Remove mandatory freq attribute of Predictor. (#2017)

* Remove freq from Predictor

Co-authored-by: Jasper Zschiegner <schjaspe@amazon.de>

Author: kashif

docs/tutorials/advanced_topics/howto_pytorch_lightning.md

@@ -71,7 +71,6 @@ def mean_abs_scaling(context, min_scale=1e-5):
 class FeedForwardNetwork(nn.Module):
     def __init__(
         self,
-        freq: str,
         prediction_length: int,
         context_length: int,
         hidden_dimensions: List[int],
@@ -85,7 +84,6 @@ class FeedForwardNetwork(nn.Module):
         assert context_length > 0
         assert len(hidden_dimensions) > 0
 
-        self.freq = freq
         self.prediction_length = prediction_length
         self.context_length = context_length
         self.hidden_dimensions = hidden_dimensions
@@ -123,7 +121,6 @@ class FeedForwardNetwork(nn.Module):
     def get_predictor(self, input_transform, batch_size=32, device=None):
         return PyTorchPredictor(
             prediction_length=self.prediction_length,
-            freq=self.freq, 
             input_names=["past_target"],
             prediction_net=self,
             batch_size=batch_size,
@@ -168,7 +165,6 @@ We can now instantiate the training network, and explore its set of parameters.
 
 
 ```python
-freq = "1H"
 context_length = 2 * 7 * 24
 prediction_length = dataset.metadata.prediction_length
 hidden_dimensions = [96, 48]
@@ -177,7 +173,6 @@ hidden_dimensions = [96, 48]
 
 ```python
 net = LightningFeedForwardNetwork(
-    freq=freq,
     prediction_length=prediction_length,
     context_length=context_length,
     hidden_dimensions=hidden_dimensions,

docs/tutorials/advanced_topics/trainer_callbacks.md

@@ -26,7 +26,7 @@ from gluonts.mx.trainer.callback import TrainingHistory
 history = TrainingHistory()
 
 trainer = Trainer(epochs=3, callbacks=[history])
-estimator = SimpleFeedForwardEstimator(prediction_length=prediction_length, freq=freq, trainer=trainer)
+estimator = SimpleFeedForwardEstimator(prediction_length=prediction_length, trainer=trainer)
 
 predictor = estimator.train(dataset.train, num_workers=None)
 ```
@@ -49,7 +49,7 @@ warm_start = WarmStart(predictor=predictor)
 
 trainer=Trainer(epochs=3, callbacks=[history, warm_start])
 
-estimator = SimpleFeedForwardEstimator(prediction_length=prediction_length, freq=freq, trainer=trainer)
+estimator = SimpleFeedForwardEstimator(prediction_length=prediction_length, trainer=trainer)
 
 predictor = estimator.train(dataset.train, num_workers=None)
 ```
@@ -222,7 +222,7 @@ Note that we're running an extremely short number of epochs, simply to keep the
 feel free to increase the number of epochs to properly test the effectiveness of the callback.
 
 ```python
-estimator = SimpleFeedForwardEstimator(prediction_length=prediction_length, freq=freq)
+estimator = SimpleFeedForwardEstimator(prediction_length=prediction_length)
 training_network = estimator.create_training_network()
 transformation = estimator.create_transformation()
 

docs/tutorials/data_manipulation/synthetic_data_generation.md

@@ -128,7 +128,7 @@ print_dicts(
 )
 ```
 
-Note that you may create and use intermediate random varibles such as `stddev2` in the above example without including them in the output.
+Note that you may create and use intermediate random variables such as `stddev2` in the above example without including them in the output.
 
 
 ```python

docs/tutorials/forecasting/extended_tutorial.md

@@ -563,7 +563,7 @@ GluonTS comes with a number of pre-built models. All the user needs to do is con
 
 ### Configuring an estimator
 
-We will begin with GulonTS's pre-built feedforward neural network estimator, a simple but powerful forecasting model. We will use this model to demonstrate the process of training a model, producing forecasts, and evaluating the results.
+We will begin with GluonTS's pre-built feedforward neural network estimator, a simple but powerful forecasting model. We will use this model to demonstrate the process of training a model, producing forecasts, and evaluating the results.
 
 GluonTS's built-in feedforward neural network (`SimpleFeedForwardEstimator`) accepts an input window of length `context_length` and predicts the distribution of the values of the subsequent `prediction_length` values. In GluonTS parlance, the feedforward neural network model is an example of `Estimator`. In GluonTS, `Estimator` objects represent a forecasting model as well as details such as its coefficients, weights, etc.
 
@@ -583,7 +583,6 @@ estimator = SimpleFeedForwardEstimator(
     num_hidden_dimensions=[10],
     prediction_length=custom_ds_metadata['prediction_length'],
     context_length=2*custom_ds_metadata['prediction_length'],
-    freq=custom_ds_metadata['freq'],
     trainer=Trainer(
         ctx="cpu", 
         epochs=5, 
@@ -788,7 +787,7 @@ The estimator should also include the following methods:
 - `create_transformation`, defining all the data pre-processing transformations (like adding features)
 - `create_training_data_loader`, constructing the data loader that gives batches to be used in training, out of a given dataset
 - `create_training_network`, returning the training network configured with any necessary hyperparameters
-- `create_predictor`, creting the prediction network, and returning a `Predictor` object 
+- `create_predictor`, creating the prediction network, and returning a `Predictor` object 
 
 If a validation dataset is to be accepted, for some validation metric to be computed, then also the following should be defined:
 
@@ -871,15 +870,13 @@ class MyEstimator(GluonEstimator):
         self,
         prediction_length: int,
         context_length: int,
-        freq: str,
         num_cells: int,
         batch_size: int = 32,
         trainer: Trainer = Trainer()
     ) -> None:
         super().__init__(trainer=trainer, batch_size=batch_size)
         self.prediction_length = prediction_length
         self.context_length = context_length
-        self.freq = freq
         self.num_cells = num_cells
 
     def create_transformation(self):
@@ -937,7 +934,6 @@ class MyEstimator(GluonEstimator):
             input_transform=transformation + prediction_splitter,
             prediction_net=prediction_network,
             batch_size=self.trainer.batch_size,
-            freq=self.freq,
             prediction_length=self.prediction_length,
             ctx=self.trainer.ctx,
         )
@@ -950,7 +946,6 @@ After defining the training and prediction network, as well as the estimator cla
 estimator = MyEstimator(
     prediction_length=custom_ds_metadata['prediction_length'],
     context_length=2*custom_ds_metadata['prediction_length'],
-    freq=custom_ds_metadata['freq'],
     num_cells=40,
     trainer=Trainer(
         ctx="cpu",
@@ -1104,7 +1099,6 @@ class MyProbEstimator(GluonEstimator):
         self,
         prediction_length: int,
         context_length: int,
-        freq: str,
         distr_output: DistributionOutput,
         num_cells: int,
         num_sample_paths: int = 100,
@@ -1114,7 +1108,6 @@ class MyProbEstimator(GluonEstimator):
         super().__init__(trainer=trainer, batch_size=batch_size)
         self.prediction_length = prediction_length
         self.context_length = context_length
-        self.freq = freq
         self.distr_output = distr_output
         self.num_cells = num_cells
         self.num_sample_paths = num_sample_paths
@@ -1178,7 +1171,6 @@ class MyProbEstimator(GluonEstimator):
             input_transform=transformation + prediction_splitter,
             prediction_net=prediction_network,
             batch_size=self.trainer.batch_size,
-            freq=self.freq,
             prediction_length=self.prediction_length,
             ctx=self.trainer.ctx,
         )
@@ -1189,7 +1181,6 @@ class MyProbEstimator(GluonEstimator):
 estimator = MyProbEstimator(
     prediction_length=custom_ds_metadata['prediction_length'],
     context_length=2*custom_ds_metadata['prediction_length'],
-    freq=custom_ds_metadata['freq'],
     distr_output=GaussianOutput(),
     num_cells=40,
     trainer=Trainer(
@@ -1369,7 +1360,6 @@ class MyProbEstimator(GluonEstimator):
         self,
         prediction_length: int,
         context_length: int,
-        freq: str,
         distr_output: DistributionOutput,
         num_cells: int,
         num_sample_paths: int = 100,
@@ -1380,7 +1370,6 @@ class MyProbEstimator(GluonEstimator):
         super().__init__(trainer=trainer, batch_size=batch_size)
         self.prediction_length = prediction_length
         self.context_length = context_length
-        self.freq = freq
         self.distr_output = distr_output
         self.num_cells = num_cells
         self.num_sample_paths = num_sample_paths
@@ -1461,7 +1450,6 @@ class MyProbEstimator(GluonEstimator):
             input_transform=transformation + prediction_splitter,
             prediction_net=prediction_network,
             batch_size=self.trainer.batch_size,
-            freq=self.freq,
             prediction_length=self.prediction_length,
             ctx=self.trainer.ctx,
         )
@@ -1472,7 +1460,6 @@ class MyProbEstimator(GluonEstimator):
 estimator = MyProbEstimator(
     prediction_length=custom_ds_metadata['prediction_length'],
     context_length=2*custom_ds_metadata['prediction_length'],
-    freq=custom_ds_metadata['freq'],
     distr_output=GaussianOutput(),
     num_cells=40,
     trainer=Trainer(
@@ -1747,7 +1734,6 @@ class MyProbRNNEstimator(GluonEstimator):
         self,
         prediction_length: int,
         context_length: int,
-        freq: str,
         distr_output: DistributionOutput,
         num_cells: int,
         num_layers: int,
@@ -1759,7 +1745,6 @@ class MyProbRNNEstimator(GluonEstimator):
         super().__init__(trainer=trainer, batch_size=batch_size)
         self.prediction_length = prediction_length
         self.context_length = context_length
-        self.freq = freq
         self.distr_output = distr_output
         self.num_cells = num_cells
         self.num_layers = num_layers
@@ -1842,7 +1827,6 @@ class MyProbRNNEstimator(GluonEstimator):
             input_transform=transformation + prediction_splitter,
             prediction_net=prediction_network,
             batch_size=self.trainer.batch_size,
-            freq=self.freq,
             prediction_length=self.prediction_length,
             ctx=self.trainer.ctx,
         )
@@ -1853,7 +1837,6 @@ class MyProbRNNEstimator(GluonEstimator):
 estimator = MyProbRNNEstimator(
     prediction_length=24,
     context_length=48,
-    freq="1H",
     num_cells=40,
     num_layers=2,
     distr_output=GaussianOutput(),

docs/tutorials/forecasting/quick_start_tutorial.md

@@ -120,7 +120,7 @@ test_ds = ListDataset(
 
 GluonTS comes with a number of pre-built models. All the user needs to do is configure some hyperparameters. The existing models focus on (but are not limited to) probabilistic forecasting. Probabilistic forecasts are predictions in the form of a probability distribution, rather than simply a single point estimate.
 
-We will begin with GulonTS's pre-built feedforward neural network estimator, a simple but powerful forecasting model. We will use this model to demonstrate the process of training a model, producing forecasts, and evaluating the results.
+We will begin with GluonTS's pre-built feedforward neural network estimator, a simple but powerful forecasting model. We will use this model to demonstrate the process of training a model, producing forecasts, and evaluating the results.
 
 GluonTS's built-in feedforward neural network (`SimpleFeedForwardEstimator`) accepts an input window of length `context_length` and predicts the distribution of the values of the subsequent `prediction_length` values. In GluonTS parlance, the feedforward neural network model is an example of `Estimator`. In GluonTS, `Estimator` objects represent a forecasting model as well as details such as its coefficients, weights, etc.
 
@@ -140,7 +140,6 @@ estimator = SimpleFeedForwardEstimator(
     num_hidden_dimensions=[10],
     prediction_length=dataset.metadata.prediction_length,
     context_length=100,
-    freq=dataset.metadata.freq,
     trainer=Trainer(
         ctx="cpu", 
         epochs=5, 

examples/GluonTS_SageMaker_SDK_Tutorial.ipynb

@@ -406,7 +406,6 @@
    "source": [
     "my_estimator = SimpleFeedForwardEstimator(\n",
     "                    prediction_length=prediction_length,\n",
-    "                    freq=freq,\n",
     "                    trainer=Trainer(ctx=general_instance_type, epochs=5) # optional\n",
     "                )"
    ]

examples/evaluate_model.py

@@ -31,7 +31,6 @@
 
     estimator = SimpleFeedForwardEstimator(
         prediction_length=dataset.metadata.prediction_length,
-        freq=dataset.metadata.freq,
         trainer=Trainer(epochs=5, num_batches_per_epoch=10),
     )
 

examples/persist_model.py

@@ -29,7 +29,6 @@
 
     estimator = SimpleFeedForwardEstimator(
         prediction_length=dataset.metadata.prediction_length,
-        freq=dataset.metadata.freq,
         trainer=Trainer(epochs=5, num_batches_per_epoch=10),
     )
 

examples/run_rolling_forecast_backtest.py

@@ -28,7 +28,6 @@
 
     estimator = SimpleFeedForwardEstimator(
         prediction_length=dataset.metadata.prediction_length,
-        freq=dataset.metadata.freq,
         trainer=Trainer(epochs=5, num_batches_per_epoch=10),
     )
 

examples/warm_start.py

@@ -31,7 +31,6 @@
     # First train a model
     estimator = SimpleFeedForwardEstimator(
         prediction_length=dataset.metadata.prediction_length,
-        freq=dataset.metadata.freq,
         trainer=Trainer(epochs=10, num_batches_per_epoch=10),
     )
 
@@ -46,7 +45,6 @@ def copy_params(net):
 
     estimator = SimpleFeedForwardEstimator(
         prediction_length=dataset.metadata.prediction_length,
-        freq=dataset.metadata.freq,
         trainer=Trainer(
             epochs=5, num_batches_per_epoch=10, post_initialize_cb=copy_params
         ),