Merge branch 'main' into add-autoformer-v2

harshsomankar123-tech · harshsomankar123-tech · commit 630230abeae8 · 2026-08-09T20:17:38.000+05:30
diff --git a/.github/workflows/pypi_release.yml b/.github/workflows/pypi_release.yml
@@ -12,7 +12,7 @@ jobs:
     steps:
       - uses: actions/checkout@v7
 
-      - uses: actions/setup-python@v6
+      - uses: actions/setup-python@v7
         with:
           python-version: '3.11'
 
@@ -34,11 +34,12 @@ jobs:
   build_wheels:
     name: Build wheels
     runs-on: ubuntu-latest
+    needs: [check_tag]
 
     steps:
       - uses: actions/checkout@v7
 
-      - uses: actions/setup-python@v6
+      - uses: actions/setup-python@v7
         with:
           python-version: '3.11'
 
@@ -67,7 +68,7 @@ jobs:
       - uses: actions/checkout@v7
 
       - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v6
+        uses: actions/setup-python@v7
         with:
           python-version: ${{ matrix.python-version }}
 
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
@@ -22,7 +22,7 @@ jobs:
         uses: actions/checkout@v7
 
       - name: python environment step
-        uses: actions/setup-python@v6
+        uses: actions/setup-python@v7
         with:
           python-version: "3.11"
 
@@ -59,7 +59,7 @@ jobs:
     steps:
       - uses: actions/checkout@v7
       - name: Set up Python
-        uses: actions/setup-python@v6
+        uses: actions/setup-python@v7
         with:
           python-version: "3.11"
       - name: Install dependencies
@@ -88,7 +88,7 @@ jobs:
       - uses: actions/checkout@v7
 
       - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v6
+        uses: actions/setup-python@v7
         with:
           python-version: ${{ matrix.python-version }}
 
@@ -132,7 +132,7 @@ jobs:
           enable-cache: true
 
       - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v6
+        uses: actions/setup-python@v7
         with:
           python-version: ${{ matrix.python-version }}
 
@@ -172,7 +172,7 @@ jobs:
       - uses: actions/checkout@v7
 
       - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v6
+        uses: actions/setup-python@v7
         with:
           python-version: ${{ matrix.python-version }}
 
diff --git a/docs/source/m_layer_v2.rst b/docs/source/m_layer_v2.rst
@@ -47,3 +47,4 @@ See the detailed API documentation for the V2 base classes and specific model im
    models.samformer._samformer_v2.Samformer
    models.tide._tide_dsipts._tide_v2.TIDE
    models.timexer._timexer_v2.TimeXer
+   models.mlp._decodermlp_v2.DecoderMLP_v2
diff --git a/docs/source/pkg_v2.rst b/docs/source/pkg_v2.rst
@@ -99,3 +99,4 @@ See the detailed API documentation for the available V2 Package classes below:
    models.samformer._samformer_v2_pkg.Samformer_pkg_v2
    models.tide._tide_dsipts._tide_v2_pkg.TIDE_pkg_v2
    models.timexer._timexer_pkg_v2.TimeXer_pkg_v2
+   models.mlp._decodermlp_pkg_v2.DecoderMLP_pkg_v2
diff --git a/pyproject.toml b/pyproject.toml
@@ -33,7 +33,7 @@ dependencies = [
   "scipy >=1.8,<2.0",
   "pandas >=1.3.0,<3.1.0",
   "scikit-learn >=1.2,<2.0",
-  "scikit-base <1.1.0",
+  "scikit-base <1.2.0",
 ]
 
 [project.optional-dependencies]
diff --git a/pytorch_forecasting/layers/__init__.py b/pytorch_forecasting/layers/__init__.py
@@ -32,6 +32,7 @@
     Encoder,
     EncoderLayer,
 )
+from pytorch_forecasting.layers._mlp import FullyConnectedModule
 from pytorch_forecasting.layers._normalization import RevIN, SeasonalLayerNorm
 from pytorch_forecasting.layers._output._flatten_head import (
     FlattenHead,
@@ -48,6 +49,7 @@
 )
 
 __all__ = [
+    "AttentionLayer",
     "AutoCorrelation",
     "AutoCorrelationLayer",
     "AutoformerDecoder",
@@ -62,10 +64,10 @@
     "FixedEmbedding",
     "FlattenHead",
     "FullAttention",
-    "AttentionLayer",
+    "FullyConnectedModule",
     "PositionalEmbedding",
-    "RevIN",
     "ResidualBlock",
+    "RevIN",
     "SeasonalLayerNorm",
     "SeriesDecomposition",
     "TemporalEmbedding",
diff --git a/pytorch_forecasting/layers/_mlp/__init__.py b/pytorch_forecasting/layers/_mlp/__init__.py
@@ -0,0 +1,7 @@
+"""
+Fully connected (MLP) layers.
+"""
+
+from pytorch_forecasting.layers._mlp._fully_connected import FullyConnectedModule
+
+__all__ = ["FullyConnectedModule"]
diff --git a/pytorch_forecasting/layers/_mlp/_fully_connected.py b/pytorch_forecasting/layers/_mlp/_fully_connected.py
@@ -0,0 +1,52 @@
+"""
+Fully connected (MLP) module.
+"""
+
+import torch
+from torch import nn
+
+
+class FullyConnectedModule(nn.Module):
+    def __init__(
+        self,
+        input_size: int,
+        output_size: int,
+        hidden_size: int,
+        n_hidden_layers: int,
+        activation_class: nn.ReLU,
+        dropout: float = None,
+        norm: bool = True,
+    ):
+        super().__init__()
+        self.input_size = input_size
+        self.output_size = output_size
+        self.hidden_size = hidden_size
+        self.n_hidden_layers = n_hidden_layers
+        self.activation_class = activation_class
+        self.dropout = dropout
+        self.norm = norm
+
+        # input layer
+        module_list = [nn.Linear(input_size, hidden_size), activation_class()]
+        if dropout is not None:
+            module_list.append(nn.Dropout(dropout))
+        if norm:
+            module_list.append(nn.LayerNorm(hidden_size))
+        # hidden layers
+        for _ in range(n_hidden_layers):
+            module_list.extend(
+                [nn.Linear(hidden_size, hidden_size), activation_class()]
+            )
+            if dropout is not None:
+                module_list.append(nn.Dropout(dropout))
+            if norm:
+                module_list.append(nn.LayerNorm(hidden_size))
+        # output layer
+        module_list.append(nn.Linear(hidden_size, output_size))
+
+        self.sequential = nn.Sequential(*module_list)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        # x of shape: batch_size x n_timesteps_in
+        # output of shape batch_size x n_timesteps_out
+        return self.sequential(x)
diff --git a/pytorch_forecasting/metrics/base_metrics/_base_metrics.py b/pytorch_forecasting/metrics/base_metrics/_base_metrics.py
@@ -689,18 +689,22 @@ def to_quantiles(self, y_pred: torch.Tensor, **kwargs) -> torch.Tensor:
         return self._metrics[0].to_quantiles(y_pred, **kwargs)
 
     def __add__(self, metric: LightningMetric):
+        new_metrics = list(self._metrics)
+        new_weights = list(self._weights)
         if isinstance(metric, self.__class__):
-            self._metrics.extend(metric._metrics)
-            self._weights.extend(metric._weights)
+            new_metrics.extend(metric._metrics)
+            new_weights.extend(metric._weights)
         else:
-            self._metrics.append(metric)
-            self._weights.append(1.0)
+            new_metrics.append(metric)
+            new_weights.append(1.0)
 
-        return self
+        result = CompositeMetric(metrics=new_metrics, weights=new_weights)
+        return result
 
     def __mul__(self, multiplier: float):
-        self._weights = [w * multiplier for w in self._weights]
-        return self
+        new_weights = [w * multiplier for w in self._weights]
+        result = CompositeMetric(metrics=list(self._metrics), weights=new_weights)
+        return result
 
     __rmul__ = __mul__
 
diff --git a/pytorch_forecasting/models/mlp/__init__.py b/pytorch_forecasting/models/mlp/__init__.py
@@ -2,6 +2,14 @@
 
 from pytorch_forecasting.models.mlp._decodermlp import DecoderMLP
 from pytorch_forecasting.models.mlp._decodermlp_pkg import DecoderMLP_pkg
+from pytorch_forecasting.models.mlp._decodermlp_pkg_v2 import DecoderMLP_pkg_v2
+from pytorch_forecasting.models.mlp._decodermlp_v2 import DecoderMLP_v2
 from pytorch_forecasting.models.mlp.submodules import FullyConnectedModule
 
-__all__ = ["DecoderMLP", "DecoderMLP_pkg", "FullyConnectedModule"]
+__all__ = [
+    "DecoderMLP",
+    "DecoderMLP_pkg",
+    "DecoderMLP_v2",
+    "DecoderMLP_pkg_v2",
+    "FullyConnectedModule",
+]
diff --git a/pytorch_forecasting/models/mlp/_decodermlp_pkg_v2.py b/pytorch_forecasting/models/mlp/_decodermlp_pkg_v2.py
@@ -0,0 +1,81 @@
+"""DecoderMLP v2 package container."""
+
+from pytorch_forecasting.base._base_pkg import Base_pkg
+
+
+class DecoderMLP_pkg_v2(Base_pkg):
+    """DecoderMLP v2 package container."""
+
+    _tags = {
+        "info:name": "DecoderMLP_v2",
+        "info:compute": 1,
+        "authors": ["jdb78", "echo-xiao"],
+        # TODO: the v2 datamodule supports categorical inputs but not
+        # categorical targets yet; add "categorical" to y_type once
+        # EncoderDecoderTimeSeriesDataModule supports categorical targets.
+        "info:y_type": ["numeric"],
+        "capability:exogenous": True,
+        "capability:multivariate": False,
+        "capability:pred_int": True,
+        "capability:flexible_history_length": True,
+        "capability:cold_start": True,
+    }
+
+    @classmethod
+    def get_cls(cls):
+        """Get model class."""
+        from pytorch_forecasting.models.mlp._decodermlp_v2 import DecoderMLP_v2
+
+        return DecoderMLP_v2
+
+    @classmethod
+    def get_datamodule_cls(cls):
+        """Get the underlying DataModule class."""
+        from pytorch_forecasting.data.data_module import (
+            EncoderDecoderTimeSeriesDataModule,
+        )
+
+        return EncoderDecoderTimeSeriesDataModule
+
+    @classmethod
+    def get_test_train_params(cls):
+        """Return testing parameter settings for the trainer.
+
+        Returns
+        -------
+        params : list of dict
+            Parameters to create testing instances of the class. Each dict is passed
+            as ``model_cfg`` to the package constructor; the ``"datamodule_cfg"`` key
+            is forwarded to the datamodule constructor.
+        """
+        from pytorch_forecasting.metrics import MAE, RMSE, SMAPE, QuantileLoss
+
+        params = [
+            {},
+            dict(
+                hidden_size=64, n_hidden_layers=2, dropout=0.1, norm=True, loss=RMSE()
+            ),
+            dict(
+                hidden_size=128,
+                n_hidden_layers=1,
+                activation_class="ReLU",
+                loss=SMAPE(),
+                logging_metrics=[MAE()],
+            ),
+            dict(hidden_size=32, n_hidden_layers=2, norm=False, loss=MAE()),
+            dict(hidden_size=64, n_hidden_layers=1, loss=QuantileLoss()),
+            dict(
+                optimizer="adamw",
+                lr_scheduler="cosine_annealing",
+                lr_scheduler_params={"T_max": 5},
+                loss=MAE(),
+            ),
+        ]
+
+        default_dm_cfg = {"max_encoder_length": 4, "max_prediction_length": 3}
+        for param in params:
+            dm_cfg = default_dm_cfg.copy()
+            dm_cfg.update(param.get("datamodule_cfg", {}))
+            param["datamodule_cfg"] = dm_cfg
+
+        return params
diff --git a/pytorch_forecasting/models/mlp/_decodermlp_v2.py b/pytorch_forecasting/models/mlp/_decodermlp_v2.py
diff --git a/pytorch_forecasting/models/mlp/submodules.py b/pytorch_forecasting/models/mlp/submodules.py
diff --git a/tests/test_metrics.py b/tests/test_metrics.py

Original file line number	Diff line number	Diff line change
`@@ -33,7 +33,7 @@ dependencies = [`
`33`	`33`	`"scipy >=1.8,<2.0",`
`34`	`34`	`"pandas >=1.3.0,<3.1.0",`
`35`	`35`	`"scikit-learn >=1.2,<2.0",`
`36`		`- "scikit-base <1.1.0",`
	`36`	`+ "scikit-base <1.2.0",`
`37`	`37`	`]`
`38`	`38`
`39`	`39`	`[project.optional-dependencies]`