installation issue in windows 10 about easse HOT 7 CLOSED

Hi!
It seems to me that the key error message is here:

error: [WinError 2] The system cannot find the file specified msvc py_compiler msvc

Does the solution outline here help?

from easse.

Hi Fernando,
Thanks, it helped me to get clue. I had spacy version 2.2.2.& Python 3.8

So, to resolve it i created a virtual environment with python 3.6.
and install easse. it works now but can you please have a quick look on this error while running my dataset on colab link

https://colab.research.google.com/drive/1iTId7uPlMNMMeSz6QuY1CDFxU6GBv4qY?usp=sharing

you can close this issue.

from easse.

Hi!
Note that for both corpus_sari and corpus_bleu the refs_sents argument is List[List[str]]. This is to account for the general case where multiple references are available for each instance. You didn't get an error with corpus_sari because we don't make the same verification as sacrebleu (maybe we should), but the sari score you get is most likely incorrect. You should change the way you are calling the functions so that, for each instance, the refs_sents argument is a List.

from easse.

OK, so it means i cannot use easse or i need to change it files .[bleu.py, sari.py]
sys_sent: List[str], ref_sents: List[str],

But my input has will be like it. as i am doing translation
https://colab.research.google.com/drive/1iTId7uPlMNMMeSz6QuY1CDFxU6GBv4qY?usp=sharing

from easse.

It means that you need to reformat your input so that what you pass in the refs_sents argument is a List[List[str]]. For example, you could change the way you load your data into the dataframe so that each item in the Target_Text column is a list of only one element.

from easse.

ok thank you

from easse.

Hi!
Note that for both corpus_sari and corpus_bleu the refs_sents argument is List[List[str]]. This is to account for the general case where multiple references are available for each instance. You didn't get an error with corpus_sari because we don't make the same verification as sacrebleu (maybe we should), but the sari score you get is most likely incorrect. You should change the way you are calling the functions so that, for each instance, the refs_sents argument is a List.

@feralvam can you create a new issue highlighting what we need to do ideally to fix this problem in the future?

from easse.