Data loader query

[shrimonmuke0202]

If apply batch_size = 512 I get adj matrix with shape (512,46,46) but when I apply batch_size = 128 I obtain some adj matrix with shape (128,44,44) but I want all adj matrix with shape (128,46,46). How can I get this type of result?

for ,data in enumerate(train_loader):
x , edge_index, batch = data.x,data.edge_index,data.batch
x, = torch_geometric.utils.to_dense_batch(x,batch)
adj = to_dense_adj(edge_index,batch)
print(batch)
print(adj.shape)
print(data.c_size)

[rusty1s]

to_dense_adj provides an additional argument named max_num_nodes that should fix your issues.

[shrimonmuke0202]

If I fixed max_num_nodes = 43 and I want all the adjacency matrices of the same dimension then an error is shown

index 236672 is out of bounds for dimension 0 with size 236672

[rusty1s]

Can you show me an example to reproduce? For example, the following works for me:

from torch_geometric.data import DataLoader
from torch_geometric.datasets import TUDataset
from torch_geometric.utils import to_dense_batch, to_dense_adj

dataset = TUDataset('/tmp/MUTAG', name='MUTAG')
loader = DataLoader(dataset, batch_size=32)

for data in loader:
    x, _ = to_dense_batch(data.x, data.batch, max_num_nodes=40)
    adj = to_dense_adj(data.edge_index, data.batch, max_num_nodes=40)
    print(x.shape, adj.shape)

[shrimonmuke0202]

`from torch_geometric.data import DataLoader
from torch_geometric.datasets import TUDataset
from torch_geometric.utils import to_dense_batch, to_dense_adj

dataset = TUDataset('/tmp/MUTAG', name='MUTAG')
loader = DataLoader(dataset, batch_size=32)

for data in loader:
x, _ = to_dense_batch(data.x, data.batch)
adj = to_dense_adj(data.edge_index, data.batch,max_num_nodes=22)
print(x.shape, adj.shape)`

RuntimeError: index 15488 is out of bounds for dimension 0 with size 15488 this type of error occurs

[rusty1s]

max_num_nodes in to_dense_batch and to_dense_adj is expected to be larger than the maximum number of nodes in the graphs. I fixed this for to_dense_adj in master, but it is a bit trickier to implement for to_dense_batch.