Merge branch 'master' into weight-size-bug-fix

.pre-commit-config.yaml

@@ -9,19 +9,19 @@ repos:
       - id: check-yaml
       - id: check-ast
   - repo: https://github.com/pycqa/flake8
-    rev: 6.0.0
+    rev: 6.1.0
     hooks:
       - id: flake8
   - repo: https://github.com/pre-commit/mirrors-isort
     rev: v5.10.1
     hooks:
       - id: isort
   - repo: https://github.com/psf/black
-    rev: 23.1.0
+    rev: 23.9.0
     hooks:
       - id: black
   - repo: https://github.com/nbQA-dev/nbQA
-    rev: 1.6.4
+    rev: 1.7.0
     hooks:
       - id: nbqa-black
       - id: nbqa-isort

docs/requirements.txt

@@ -5,7 +5,7 @@ docutils
 pydata-sphinx-theme
 lightning>=2.0.0
 cloudpickle
-torch>=2.0
+torch>=2.0,!=2.0.1
 optuna>=3.1.0
 scipy
 pandas>=1.3

docs/source/conf.py

@@ -172,3 +172,7 @@ def setup(app: Sphinx):
     "../../CHANGELOG.md",
     "CHANGELOG.md",
 )
+
+intersphinx_mapping = {
+    "sklearn": ("https://scikit-learn.org/stable/", None),
+}

docs/source/tutorials/ar.ipynb

@@ -231,7 +231,8 @@
    ],
    "source": [
     "# calculate baseline absolute error\n",
-    "actuals = torch.cat([y[0] for x, y in iter(val_dataloader)])\n",
+    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
+    "actuals = torch.cat([y[0] for x, y in iter(val_dataloader)]).to(device)\n",
     "baseline_predictions = Baseline().predict(val_dataloader)\n",
     "SMAPE()(baseline_predictions, actuals)"
    ]
@@ -519,7 +520,7 @@
     }
    ],
    "source": [
-    "actuals = torch.cat([y[0] for x, y in iter(val_dataloader)])\n",
+    "actuals = torch.cat([y[0] for x, y in iter(val_dataloader)]).to(device)\n",
     "predictions = best_model.predict(val_dataloader)\n",
     "(actuals - predictions).abs().mean()"
    ]
@@ -538,7 +539,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "raw_predictions, x = best_model.predict(val_dataloader, mode=\"raw\", return_x=True)"
+    "raw_predictions = best_model.predict(val_dataloader, mode=\"raw\", return_x=True)"
    ]
   },
   {
@@ -649,7 +650,7 @@
    ],
    "source": [
     "for idx in range(10):  # plot 10 examples\n",
-    "    best_model.plot_prediction(x, raw_predictions, idx=idx, add_loss_to_title=True)"
+    "    best_model.plot_prediction(raw_predictions.x, raw_predictions.output, idx=idx, add_loss_to_title=True)"
    ]
   },
   {
@@ -797,7 +798,7 @@
    ],
    "source": [
     "for idx in range(10):  # plot 10 examples\n",
-    "    best_model.plot_interpretation(x, raw_predictions, idx=idx)"
+    "    best_model.plot_interpretation(raw_predictions.x, raw_predictions.output, idx=idx)"
    ]
   },
   {

pyproject.toml

@@ -28,24 +28,25 @@ black = 1
 
 [tool.poetry]
 name = "pytorch_forecasting"
-readme = "README.md"  # Markdown files are supported
-version = "0.0.0"  # is being replaced automatically
+readme = "README.md"         # Markdown files are supported
+version = "0.0.0"            # is being replaced automatically
 
 authors = ["Jan Beitner"]
 classifiers = [
-    "Intended Audience :: Developers",
-    "Intended Audience :: Science/Research",
-    "Programming Language :: Python :: 3",
-    "Programming Language :: Python :: 3.6",
-    "Programming Language :: Python :: 3.7",
-    "Programming Language :: Python :: 3.8",
-    "Topic :: Scientific/Engineering",
-    "Topic :: Scientific/Engineering :: Mathematics",
-    "Topic :: Scientific/Engineering :: Artificial Intelligence",
-    "Topic :: Software Development",
-    "Topic :: Software Development :: Libraries",
-    "Topic :: Software Development :: Libraries :: Python Modules",
-    "License :: OSI Approved :: MIT License"]
+  "Intended Audience :: Developers",
+  "Intended Audience :: Science/Research",
+  "Programming Language :: Python :: 3",
+  "Programming Language :: Python :: 3.6",
+  "Programming Language :: Python :: 3.7",
+  "Programming Language :: Python :: 3.8",
+  "Topic :: Scientific/Engineering",
+  "Topic :: Scientific/Engineering :: Mathematics",
+  "Topic :: Scientific/Engineering :: Artificial Intelligence",
+  "Topic :: Software Development",
+  "Topic :: Software Development :: Libraries",
+  "Topic :: Software Development :: Libraries :: Python Modules",
+  "License :: OSI Approved :: MIT License",
+]
 description = "Forecasting timeseries with PyTorch - dataloaders, normalizers, metrics and models"
 repository = "https://github.com/jdb78/pytorch-forecasting"
 documentation = "https://pytorch-forecasting.readthedocs.io"
@@ -54,7 +55,7 @@ homepage = "https://pytorch-forecasting.readthedocs.io"
 [tool.poetry.dependencies]
 python = ">=3.8,<3.11"
 
-torch = "^2.0.0"
+torch = "^2.0.0,!=2.0.1"
 lightning = "^2.0.0"
 optuna = "^3.1.0"
 scipy = "^1.8"
@@ -63,9 +64,9 @@ scikit-learn = "^1.2"
 matplotlib = "*"
 statsmodels = "*"
 
-pytest-github-actions-annotate-failures = {version = "*", optional = true}
-networkx = {version = "^3.0.0", optional = true}
-cpflows = {version="^0.1.2", optional = true}
+pytest-github-actions-annotate-failures = { version = "*", optional = true }
+networkx = { version = "^3.0.0", optional = true }
+cpflows = { version = "^0.1.2", optional = true }
 fastapi = ">=0.80"
 pytorch-optimizer = "^2.5.1"
 
@@ -104,7 +105,7 @@ tensorboard = "^2.12.1"
 pandoc = "^2.3"
 
 
-[tool.poetry.extras]  # extras
+[tool.poetry.extras] # extras
 github-actions = ["pytest-github-actions-annotate-failures"]
 graph = ["networkx"]
 mqf2 = ["cpflows"]
@@ -113,8 +114,8 @@ mqf2 = ["cpflows"]
 enable = true
 vcs = "git"
 dirty = false
-style = "semver"  # semantic versioning
+style = "semver" # semantic versioning
 
-[build-system]  # make the package pip installable
+[build-system] # make the package pip installable
 requires = ["poetry-core>=1.0.7", "poetry-dynamic-versioning>=0.13.1"]
 build-backend = "poetry_dynamic_versioning.backend"

pytorch_forecasting/data/timeseries.py

@@ -197,7 +197,7 @@ def __init__(
         add_relative_time_idx: bool = False,
         add_target_scales: bool = False,
         add_encoder_length: Union[bool, str] = "auto",
-        target_normalizer: Union[NORMALIZER, str, List[NORMALIZER], Tuple[NORMALIZER]] = "auto",
+        target_normalizer: Union[NORMALIZER, str, List[NORMALIZER], Tuple[NORMALIZER], None] = "auto",
         categorical_encoders: Dict[str, NaNLabelEncoder] = {},
         scalers: Dict[str, Union[StandardScaler, RobustScaler, TorchNormalizer, EncoderNormalizer]] = {},
         randomize_length: Union[None, Tuple[float, float], bool] = False,
@@ -291,7 +291,7 @@ def __init__(
                 the index will range from -encoder_length to prediction_length)
             add_target_scales (bool): if to add scales for target to static real features (i.e. add the center and scale
                 of the unnormalized timeseries as features)
-            add_encoder_length (bool): if to add decoder length to list of static real variables.
+            add_encoder_length (bool): if to add encoder length to list of static real variables.
                 Defaults to "auto", i.e. ``True`` if ``min_encoder_length != max_encoder_length``.
             target_normalizer (Union[TorchNormalizer, NaNLabelEncoder, EncoderNormalizer, str, list, tuple]):
                 transformer that take group_ids, target and time_idx to normalize targets.
@@ -1559,6 +1559,8 @@ def __getitem__(self, idx: int) -> Tuple[Dict[str, torch.Tensor], torch.Tensor]:
                 data_cat = data_cat[encoder_length - new_encoder_length : encoder_length + new_decoder_length]
                 data_cont = data_cont[encoder_length - new_encoder_length : encoder_length + new_decoder_length]
                 target = [t[encoder_length - new_encoder_length : encoder_length + new_decoder_length] for t in target]
+                if weight is not None:
+                    weight = weight[encoder_length - new_encoder_length : encoder_length + new_decoder_length]
                 encoder_length = new_encoder_length
                 decoder_length = new_decoder_length
 

pytorch_forecasting/models/base_model.py

@@ -204,7 +204,7 @@ def __init__(
     def _reset_data(self, result: bool = True):
         # reset data objects to save results into
         self._output = []
-        self._decode_lenghts = []
+        self._decode_lengths = []
         self._x_list = []
         self._index = []
         self._y = []
@@ -267,8 +267,8 @@ def on_predict_batch_end(
             self._index.append(trainer.predict_dataloaders.dataset.x_to_index(x))
             out["index"] = self._index[-1]
         if self.return_decoder_lengths:
-            self._decode_lenghts.append(lengths)
-            out["decoder_lengths"] = self._decode_lenghts[-1]
+            self._decode_lengths.append(lengths)
+            out["decoder_lengths"] = self._decode_lengths[-1]
         if self.return_y:
             self._y.append(batch[1])
             out["y"] = self._y[-1]
@@ -312,7 +312,7 @@ def on_predict_epoch_end(self, trainer: "pl.Trainer", pl_module: "pl.LightningMo
             if self.return_index:
                 output["index"] = pd.concat(self._index, axis=0, ignore_index=True)
             if self.return_decoder_lengths:
-                output["decoder_lengths"] = torch.cat(self._decode_lenghts, dim=0)
+                output["decoder_lengths"] = torch.cat(self._decode_lengths, dim=0)
             if self.return_y:
                 y = concat_sequences([yi[0] for yi in self._y])
                 if self._y[-1][1] is None:
@@ -1477,13 +1477,14 @@ def predict_dependency(
             # set values
             data.set_overwrite_values(variable=variable, values=value, target=target)
             # predict
-            kwargs.setdefault("mode", "prediction")
+            pred_kwargs = deepcopy(kwargs)
+            pred_kwargs.setdefault("mode", "prediction")
 
             if idx == 0 and mode == "dataframe":  # need index for returning as dataframe
-                res = self.predict(data, return_index=True, **kwargs)
+                res = self.predict(data, return_index=True, **pred_kwargs)
                 results.append(res.output)
             else:
-                results.append(self.predict(data, **kwargs))
+                results.append(self.predict(data, **pred_kwargs))
             # increment progress
             progress_bar.update()
 

pytorch_forecasting/models/nhits/sub_modules.py

@@ -61,7 +61,7 @@ def forward(
 
 
 def init_weights(module, initialization):
-    if type(module) == torch.nn.Linear:
+    if type(module) is torch.nn.Linear:
         if initialization == "orthogonal":
             torch.nn.init.orthogonal_(module.weight)
         elif initialization == "he_uniform":