Raise 'IndexError: index out of range in self' when I use HGTConv with my own dataset.

[rickyqiao]

In my work, I build the Hetergouenous graph firstly, then I apply the HGTConv as the tutorial shown.

The custom KG is shown as:
HeteroData(
symptom={ x=[39, 128] },
component={ x=[19, 128] },
reason={ x=[17, 128] },
solution={ x=[18, 128] },
(symptom, take_place, component)={ edge_index=[2, 38] },
(symptom, cause_by, reason)={ edge_index=[2, 33] },
(symptom, how_to_fit, solution)={ edge_index=[2, 33] },
(component, component_parallel, component)={ edge_index=[2, 17] }
)

And my code as follow:

---

class HGT(torch.nn.Module):
def init(self, hidden_channels, out_channels, num_heads, num_layers):
super().init()

    self.lin_dict = torch.nn.ModuleDict()
    for node_type in data_IKG.node_types:
        self.lin_dict[node_type] = Linear(-1, hidden_channels)

    self.convs = torch.nn.ModuleList()
    for _ in range(num_layers):
        conv = HGTConv(hidden_channels, hidden_channels, data_IKG.metadata(),
                       num_heads, group='sum',cached=False)
    self.convs.append(conv)

    self.lin = Linear(hidden_channels, out_channels)

def forward(self, x_dict, edge_index_dict):
    for node_type, x in x_dict.items():
        x_dict[node_type] = self.lin_dict[node_type](x).relu_()

    for conv in self.convs:
        print(1)
        x_dict = conv(x_dict, edge_index_dict)
        print(2)
    return self.lin(x_dict['symptom'])

model_IKG = HGT(hidden_channels=64, out_channels=5,
num_heads=1, num_layers=1)

with torch.no_grad(): # Initialize lazy modules.
out = model_IKG(data_IKG.x_dict, data_IKG.edge_index_dict)

Then it raises:
Input In [324], in HGT.forward(self, x_dict, edge_index_dict)
29 for conv in self.convs:
30 print(1)
---> 31 x_dict = conv(x_dict, edge_index_dict)
32 print(2)
33 return self.lin(x_dict['symptom'])

---

Then the error message is:

---

File ~/.virtualenvs/xlq/lib/python3.8/site-packages/torch/nn/modules/module.py:1102, in Module._call_impl(self, *input, **kwargs)
1098 # If we don't have any hooks, we want to skip the rest of the logic in
1099 # this function, and just call forward.
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []

File ~/.virtualenvs/xlq/lib/python3.8/site-packages/torch_geometric/nn/conv/hgt_conv.py:159, in HGTConv.forward(self, x_dict, edge_index_dict)
156 v = (v_dict[src_type].transpose(0, 1) @ m_rel).transpose(1, 0)
158 # propagate_type: (k: Tensor, q: Tensor, v: Tensor, rel: Tensor)
--> 159 out = self.propagate(edge_index, k=k, q=q_dict[dst_type], v=v,
160 rel=self.p_rel[edge_type], size=None)
161 out_dict[dst_type].append(out)
163 # Iterate over node-types:

File ~/.virtualenvs/xlq/lib/python3.8/site-packages/torch_geometric/nn/conv/message_passing.py:309, in MessagePassing.propagate(self, edge_index, size, **kwargs)
306 for arg in decomp_args:
307 kwargs[arg] = decomp_kwargs[arg][i]
--> 309 coll_dict = self.collect(self.user_args, edge_index,
310 size, kwargs)
312 msg_kwargs = self.inspector.distribute('message', coll_dict)
313 for hook in self._message_forward_pre_hooks.values():

File ~/.virtualenvs/xlq/lib/python3.8/site-packages/torch_geometric/nn/conv/message_passing.py:202, in MessagePassing.collect(self, args, edge_index, size, kwargs)
200 if isinstance(data, Tensor):
201 self.set_size(size, dim, data)
--> 202 data = self.lift(data, edge_index, dim)
204 out[arg] = data
206 if isinstance(edge_index, Tensor):

File ~/.virtualenvs/xlq/lib/python3.8/site-packages/torch_geometric/nn/conv/message_passing.py:172, in MessagePassing.lift(self, src, edge_index, dim)
170 if isinstance(edge_index, Tensor):
171 index = edge_index[dim]
--> 172 return src.index_select(self.node_dim, index)
173 elif isinstance(edge_index, SparseTensor):
174 if dim == 1:

IndexError: index out of range in self

---

Please help me to fix the problem, much appreciate.

[cold-rivers]

Did you solve the problem later? If you use the pyg data format, how should the
fine-turn code be modified?