[Fix] Update ogb examples according to new API. (#1922)

* upd

* upd
Co-authored-by: Ubuntu <ubuntu@ip-172-31-29-3.us-east-2.compute.internal>

examples/pytorch/ogb/cluster-gat/main.py

@@ -96,7 +96,7 @@ class GAT(nn.Module):
                     num_workers=args.num_workers)
 
             for input_nodes, output_nodes, blocks in tqdm.tqdm(dataloader):
-                block = blocks[0].to(device)
+                block = blocks[0].int().to(device)
                 h = x[input_nodes].to(device)
                 h_dst = h[:block.number_of_dst_nodes()].to(device)
                 if l < self.n_layers - 1:
@@ -164,7 +164,7 @@ def run(args, device, data):
         # blocks.
         tic_start = time.time()
         for step, cluster in enumerate(cluster_iterator):
-            cluster = cluster.to(device)
+            cluster = cluster.int().to(device)
             mask = cluster.ndata['train_mask']
             if mask.sum() == 0:
                 continue
@@ -265,7 +265,7 @@ if __name__ == '__main__':
 
     cluster_iter_data = ClusterIter(
             'ogbn-products', graph, args.num_partitions, args.batch_size)
-    cluster_iterator = DataLoader(cluster_iter_data, batch_size=args.batch_size, shuffle=True, pin_memory=True, num_workers=0, collate_fn=partial(subgraph_collate_fn, graph))
+    cluster_iterator = DataLoader(cluster_iter_data, batch_size=args.batch_size, shuffle=True, pin_memory=True, num_workers=4, collate_fn=partial(subgraph_collate_fn, graph))
 
     in_feats = graph.ndata['feat'].shape[1]
     n_classes = (labels.max() + 1).item()

examples/pytorch/ogb/cluster-sage/README.md

@@ -7,4 +7,4 @@ We use builtin metris to do the graph partition.
 
 Run `main.py` and you should directly see the result.
 
-Accuracy over 10 runs: 0.7565994 ± 0.00654619
+Accuracy over 10 runs: 0.7830701 ± 0.0035093208

examples/pytorch/ogb/cluster-sage/main.py

@@ -133,8 +133,7 @@ def run(args, device, data):
         # blocks.
         tic_start = time.time()
         for step, cluster in enumerate(cluster_iterator):
-            #cluster.copy_from_parent()
-            #cluster.ndata['train_mask'] = g.ndata['train_mask'][cluster.ndata[dgl.NID]]
+            cluster = cluster.int().to(device)
             mask = cluster.ndata['train_mask'].to(device)
             if mask.sum() == 0:
                 continue
@@ -184,14 +183,14 @@ if __name__ == '__main__':
     argparser = argparse.ArgumentParser("multi-gpu training")
     argparser.add_argument('--gpu', type=int, default=0,
         help="GPU device ID. Use -1 for CPU training")
-    argparser.add_argument('--num-epochs', type=int, default=20)
+    argparser.add_argument('--num-epochs', type=int, default=30)
     argparser.add_argument('--num-hidden', type=int, default=256)
     argparser.add_argument('--num-layers', type=int, default=3)
     argparser.add_argument('--batch-size', type=int, default=32)
     argparser.add_argument('--val-batch-size', type=int, default=10000)
     argparser.add_argument('--log-every', type=int, default=20)
     argparser.add_argument('--eval-every', type=int, default=1)
-    argparser.add_argument('--lr', type=float, default=0.003)
+    argparser.add_argument('--lr', type=float, default=0.001)
     argparser.add_argument('--dropout', type=float, default=0.5)
     argparser.add_argument('--save-pred', type=str, default='')
     argparser.add_argument('--wd', type=float, default=0)
@@ -229,7 +228,7 @@ if __name__ == '__main__':
     cluster_iter_data = ClusterIter(
                     'ogbn-products', graph, args.num_partitions, args.batch_size, th.cat([train_idx, val_idx, test_idx]))
     idx = th.arange(args.num_partitions // args.batch_size)
-    cluster_iterator = DataLoader(cluster_iter_data, batch_size=32, shuffle=True, pin_memory=True, num_workers=0, collate_fn=partial(subgraph_collate_fn, graph))
+    cluster_iterator = DataLoader(cluster_iter_data, batch_size=32, shuffle=True, pin_memory=True, num_workers=4, collate_fn=partial(subgraph_collate_fn, graph))
 
     in_feats = graph.ndata['feat'].shape[1]
     print(in_feats)

examples/pytorch/ogb/deepwalk/reading_data.py

@@ -116,10 +116,7 @@ def make_undirected(G):
     return G
 
 def find_connected_nodes(G):
-    nodes = []
-    for n in G.nodes():
-        if G.out_degree(n) > 0:
-            nodes.append(n.item())
+    nodes = G.out_degrees().nonzero().squeeze(-1)
     return nodes
 
 class DeepwalkDataset:
@@ -188,7 +185,7 @@ class DeepwalkDataset:
 
         # negative table for true negative sampling
         if not fast_neg:
-            node_degree = np.array(list(map(lambda x: self.G.out_degree(x), self.valid_seeds)))
+            node_degree = self.G.out_degrees(self.valid_seeds).numpy()
             node_degree = np.power(node_degree, 0.75)
             node_degree /= np.sum(node_degree)
             node_degree = np.array(node_degree * 1e8, dtype=np.int)
@@ -224,6 +221,5 @@ class DeepwalkSampler(object):
         self.walk_length = walk_length
     
     def sample(self, seeds):
-        walks = dgl.contrib.sampling.random_walk(self.G, seeds, 
-            1, self.walk_length-1)
+        walks = dgl.sampling.random_walk(self.G, seeds, length=self.walk_length-1)[0]
         return walks

examples/pytorch/ogb/ogbn-products/gat/main.py

@@ -108,7 +108,7 @@ class GAT(nn.Module):
                 num_workers=args.num_workers)
 
             for input_nodes, output_nodes, blocks in tqdm.tqdm(dataloader):
-                block = blocks[0].to(device)
+                block = blocks[0].int().to(device)
 
                 h = x[input_nodes].to(device)
                 h_dst = h[:block.number_of_dst_nodes()]
@@ -124,17 +124,6 @@ class GAT(nn.Module):
             x = y
         return y
 
-def prepare_mp(g):
-    """
-    Explicitly materialize the CSR, CSC and COO representation of the given graph
-    so that they could be shared via copy-on-write to sampler workers and GPU
-    trainers.
-    This is a workaround before full shared memory support on heterogeneous graphs.
-    """
-    g.request_format('csr')
-    g.request_format('coo')
-    g.request_format('csc')
-
 def compute_acc(pred, labels):
     """
     Compute the accuracy of prediction given the labels.
@@ -277,7 +266,7 @@ if __name__ == '__main__':
 
     in_feats = graph.ndata['feat'].shape[1]
     n_classes = (labels.max() + 1).item()
-    prepare_mp(graph)
+    graph.create_format_()
     # Pack data
     data = train_idx, val_idx, test_idx, in_feats, labels, n_classes, graph, args.head
 

examples/pytorch/ogb/ogbn-products/graphsage/main.py

@@ -81,7 +81,7 @@ class SAGE(nn.Module):
                 num_workers=args.num_workers)
 
             for input_nodes, output_nodes, blocks in tqdm.tqdm(dataloader):
-                block = blocks[0].to(device)
+                block = blocks[0].int().to(device)
 
                 h = x[input_nodes].to(device)
                 h_dst = h[:block.number_of_dst_nodes()]
@@ -95,17 +95,6 @@ class SAGE(nn.Module):
             x = y
         return y
 
-def prepare_mp(g):
-    """
-    Explicitly materialize the CSR, CSC and COO representation of the given graph
-    so that they could be shared via copy-on-write to sampler workers and GPU
-    trainers.
-    This is a workaround before full shared memory support on heterogeneous graphs.
-    """
-    g.in_degree(0)
-    g.out_degree(0)
-    g.find_edges([0])
-
 def compute_acc(pred, labels):
     """
     Compute the accuracy of prediction given the labels.
@@ -175,7 +164,7 @@ def run(args, device, data):
             tic_step = time.time()
 
             # copy block to gpu
-            blocks = [blk.to(device) for blk in blocks]
+            blocks = [blk.int().to(device) for blk in blocks]
 
             # Load the input features as well as output labels
             batch_inputs, batch_labels = load_subtensor(g, labels, seeds, input_nodes, device)
@@ -245,7 +234,7 @@ if __name__ == '__main__':
 
     in_feats = graph.ndata['feat'].shape[1]
     n_classes = (labels.max() + 1).item()
-    prepare_mp(graph)
+    graph.create_format_()
     # Pack data
     data = train_idx, val_idx, test_idx, in_feats, labels, n_classes, graph
 

python/dgl/nn/mxnet/conv/gatconv.py

@@ -27,13 +27,8 @@ class GATConv(nn.Block):
 
     Parameters
     ----------
-    in_feats : int or pair of ints
-        Input feature size.
-
-        If the layer is to be applied to a unidirectional bipartite graph, ``in_feats``
-        specifies the input feature size on both the source and destination nodes.  If
-        a scalar is given, the source and destination node feature size would take the
-        same value.
+    in_feats : int
+        Number of input features.
     out_feats : int
         Output feature size.
     num_heads : int
@@ -65,14 +60,6 @@ class GATConv(nn.Block):
         self._in_feats = in_feats
         self._out_feats = out_feats
         with self.name_scope():
-            if isinstance(in_feats, tuple):
-                self.fc_src = nn.Dense(out_feats * num_heads, use_bias=False,
-                                       weight_initializer=mx.init.Xavier(magnitude=math.sqrt(2.0)),
-                                       in_units=self._in_src_feats)
-                self.fc_dst = nn.Dense(out_feats * num_heads, use_bias=False,
-                                       weight_initializer=mx.init.Xavier(magnitude=math.sqrt(2.0)),
-                                       in_units=self._in_dst_feats)
-            else:
             self.fc = nn.Dense(out_feats * num_heads, use_bias=False,
                                weight_initializer=mx.init.Xavier(magnitude=math.sqrt(2.0)),
                                in_units=in_feats)
@@ -121,9 +108,9 @@ class GATConv(nn.Block):
             if isinstance(feat, tuple):
                 h_src = self.feat_drop(feat[0])
                 h_dst = self.feat_drop(feat[1])
-                feat_src = self.fc_src(h_src).reshape(
+                feat_src = self.fc(h_src).reshape(
                     -1, self._num_heads, self._out_feats)
-                feat_dst = self.fc_dst(h_dst).reshape(
+                feat_dst = self.fc(h_dst).reshape(
                     -1, self._num_heads, self._out_feats)
             else:
                 h_src = h_dst = self.feat_drop(feat)

python/dgl/nn/tensorflow/conv/gatconv.py

@@ -67,12 +67,6 @@ class GATConv(layers.Layer):
         self._out_feats = out_feats
         xinit = tf.keras.initializers.VarianceScaling(scale=np.sqrt(
             2), mode="fan_avg", distribution="untruncated_normal")
-        if isinstance(in_feats, tuple):
-            self.fc_src = layers.Dense(
-                out_feats * num_heads, use_bias=False, kernel_initializer=xinit)
-            self.fc_dst = layers.Dense(
-                out_feats * num_heads, use_bias=False, kernel_initializer=xinit)
-        else:
         self.fc = layers.Dense(
             out_feats * num_heads, use_bias=False, kernel_initializer=xinit)
         self.attn_l = tf.Variable(initial_value=xinit(
@@ -116,8 +110,8 @@ class GATConv(layers.Layer):
             if isinstance(feat, tuple):
                 h_src = self.feat_drop(feat[0])
                 h_dst = self.feat_drop(feat[1])
-                feat_src = tf.reshape(self.fc_src(h_src), (-1, self._num_heads, self._out_feats))
-                feat_dst = tf.reshape(self.fc_dst(h_dst), (-1, self._num_heads, self._out_feats))
+                feat_src = tf.reshape(self.fc(h_src), (-1, self._num_heads, self._out_feats))
+                feat_dst = tf.reshape(self.fc(h_dst), (-1, self._num_heads, self._out_feats))
             else:
                 h_src = h_dst = self.feat_drop(feat)
                 feat_src = feat_dst = tf.reshape(

tests/mxnet/test_nn.py

@@ -173,9 +173,9 @@ def test_gat_conv(g, idtype):
 def test_gat_conv_bi(g, idtype):
     g = g.astype(idtype).to(F.ctx())
     ctx = F.ctx()
-    gat = nn.GATConv((5, 10), 2, 4)
+    gat = nn.GATConv(5, 2, 4)
     gat.initialize(ctx=ctx)
-    feat = (F.randn((g.number_of_src_nodes(), 5)), F.randn((g.number_of_dst_nodes(), 10)))
+    feat = (F.randn((g.number_of_src_nodes(), 5)), F.randn((g.number_of_dst_nodes(), 5)))
     h = gat(g, feat)
     assert h.shape == (g.number_of_dst_nodes(), 4, 2)
 

tests/tensorflow/test_nn.py

@@ -371,8 +371,8 @@ def test_gat_conv(g, idtype):
 def test_gat_conv_bi(g, idtype):
     g = g.astype(idtype).to(F.ctx())
     ctx = F.ctx()
-    gat = nn.GATConv((5, 10), 2, 4)
-    feat = (F.randn((g.number_of_src_nodes(), 5)), F.randn((g.number_of_dst_nodes(), 10)))
+    gat = nn.GATConv(5, 2, 4)
+    feat = (F.randn((g.number_of_src_nodes(), 5)), F.randn((g.number_of_dst_nodes(), 5)))
     h = gat(g, feat)
     assert h.shape == (g.number_of_dst_nodes(), 4, 2)