[Graphbolt] Change the temporal filter condition of temporal sampler. (#6893)

graphbolt/include/graphbolt/fused_csc_sampling_graph.h

@@ -325,7 +325,7 @@ class FusedCSCSamplingGraph : public torch::CustomClassHolder {
    * @brief Sample neighboring edges of the given nodes with a temporal
    * constraint. If `node_timestamp_attr_name` or `edge_timestamp_attr_name` is
    * given, the sampled neighbors or edges of an input node must have a
-   * timestamp that is no later than that of the input node.
+   * timestamp that is smaller than that of the input node.
    *
    * @param nodes The nodes from which to sample neighbors.
    * @param input_nodes_timestamp The timestamp of the nodes.

graphbolt/src/fused_csc_sampling_graph.cc

@@ -784,10 +784,10 @@ torch::Tensor TemporalMask(
   if (node_timestamp.has_value()) {
     auto neighbor_timestamp =
         node_timestamp.value().index_select(0, csc_indices.slice(0, l, r));
-    mask &= neighbor_timestamp <= seed_timestamp;
+    mask &= neighbor_timestamp < seed_timestamp;
   }
   if (edge_timestamp.has_value()) {
-    mask &= edge_timestamp.value().slice(0, l, r) <= seed_timestamp;
+    mask &= edge_timestamp.value().slice(0, l, r) < seed_timestamp;
   }
   if (probs_or_mask.has_value()) {
     mask &= probs_or_mask.value().slice(0, l, r) != 0;

python/dgl/graphbolt/impl/fused_csc_sampling_graph.py

@@ -773,8 +773,8 @@ class FusedCSCSamplingGraph(SamplingGraph):
         subgraph.
 
         If `node_timestamp_attr_name` or `edge_timestamp_attr_name` is given,
-        the sampled neighbors or edges of an input node must have a timestamp
-        that is no later than that of the input node.
+        the sampled neighbor or edge of an input node must have a timestamp
+        that is smaller than that of the input node.
 
         Parameters
         ----------

tests/python/pytorch/graphbolt/impl/test_fused_csc_sampling_graph.py

@@ -898,12 +898,12 @@ def test_temporal_sample_neighbors_homo(
                 neighbor = indices[j].item()
                 if (
                     use_node_timestamp
-                    and (node_timestamp[neighbor] > seed_timestamp[i]).item()
+                    and (node_timestamp[neighbor] >= seed_timestamp[i]).item()
                 ):
                     continue
                 if (
                     use_edge_timestamp
-                    and (edge_timestamp[j] > seed_timestamp[i]).item()
+                    and (edge_timestamp[j] >= seed_timestamp[i]).item()
                 ):
                     continue
                 neighbors.append(neighbor)
@@ -1035,13 +1035,13 @@ def test_temporal_sample_neighbors_hetero(
                 if (
                     use_node_timestamp
                     and (
-                        node_timestamp[neighbor] > homo_seed_timestamp[i]
+                        node_timestamp[neighbor] >= homo_seed_timestamp[i]
                     ).item()
                 ):
                     continue
                 if (
                     use_edge_timestamp
-                    and (edge_timestamp[j] > homo_seed_timestamp[i]).item()
+                    and (edge_timestamp[j] >= homo_seed_timestamp[i]).item()
                 ):
                     continue
                 neighbors.append(neighbor)

tests/python/pytorch/graphbolt/test_subgraph_sampler.py

@@ -525,7 +525,7 @@ def test_SubgraphSampler_without_dedpulication_Homo(sampler_type):
         graph.edge_attributes = {
             "timestamp": torch.zeros(graph.indices.numel()).to(F.ctx())
         }
-        items = (items, torch.randint(0, 10, (3,)))
+        items = (items, torch.randint(1, 10, (3,)))
         names = (names, "timestamp")
 
     itemset = gb.ItemSet(items, names=names)
@@ -583,7 +583,7 @@ def test_SubgraphSampler_without_dedpulication_Hetero(sampler_type):
         graph.edge_attributes = {
             "timestamp": torch.zeros(graph.indices.numel()).to(F.ctx())
         }
-        items = (items, torch.randint(0, 10, (2,)))
+        items = (items, torch.randint(1, 10, (2,)))
         names = (names, "timestamp")
     itemset = gb.ItemSetDict({"n2": gb.ItemSet(items, names=names)})
     item_sampler = gb.ItemSampler(itemset, batch_size=2).copy_to(F.ctx())