[ruff] Fix first line doc-strings [5/n] (pyg-team#8334)

examples/contrib/pgm_explainer_graph_classification.py

@@ -1,5 +1,5 @@
-"""This is an example of using the PGM explainer algorithm
-on a graph classification task
+"""This is an example of using the PGM explainer algorithm on a graph
+classification task.
 """
 import os.path as osp
 

examples/contrib/pgm_explainer_node_classification.py

@@ -1,5 +1,5 @@
-"""This is an example of using the PGM explainer algorithm
-on a node classification task
+"""This is an example of using the PGM explainer algorithm on a node
+classification task.
 """
 import os.path as osp
 

examples/equilibrium_median.py

@@ -1,7 +1,6 @@
 r"""Replicates the experiment from `"Deep Graph Infomax"
-<https://arxiv.org/abs/1809.10341>`_ to try and teach
-`EquilibriumAggregation` to learn to take the median of
-a set of numbers
+<https://arxiv.org/abs/1809.10341>`_ to try and teach `EquilibriumAggregation`
+to learn to take the median of a set of numbers.
 
 This example converges slowly to being able to predict the
 median similar to what is observed in the paper.

examples/multi_gpu/multinode_multigpu_papers100m_gcn.py

@@ -1,4 +1,6 @@
-"""To run:
+"""Multi-node multi-GPU example on ogbn-papers100m.
+
+To run:
 srun -l -N<num_nodes> --ntasks-per-node=<ngpu_per_node> \
 --container-name=cont --container-image=<image_url> \
 --container-mounts=/ogb-papers100m/:/workspace/dataset

examples/randlanet_classification.py

@@ -1,7 +1,6 @@
 """An adaptation of RandLA-Net to the classification task, which was not
-addressed in the paper:
-RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds
-Reference: https://arxiv.org/abs/1911.11236
+addressed in the `"RandLA-Net: Efficient Semantic Segmentation of Large-Scale
+Point Clouds" <https://arxiv.org/abs/1911.11236>`_ paper.
 """
 import os.path as osp
 

examples/randlanet_segmentation.py

@@ -1,6 +1,6 @@
-"""An implementation of RandLA-Net based on the paper:
-RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds
-Reference: https://arxiv.org/abs/1911.11236
+"""An implementation of RandLA-Net based on the `"RandLA-Net: Efficient
+Semantic Segmentation of Large-Scale Point Clouds"
+<https://arxiv.org/abs/1911.11236>`_ paper.
 """
 import os.path as osp
 

graphgym/custom_graphgym/head/example.py

@@ -5,7 +5,7 @@
 
 @register_head('head')
 class ExampleNodeHead(nn.Module):
-    """Head of GNN, node prediction"""
+    r"""Head of GNN for node prediction."""
     def __init__(self, dim_in, dim_out):
         super().__init__()
         self.layer_post_mp = nn.Linear(dim_in, dim_out, bias=True)

graphgym/custom_graphgym/layer/example.py

@@ -15,9 +15,7 @@
 # take 'batch' as input and 'batch' as output
 @register_layer('exampleconv1')
 class ExampleConv1(MessagePassing):
-    r"""Example GNN layer
-
-    """
+    r"""Example GNN layer."""
     def __init__(self, in_channels, out_channels, bias=True, **kwargs):
         super().__init__(aggr=cfg.gnn.agg, **kwargs)
 
@@ -57,9 +55,7 @@ def update(self, aggr_out):
 # Example 2: First define a PyG format Conv layer
 # Then wrap it to become GraphGym format
 class ExampleConv2Layer(MessagePassing):
-    r"""Example GNN layer
-
-    """
+    r"""Example GNN layer."""
     def __init__(self, in_channels, out_channels, bias=True, **kwargs):
         super().__init__(aggr=cfg.gnn.agg, **kwargs)
 

graphgym/custom_graphgym/stage/example.py

@@ -12,7 +12,7 @@ def GNNLayer(dim_in, dim_out, has_act=True):
 
 @register_stage('example')
 class GNNStackStage(nn.Module):
-    """Simple Stage that stack GNN layers"""
+    r"""Simple stage that stacks GNN layers."""
     def __init__(self, dim_in, dim_out, num_layers):
         super().__init__()
         for i in range(num_layers):

pyproject.toml

@@ -131,7 +131,6 @@ ignore = [
     "D107",  # TODO: Don't ignore "Missing docstring in __init__"
     "D200",  # Ignore "One-line docstring should fit on one line"
     "D205",  # Ignore "1 blank line required between summary line and description"
-    "D415",  # Ignore "First line should end with a period, question mark, or exclamation point"
 
 ]
 src = ["torch_geometric"]

torch_geometric/config_store.py

@@ -185,7 +185,9 @@ def to_dataclass(
     strict: bool = False,
 ) -> Any:
     r"""Converts the input arguments of a given class :obj:`cls` to a
-    :obj:`dataclass` schema, *e.g.*,
+    :obj:`dataclass` schema.
+
+    For example,
 
     .. code-block:: python
 

torch_geometric/contrib/nn/models/rbcd_attack.py

@@ -320,7 +320,7 @@ def _forward_and_gradient(self, x: Tensor, labels: Tensor,
     def _get_modified_adj(self, edge_index: Tensor, edge_weight: Tensor,
                           block_edge_index: Tensor,
                           block_edge_weight: Tensor) -> Tuple[Tensor, Tensor]:
-        """Merges adjacency matrix with current block (incl. weights)"""
+        """Merges adjacency matrix with current block (incl. weights)."""
         if self.is_undirected:
             block_edge_index, block_edge_weight = to_undirected(
                 block_edge_index, block_edge_weight, num_nodes=self.num_nodes,
@@ -539,7 +539,7 @@ def _linear_to_full_idx(n: int, lin_idx: Tensor) -> Tensor:
     def _margin_loss(score: Tensor, labels: Tensor,
                      idx_mask: Optional[Tensor] = None,
                      reduce: Optional[str] = None) -> Tensor:
-        r"""Margin loss between true score and highest non-target score:
+        r"""Margin loss between true score and highest non-target score.
 
         .. math::
             m = - s_{y} + max_{y' \ne y} s_{y'}

torch_geometric/data/database.py

@@ -36,6 +36,7 @@ def maybe_cast_to_tensor_info(value: Any) -> Union[Any, TensorInfo]:
 
 class Database(ABC):
     r"""Base class for inserting and retrieving data from a database.
+
     A database acts as a persisted, out-of-memory and index-based key/value
     store for tensor and custom data:
 

torch_geometric/data/feature_store.py

@@ -124,6 +124,7 @@ def __init__(self, store: 'FeatureStore', attr: TensorAttr):
     def __getattr__(self, key: Any) -> Union['AttrView', FeatureTensorType]:
         r"""Sets the first unset field of the backing :class:`TensorAttr`
         object to the attribute.
+
         This allows for :class:`AttrView` to be indexed by different values of
         attributes, in order.
         In particular, for a feature store that we want to index by
@@ -159,6 +160,7 @@ def __getattr__(self, key: Any) -> Union['AttrView', FeatureTensorType]:
     def __getitem__(self, key: Any) -> Union['AttrView', FeatureTensorType]:
         r"""Sets the first unset field of the backing :class:`TensorAttr`
         object to the attribute via indexing.
+
         This allows for :class:`AttrView` to be indexed by different values of
         attributes, in order.
         In particular, for a feature store that we want to index by
@@ -176,9 +178,12 @@ def __getitem__(self, key: Any) -> Union['AttrView', FeatureTensorType]:
 
     def __setattr__(self, key: str, value: Any):
         r"""Supports attribute assignment to the backing :class:`TensorAttr` of
-        an :class:`AttrView`. This allows for :class:`AttrView` objects to set
-        their backing attribute values. In particular, the following operation
-        sets the :obj:`index` of an :class:`AttrView`:
+        an :class:`AttrView`.
+
+        This allows for :class:`AttrView` objects to set their backing
+        attribute values.
+        In particular, the following operation sets the :obj:`index` of an
+        :class:`AttrView`:
 
         .. code-block:: python
 
@@ -194,9 +199,12 @@ def __setattr__(self, key: str, value: Any):
 
     def __setitem__(self, key: str, value: Any):
         r"""Supports attribute assignment to the backing :class:`TensorAttr` of
-        an :class:`AttrView` via indexing. This allows for :class:`AttrView`
-        objects to set their backing attribute values. In particular, the
-        following operation sets the `index` of an :class:`AttrView`:
+        an :class:`AttrView` via indexing.
+
+        This allows for :class:`AttrView` objects to set their backing
+        attribute values.
+        In particular, the following operation sets the `index` of an
+        :class:`AttrView`:
 
         .. code-block:: python
 
@@ -209,9 +217,11 @@ def __setitem__(self, key: str, value: Any):
 
     def __call__(self) -> FeatureTensorType:
         r"""Supports :class:`AttrView` as a callable to force retrieval from
-        the currently specified attributes. In particular, this passes the
-        current :class:`TensorAttr` object to a GET call, regardless of whether
-        all attributes have been specified. It returns the result of this call.
+        the currently specified attributes.
+
+        In particular, this passes the current :class:`TensorAttr` object to a
+        GET call, regardless of whether all attributes have been specified.
+        It returns the result of this call.
         In particular, the following operation returns a tensor by performing a
         GET operation on the backing feature store:
 
@@ -487,6 +497,7 @@ def __setitem__(self, key: TensorAttr, value: FeatureTensorType):
 
     def __getitem__(self, key: TensorAttr) -> Any:
         r"""Supports pythonic indexing into the :class:`FeatureStore`.
+
         In particular, the following rules are followed for indexing:
 
         * A fully-specified :obj:`key` will produce a tensor output.

torch_geometric/data/hypergraph_data.py

@@ -13,6 +13,7 @@
 
 class HyperGraphData(Data):
     r"""A data object describing a hypergraph.
+
     The data object can hold node-level, link-level and graph-level attributes.
     This object differs from a standard :obj:`~torch_geometric.data.Data`
     object by having hyperedges, i.e. edges that connect more

torch_geometric/data/storage.py

@@ -288,6 +288,7 @@ def record_stream(self, stream: torch.cuda.Stream, *args: str):
 
 
 class NodeStorage(BaseStorage):
+    r"""A storage for node-level information."""
     @property
     def _key(self) -> NodeType:
         key = self.__dict__.get('_key', None)
@@ -412,7 +413,9 @@ def node_attrs(self) -> List[str]:
 
 
 class EdgeStorage(BaseStorage):
-    r"""We support multiple ways to store edge connectivity in a
+    r"""A storage for edge-level information.
+
+    We support multiple ways to store edge connectivity in a
     :class:`EdgeStorage` object:
 
     * :obj:`edge_index`: A :class:`torch.LongTensor` holding edge indices in
@@ -624,6 +627,7 @@ def is_bipartite(self) -> bool:
 
 
 class GlobalStorage(NodeStorage, EdgeStorage):
+    r"""A storage for both node-level and edge-level information."""
     @property
     def _key(self) -> Any:
         return None

torch_geometric/datasets/ba_multi_shapes.py

@@ -12,6 +12,7 @@ class BAMultiShapesDataset(InMemoryDataset):
     evaluating explainabilty algorithms, as described in the
     `"Global Explainability of GNNs via Logic Combination of Learned Concepts"
     <https://arxiv.org/abs/2210.07147>`_ paper.
+
     Given three atomic motifs, namely House (H), Wheel (W), and Grid (G),
     :class:`~torch_geometric.datasets.BAMultiShapesDataset` contains 1,000
     graphs where each graph is obtained by attaching the motifs to a random

torch_geometric/datasets/ged_dataset.py

@@ -19,6 +19,7 @@
 class GEDDataset(InMemoryDataset):
     r"""The GED datasets from the `"Graph Edit Distance Computation via Graph
     Neural Networks" <https://arxiv.org/abs/1808.05689>`_ paper.
+
     GEDs can be accessed via the global attributes :obj:`ged` and
     :obj:`norm_ged` for all train/train graph pairs and all train/test graph
     pairs:

torch_geometric/distributed/partition.py

@@ -15,6 +15,7 @@ class Partitioner:
     r"""Partition the graph structure and its features of a
     :class:`~torch_geometric.data.Data` or
     :class:`~torch_geometric.data.HeteroData` object.
+
     Partitioned data output will be structured like this:
 
     **Homogeneous graphs:**

torch_geometric/distributed/utils.py

@@ -16,7 +16,7 @@ class NodeDict:
     r"""Class used during heterogeneous sampling:
     1) The nodes to serve as source nodes in the next layer
     2) The nodes with duplicates that are further needed to create COO output
-    3) The output nodes without duplicates
+    3) The output nodes without duplicates.
     """
     src: Dict[NodeType, Tensor] = field(default_factory=dict)
     with_dupl: Dict[NodeType, Tensor] = field(default_factory=dict)
@@ -30,7 +30,7 @@ class BatchDict:
        layer
     2) The subgraph IDs with duplicates that are further needed to create COO
        output
-    3) The output subgraph IDs without duplicates
+    3) The output subgraph IDs without duplicates.
     """
     src: Dict[NodeType, Tensor] = field(default_factory=dict)
     with_duple: Dict[NodeType, Tensor] = field(default_factory=dict)

torch_geometric/explain/algorithm/pg_explainer.py

@@ -21,6 +21,7 @@
 class PGExplainer(ExplainerAlgorithm):
     r"""The PGExplainer model from the `"Parameterized Explainer for Graph
     Neural Network" <https://arxiv.org/abs/2011.04573>`_ paper.
+
     Internally, it utilizes a neural network to identify subgraph structures
     that play a crucial role in the predictions made by a GNN.
     Importantly, the :class:`PGExplainer` needs to be trained via

torch_geometric/explain/metric/fidelity.py

@@ -108,7 +108,7 @@ def characterization_score(
 ) -> Tensor:
     r"""Returns the componentwise characterization score as described in the
     `"GraphFramEx: Towards Systematic Evaluation of Explainability Methods for
-    Graph Neural Networks" <https://arxiv.org/abs/2206.09677>`_ paper:
+    Graph Neural Networks" <https://arxiv.org/abs/2206.09677>`_ paper.
 
     ..  math::
        \textrm{charact} = \frac{w_{+} + w_{-}}{\frac{w_{+}}{\textrm{fid}_{+}} +

torch_geometric/graphgym/config.py

@@ -22,12 +22,14 @@
 
 def set_cfg(cfg):
     r"""This function sets the default config value.
+
     1) Note that for an experiment, only part of the arguments will be used
-    The remaining unused arguments won't affect anything.
-    So feel free to register any argument in graphgym.contrib.config
-    2) We support *at most* two levels of configs, e.g., cfg.dataset.name
+       The remaining unused arguments won't affect anything.
+       So feel free to register any argument in graphgym.contrib.config
+    2) We support *at most* two levels of configs, *e.g.*,
+       :obj:`cfg.dataset.name`.
 
-    :return: configuration use by the experiment.
+    :return: Configuration use by the experiment.
     """
     if cfg is None:
         return cfg
@@ -449,7 +451,7 @@ def set_cfg(cfg):
 
 
 def assert_cfg(cfg):
-    r"""Checks config values, do necessary post processing to the configs"""
+    r"""Checks config values, do necessary post processing to the configs."""
     if cfg.dataset.task not in ['node', 'edge', 'graph', 'link_pred']:
         raise ValueError(f"Task '{cfg.dataset.task}' not supported. Must be "
                          f"one of node, edge, graph, link_pred")
@@ -476,7 +478,7 @@ def assert_cfg(cfg):
 
 def dump_cfg(cfg):
     r"""Dumps the config to the output directory specified in
-    :obj:`cfg.out_dir`
+    :obj:`cfg.out_dir`.
 
     Args:
         cfg (CfgNode): Configuration node
@@ -488,7 +490,7 @@ def dump_cfg(cfg):
 
 
 def load_cfg(cfg, args):
-    r"""Load configurations from file system and command line
+    r"""Load configurations from file system and command line.
 
     Args:
         cfg (CfgNode): Configuration node
@@ -506,7 +508,7 @@ def makedirs_rm_exist(dir):
 
 
 def get_fname(fname):
-    r"""Extract filename from file name path
+    r"""Extract filename from file name path.
 
     Args:
         fname (str): Filename for the yaml format configuration file
@@ -520,7 +522,7 @@ def get_fname(fname):
 
 
 def set_out_dir(out_dir, fname):
-    r"""Create the directory for full experiment run
+    r"""Create the directory for full experiment run.
 
     Args:
         out_dir (str): Directory for output, specified in :obj:`cfg.out_dir`
@@ -536,7 +538,7 @@ def set_out_dir(out_dir, fname):
 
 
 def set_run_dir(out_dir):
-    r"""Create the directory for each random seed experiment run
+    r"""Create the directory for each random seed experiment run.
 
     Args:
         out_dir (str): Directory for output, specified in :obj:`cfg.out_dir`

torch_geometric/graphgym/contrib/layer/generalconv.py

@@ -112,7 +112,7 @@ def __repr__(self):
 
 
 class GeneralEdgeConvLayer(MessagePassing):
-    r"""General GNN layer, with edge features
+    r"""General GNN layer, with edge features.
     """
     def __init__(self, in_channels, out_channels, edge_dim, improved=False,
                  cached=False, bias=True, **kwargs):

torch_geometric/graphgym/init.py

@@ -2,7 +2,7 @@
 
 
 def init_weights(m):
-    r"""Performs weight initialization
+    r"""Performs weight initialization.
 
     Args:
         m (nn.Module): PyTorch module