RuntimeError: expected scalar type Long but found Float

[bakszero]

After training the model successfully, I ran the translate command after which I get the following runtime error:

  File "translate.py", line 36, in <module>
    main(opt)
  File "translate.py", line 21, in main
    attn_debug=opt.attn_debug
  File "/mnt/disks/disk-huge/bakhtiyar/ITDD/onmt/translate/translator.py", line 205, in translate
    batch, data, attn_debug, fast=self.fast
  File "/mnt/disks/disk-huge/bakhtiyar/ITDD/onmt/translate/translator.py", line 309, in translate_batch
    return self._translate_batch(batch, data)
  File "/mnt/disks/disk-huge/bakhtiyar/ITDD/onmt/translate/translator.py", line 706, in _translate_batch
    beam_attn.data[j, :, :memory_lengths[j]])
  File "/mnt/disks/disk-huge/bakhtiyar/ITDD/onmt/translate/beam.py", line 139, in advance
    self.attn.append(attn_out.index_select(0, prev_k))
RuntimeError: expected scalar type Long but found Float

I'm using Pytorch version 1.7.1 and torchtext 0.4.0, on Ubuntu 18.04.

I think the prev_k is a list of float values in this case which creates the error. If I change it to just self.attn.append(attn_out), I then get an error at ITDD/onmt/translate/translator.py", line 712, in <lambda> lambda state, dim: state.index_select(dim, select_indices)) RuntimeError: expected scalar type Long but found Float

Please could you help resolve the issue? Thanks!

[lizekang]

You can try torch 1.0.0. We use the version two years ago.

[bakszero]

Looks like it, retraining with 1.0.0 now. Thanks 👍