uhauha2929 / examples Goto Github PK

model = AttnClassifier(len(TEXT.vocab), embedding_dim, hidden_dim).to(device)

请问这一句报错的原因是什么：
pytorch1.2

File "E:\软件\python\lib\site-packages\torch\nn\modules\module.py", line 432, in to
return self._apply(convert)
File "E:\软件\python\lib\site-packages\torch\nn\modules\module.py", line 208, in _apply
module._apply(fn)
File "E:\软件\python\lib\site-packages\torch\nn\modules\rnn.py", line 124, in apply
self.flatten_parameters()
File "E:\软件\python\lib\site-packages\torch\nn\modules\rnn.py", line 97, in flatten_parameters
if not any_param.is_cuda or not torch.backends.cudnn.is_acceptable(any_param):
File "E:\软件\python\lib\site-packages\torch\backends\cudnn_init.py", line 118, in is_acceptable
if libcudnn() is None:
File "E:\软件\python\lib\site-packages\torch\backends\cudnn_init.py", line 61, in libcudnn
lib = find_cudnn_windows_lib()
File "E:\软件\python\lib\site-packages\torch\backends\cudnn_init.py", line 45, in find_cudnn_windows_lib
out = out.decode().strip()
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xc8 in position 3: invalid continuation byte

Hi,I have a question in self-attention.ipynb about this code as below:
out = out[:, :, :self.hidden_dim] + out[:, :, self.hidden_dim:]

Why do you add the two hidden states of Bidirectional LSTM.
Why not use the torch.cat()?

thanks

您好，我试着用别的数据集跑了下textual entailment部分的代码，与SNLI对比数据集的格式是一致，在最后一步运行train()的时候报错_IndexError: too many indices for tensor of dimension 1_，我不太清楚是哪里处理问题，另外在运行这行代码iter(train_iter).__next__()时我的输出与与您不太一样，结果如下：
[torchtext.data.batch.Batch of size 256]
[.premise]:[torch.cuda.LongTensor of size 44x256 (GPU 0)]
[.hypothesis]:[torch.cuda.LongTensor of size 25x256 (GPU 0)]
[.label]:[torch.cuda.LongTensor of size 256 (GPU 0)]
请问是否是这一步出现了问题呢，十分感谢您的帮助，谢谢！