Input feature mismatch when using a heterogeneous dgl graph with Source node having multiple edges to target node

[kai5gabriel]

I have a graph where the number of source nodes (users) is not the same as the number of item nodes. Since multiple users can share the same item. When doing the forward pass "Many to One " edges and the mismatched number of features are causing this error. What is the standard way to handle this Many to One edge relationship scenario. As the number of items are less..I have assigned it to 8 rows with 3 features each.

`import dgl import dgl.nn as dglnn import dgl.function as fn import torch import torch.nn as nn num_features = 3

Define the nodes and edges

num_users = 10 num_items = 8 g = dgl.heterograph({ ('user', 'connects', 'item'): (range(num_users), range(num_users, num_users + num_items)) })

Set node features for user and item nodes

g.nodes['user'].data['user_features'] = torch.randn(num_users, num_features) g.nodes['item'].data['item_features'] = torch.randn(num_items, num_features)

print(g)

class HAN(nn.Module): def init(self, g, in_size, hidden_size, out_size, num_heads): super().init()

    # Define the layers for each node type
    self.layers = nn.ModuleDict({
        'user': dglnn.SAGEConv(in_size, out_size, 'mean'),
        'item': dglnn.SAGEConv(in_size, out_size, 'mean')
    })

def forward(self, g, inputs):
    h_dict = inputs
    for ntype in g.ntypes:
        layer = self.layers[ntype]
        g.nodes[ntype].data['h'] = h_dict[ntype]
        g.update_all(fn.copy_u('h', 'm'), fn.mean('m', 'h_agg'))
        h_dict[ntype] = layer(g.nodes[ntype], h_dict[ntype])
    return h_dict

Example usage

model = HAN(g, num_features, 16, 16, 2) output = model(g, {'user': g.nodes['user'].data, 'item': g.nodes['item'].data}) `

`--------------------------------------------------------------------------- DGLError Traceback (most recent call last) in <cell line: 11>() 9 num_users = 10 10 num_items = 8 ---> 11 g = dgl.heterograph({ 12 ('user', 'connects', 'item'): (range(num_users), range(num_users, num_users + num_items)) 13 })

2 frames /usr/local/lib/python3.10/dist-packages/dgl/heterograph_index.py in create_unitgraph_from_coo(num_ntypes, num_src, num_dst, row, col, formats, row_sorted, col_sorted) 1290 if isinstance(formats, str): 1291 formats = [formats] -> 1292 return _CAPI_DGLHeteroCreateUnitGraphFromCOO( 1293 int(num_ntypes), 1294 int(num_src),

dgl/_ffi/_cython/./function.pxi in dgl._ffi._cy3.core.FunctionBase.call()

dgl/_ffi/_cython/./function.pxi in dgl._ffi._cy3.core.FuncCall()

DGLError: [18:56:58] /opt/dgl/src/graph/unit_graph.cc:71: Check failed: src->shape[0] == dst->shape[0] (10 vs. 8) : Input arrays should have the same length.`

[rudongyu]

Hi, the number of source nodes and the number of target nodes should always keep the same. You can repeat the node IDs to represent multiple edges from or to them, e.g.,

src_nodes = [0, 1]
tgt_nodes = [1, 1]

It means both node 0 and node 1 are connected to node 1.

[github-actions[bot]]

This issue has been automatically marked as stale due to lack of activity. It will be closed if no further activity occurs. Thank you