RuntimeError: index out of range at /pytorch/aten/src/TH/generic/THTensorEvenMoreMath.cpp:193

[xiaolibird]

When I feed my own data embedded by LINE
it happens:

And the problem arises at GrpahSGAN.py line 32
self.embedding_layer = nn.Embedding(dataset.n, dataset.d)

After searching, I found that the first parameter is actually related to vocabulary size, which I have no idea and could not find in LINE codes.

[Sleepychord]

I cannot not fully understand your problem. The first parameter is ``the number of nodes'' in the graph. You should build the dataset following the description in README.

[xiaolibird]

Sorry for my poor expression.
But I am building the dataset according to the description in README
dataset.n == dataset.embeddings.shape[0] == #nodes in graph
And there is something wrong with nn.Embedding(num_embeddings, embedding_dim) when processing inappropriate num_embeddings

And here is the traceback
Traceback (most recent call last):
File "GraphSGAN.py", line 224, in
gan.train()
File "GraphSGAN.py", line 109, in train
ll, lu, acc = self.trainD((id0, xf), y, idf_unlabel1)
File "GraphSGAN.py", line 48, in trainD
x_label, x_unlabel, y = self.make_input(*idf_label), self.make_input(*idf_unlabel), Variable(y, requires_grad = False)
File "GraphSGAN.py", line 91, in make_input
embedding = self.embedding_layer(Variable(ids, volatile = volatile)).detach() # detach temporarily
File "/path/to/python/site-packages/torch/nn/modules/module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "/path/to/python/site-packages/torch/nn/modules/sparse.py", line 117, in forward
self.norm_type, self.scale_grad_by_freq, self.sparse)
File "/path/to/python/site-packages/torch/nn/functional.py", line 1506, in embedding
return torch.embedding(weight, input, padding_idx, scale_grad_by_freq, sparse)
RuntimeError: index out of range at /pytorch/aten/src/TH/generic/THTensorEvenMoreMath.cpp:193

===
It seems that this issue arises at many embedding tasks, and some say it involves the vocabulary size of your embedding and it confuses me since I was using LINE for embedding and I am not doing a word embedding task.

chenxijun1029/DeepFM_with_PyTorch#1
https://stackoverflow.com/questions/51456059/pytorch-nn-embedding-error

[Sleepychord]

this error just means that you are asking for an embedding of a node whose Id is larger than vocabulary size...
you'd better double check all the parameters relevant to the number of nodes. I think this is not a big problem.

来自GitHub小程序客户端：

[xiaolibird]

Found it
Seems it load pretrianed network as default
Thanks a lot for your patience!