An error was encountered while using .to_homogeneous()

[RENEK-bool]

I am currently planning to use HEATConv. I have read some discussions in the past (#6869) and found that I need to perform hetero_data.to_homogeneous() first to get the correct input format, but when I use .to_homogeneous() an error occurs.

wire_train_sample = train_dataset[0].to_homogeneous()
RuntimeError: Sizes of tensors must match except in dimension 0. Expected size 1 but got size 5 for tensor number 1 in the list.

This is my HeteroData format

HeteroData(
  pin={ x=[66704, 1] },
  (pin, cell_arc, pin)={
    edge_index=[2, 43152],
    edge_attr=[43152, 1],
  },
  (pin, net_arc, pin)={
    edge_index=[2, 42662],
    edge_attr=[42662, 5],
    y=[42662],
  }
)

Is it because I put the label y on the edge (pin, net_arc, pin) that caused the error?
In addition, I saw that HeteroData.to_homogenous() will change the edge order.(#8707) Will this cause my edge labels to not correspond? What I'm currently trying to achieve is edge regression.

[rusty1s]

This is probably because edge_attr is of different size in the two edge types. What you can do is to "pad" each edge_attr to size [num_edges, 6].