WMT German to English news translation - Missing Metrics and Module 'apply_bpe' not found

I have two problems related with the WMT German to English news translation tutorial.

1. After training the WMT Model from the Tutorial 2 https://awslabs.github.io/sockeye/tutorials/wmt.html I noticed, that the ‘metrics’ folder was not generated inside the wmt_model folder. Is the metrics folder only generated when I set --optimized-metric bleu as parameter or is an error happened while training?

2. If I want to do the Translation as described with the code

   echo "er ist so ein toller Kerl und ein Familienvater ." | \
           python -m apply_bpe -c bpe.codes --vocabulary bpe.vocab.en \
           --vocabulary-threshold 50 | \
           python -m sockeye.translate -m wmt_model 2>/dev/null | \
           sed -r 's/@@( |$)//g'  
   

   I get the error message ModuleNotFoundError: No module named 'apply_bpe'. After installing subwort-nmt via pip, it was working, but I don’t receive nothing. What could be the source of the error?

[fhieber]

Add --use-cpu to the Sockeye command. I’ll update the tutorial to conform to the assumption made at the beginning of the tutorial regarding commands on CPU/GPU.

[RamoramaInteractive]

I think I found a solution for my problem. If I rewrite the Sockeye code like in the tutorial https://awslabs.github.io/sockeye/tutorials/wmt_large.html then it starts to train as if by magic.

Thank you for you help and I think this issue can be closed.