[Doc/Feature] Refactor, doc update and behavior fix for graphs (#1983)

* Update graph

* Fix for dgl.graph

* from_scipy

* Replace canonical_etypes with relations

* from_networkx

* Update for hetero_from_relations

* Roll back the change of canonical_etypes to relations

* heterograph

* bipartite

* Update doc

* Fix lint

* Fix lint

* Fix test cases

* Fix

* Fix

* Fix

* Fix

* Fix

* Fix

* Update

* Fix test

* Fix

* Update

* Use DGLError

* Update

* Update

* Update

* Update

* Fix

* Fix

* Fix

* Fix

* Fix

* Fix

* Fix

* Fix

* Update

* Fix

* Update

* Fix

* Fix

* Fix

* Update

* Fix

* Update

* Fix

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Fix

* Fix

* Update

* Update

* Update

* Update

* Update

* Update

* rewrite sanity checks

* delete unnecessary checks

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Fix

* Update

* Update

* Update

* Fix

* Fix

* Fix

* Update

* Fix

* Update

* Fix

* Fix

* Update

* Fix

* Update

* Fix
Co-authored-by: xiang song(charlie.song) <classicxsong@gmail.com>
Co-authored-by: Minjie Wang <wmjlyjemaine@gmail.com>
Co-authored-by: Quan Gan <coin2028@hotmail.com>

docs/source/api/python/dgl.DGLGraph.rst

 .. _apigraph:
 
-dgl.DGLHeteroGraph
+dgl.DGLGraph
 =====================================================
 
 .. currentmodule:: dgl
-.. autoclass:: DGLHeteroGraph
+.. class:: DGLGraph
 
 Querying metagraph structure
 ----------------------------
@@ -12,13 +12,15 @@ Querying metagraph structure
 .. autosummary::
     :toctree: ../../generated/
 
-    DGLHeteroGraph.ntypes
-    DGLHeteroGraph.etypes
-    DGLHeteroGraph.canonical_etypes
-    DGLHeteroGraph.metagraph
-    DGLHeteroGraph.to_canonical_etype
-    DGLHeteroGraph.get_ntype_id
-    DGLHeteroGraph.get_etype_id
+    DGLGraph.ntypes
+    DGLGraph.etypes
+    DGLGraph.srctypes
+    DGLGraph.dsttypes
+    DGLGraph.canonical_etypes
+    DGLGraph.metagraph
+    DGLGraph.to_canonical_etype
+    DGLGraph.get_ntype_id
+    DGLGraph.get_etype_id
 
 Querying graph structure
 ------------------------
@@ -26,20 +28,27 @@ Querying graph structure
 .. autosummary::
     :toctree: ../../generated/
 
-    DGLHeteroGraph.number_of_nodes
-    DGLHeteroGraph.number_of_edges
-    DGLHeteroGraph.is_multigraph
-    DGLHeteroGraph.has_nodes
-    DGLHeteroGraph.has_edges_between
-    DGLHeteroGraph.predecessors
-    DGLHeteroGraph.successors
-    DGLHeteroGraph.edge_ids
-    DGLHeteroGraph.find_edges
-    DGLHeteroGraph.in_edges
-    DGLHeteroGraph.out_edges
-    DGLHeteroGraph.all_edges
-    DGLHeteroGraph.in_degrees
-    DGLHeteroGraph.out_degrees
+    DGLGraph.num_nodes
+    DGLGraph.number_of_nodes
+    DGLGraph.num_edges
+    DGLGraph.number_of_edges
+    DGLGraph.num_src_nodes
+    DGLGraph.number_of_src_nodes
+    DGLGraph.num_dst_nodes
+    DGLGraph.number_of_dst_nodes
+    DGLGraph.is_unibipartite
+    DGLGraph.is_multigraph
+    DGLGraph.is_homogeneous
+    DGLGraph.has_nodes
+    DGLGraph.has_edges_between
+    DGLGraph.predecessors
+    DGLGraph.successors
+    DGLGraph.edge_ids
+    DGLGraph.find_edges
+    DGLGraph.in_edges
+    DGLGraph.out_edges
+    DGLGraph.in_degrees
+    DGLGraph.out_degrees
 
 Querying and manipulating sparse format
 ---------------------------------------
@@ -47,9 +56,8 @@ Querying and manipulating sparse format
 .. autosummary::
     :toctree: ../../generated/
 
-    DGLHeteroGraph.format_in_use
-    DGLHeteroGraph.restrict_format
-    DGLHeteroGraph.to_format
+    DGLGraph.formats
+    DGLGraph.create_format_
 
 Querying and manipulating index data type
 -----------------------------------------
@@ -57,9 +65,9 @@ Querying and manipulating index data type
 .. autosummary::
     :toctree: ../../generated/
 
-    DGLHeteroGraph.idtype
-    DGLHeteroGraph.long
-    DGLHeteroGraph.int
+    DGLGraph.idtype
+    DGLGraph.long
+    DGLGraph.int
 
 Using Node/edge features
 ------------------------
@@ -67,59 +75,86 @@ Using Node/edge features
 .. autosummary::
     :toctree: ../../generated/
 
-    DGLHeteroGraph.nodes
-    DGLHeteroGraph.ndata
-    DGLHeteroGraph.edges
-    DGLHeteroGraph.edata
-    DGLHeteroGraph.node_attr_schemes
-    DGLHeteroGraph.edge_attr_schemes
-    DGLHeteroGraph.set_n_initializer
-    DGLHeteroGraph.set_e_initializer
-    DGLHeteroGraph.local_var
-    DGLHeteroGraph.local_scope
+    DGLGraph.nodes
+    DGLGraph.ndata
+    DGLGraph.edges
+    DGLGraph.edata
+    DGLGraph.node_attr_schemes
+    DGLGraph.edge_attr_schemes
+    DGLGraph.srcnodes
+    DGLGraph.dstnodes
+    DGLGraph.srcdata
+    DGLGraph.dstdata
+    DGLGraph.local_var
+    DGLGraph.local_scope
 
-Using Node/edge features for blocks
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
-Please refer to :ref:`guide-minibatch` for the definition of blocks.
+Transforming graph
+------------------
 
 .. autosummary::
     :toctree: ../../generated/
 
-    DGLHeteroGraph.number_of_src_nodes
-    DGLHeteroGraph.number_of_dst_nodes
-    DGLHeteroGraph.srcnodes
-    DGLHeteroGraph.srcdata
-    DGLHeteroGraph.dstnodes
-    DGLHeteroGraph.dstdata
+    DGLGraph.subgraph
+    DGLGraph.edge_subgraph
+    DGLGraph.node_type_subgraph
+    DGLGraph.edge_type_subgraph
+    DGLGraph.__getitem__
 
-Transforming graph
-------------------
+Converting to other formats
+---------------------------
 
 .. autosummary::
     :toctree: ../../generated/
 
-    DGLHeteroGraph.subgraph
-    DGLHeteroGraph.edge_subgraph
-    DGLHeteroGraph.node_type_subgraph
-    DGLHeteroGraph.edge_type_subgraph
-    DGLHeteroGraph.__getitem__
+    DGLGraph.adj
+    DGLGraph.adjacency_matrix
+    DGLGraph.inc
+    DGLGraph.incidence_matrix
 
-Computing with DGLHeteroGraph
+Computing with DGLGraph
 -----------------------------
 
 .. autosummary::
     :toctree: ../../generated/
 
-    DGLHeteroGraph.apply_nodes
-    DGLHeteroGraph.apply_edges
-    DGLHeteroGraph.send_and_recv
-    DGLHeteroGraph.pull
-    DGLHeteroGraph.push
-    DGLHeteroGraph.update_all
-    DGLHeteroGraph.multi_update_all
-    DGLHeteroGraph.prop_nodes
-    DGLHeteroGraph.prop_edges
-    DGLHeteroGraph.filter_nodes
-    DGLHeteroGraph.filter_edges
-    DGLHeteroGraph.to
+    DGLGraph.apply_nodes
+    DGLGraph.apply_edges
+    DGLGraph.send_and_recv
+    DGLGraph.pull
+    DGLGraph.push
+    DGLGraph.update_all
+    DGLGraph.multi_update_all
+    DGLGraph.prop_nodes
+    DGLGraph.prop_edges
+    DGLGraph.filter_nodes
+    DGLGraph.filter_edges
+
+Querying batch summary
+----------------------
+
+.. autosummary::
+    :toctree: ../../generated/
+
+    DGLGraph.batch_size
+    DGLGraph.batch_num_nodes
+    DGLGraph.batch_num_edges
+
+Mutating topology
+-----------------
+
+.. autosummary::
+    :toctree: ../../generated/
+
+    DGLGraph.add_nodes
+    DGLGraph.add_edges
+    DGLGraph.remove_nodes
+    DGLGraph.remove_edges
+
+Device Control
+--------------
+
+.. autosummary::
+    :toctree: ../../generated/
+
+    DGLGraph.to
+    DGLGraph.device

docs/source/api/python/dgl.ops.rst

@@ -81,7 +81,7 @@ here, you can enjoy the same convenience on other frameworks by similar usage):
    >>> F.u_mul_e_sum(g, x, y1).shape  # (2,), (4, 2) -> (4, 2)
    torch.Size([3, 4, 2])
 
-For all operators, the input graph could either be a homograph or a bipartite
+For all operators, the input graph could either be a homogeneous or a bipartite
 graph.
 
 .. autosummary::
@@ -228,7 +228,7 @@ The following is an example showing how GSDDMM works:
     copy_u
     copy_v
 
-Like GSpMM, GSDDMM operators support both homograph and bipartite graph.
+Like GSpMM, GSDDMM operators support both homogeneous and bipartite graph.
 
 Edge Softmax module
 -------------------

docs/source/api/python/dgl.rst

@@ -62,8 +62,8 @@ Graph Transform Routines
     to_simple
     to_block
     compact_graphs
-    to_hetero
-    to_homo
+    to_heterogeneous
+    to_homogeneous
     to_networkx
     line_graph
     khop_graph

docs/source/api/python/nn.pytorch.rst

@@ -12,6 +12,8 @@ please `create an issue <https://github.com/dmlc/dgl/issues>`_ started with "[Fe
 If you want to contribute a NN module, please `create a pull request <https://github.com/dmlc/dgl/pulls>`_ started
 with "[NN] XXXModel in PyTorch NN Modules" and our team member would review this PR.
 
+.. _apinn-pytorch-conv:
+
 Conv Layers
 ----------------------------------------
 
@@ -136,6 +138,7 @@ DotGatConv
     :members: forward
     :show-inheritance:
 
+.. _apinn-pytorch-dense-conv:
 
 Dense Conv Layers
 ----------------------------------------
@@ -161,6 +164,8 @@ DenseChebConv
     :members: forward
     :show-inheritance:
 
+.. _apinn-pytorch-pooling:
+
 Global Pooling Layers
 ----------------------------------------
 
@@ -232,6 +237,8 @@ HeteroGraphConv
    :members:
    :show-inheritance:
 
+.. _apinn-pytorch-util:
+
 Utility Modules
 ----------------------------------------
 

docs/source/guide/message.rst

@@ -267,7 +267,7 @@ The message passing on heterographs can be split into two parts:
    relationships.
 
 DGL’s interface to call message passing on heterographs is
-`multi_update_all() <https://docs.dgl.ai/generated/dgl.DGLHeteroGraph.multi_update_all.html>`__.
+:meth:`~dgl.DGLGraph.multi_update_all`.
 ``multi_update_all`` takes a dictionary containing the parameters for
 ``update_all`` within each relation using relation as the key, and a
 string represents the cross type reducer. The reducer can be one of

docs/source/guide/nn.rst

@@ -13,15 +13,12 @@ framework code. The major difference lies in the message passing
 operations that are unique in DGL.
 
 DGL has integrated many commonly used
-`Sparse_GraphConvs <https://docs.dgl.ai/api/python/nn.pytorch.html#module-dgl.nn.pytorch.conv>`__,
-`Dense_GraphConvs <https://docs.dgl.ai/api/python/nn.pytorch.html#dense-conv-layers>`__,
-`Graph_Poolings <https://docs.dgl.ai/api/python/nn.pytorch.html#module-dgl.nn.pytorch.glob>`__,
+:ref:`apinn-pytorch-conv`, :ref:`apinn-pytorch-dense-conv`, :ref:`apinn-pytorch-pooling`,
 and
-`Utility <https://docs.dgl.ai/api/python/nn.pytorch.html#utility-modules>`__
-NN modules. We welcome your contribution!
+:ref:`apinn-pytorch-util`. We welcome your contribution!
 
 In this section, we will use
-`dgl.nn.conv.SAGEConv <https://github.com/sneakerkg/dgl/blob/nn_doc_refactor/python/dgl/nn/pytorch/conv/sageconv.py>`__
+:class:`~dgl.nn.pytorch.conv.SAGEConv`
 with Pytorch backend as an example to introduce how to build your own
 DGL NN Module.
 
@@ -289,7 +286,7 @@ object construction phase.
 Heterogeneous GraphConv Module
 ------------------------------
 
-`HeteroGraphConv <https://github.com/dmlc/dgl/blob/master/python/dgl/nn/pytorch/hetero.py>`__
+:class:`dgl.nn.pytorch.HeteroGraphConv`
 is a module-level encapsulation to run DGL NN module on heterogeneous
 graph. The implementation logic is the same as message passing level API
 ``multi_update_all()``:

examples/mxnet/gcmc/data.py

@@ -237,21 +237,19 @@ class MovieLens(object):
     def _generate_enc_graph(self, rating_pairs, rating_values, add_support=False):
         user_movie_R = np.zeros((self._num_user, self._num_movie), dtype=np.float32)
         user_movie_R[rating_pairs] = rating_values
-        movie_user_R = user_movie_R.transpose()
 
-        rating_graphs = []
+        data_dict = dict()
+        num_nodes_dict = {'user': self._num_user, 'movie': self._num_movie}
         rating_row, rating_col = rating_pairs
         for rating in self.possible_rating_values:
             ridx = np.where(rating_values == rating)
             rrow = rating_row[ridx]
             rcol = rating_col[ridx]
-            bg = dgl.bipartite((rrow, rcol), 'user', str(rating), 'movie',
-                               num_nodes=(self._num_user, self._num_movie))
-            rev_bg = dgl.bipartite((rcol, rrow), 'movie', 'rev-%s' % str(rating), 'user',
-                               num_nodes=(self._num_movie, self._num_user))
-            rating_graphs.append(bg)
-            rating_graphs.append(rev_bg)
-        graph = dgl.hetero_from_relations(rating_graphs)
+            data_dict.update({
+                ('user', str(rating), 'movie'): (rrow, rcol),
+                ('movie', 'rev-%s' % str(rating), 'user'): (rcol, rrow)
+            })
+        graph = dgl.heterograph(data_dict, num_nodes_dict=num_nodes_dict)
 
         # sanity check
         assert len(rating_pairs[0]) == sum([graph.number_of_edges(et) for et in graph.etypes]) // 2
@@ -294,7 +292,8 @@ class MovieLens(object):
         user_movie_ratings_coo = sp.coo_matrix(
             (ones, rating_pairs),
             shape=(self.num_user, self.num_movie), dtype=np.float32)
-        return dgl.bipartite(user_movie_ratings_coo, 'user', 'rate', 'movie')
+        g = dgl.bipartite_from_scipy(user_movie_ratings_coo, utype='_U', etype='_E', vtype='_V')
+        return dgl.heterograph({('user', 'rate', 'movie'): g.edges()})
 
     @property
     def num_links(self):

examples/mxnet/rgcn/entity_classify.py

@@ -55,7 +55,6 @@ def main(args):
     hg = dataset[0]
 
     num_rels = len(hg.canonical_etypes)
-    num_of_ntype = len(hg.ntypes)
     category = dataset.predict_category
     num_classes = dataset.num_classes
     train_mask = hg.nodes[category].data.pop('train_mask')
@@ -86,7 +85,7 @@ def main(args):
         if ntype == category:
             category_id = i
 
-    g = dgl.to_homo(hg)
+    g = dgl.to_homogeneous(hg, edata=['norm'])
     num_nodes = g.number_of_nodes()
     node_ids = mx.nd.arange(num_nodes)
     edge_norm = g.edata['norm']

examples/mxnet/sampling/run_store_server.py

@@ -10,7 +10,6 @@ from dgl.data import register_data_args, load_data
 class GraphData:
     def __init__(self, csr, num_feats, graph_name):
         num_nodes = csr.shape[0]
-        num_edges = mx.nd.contrib.getnnz(csr).asnumpy()[0]
         self.graph = dgl.graph_index.from_csr(csr.indptr, csr.indices, False, 'in')
         self.graph = self.graph.copyto_shared_mem(dgl.contrib.graph_store._get_graph_path(graph_name))
         self.features = mx.nd.random.normal(shape=(csr.shape[0], num_feats))

examples/pytorch/GATNE-T/src/main.py

@@ -32,7 +32,10 @@ def get_graph(network_data, vocab):
         a heterogenous graph, with one node type and different edge types
     '''
     graphs = []
-    num_nodes = len(vocab)
+
+    node_type = '_N' # '_N' can be replaced by an arbitrary name
+    data_dict = dict()
+    num_nodes_dict = {node_type: len(vocab)}
 
     for edge_type in network_data:
         tmp_data = network_data[edge_type]
@@ -40,9 +43,8 @@ def get_graph(network_data, vocab):
         for edge in tmp_data:
             edges.append((vocab[edge[0]], vocab[edge[1]]))
             edges.append((vocab[edge[1]], vocab[edge[0]]))
-        g = dgl.graph(edges, etype=edge_type, num_nodes=num_nodes)
-        graphs.append(g)
-    graph = dgl.hetero_from_relations(graphs)
+        data_dict[(node_type, edge_type, node_type)] = zip(*edges)
+    graph = dgl.heterograph(data_dict, num_nodes_dict)
     
     return graph
 

examples/pytorch/_deprecated/adaptive_sampling/adaptive_sampling.py

@@ -76,8 +76,9 @@ class NodeSampler(object):
                 yield self.seeds[batches[i]], batches[i]
 
 
-def create_nodeflow(layer_mappings, block_mappings, block_aux_data, rel_graphs, seed_map):
-    hg = dgl.hetero_from_relations(rel_graphs)
+def create_nodeflow(layer_mappings, block_mappings, block_aux_data,
+                    data_dict, num_nodes_dict, seed_map):
+    hg = dgl.heterograph(data_dict, num_nodes_dict=num_nodes_dict)
     hg.layer_mappings = layer_mappings
     hg.block_mappings = block_mappings
     hg.block_aux_data = block_aux_data
@@ -135,7 +136,8 @@ class AdaptGenerator(object):
         layer_mappings = []  # Mapping from layer node ID to parent node ID
         block_mappings = []  # Mapping from block edge ID to parent edge ID, or -1 if nonexistent
         block_aux_data = []
-        rel_graphs = []
+        data_dict = dict()
+        num_nodes_dict = dict()
 
         if self.coalesce:
             curr_frontier = torch.LongTensor(np.unique(seeds.numpy()))
@@ -163,11 +165,13 @@ class AdaptGenerator(object):
             aux_result = aux_result[[nodes_idx_map[i] for i in prev_frontier]]
             block_dsts = np.arange(len(curr_frontier)).repeat(num_neighbors)
 
-            rel_graphs.insert(0, dgl.bipartite(
-                (block_srcs, block_dsts),
-                'layer%d' % i, 'block%d' % i, 'layer%d' % (i + 1),
-                (len(prev_frontier), len(curr_frontier))
-            ))
+            data_dict.update({
+                ('layer%d' % i, 'block%d' % i, 'layer%d' % (i + 1)): (block_srcs, block_dsts)
+            })
+            num_nodes_dict.update({
+                'layer%d' % i: len(prev_frontier),
+                'layer%d' % (i + 1): len(curr_frontier)
+            })
 
             layer_mappings.insert(0, prev_frontier)
             block_mappings.insert(0, prev_frontier_edges)
@@ -179,7 +183,8 @@ class AdaptGenerator(object):
             layer_mappings=layer_mappings,
             block_mappings=block_mappings,
             block_aux_data=block_aux_data,
-            rel_graphs=rel_graphs,
+            data_dict=data_dict,
+            num_nodes_dict=num_nodes_dict,
             seed_map=seed_map)
 
     def stepback(self, curr_frontier, layer_index, *auxiliary):

examples/pytorch/gcmc/data.py

@@ -245,22 +245,20 @@ class MovieLens(object):
     def _generate_enc_graph(self, rating_pairs, rating_values, add_support=False):
         user_movie_R = np.zeros((self._num_user, self._num_movie), dtype=np.float32)
         user_movie_R[rating_pairs] = rating_values
-        movie_user_R = user_movie_R.transpose()
 
-        rating_graphs = []
+        data_dict = dict()
+        num_nodes_dict = {'user': self._num_user, 'movie': self._num_movie}
         rating_row, rating_col = rating_pairs
         for rating in self.possible_rating_values:
             ridx = np.where(rating_values == rating)
             rrow = rating_row[ridx]
             rcol = rating_col[ridx]
             rating = to_etype_name(rating)
-            bg = dgl.bipartite((rrow, rcol), 'user', rating, 'movie',
-                               num_nodes=(self._num_user, self._num_movie))
-            rev_bg = dgl.bipartite((rcol, rrow), 'movie', 'rev-%s' % rating, 'user',
-                               num_nodes=(self._num_movie, self._num_user))
-            rating_graphs.append(bg)
-            rating_graphs.append(rev_bg)
-        graph = dgl.hetero_from_relations(rating_graphs)
+            data_dict.update({
+                ('user', str(rating), 'movie'): (rrow, rcol),
+                ('movie', 'rev-%s' % str(rating), 'user'): (rcol, rrow)
+            })
+        graph = dgl.heterograph(data_dict, num_nodes_dict=num_nodes_dict)
 
         # sanity check
         assert len(rating_pairs[0]) == sum([graph.number_of_edges(et) for et in graph.etypes]) // 2
@@ -303,7 +301,8 @@ class MovieLens(object):
         user_movie_ratings_coo = sp.coo_matrix(
             (ones, rating_pairs),
             shape=(self.num_user, self.num_movie), dtype=np.float32)
-        return dgl.bipartite(user_movie_ratings_coo, 'user', 'rate', 'movie')
+        g = dgl.bipartite_from_scipy(user_movie_ratings_coo, utype='_U', etype='_E', vtype='_V')
+        return dgl.heterograph({('user', 'rate', 'movie'): g.edges()})
 
     @property
     def num_links(self):

examples/pytorch/han/utils.py

@@ -136,8 +136,8 @@ def load_acm(remove_self_loop):
 
     # Adjacency matrices for meta path based neighbors
     # (Mufei): I verified both of them are binary adjacency matrices with self loops
-    author_g = dgl.graph(data['PAP'], ntype='paper', etype='author')
-    subject_g = dgl.graph(data['PLP'], ntype='paper', etype='subject')
+    author_g = dgl.from_scipy(data['PAP'])
+    subject_g = dgl.from_scipy(data['PLP'])
     gs = [author_g, subject_g]
 
     train_idx = torch.from_numpy(data['train_idx']).long().squeeze(0)
@@ -186,11 +186,12 @@ def load_acm_raw(remove_self_loop):
     p_vs_t = p_vs_t[p_selected]
     p_vs_c = p_vs_c[p_selected]
 
-    pa = dgl.bipartite(p_vs_a, 'paper', 'pa', 'author')
-    ap = dgl.bipartite(p_vs_a.transpose(), 'author', 'ap', 'paper')
-    pl = dgl.bipartite(p_vs_l, 'paper', 'pf', 'field')
-    lp = dgl.bipartite(p_vs_l.transpose(), 'field', 'fp', 'paper')
-    hg = dgl.hetero_from_relations([pa, ap, pl, lp])
+    hg = dgl.heterograph({
+        ('paper', 'pa', 'author'): p_vs_a.nonzero(),
+        ('author', 'ap', 'paper'): p_vs_a.transpose.nonzero(),
+        ('paper', 'pf', 'field'): p_vs_l.nonzero(),
+        ('field', 'fp', 'paper'): p_vs_l.transpose().nonzero()
+    })
 
     features = torch.FloatTensor(p_vs_t.toarray())
 

examples/pytorch/hgt/train_acm.py

@@ -80,12 +80,12 @@ def train(model, G):
 device = torch.device("cuda:0")
 
 G = dgl.heterograph({
-        ('paper', 'written-by', 'author') : data['PvsA'],
-        ('author', 'writing', 'paper') : data['PvsA'].transpose(),
-        ('paper', 'citing', 'paper') : data['PvsP'],
-        ('paper', 'cited', 'paper') : data['PvsP'].transpose(),
-        ('paper', 'is-about', 'subject') : data['PvsL'],
-        ('subject', 'has', 'paper') : data['PvsL'].transpose(),
+        ('paper', 'written-by', 'author') : data['PvsA'].nonzero(),
+        ('author', 'writing', 'paper') : data['PvsA'].transpose().nonzero(),
+        ('paper', 'citing', 'paper') : data['PvsP'].nonzero(),
+        ('paper', 'cited', 'paper') : data['PvsP'].transpose().nonzero(),
+        ('paper', 'is-about', 'subject') : data['PvsL'].nonzero(),
+        ('subject', 'has', 'paper') : data['PvsL'].transpose().nonzero(),
     })
 print(G)
 

examples/pytorch/model_zoo/geometric/mnist.py

@@ -32,7 +32,7 @@ A = grid_graph(28, 8, metric)
 
 coarsening_levels = 4
 L, perm = coarsen(A, coarsening_levels)
-g_arr = [dgl.graph(csr) for csr in L]
+g_arr = [dgl.from_scipy(csr) for csr in L]
 
 coordinate_arr = get_coordinates(g_arr, grid_side, coarsening_levels, perm)
 for g, coordinate_arr in zip(g_arr, coordinate_arr):

examples/pytorch/rgcn/entity_classify.py

@@ -59,7 +59,6 @@ def main(args):
     hg = dataset[0]
 
     num_rels = len(hg.canonical_etypes)
-    num_of_ntype = len(hg.ntypes)
     category = dataset.predict_category
     num_classes = dataset.num_classes
     train_mask = hg.nodes[category].data.pop('train_mask')
@@ -90,7 +89,7 @@ def main(args):
         if ntype == category:
             category_id = i
 
-    g = dgl.to_homo(hg)
+    g = dgl.to_homogeneous(hg, edata=['norm'])
     num_nodes = g.number_of_nodes()
     node_ids = torch.arange(num_nodes)
     edge_norm = g.edata['norm']

examples/pytorch/rgcn/entity_classify_mp.py

@@ -484,7 +484,7 @@ def main(args, devices):
         if ntype == category:
             category_id = i
 
-    g = dgl.to_homo(hg)
+    g = dgl.to_homogeneous(hg, edata=['norm'])
     if args.global_norm:
         u, v, eid = g.all_edges(form='all')
         _, inverse_index, count = th.unique(v, return_inverse=True, return_counts=True)

examples/pytorch/rgcn/utils.py

@@ -135,7 +135,7 @@ def build_graph_from_triplets(num_nodes, num_rels, triplets):
         This function also generates edge type and normalization factor
         (reciprocal of node incoming degree)
     """
-    g = dgl.graph([])
+    g = dgl.graph(([], []))
     g.add_nodes(num_nodes)
     src, rel, dst = triplets
     src, dst = np.concatenate((src, dst)), np.concatenate((dst, src))

examples/pytorch/transformer/dataset/graph.py

@@ -19,7 +19,7 @@ class GraphPool:
         print('start creating graph pool...')
         tic = time.time()
         self.n, self.m = n, m
-        g_pool = [[dgl.graph([]) for _ in range(m)] for _ in range(n)]
+        g_pool = [[dgl.graph(([], [])) for _ in range(m)] for _ in range(n)]
         num_edges = {
             'ee': np.zeros((n, n)).astype(int),
             'ed': np.zeros((n, m)).astype(int),

examples/tensorflow/rgcn/README.md

@@ -5,7 +5,7 @@
 * Author's code for link prediction: [https://github.com/MichSchli/RelationPrediction](https://github.com/MichSchli/RelationPrediction)
 
 ### Dependencies
-* Tensorflow 2.1+
+* Tensorflow 2.2+
 * requests
 * rdflib
 * pandas