Merge branch 'pyg-team:master' into master

Author: LeonResearch

docs/source/tutorial/graph_transformer.rst

@@ -3,7 +3,7 @@ Graph Transformer
 
 `Transformer <https://arxiv.org/abs/1706.03762>`_ is an effictive architecture in `natural language processing <https://arxiv.org/abs/1810.04805>`_ and `computer vision <https://arxiv.org/abs/2010.11929>`_.
 Recently, there have been some applications(`Grover <https://arxiv.org/abs/2007.02835>`_, `GraphGPS <https://arxiv.org/abs/2205.12454>`_, etc) that combine transformers on graphs.
-In this tutorial, we will present how to build a graph transformer model via :pyg:`PyG`.
+In this tutorial, we will present how to build a graph transformer model via :pyg:`PyG`. See `<our webinar https://youtu.be/wAYryx3GjLw?si=2vB7imfenP5tUvqd>` for in-depth learning on this topic.
 
 .. note::
     Click `here <https://github.com/pyg-team/pytorch_geometric/blob/master/examples/graph_gps.py>`_ to download the full example code

torch_geometric/metrics/link_pred.py

@@ -221,8 +221,8 @@ def __init__(self, k: int) -> None:
             self.add_state('accum', torch.tensor(0.), dist_reduce_fx='sum')
             self.add_state('total', torch.tensor(0), dist_reduce_fx='sum')
         else:
-            self.register_buffer('accum', torch.tensor(0.))
-            self.register_buffer('total', torch.tensor(0))
+            self.register_buffer('accum', torch.tensor(0.), persistent=False)
+            self.register_buffer('total', torch.tensor(0), persistent=False)
 
     def update(
         self,
@@ -523,10 +523,11 @@ def __init__(self, k: int, weighted: bool = False):
         discount = torch.arange(2, k + 2, dtype=dtype).log2()
 
         self.discount: Tensor
-        self.register_buffer('discount', discount)
+        self.register_buffer('discount', discount, persistent=False)
 
         if not weighted:
-            self.register_buffer('idcg', cumsum(1.0 / discount))
+            self.register_buffer('idcg', cumsum(1.0 / discount),
+                                 persistent=False)
         else:
             self.idcg = None
 
@@ -617,7 +618,7 @@ def __init__(self, k: int, num_dst_nodes: int) -> None:
         if WITH_TORCHMETRICS:
             self.add_state('mask', mask, dist_reduce_fx='max')
         else:
-            self.register_buffer('mask', mask)
+            self.register_buffer('mask', mask, persistent=False)
 
     def update(
         self,
@@ -673,11 +674,11 @@ def __init__(self, k: int, category: Tensor) -> None:
             self.add_state('accum', torch.tensor(0.), dist_reduce_fx='sum')
             self.add_state('total', torch.tensor(0), dist_reduce_fx='sum')
         else:
-            self.register_buffer('accum', torch.tensor(0.))
-            self.register_buffer('total', torch.tensor(0))
+            self.register_buffer('accum', torch.tensor(0.), persistent=False)
+            self.register_buffer('total', torch.tensor(0), persistent=False)
 
         self.category: Tensor
-        self.register_buffer('category', category)
+        self.register_buffer('category', category, persistent=False)
 
     def update(
         self,
@@ -740,7 +741,7 @@ def __init__(
             self.add_state('total', torch.tensor(0), dist_reduce_fx='sum')
         else:
             self.preds: List[Tensor] = []
-            self.register_buffer('total', torch.tensor(0))
+            self.register_buffer('total', torch.tensor(0), persistent=False)
 
     def update(
         self,
@@ -829,11 +830,11 @@ def __init__(self, k: int, popularity: Tensor) -> None:
             self.add_state('accum', torch.tensor(0.), dist_reduce_fx='sum')
             self.add_state('total', torch.tensor(0), dist_reduce_fx='sum')
         else:
-            self.register_buffer('accum', torch.tensor(0.))
-            self.register_buffer('total', torch.tensor(0))
+            self.register_buffer('accum', torch.tensor(0.), persistent=False)
+            self.register_buffer('total', torch.tensor(0), persistent=False)
 
         self.popularity: Tensor
-        self.register_buffer('popularity', popularity)
+        self.register_buffer('popularity', popularity, persistent=False)
 
     def update(
         self,