[ENH] rename "model metadata" classes to "model package" classes (#1892)

Proposed solution for, and fixes
#1886

This PR contains changes naming convention of "model metadata" class to
"model package container" nomenclature

Not breaking as the v2 package/metadata layer has not been released yet.

Also contains:

* minor docstring fixes
* tests for the naming pattern
* a small bugfix to the test framework, for the above test

Author: fkiraly

pytorch_forecasting/models/base/_base_object.py

@@ -12,7 +12,7 @@ class _BaseObject(_SkbaseBaseObject):
 
 
 class _BasePtForecaster(_BaseObject):
-    """Base class for all PyTorch Forecasting forecaster metadata.
+    """Base class for all PyTorch Forecasting forecaster packages.
 
     This class points to model objects and contains metadata as tags.
     """

pytorch_forecasting/models/deepar/__init__.py

@@ -1,6 +1,6 @@
 """DeepAR: Probabilistic forecasting with autoregressive recurrent networks."""
 
 from pytorch_forecasting.models.deepar._deepar import DeepAR
-from pytorch_forecasting.models.deepar._deepar_metadata import DeepARMetadata
+from pytorch_forecasting.models.deepar._deepar_pkg import DeepAR_pkg
 
-__all__ = ["DeepAR", "DeepARMetadata"]
+__all__ = ["DeepAR", "DeepAR_pkg"]

pytorch_forecasting/models/deepar/_deepar.py

@@ -35,6 +35,8 @@
 
 
 class DeepAR(AutoRegressiveBaseModelWithCovariates):
+    """DeepAR: Probabilistic forecasting with autoregressive recurrent networks."""
+
     def __init__(
         self,
         cell_type: str = "LSTM",

pytorch_forecasting/models/deepar/_deepar_metadata.py renamed to pytorch_forecasting/models/deepar/_deepar_pkg.py

@@ -1,10 +1,10 @@
-"""DeepAR metadata container."""
+"""DeepAR package container."""
 
 from pytorch_forecasting.models.base._base_object import _BasePtForecaster
 
 
-class DeepARMetadata(_BasePtForecaster):
-    """DeepAR metadata container."""
+class DeepAR_pkg(_BasePtForecaster):
+    """DeepAR package container."""
 
     _tags = {
         "info:name": "DeepAR",

pytorch_forecasting/models/mlp/__init__.py

@@ -1,7 +1,7 @@
 """Simple models based on fully connected networks."""
 
 from pytorch_forecasting.models.mlp._decodermlp import DecoderMLP
-from pytorch_forecasting.models.mlp._decodermlp_metadata import DecoderMLPMetadata
+from pytorch_forecasting.models.mlp._decodermlp_pkg import DecoderMLP_pkg
 from pytorch_forecasting.models.mlp.submodules import FullyConnectedModule
 
-__all__ = ["DecoderMLP", "DecoderMLPMetadata", "FullyConnectedModule"]
+__all__ = ["DecoderMLP", "DecoderMLP_pkg", "FullyConnectedModule"]

pytorch_forecasting/models/mlp/_decodermlp.py

@@ -24,8 +24,7 @@
 
 
 class DecoderMLP(BaseModelWithCovariates):
-    """
-    MLP on the decoder.
+    """MLP on the decoder.
 
     MLP that predicts output only based on information available in the decoder.
     """

pytorch_forecasting/models/mlp/_decodermlp_metadata.py renamed to pytorch_forecasting/models/mlp/_decodermlp_pkg.py

@@ -1,10 +1,10 @@
-"""DecoderMLP metadata container."""
+"""DecoderMLP package container."""
 
 from pytorch_forecasting.models.base._base_object import _BasePtForecaster
 
 
-class DecoderMLPMetadata(_BasePtForecaster):
-    """DecoderMLP metadata container."""
+class DecoderMLP_pkg(_BasePtForecaster):
+    """DecoderMLP package container."""
 
     _tags = {
         "info:name": "DecoderMLP",

pytorch_forecasting/models/nbeats/__init__.py

@@ -1,7 +1,7 @@
 """N-Beats model for timeseries forecasting without covariates."""
 
 from pytorch_forecasting.models.nbeats._nbeats import NBeats
-from pytorch_forecasting.models.nbeats._nbeats_metadata import NBeatsMetadata
+from pytorch_forecasting.models.nbeats._nbeats_pkg import NBeats_pkg
 from pytorch_forecasting.models.nbeats.sub_modules import (
     NBEATSGenericBlock,
     NBEATSSeasonalBlock,
@@ -11,7 +11,7 @@
 __all__ = [
     "NBeats",
     "NBEATSGenericBlock",
-    "NBeatsMetadata",
+    "NBeats_pkg",
     "NBEATSSeasonalBlock",
     "NBEATSTrendBlock",
 ]

pytorch_forecasting/models/nbeats/_nbeats.py

@@ -20,6 +20,8 @@
 
 
 class NBeats(BaseModel):
+    """N-Beats model for timeseries forecasting without covariates."""
+
     def __init__(
         self,
         stack_types: Optional[list[str]] = None,

pytorch_forecasting/models/nbeats/_nbeats_metadata.py renamed to pytorch_forecasting/models/nbeats/_nbeats_pkg.py

@@ -1,10 +1,10 @@
-"""NBeats metadata container."""
+"""NBeats package container."""
 
 from pytorch_forecasting.models.base._base_object import _BasePtForecaster
 
 
-class NBeatsMetadata(_BasePtForecaster):
-    """NBeats metadata container."""
+class NBeats_pkg(_BasePtForecaster):
+    """NBeats package container."""
 
     _tags = {
         "info:name": "NBeats",