Encountering ValueError with Tensor Dimension

[isolated-paul]

Hi,
Thank you for sharing the code. I am trying to run the experimentation on Google Colab. But I keep running into a problem when I am trying to train or test the model.

Could you please help me understand what is the source of this problem because I am having difficulty understanding it.

Basically, Colab gives me the following error:

`ValueError Traceback (most recent call last)

in <cell line: 5>()
5 with torch.no_grad():
6 for i, sample in enumerate(tqdm.tqdm(test_loader, desc="Visualising")):
----> 7 prediction = model.to(device)(sample.to(device)) ## Don't understand this line
8 # target = sample.y.clone()
9 # print('\n')

6 frames

/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py in _call_impl(self, *args, **kwargs)
1499 or _global_backward_pre_hooks or _global_backward_hooks
1500 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1501 return forward_call(*args, **kwargs)
1502 # Do not call functions when jit is used
1503 full_backward_hooks, non_full_backward_hooks = [], []

in forward(self, data)
95 # Downstream
96 copy0 = x.clone()
---> 97 x = self.pool1(x, data)
98 x = self.conv11(x, *scale_attr[1])
99 x = self.conv12(x, *scale_attr[1])

/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py in _call_impl(self, *args, **kwargs)
1499 or _global_backward_pre_hooks or _global_backward_hooks
1500 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1501 return forward_call(*args, **kwargs)
1502 # Do not call functions when jit is used
1503 full_backward_hooks, non_full_backward_hooks = [], []

/usr/local/lib/python3.10/dist-packages/gem_cnn/nn/pool.py in forward(self, x, data)
44 size = (num_fine, num_coarse)
45
---> 46 out = self.propagate(edge_index=edge_index, x=x, connection=connection, size=size)
47 return out
48

/usr/local/lib/python3.10/dist-packages/torch_geometric/nn/conv/message_passing.py in propagate(self, edge_index, size, **kwargs)
457 kwargs[arg] = decomp_kwargs[arg][i]
458
--> 459 coll_dict = self._collect(self._user_args, edge_index, size,
460 kwargs)
461

/usr/local/lib/python3.10/dist-packages/torch_geometric/nn/conv/message_passing.py in _collect(self, args, edge_index, size, kwargs)
333
334 if isinstance(data, Tensor):
--> 335 self._set_size(size, dim, data)
336 data = self._lift(data, edge_index, dim)
337

/usr/local/lib/python3.10/dist-packages/torch_geometric/nn/conv/message_passing.py in _set_size(self, size, dim, src)
243 size[dim] = src.size(self.node_dim)
244 elif the_size != src.size(self.node_dim):
--> 245 raise ValueError(
246 (f'Encountered tensor with size {src.size(self.node_dim)} in '
247 f'dimension {self.node_dim}, but expected size {the_size}.'))

ValueError: Encountered tensor with size 24269 in dimension 0, but expected size 9708.`

[isolated-paul]

Is the problem related to how the dataset is being transformed under the "pre-computed graph transform" set up and how the datasets are being transformed and loaded?

The ValueError shown above is with respect to the first dimension, in this case the dimension 0 value 24269.

Here are two samples from the "test_loader" :

(device = 'cpu')

Visualising: 100%|██████████| 5/5 [00:00<00:00, 77.37it/s]

sample 0
MultiscaleDataBatch(y=[24269, 3], pos=[24269, 3], face=[3, 48534], inlet_index=[1626], normal=[24269, 3], edge_index=[2, 271008], geo=[24269], mask=[24269], node_mask=[24269], edge_mask=[271008], edge_coords=[271008, 2], connection=[271008], weight=[271008], frame=[24269, 3, 3], matrix_features=[24269, 7, 5], batch=[24269], ptr=[2])

sample 1
MultiscaleDataBatch(y=[18737, 3], pos=[18737, 3], face=[3, 37470], inlet_index=[923], normal=[18737, 3], edge_index=[2, 207999], geo=[18737], mask=[18737], node_mask=[18737], edge_mask=[207999], edge_coords=[207999, 2], connection=[207999], weight=[207999], frame=[18737, 3, 3], matrix_features=[18737, 7, 5], batch=[18737], ptr=[2])

[sukjulian]

Hi! It might be that the version of PyG that Google colab is using is incompatible with gem_cnn (see here). Could you try resetting PyG to version 2.0.3?