Closed ChiZhangRIT closed 6 years ago
hi, this code was based on a previous version of tensor2tensor -- it was merged into main repo so you can run the translation experiment there.
Default tensor2tensor tokenizer was used, based on subword encoding (sennrich et al)
Thanks for sharing the code.
I was trying to run the code, but this is what I got:
BTW, how did you run BLEU evaluation on Chinese sentences? Did you use any tokenizer to get BLEU score?