pyg-nightly 2.6.0.dev20240704__py3-none-any.whl → 2.8.0.dev20251207__py3-none-any.whl

torch_geometric/loader/prefetch.py

@@ -27,8 +27,9 @@ class DeviceHelper:
 
         if ((self.device.type == 'cuda' and not with_cuda)
                 or (self.device.type == 'xpu' and not with_xpu)):
-            warnings.warn(f"Requested device '{self.device.type}' is not "
-                          f"available, falling back to CPU")
+            warnings.warn(
+                f"Requested device '{self.device.type}' is not "
+                f"available, falling back to CPU", stacklevel=2)
             self.device = torch.device('cpu')
 
         self.stream = None
@@ -73,7 +74,7 @@ class PrefetchLoader:
         if isinstance(batch, dict):
             return {k: self.non_blocking_transfer(v) for k, v in batch.items()}
 
-        batch = batch.pin_memory(self.device_helper.device)
+        batch = batch.pin_memory()
         return batch.to(self.device_helper.device, non_blocking=True)
 
     def __iter__(self) -> Any:

torch_geometric/loader/temporal_dataloader.py

@@ -6,7 +6,7 @@ from torch_geometric.data import TemporalData
 
 
 class TemporalDataLoader(torch.utils.data.DataLoader):
-    r"""A data loader which merges succesive events of a
+    r"""A data loader which merges successive events of a
     :class:`torch_geometric.data.TemporalData` to a mini-batch.
 
     Args:
@@ -15,7 +15,7 @@ class TemporalDataLoader(torch.utils.data.DataLoader):
         batch_size (int, optional): How many samples per batch to load.
             (default: :obj:`1`)
         neg_sampling_ratio (float, optional): The ratio of sampled negative
-            destination nodes to the number of postive destination nodes.
+            destination nodes to the number of positive destination nodes.
             (default: :obj:`0.0`)
         **kwargs (optional): Additional arguments of
             :class:`torch.utils.data.DataLoader`.

torch_geometric/loader/utils.py

@@ -178,7 +178,7 @@ def filter_hetero_data(
     out = copy.copy(data)
 
     for node_type in out.node_types:
-        # Handle the case of disconneted graph sampling:
+        # Handle the case of disconnected graph sampling:
         if node_type not in node_dict:
             node_dict[node_type] = torch.empty(0, dtype=torch.long)
 
@@ -186,7 +186,7 @@ def filter_hetero_data(
                            node_dict[node_type])
 
     for edge_type in out.edge_types:
-        # Handle the case of disconneted graph sampling:
+        # Handle the case of disconnected graph sampling:
         if edge_type not in row_dict:
             row_dict[edge_type] = torch.empty(0, dtype=torch.long)
         if edge_type not in col_dict:
@@ -256,14 +256,6 @@ def filter_custom_hetero_store(
     # Construct a new `HeteroData` object:
     data = custom_cls() if custom_cls is not None else HeteroData()
 
-    # Filter edge storage:
-    # TODO support edge attributes
-    for attr in graph_store.get_all_edge_attrs():
-        key = attr.edge_type
-        if key in row_dict and key in col_dict:
-            edge_index = torch.stack([row_dict[key], col_dict[key]], dim=0)
-            data[attr.edge_type].edge_index = edge_index
-
     # Filter node storage:
     required_attrs = []
     for attr in feature_store.get_all_tensor_attrs():
@@ -280,6 +272,14 @@ def filter_custom_hetero_store(
     for i, attr in enumerate(required_attrs):
         data[attr.group_name][attr.attr_name] = tensors[i]
 
+    # Filter edge storage:
+    # TODO support edge attributes
+    for attr in graph_store.get_all_edge_attrs():
+        key = attr.edge_type
+        if key in row_dict and key in col_dict:
+            edge_index = torch.stack([row_dict[key], col_dict[key]], dim=0)
+            data[attr.edge_type].edge_index = edge_index
+
     return data
 
 

torch_geometric/metrics/__init__.py

@@ -1,14 +1,35 @@
 # flake8: noqa
 
-from .link_pred import (LinkPredPrecision, LinkPredRecall, LinkPredF1,
-                        LinkPredMAP, LinkPredNDCG)
+from .link_pred import (
+    LinkPredMetric,
+    LinkPredMetricCollection,
+    LinkPredPrecision,
+    LinkPredRecall,
+    LinkPredF1,
+    LinkPredMAP,
+    LinkPredNDCG,
+    LinkPredMRR,
+    LinkPredHitRatio,
+    LinkPredCoverage,
+    LinkPredDiversity,
+    LinkPredPersonalization,
+    LinkPredAveragePopularity,
+)
 
 link_pred_metrics = [
+    'LinkPredMetric',
+    'LinkPredMetricCollection',
     'LinkPredPrecision',
     'LinkPredRecall',
     'LinkPredF1',
     'LinkPredMAP',
     'LinkPredNDCG',
+    'LinkPredMRR',
+    'LinkPredHitRatio',
+    'LinkPredCoverage',
+    'LinkPredDiversity',
+    'LinkPredPersonalization',
+    'LinkPredAveragePopularity',
 ]
 
 __all__ = link_pred_metrics