CUDA error: an illegal memory access was encountered

[tjusxh]

I implement the vae (gcn) according to your code. But when I run the code, I encounter the following error in the second iteration. Do you encounter the problem?

step0 tensor(0, device='cuda:0')
step1 tensor(0, device='cuda:0')
step2 Traceback (most recent call last):
File "train.py", line 235, in
train_loss = train(epoch)
File "train.py", line 156, in train
z = model.encode(x, edge_idxs)
File "/root/hair_vae/vgae/model/vgae.py", line 108, in encode
self.mu, self.logvar = self.encoder(*args, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 550, in call
result = self.forward(*input, **kwargs)
File "/root/hair_vae/vgae/model/vgae.py", line 44, in forward
print("step2", torch.isnan(x).sum())
File "/usr/local/lib/python3.6/dist-packages/torch/tensor.py", line 162, in repr
return torch._tensor_str._str(self)
File "/usr/local/lib/python3.6/dist-packages/torch/_tensor_str.py", line 315, in _str
tensor_str = _tensor_str(self, indent)
File "/usr/local/lib/python3.6/dist-packages/torch/_tensor_str.py", line 213, in _tensor_str
formatter = _Formatter(get_summarized_data(self) if summarize else self)
File "/usr/local/lib/python3.6/dist-packages/torch/_tensor_str.py", line 84, in init
value_str = '{}'.format(value)
File "/usr/local/lib/python3.6/dist-packages/torch/tensor.py", line 418, in format
return self.item().format(format_spec)
RuntimeError: CUDA error: an illegal memory access was encountered

x = F.elu(self.conv1(x, edge_indexs[0]))
print("step0", torch.isnan(x).sum())
x = Pool(x, self.down_samples[0])
print("step1", torch.isnan(x).sum())
x= self.conv2(x, edge_indexs[1])
print("step2", torch.isnan(x).sum())