Question: How to set the encoding to prevent unexpected conversion? about nmt HOT 8 CLOSED

我在pycharm上输出时加了decode（’utf-8‘）解决了，但是看其它人都没遇到这个问题很奇怪

from nmt.

For 1, I think it is because the tutorial logs are generated by python2. I will investigate later to make the logging works correctly for both python2 and 3.

For 2, We recently updated the flags. Change model_dir to out_dir should fix the problem. For example

python -m nmt.nmt \
    --out_dir=/tmp/nmt_model \
    --inference_input_file=/tmp/my_infer_file.vi \
    --inference_output_file=/tmp/nmt_model/output_infer

from nmt.

Thank you @chuangfortytwo @oahziur! I would try today and share you the result later.

from nmt.

@oahziur @lmthang @yanl10n I tried the vietnamese to english translation in the tutorial and got decoder output all in byte string format, i.e., string with a 'b' in front of it. Is this the expected behavior? Should I manually convert the output strings into utf-8? Thanks!

from nmt.

@chuangfortytwo ,那能不能在输出到文件的时候就给转码了（抱歉，pycharm没有接触过），还有如果加载到一个服务上，是否可以输出转码的内容

from nmt.

@sdlmw 最后计算bleu值的时候就是从文件读取的，可以在输出到文件时转码。服务这个没试过

from nmt.

sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8')

from nmt.

Thank you All for the supports! Confirmed all issues had been fixed now.

from nmt.