Running build_vocab.py for wmt17_bpe with or without --joint? about joeynmt

[AlexWang1900]

I think I found the bug:
in wmt17_ende_bpe.py:

src:
        lang: "en"
        max_length: 512
        lowercase: False
        normalize: True
        level: "bpe"
        voc_min_freq: 1
        **voc_limit: 32000**
        voc_file: "subwords/wmt17_bpe.vocab"
        tokenizer_type: "subword-nmt"
        tokenizer_cfg:
            num_merges: 32000
            codes: "subwords/wmt17_bpe.codes"
            dropout: 0.1
            pretokenizer: "moses"

left a line : voc_limit: 32000

[may-]

Hi @AlexWang1900 ,

you are right, for joined vocab, src and trg config should be identical.
thank you for letting us know this! we will fix the config file for wmt17 ende.

[may-]

fixed in #211