sequence["Differences"]) -> "Differences":  # type: ignore[ov
         diffs = first_scaler.differences
         out = Differences(diffs)
         out.fitted_ = []
-        for i in range(len(scalers[0].fitted_)):
-            out.fitted_.append(np.hstack([sc.fitted_[i] for sc in scalers]))
-        if first_scaler.fitted_indptr_ is not None:
+        if first_scaler.fitted_indptr_ is None:
+            out.fitted_indptr_ = None
+        else:
+            for i in range(len(scalers[0].fitted_)):
+                out.fitted_.append(np.hstack([sc.fitted_[i] for sc in scalers]))
             sizes = np.hstack([np.diff(sc.fitted_indptr_) for sc in scalers])
             out.fitted_indptr_ = np.append(0, sizes.cumsum())
         out.scalers_ = [

nbs/target_transforms.ipynb

@@ -201,14 +201,14 @@
     "    def inverse_transform_fitted(self, ga: GroupedArray) -> GroupedArray:\n",
     "        if self.fitted_[0].size < ga.data.size:\n",
     "            raise ValueError(\"fitted differences are smaller than provided target.\")\n",
+    "        transformed = ga.data\n",
     "        for d, fitted in zip(reversed(self.differences), reversed(self.fitted_)):\n",
     "            fitted_ga = CoreGroupedArray(fitted, self.fitted_indptr_)\n",
     "            adds = fitted_ga._lag(d)\n",
     "            if adds.size > ga.data.size:\n",
     "                adds = CoreGroupedArray(adds, self.fitted_indptr_)._tails(ga.indptr)\n",
-    "            transformed = ga.data + adds\n",
-    "            ga = GroupedArray(transformed, ga.indptr)\n",
-    "        return ga\n",
+    "            transformed = transformed + adds\n",
+    "        return GroupedArray(transformed, ga.indptr)\n",
     "\n",
     "    def take(self, idxs: np.ndarray) -> \"Differences\":\n",
     "        out = Differences(self.differences)\n",
@@ -237,9 +237,11 @@
     "        diffs = first_scaler.differences\n",
     "        out = Differences(diffs)\n",
     "        out.fitted_ = []\n",
-    "        for i in range(len(scalers[0].fitted_)):\n",
-    "            out.fitted_.append(np.hstack([sc.fitted_[i] for sc in scalers]))\n",
-    "        if first_scaler.fitted_indptr_ is not None:\n",
+    "        if first_scaler.fitted_indptr_ is None:\n",
+    "            out.fitted_indptr_ = None\n",
+    "        else:\n",
+    "            for i in range(len(scalers[0].fitted_)):\n",
+    "                out.fitted_.append(np.hstack([sc.fitted_[i] for sc in scalers]))\n",
     "            sizes = np.hstack([np.diff(sc.fitted_indptr_) for sc in scalers])\n",
     "            out.fitted_indptr_ = np.append(0, sizes.cumsum())\n",
     "        out.scalers_ = [\n",

settings.ini

@@ -9,13 +9,13 @@ author_email = [email protected]
 copyright = Nixtla
 branch = main
 version = 0.15.0
-min_python = 3.8
+min_python = 3.9
 audience = Developers
 language = English
 custom_sidebar = True
 license = apache2
-status = 3
-requirements = cloudpickle coreforecast>=0.0.11 fsspec numba optuna packaging pandas scikit-learn utilsforecast>=0.2.3 window-ops
+status = 4
+requirements = cloudpickle coreforecast>=0.0.15 fsspec numba optuna packaging pandas scikit-learn utilsforecast>=0.2.3 window-ops
 dask_requirements = fugue dask[complete] lightgbm xgboost
 ray_requirements = fugue[ray] lightgbm_ray numpy<2 pandas<2.2 ray<2.8 setuptools<70 xgboost<2 xgboost_ray
 spark_requirements = fugue pyspark>=3.3 lightgbm xgboost

setup.py

@@ -22,7 +22,7 @@
 }
 statuses = [ '1 - Planning', '2 - Pre-Alpha', '3 - Alpha',
     '4 - Beta', '5 - Production/Stable', '6 - Mature', '7 - Inactive' ]
-py_versions = '3.8 3.9 3.10 3.11'.split()
+py_versions = '3.9 3.10 3.11 3.12'.split()
 
 requirements = cfg['requirements'].split()
 dask_requirements = cfg['dask_requirements'].split()