ULMFit for non-English Languages (NLP Video 10)

[@OleksandrFialko]

Jeremy is an amazing educator !

[@yulipurwanti6980]

Giededwie

[@sheikhakbar2067]

Thanks ... I will be watching the rest of the video later.

[@SupeHero00]

This was amazing! Thanks :)

[@dr.flintwestwood7004]

Using ensembles is sure a good thing, but it seems like for the Thai sentiment tasks, the F1 is identical for forward-only and for ensemble. Also Jeremy loads the (forward) encoder at 56:06 for the backwards classifier. Is this a bug?!

[@nicooteiza]

Looks like a bug, yes

[@klingefjord]

Good spot!

[@ToukoApajalahti]

Is there a jupyter notebook with the code for the vietnamese/turkish part of the video? The one I found ends before that part and only has an appendix "language zoo". Basically I'm searching for some examples how to use the pretrained models that I can find from the "language zoo".

[@cansencaglayan]

Hello I'm having trouble with nlputils. get_wiki keeps giving the following error: No such file or directory: '/root/.fastai/data/trwiki/text/AA/wiki_00' Any suggestions?

[@noahdrisort2005]

Việt nam lên sóng

[@mohammedhasan6522]

In Notebook nn-vietneme the following code is not working get_wiki(path,lang) is giving error. The error says [Errno 2] No such file or directory: 'C:\\Users\\hasan\\.fastai\\data\\trwiki\\text\\AA\\wiki_00'. I tried to see what's going on? This error occurs because of this line os.system("python wikiextractor/WikiExtractor.py --processes 4 --no_templates " + f"--min_text_length 1800 --filter_disambig_pages --log_file log -b 100G -q {xml_fn}"). It is not converting the xml file to text file, therefore when shifting file through shutil it founds no file and creates problem. I don't know why is this. It would be helpful if anyone can help me in this manner

[@rahuluppari732]

Even Im getting the same error... Have you able to solve it?

[@mohammedhasan6522]

@@rahuluppari732 No, But the problem is the relevant package we are using on that. I assume it is because of the windows system. Maybe they have written it for Linux system (could be wrong). You have to search for another package that can process the XML file and clean it.

[@rahuluppari732]

@@mohammedhasan6522 brother have you tried the same thing in linux?? If yes did it worked for you??

[@dalissonfigueiredo]

anybody else having trouble with half precision despite using dynamic scaling.

[@31.nguyenthisongthuong3]

46:30