[GraphBolt] remove num_workers=0 from doc (#6803)

docs/source/guide/minibatch-custom-sampler.rst

@@ -48,7 +48,7 @@ To use this sampler with :class:`~dgl.graphbolt.DataLoader`:
     datapipe = datapipe.customized_sample_neighbor(g, [10, 10]) # 2 layers.
     datapipe = datapipe.fetch_feature(feature, node_feature_keys=["feat"])
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
     for data in dataloader:
         input_features = data.node_features["feat"]
@@ -93,7 +93,7 @@ can be used on heterogeneous graphs:
         feature, node_feature_keys={"user": ["feat"], "item": ["feat"]}
     )
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
     for data in dataloader:
         input_features = {

docs/source/guide/minibatch-edge.rst

@@ -39,7 +39,7 @@ edges(namely, node pairs) in the training set instead of the nodes.
     # datapipe = gb.NeighborSampler(datapipe, g, [10, 10])
     datapipe = datapipe.fetch_feature(feature, node_feature_keys=["feat"])
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
 Iterating over the DataLoader will yield :class:`~dgl.graphbolt.MiniBatch`
 which contains a list of specially created graphs representing the computation
@@ -92,7 +92,7 @@ You can use :func:`~dgl.graphbolt.exclude_seed_edges` alongside with
     datapipe = datapipe.transform(exclude_seed_edges)
     datapipe = datapipe.fetch_feature(feature, node_feature_keys=["feat"])
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
     
 
 Adapt your model for minibatch training
@@ -273,7 +273,7 @@ only difference is that the train_set is now an instance of
         feature, node_feature_keys={"item": ["feat"], "user": ["feat"]}
     )
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
 Things become a little different if you wish to exclude the reverse
 edges on heterogeneous graphs. On heterogeneous graphs, reverse edges

docs/source/guide/minibatch-inference.rst

@@ -48,7 +48,7 @@ only one layer at a time.
     datapipe = datapipe.sample_neighbor(g, [-1]) # 1 layers.
     datapipe = datapipe.fetch_feature(feature, node_feature_keys=["feat"])
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
 
 Note that offline inference is implemented as a method of the GNN module

docs/source/guide/minibatch-link.rst

@@ -28,7 +28,7 @@ The whole data loader pipeline is as follows:
     datapipe = datapipe.transform(gb.exclude_seed_edges)
     datapipe = datapipe.fetch_feature(feature, node_feature_keys=["feat"])
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
 
 For the details about the builtin uniform negative sampler please see
@@ -213,7 +213,7 @@ only difference is that you need to give edge types for feature fetching.
         node_feature_keys={"user": ["feat"], "item": ["feat"]}
     )
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
 If you want to give your own negative sampling function, just inherit from the
 :class:`~dgl.graphbolt.NegativeSampler` class and override the

docs/source/guide/minibatch-node.rst

@@ -51,7 +51,7 @@ putting the list of generated MFGs onto GPU.
     # datapipe = gb.NeighborSampler(datapipe, g, [10, 10])
     datapipe = datapipe.fetch_feature(feature, node_feature_keys=["feat"])
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
 
 Iterating over the DataLoader will yield :class:`~dgl.graphbolt.MiniBatch`
@@ -216,7 +216,7 @@ of node types to node IDs.
         feature, node_feature_keys={"author": ["feat"], "paper": ["feat"]}
     )
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
 The training loop is almost the same as that of homogeneous graphs,
 except for the implementation of ``compute_loss`` that will take in two

docs/source/guide/minibatch-parallelism.rst

@@ -22,7 +22,7 @@ generate a minibatch, including:
     datapipe = datapipe.transform(gb.exclude_seed_edges)
     datapipe = datapipe.fetch_feature(feature, node_feature_keys=["feat"])
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
 
 All these stages are implemented in separate
 `IterableDataPipe <https://pytorch.org/data/main/torchdata.datapipes.iter.html>`__

tutorials/multi/2_node_classification.py

@@ -119,7 +119,7 @@ def create_dataloader(
     datapipe = datapipe.sample_neighbor(graph, [10, 10, 10])
     datapipe = datapipe.fetch_feature(features, node_feature_keys=["feat"])
     datapipe = datapipe.copy_to(device)
-    dataloader = gb.DataLoader(datapipe, num_workers=0)
+    dataloader = gb.DataLoader(datapipe)
     return dataloader