where I tried the supervised.py, I got "ValueError: The input must have at least 3 entries!" about muse HOT 11 CLOSED

The problem is solved for now.
I just re-built the docker image with pytorch. Then everything goes fine. And the difference between the two images is the python version. So I think the problem may be the adaption of pytorch with python3.6.

from muse.

Hello,

Did you run the get_evaluation.sh script inside the data folder?

from muse.

@glample Thank you.
And yes, this script is to download all the corpus, wordvec models and dictionaries. You mean the problem maybe in these data?

from muse.

I'm not sure what the problem is, and I was wondering whether the evaluation files were properly downloaded.

Can you comment out this line: https://github.com/facebookresearch/MUSE/blob/master/src/evaluation/evaluator.py#L188 and see what happens?

from muse.

It seems worked.
here is the log message for now
logs.txt

I wonder why?

from muse.

I'm not sure this is working. The translation accuracy is 0%. What are the embeddings you are using?

from muse.

@glample
Oh, THE ACCURACY is 0!!
But, I downloaded them with the url given on the page fastText Wikipedia embeddings, . and copy them into the docker image. Do you think the environment may cause these trouble?

from muse.

Can you try to re-download the Spanish embeddings using this command?

curl -Lo data/wiki.es.vec https://s3-us-west-1.amazonaws.com/fasttext-vectors/wiki.es.vec

And see if this helps?

from muse.

@glample sorry to trouble so much.
I think the problem is in the environment. Yestoday, I reinstall the environment with the pytorch/pytorch docker image which used the python3.5. Since there says pytorch's system requirement is python3.5 or python2.7, and my former py is python3.6.
Thank you very much. ^^.

from muse.

Not sure I get it. Did you manage to fix the problem?

from muse.

Sounds good, thank you. I'll close the issue then

from muse.