Deprecate (#4864)

rename DGLHeteroGraph to DGLGraph

benchmarks/benchmarks/model_speed/bench_pinsage.py

@@ -133,7 +133,7 @@ class WeightedSAGEConv(nn.Module):
 class SAGENet(nn.Module):
     def __init__(self, hidden_dims, n_layers):
         """
-        g : DGLHeteroGraph
+        g : DGLGraph
             The user-item interaction graph.
             This is only for finding the range of categorical variables.
         item_textsets : torchtext.data.Dataset
@@ -357,7 +357,7 @@ def assign_textual_node_features(ndata, textset, ntype):
     The length would be stored as another node feature with name
     ``field_name + '__len'``.
 
-    block : DGLHeteroGraph
+    block : DGLGraph
         First element of the compacted blocks, with "dgl.NID" as the
         corresponding node ID in the original graph, hence the index to the
         text dataset.

benchmarks/benchmarks/model_speed/bench_rgcn_hetero_ns.py

@@ -88,7 +88,7 @@ class RelGraphConvLayer(nn.Module):
 
         Parameters
         ----------
-        g : DGLHeteroGraph
+        g : DGLGraph
             Input graph.
         inputs : dict[str, torch.Tensor]
             Node feature for each node type.
@@ -164,14 +164,14 @@ class RelGraphEmbed(nn.Module):
 
         Parameters
         ----------
-        block : DGLHeteroGraph, optional
+        block : DGLGraph, optional
             If not specified, directly return the full graph with embeddings stored in
             :attr:`embed_name`. Otherwise, extract and store the embeddings to the block
             graph and return.
 
         Returns
         -------
-        DGLHeteroGraph
+        DGLGraph
             The block graph fed with embeddings.
         """
         embeds = {}

docs/source/api/python/dgl.rst

@@ -67,8 +67,6 @@ Operators for generating new graphs by manipulating the structure of the existin
     add_self_loop
     adj_product_graph
     adj_sum_graph
-    as_heterograph
-    as_immutable_graph
     compact_graphs
     khop_adj
     khop_graph

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

@@ -28,7 +28,7 @@ def get_graph(network_data, vocab):
         mapping node IDs to node indices
     Output
     ------
-    DGLHeteroGraph
+    DGLGraph
         a heterogenous graph, with one node type and different edge types
     """
     graphs = []

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

@@ -28,7 +28,7 @@ def get_graph(network_data, vocab):
         mapping node IDs to node indices
     Output
     ------
-    DGLHeteroGraph
+    DGLGraph
         a heterogenous graph, with one node type and different edge types
     """
     graphs = []

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

@@ -39,7 +39,7 @@ def get_graph(network_data, vocab):
         mapping node IDs to node indices
     Output
     ------
-    DGLHeteroGraph
+    DGLGraph
         a heterogenous graph, with one node type and different edge types
     """
     graphs = []

examples/pytorch/gcmc/data.py

@@ -46,25 +46,25 @@ class MovieLens(object):
 
     Attributes
     ----------
-    train_enc_graph : dgl.DGLHeteroGraph
+    train_enc_graph : dgl.DGLGraph
         Encoder graph for training.
-    train_dec_graph : dgl.DGLHeteroGraph
+    train_dec_graph : dgl.DGLGraph
         Decoder graph for training.
     train_labels : torch.Tensor
         The categorical label of each user-movie pair
     train_truths : torch.Tensor
         The actual rating values of each user-movie pair
-    valid_enc_graph : dgl.DGLHeteroGraph
+    valid_enc_graph : dgl.DGLGraph
         Encoder graph for validation.
-    valid_dec_graph : dgl.DGLHeteroGraph
+    valid_dec_graph : dgl.DGLGraph
         Decoder graph for validation.
     valid_labels : torch.Tensor
         The categorical label of each user-movie pair
     valid_truths : torch.Tensor
         The actual rating values of each user-movie pair
-    test_enc_graph : dgl.DGLHeteroGraph
+    test_enc_graph : dgl.DGLGraph
         Encoder graph for test.
-    test_dec_graph : dgl.DGLHeteroGraph
+    test_dec_graph : dgl.DGLGraph
         Decoder graph for test.
     test_labels : torch.Tensor
         The categorical label of each user-movie pair

examples/pytorch/gcmc/model.py

@@ -233,7 +233,7 @@ class GCMCLayer(nn.Module):
 
         Parameters
         ----------
-        graph : DGLHeteroGraph
+        graph : DGLGraph
             User-movie rating graph. It should contain two node types: "user"
             and "movie" and many edge types each for one rating value.
         ufeat : torch.Tensor, optional
@@ -320,7 +320,7 @@ class BiDecoder(nn.Module):
 
         Parameters
         ----------
-        graph : DGLHeteroGraph
+        graph : DGLGraph
             "Flattened" user-movie graph with only one edge type.
         ufeat : th.Tensor
             User embeddings. Shape: (|V_u|, D)

examples/pytorch/han/model_hetero.py

@@ -47,7 +47,7 @@ class HANLayer(nn.Module):
 
     Inputs
     ------
-    g : DGLHeteroGraph
+    g : DGLGraph
         The heterogeneous graph
     h : tensor
         Input features

examples/pytorch/han/train_sampling.py

@@ -34,7 +34,7 @@ class HANLayer(torch.nn.Module):
 
     Inputs
     ------
-    g : DGLHeteroGraph
+    g : DGLGraph
         The heterogeneous graph
     h : tensor
         Input features

examples/pytorch/lda/lda_model.py

@@ -469,7 +469,7 @@ def doc_subgraph(G, doc_ids):
     _, _, (block,) = sampler.sample(
         G.reverse(), {"doc": torch.as_tensor(doc_ids)}
     )
-    B = dgl.DGLHeteroGraph(
+    B = dgl.DGLGraph(
         block._graph, ["_", "word", "doc", "_"], block.etypes
     ).reverse()
     B.nodes["word"].data["_ID"] = block.nodes["word"].data["_ID"]

examples/pytorch/ogb/ogbn-mag/hetero_rgcn.py

@@ -113,7 +113,7 @@ class RelGraphConvLayer(nn.Module):
         """
         Parameters
         ----------
-        g : DGLHeteroGraph
+        g : DGLGraph
             Input graph.
         inputs : dict[str, torch.Tensor]
             Node feature for each node type.

examples/pytorch/pinsage/layers.py

@@ -136,7 +136,7 @@ class WeightedSAGEConv(nn.Module):
 class SAGENet(nn.Module):
     def __init__(self, hidden_dims, n_layers):
         """
-        g : DGLHeteroGraph
+        g : DGLGraph
             The user-item interaction graph.
             This is only for finding the range of categorical variables.
         item_textsets : torchtext.data.Dataset

examples/pytorch/pinsage/sampler.py

@@ -147,7 +147,7 @@ def assign_textual_node_features(ndata, textset, ntype):
     The length would be stored as another node feature with name
     ``field_name + '__len'``.
 
-    block : DGLHeteroGraph
+    block : DGLGraph
         First element of the compacted blocks, with "dgl.NID" as the
         corresponding node ID in the original graph, hence the index to the
         text dataset.

examples/pytorch/rgcn-hetero/model.py

@@ -101,7 +101,7 @@ class RelGraphConvLayer(nn.Module):
 
         Parameters
         ----------
-        g : DGLHeteroGraph
+        g : DGLGraph
             Input graph.
         inputs : dict[str, torch.Tensor]
             Node feature for each node type.
@@ -224,7 +224,7 @@ class RelGraphConvLayerHeteroAPI(nn.Module):
 
         Parameters
         ----------
-        g : DGLHeteroGraph
+        g : DGLGraph
             Input graph.
         inputs : dict[str, torch.Tensor]
             Node feature for each node type.
@@ -302,14 +302,14 @@ class RelGraphEmbed(nn.Module):
 
         Parameters
         ----------
-        block : DGLHeteroGraph, optional
+        block : DGLGraph, optional
             If not specified, directly return the full graph with embeddings stored in
             :attr:`embed_name`. Otherwise, extract and store the embeddings to the block
             graph and return.
 
         Returns
         -------
-        DGLHeteroGraph
+        DGLGraph
             The block graph fed with embeddings.
         """
         return self.embeds

examples/pytorch/rgcn/experimental/entity_classify_dist.py

@@ -105,7 +105,7 @@ class RelGraphConvLayer(nn.Module):
         """Forward computation
         Parameters
         ----------
-        g : DGLHeteroGraph
+        g : DGLGraph
             Input graph.
         inputs : dict[str, torch.Tensor]
             Node feature for each node type.

python/dgl/__init__.py

@@ -36,8 +36,8 @@ from .readout import *
 from .batch import *
 from .convert import *
 from .generators import *
-from .heterograph import DGLHeteroGraph
-from .heterograph import DGLHeteroGraph as DGLGraph  # pylint: disable=reimported
+from .heterograph import DGLGraph
+from .heterograph import DGLGraph as DGLHeteroGraph  # pylint: disable=reimported
 from .dataloading import set_src_lazy_features, set_dst_lazy_features, set_edge_lazy_features, \
     set_node_lazy_features
 from .merge import *

python/dgl/_dataloading/dataloader.py

@@ -10,7 +10,7 @@ from .. import backend as F
 from .. import utils
 from ..batch import batch
 from ..convert import heterograph
-from ..heterograph import DGLHeteroGraph as DGLGraph
+from ..heterograph import DGLGraph
 from ..distributed.dist_graph import DistGraph
 from ..utils import to_device
 

python/dgl/_deprecate/graph.py

@@ -40,7 +40,7 @@ class DGLBaseGraph(object):
         Data to initialize graph.
     """
 
-    is_block = False        # for compatibility with DGLHeteroGraph
+    is_block = False        # for compatibility with DGLGraph
 
     def __init__(self, graph):
         self._graph = graph

python/dgl/batch.py

@@ -4,12 +4,12 @@ from collections.abc import Mapping
 from . import backend as F
 from .base import ALL, is_all, DGLError, dgl_warning, NID, EID
 from .heterograph_index import disjoint_union, slice_gidx
-from .heterograph import DGLHeteroGraph
+from .heterograph import DGLGraph
 from . import convert
 from . import utils
 
 
-__all__ = ['batch', 'unbatch', 'slice_batch', 'batch_hetero', 'unbatch_hetero']
+__all__ = ['batch', 'unbatch', 'slice_batch']
 
 def batch(graphs, ndata=ALL, edata=ALL, *,
           node_attrs=None, edge_attrs=None):
@@ -175,7 +175,7 @@ def batch(graphs, ndata=ALL, edata=ALL, *,
     etypes = [etype for _, etype, _ in relations]
 
     gidx = disjoint_union(graphs[0]._graph.metagraph, [g._graph for g in graphs])
-    retg = DGLHeteroGraph(gidx, ntypes, etypes)
+    retg = DGLGraph(gidx, ntypes, etypes)
 
     # Compute batch num nodes
     bnn = {}
@@ -479,7 +479,7 @@ def slice_batch(g, gid, store_ids=False):
     # Slice graph structure
     gidx = slice_gidx(g._graph, utils.toindex(num_nodes), utils.toindex(start_nid),
                       utils.toindex(num_edges), utils.toindex(start_eid))
-    retg = DGLHeteroGraph(gidx, g.ntypes, g.etypes)
+    retg = DGLGraph(gidx, g.ntypes, g.etypes)
 
     # Slice node features
     for ntid, ntype in enumerate(g.ntypes):
@@ -504,16 +504,3 @@ def slice_batch(g, gid, store_ids=False):
                                                    retg.idtype, retg.device)
 
     return retg
-
-#### DEPRECATED APIS ####
-def batch_hetero(*args, **kwargs):
-    """DEPREACTED: please use dgl.batch """
-    dgl_warning('From v0.5, DGLHeteroGraph is merged into DGLGraph. You can safely'
-                ' replace dgl.batch_hetero with dgl.batch')
-    return batch(*args, **kwargs)
-
-def unbatch_hetero(*args, **kwargs):
-    """DEPREACTED: please use dgl.unbatch """
-    dgl_warning('From v0.5, DGLHeteroGraph is merged into DGLGraph. You can safely'
-                ' replace dgl.unbatch_hetero with dgl.unbatch')
-    return unbatch(*args, **kwargs)