Backports v0.15.1 (reprise) (#3191)

*Description of changes:* backporting fixes
- #3188 
- #3189


By submitting this pull request, I confirm that you can use, modify,
copy, and redistribute this contribution, under the terms of your
choice.


**Please tag this pr with at least one of these labels to make our
release process faster:** BREAKING, new feature, bug fix, other change,
dev setup

---------

Co-authored-by: Oleksandr Shchur <[email protected]>

docs/tutorials/advanced_topics/howto_pytorch_lightning.md.template

@@ -110,10 +110,10 @@ class FeedForwardNetwork(nn.Module):
         torch.nn.init.zeros_(lin.bias)
         return lin
 
-    def forward(self, context):
-        scale = self.scaling(context)
-        scaled_context = context / scale
-        nn_out = self.nn(scaled_context)
+    def forward(self, past_target):
+        scale = self.scaling(past_target)
+        scaled_past_target = past_target / scale
+        nn_out = self.nn(scaled_past_target)
         nn_out_reshaped = nn_out.reshape(-1, self.prediction_length, self.hidden_dimensions[-1])
         distr_args = self.args_proj(nn_out_reshaped)
         return distr_args, torch.zeros_like(scale), scale
@@ -143,15 +143,15 @@ class LightningFeedForwardNetwork(FeedForwardNetwork, pl.LightningModule):
         super().__init__(*args, **kwargs)
 
     def training_step(self, batch, batch_idx):
-        context = batch["past_target"]
-        target = batch["future_target"]
+        past_target = batch["past_target"]
+        future_target = batch["future_target"]
 
-        assert context.shape[-1] == self.context_length
-        assert target.shape[-1] == self.prediction_length
+        assert past_target.shape[-1] == self.context_length
+        assert future_target.shape[-1] == self.prediction_length
 
-        distr_args, loc, scale = self(context)
+        distr_args, loc, scale = self(past_target)
         distr = self.distr_output.distribution(distr_args, loc, scale)
-        loss = -distr.log_prob(target)
+        loss = -distr.log_prob(future_target)
 
         return loss.mean()
 

src/gluonts/model/forecast_generator.py

@@ -83,12 +83,15 @@ def make_distribution_forecast(distr, *args, **kwargs) -> Forecast:
 
 
 def make_predictions(prediction_net, inputs: dict):
-    # MXNet predictors only support positional arguments
-    class_name = prediction_net.__class__.__module__
-    if class_name.startswith("gluonts.mx") or class_name.startswith("mxnet"):
-        return prediction_net(*inputs.values())
-    else:
-        return prediction_net(**inputs)
+    try:
+        # Feed inputs as positional arguments for MXNet block predictors
+        import mxnet as mx
+
+        if isinstance(prediction_net, mx.gluon.Block):
+            return prediction_net(*inputs.values())
+    except ImportError:
+        pass
+    return prediction_net(**inputs)
 
 
 class ForecastGenerator:

src/gluonts/torch/distributions/distribution_output.py

@@ -90,14 +90,6 @@ def loss(
             nll = nll * (variance.detach() ** self.beta)
         return nll
 
-    @property
-    def event_shape(self) -> Tuple:
-        r"""
-        Shape of each individual event contemplated by the distributions that
-        this object constructs.
-        """
-        raise NotImplementedError()
-
     @property
     def event_dim(self) -> int:
         r"""

src/gluonts/torch/distributions/output.py

@@ -105,6 +105,13 @@ def loss(
         """
         raise NotImplementedError()
 
+    @property
+    def event_shape(self) -> Tuple:
+        r"""
+        Shape of each individual event compatible with the output object.
+        """
+        raise NotImplementedError()
+
     @property
     def forecast_generator(self) -> ForecastGenerator:
         raise NotImplementedError()

src/gluonts/torch/distributions/quantile_output.py

@@ -37,6 +37,10 @@ def __init__(self, quantiles: List[float]) -> None:
     def forecast_generator(self) -> ForecastGenerator:
         return QuantileForecastGenerator(quantiles=self.quantiles)
 
+    @property
+    def event_shape(self) -> Tuple:
+        return ()
+
     def domain_map(self, *args: torch.Tensor) -> Tuple[torch.Tensor, ...]:
         return args
 

src/gluonts/torch/model/tide/estimator.py

@@ -21,7 +21,6 @@
 from gluonts.dataset.field_names import FieldName
 from gluonts.dataset.loader import as_stacked_batches
 from gluonts.itertools import Cyclic
-from gluonts.model.forecast_generator import DistributionForecastGenerator
 from gluonts.time_feature import (
     minute_of_hour,
     hour_of_day,
@@ -49,10 +48,7 @@
 
 from gluonts.torch.model.estimator import PyTorchLightningEstimator
 from gluonts.torch.model.predictor import PyTorchPredictor
-from gluonts.torch.distributions import (
-    DistributionOutput,
-    StudentTOutput,
-)
+from gluonts.torch.distributions import Output, StudentTOutput
 
 from .lightning_module import TiDELightningModule
 
@@ -174,7 +170,7 @@ def __init__(
         weight_decay: float = 1e-8,
         patience: int = 10,
         scaling: Optional[str] = "mean",
-        distr_output: DistributionOutput = StudentTOutput(),
+        distr_output: Output = StudentTOutput(),
         batch_size: int = 32,
         num_batches_per_epoch: int = 50,
         trainer_kwargs: Optional[Dict[str, Any]] = None,
@@ -403,9 +399,7 @@ def create_predictor(
             input_transform=transformation + prediction_splitter,
             input_names=PREDICTION_INPUT_NAMES,
             prediction_net=module,
-            forecast_generator=DistributionForecastGenerator(
-                self.distr_output
-            ),
+            forecast_generator=self.distr_output.forecast_generator,
             batch_size=self.batch_size,
             prediction_length=self.prediction_length,
             device="auto",

src/gluonts/torch/model/tide/module.py

@@ -19,7 +19,7 @@
 from gluonts.core.component import validated
 from gluonts.torch.modules.feature import FeatureEmbedder
 from gluonts.model import Input, InputSpec
-from gluonts.torch.distributions import DistributionOutput
+from gluonts.torch.distributions import Output
 from gluonts.torch.scaler import StdScaler, MeanScaler, NOPScaler
 from gluonts.torch.model.simple_feedforward import make_linear_layer
 from gluonts.torch.util import weighted_average
@@ -242,7 +242,7 @@ def __init__(
         num_layers_encoder: int,
         num_layers_decoder: int,
         layer_norm: bool,
-        distr_output: DistributionOutput,
+        distr_output: Output,
         scaling: str,
     ) -> None:
         super().__init__()

test/torch/model/test_estimators.py

@@ -148,6 +148,14 @@
             num_batches_per_epoch=3,
             trainer_kwargs=dict(max_epochs=2),
         ),
+        lambda dataset: TiDEEstimator(
+            freq=dataset.metadata.freq,
+            prediction_length=dataset.metadata.prediction_length,
+            distr_output=QuantileOutput(quantiles=[0.1, 0.6, 0.85]),
+            batch_size=4,
+            num_batches_per_epoch=3,
+            trainer_kwargs=dict(max_epochs=2),
+        ),
         lambda dataset: WaveNetEstimator(
             freq=dataset.metadata.freq,
             prediction_length=dataset.metadata.prediction_length,