exclude PLForecastingModule docs from custom rnn and blockrnn modules

darts/models/forecasting/block_rnn_model.py

@@ -65,16 +65,6 @@ def __init__(
             The fraction of neurons that are dropped in all-but-last RNN layers.
         **kwargs
             all parameters required for :class:`darts.model.forecasting_models.PLForecastingModule` base class.
-
-        Inputs
-        ------
-        x of shape `(batch_size, input_chunk_length, input_size, nr_params)`
-            Tensor containing the features of the input sequence.
-
-        Outputs
-        -------
-        y of shape `(batch_size, output_chunk_length, target_size, nr_params)`
-            Tensor containing the prediction at the last time step of the sequence.
         """
         super().__init__(**kwargs)
 
@@ -91,6 +81,21 @@ def __init__(
     @io_processor
     @abstractmethod
     def forward(self, x_in: Tuple) -> torch.Tensor:
+        """BlockRNN Module forward.
+
+        Parameters
+        ----------
+        x_in
+            Tuple of Tensors containing the features of the input sequence. The tuple has elements
+            (past target, historic future covariates, future covariates, static covariates).
+            The shape of the past target is `(batch_size, input_length, input_size)`.
+
+        Returns
+        -------
+        torch.Tensor
+            The BlockRNN output Tensor with shape `(batch_size, output_chunk_length, target_size, nr_params)`.
+            It contains the prediction at the last time step of the sequence.
+        """
         pass
 
 

darts/models/forecasting/pl_forecasting_module.py

@@ -3,6 +3,7 @@
 """
 
 from abc import ABC, abstractmethod
+from functools import wraps
 from typing import Any, Dict, Optional, Sequence, Tuple, Union
 
 import pytorch_lightning as pl
@@ -42,6 +43,7 @@ def forward(self, *args, **kwargs)
         normalizes batch input target features, and inverse transform the forward output back to the original scale
     """
 
+    @wraps(forward)
     def forward_wrapper(self, *args, **kwargs):
         if not self.use_reversible_instance_norm:
             return forward(self, *args, **kwargs)

darts/models/forecasting/rnn_model.py

@@ -63,18 +63,6 @@ def __init__(
             The fraction of neurons that are dropped in all-but-last RNN layers.
         **kwargs
             all parameters required for :class:`darts.model.forecasting_models.PLForecastingModule` base class.
-
-        Inputs
-        ------
-        x of shape `(batch_size, input_length, input_size)`
-            Tensor containing the features of the input sequence. The `input_length` is not fixed.
-
-        Outputs
-        -------
-        y of shape `(batch_size, output_chunk_length, target_size, nr_params)`
-            Tensor containing the outputs of the RNN at every time step of the input sequence.
-            During training the whole tensor is used as output, whereas during prediction we only use y[:, -1, :].
-            However, this module always returns the whole Tensor.
         """
         # RNNModule doesn't really need input and output_chunk_length for PLModule
         super().__init__(**kwargs)
@@ -92,6 +80,25 @@ def __init__(
     def forward(
         self, x_in: Tuple, h: Optional[torch.Tensor] = None
     ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """RNN Module forward.
+
+        Parameters
+        ----------
+        x_in
+            Tuple of Tensors containing the features of the input sequence. The tuple has elements (past target,
+            historic future covariates, future covariates, static covariates). The shape of the past target is
+            `(batch_size, input_length, input_size)`.
+        h
+            Optionally, the hidden state.
+
+        Returns
+        -------
+        Tuple[torch.Tensor, torch.Tensor]
+            Tuple of Tensors with elements (RNN output, hidden state). The RNN output Tensor has shape
+            `(batch_size, output_chunk_length, target_size, nr_params)`. It contains the outputs at every
+            time step of the input sequence. During training the whole tensor is used as output, whereas during
+            prediction we only use y[:, -1, :]. However, this module always returns the whole Tensor.
+        """
         pass
 
     def _produce_train_output(self, input_batch: Tuple) -> torch.Tensor:

docs/Makefile

@@ -7,6 +7,7 @@ SPHINXOPTS    ?=
 SPHINXBUILD   ?= sphinx-build
 SOURCEDIR     = source
 USERGUIDEDIR  = userguide
+GENAPIDIR     = generated_api
 BUILDDIR      = build
 
 # Put it first so that "make" without argument is like "make help".
@@ -50,6 +51,7 @@ generate-userguide:
 generate-api:
 	@echo "[Makefile] generating API using sphinx-apidoc..."
 	@sphinx-apidoc -e -f --templatedir templates -o "$(SOURCEDIR)/generated_api" ../darts ../darts/logging.py ../darts/tests/*
+	@cp -r $(GENAPIDIR)/*.rst "$(SOURCEDIR)/generated_api/"
 
 html:
 	# Note: this target has to be called "html" because its name is given as an argument to Sphinx

docs/generated_api/darts.models.forecasting.block_rnn_model.rst

@@ -0,0 +1,10 @@
+Block Recurrent Neural Networks
+-------------------------------
+
+.. autoclass:: darts.models.forecasting.block_rnn_model.BlockRNNModel
+   :members:
+   :exclude-members:
+
+.. autoclass:: darts.models.forecasting.block_rnn_model.CustomBlockRNNModule
+   :members:
+   :no-inherited-members:

docs/generated_api/darts.models.forecasting.rnn_model.rst

@@ -0,0 +1,10 @@
+Recurrent Neural Networks
+-------------------------
+
+.. autoclass:: darts.models.forecasting.rnn_model.RNNModel
+   :members:
+   :exclude-members:
+
+.. autoclass:: darts.models.forecasting.rnn_model.CustomRNNModule
+   :members:
+   :no-inherited-members:

docs/templates/module.rst_t

@@ -1,4 +1,3 @@
-
 .. automodule:: {{ qualname }}
    :members:
    :undoc-members: