Runtime Error: 'lengths' argument should be a 1D CPU int64 tensor, but got 1D cuda:0 Long tensor

[goobill]

I've been following the chatbot tutorial (https://pytorch.org/tutorials/beginner/chatbot_tutorial.html#run-model) and discovered the code does not run out of the box for me.

After training the model and inputting custom test data for the bot to respond I get the error:

RuntimeError: 'lengths' argument should be a 1D CPU int64 tensor, but got 1D cuda:0 Long tensor

I found a similar issue with a fix on the Pytorch Github issues, pytorch/pytorch#43227

Their solution was to use the .cpu() on the length argument.

On line 683, of https://github.com/pytorch/tutorials/blob/master/beginner_source/chatbot_tutorial.py, I applied the .cpu() method to the variable input_sequence and that resolved my issue.

[chaojihuihui]

well

[krkv]

Ran into the same issue, but the variable input_sequence is not there anymore, so the provided solution is not applicable.

[holly1238]

Closing this issue. This issue was fixed and merged in #1535