rashad101 / rome Goto Github PK

While trying to install RoMe, I came across a few more things I needed to do to get it to work on my machine:

1. Download stanza model, by opening up a python interpreter, and:

import stanza
stanza.download('en')
stanza_nlp = stanza.Pipeline('en')

2. Download pretrained fasttext bin and vec files

• download it from https://dl.fbaipublicfiles.com/fasttext/vectors-wiki/wiki.simple.zip
• extract contents into data/

Maybe worthwhile putting this into the README file. :)

Hello,

thank you very much for releasing the code for this work!

using the default setting of RoMe, the program got aborted due to some memory allocation issues like “*** Error in python': corrupted size vs. prev_size: 0x000055ee10091dd0 ***” or “python *** Error in python': free(): invalid next size (fast)”.

I noticed that it occurs only when evaluating some specific texts, for example:

hyp = "Cheb's neighbors from the German side of the border gave performances including some with reconstructed Jak 18T aircraft and three-member formations using Bücker Jungmann biplanes"
ref = "Germany's neighbors in Chisinau demonstrated with reconstructed aircraft How 18t or three formations dvouplošníků Bücker Jungmann." (for this one, when removing the word "dvouplošníků", RoMe could work.)

hyp = "Since then, the physicists have referred to the possibility of additional small spatial dimensions as the Kaluza-Klein Theory."
ref = "Since then physicists call the possibility of additional small spatial dimensions Kaluzovou-Kleinovou theories."

I recreated exactly the same environment as required, but it still doesn't work.

Could you please help me out? Thanks a lot!

Kind regards,