NLP challenge, GloVe+FastText


#1

Good day. Thank you for GloVe in public directory. Can you add FastText by Facebook?


#2

It is already added on the dev branch along with the word2vec from google.
I also added architectures that involve everything that was mentioned in the “best auc score single models” thread like weighted average attention, concatenating max pool and mean pool and other tricks on the dev-architectures.
I will have it all on master by the end of this week. In the meantime feel free to use dev branch .


#3

Sorry, tell me please how can i add FastText for my Neptune notebook?
Maybe it is possible to add this set in: ! ls /public/…?
Or tell me please how can i upload this set (i have Windows).


#4

We will add those corpora to /public but most likely you will be able to use it on Monday.
In the meantime I would simply upload it using https://docs.neptune.ml/advanced-topics/storage/ :
neptunde data upload fasttext_corpus.txt
and then when starting new notebook just add fasttext_corpus.txt as explained here: https://docs.neptune.ml/advanced-topics/notebooks/


#5

Thank you for help ))

Unfortunatelly i can’t use ‘neptunde data upload fasttext_corpus.txt’…
File “”, line 1
neptunde data upload fasttext_corpus.txt
^
SyntaxError: invalid syntax

I believe that in Monday i can take this data from ls /public/


#6

I am sorry for the typo it was supposed to be neptune data upload fasttext_corpus.txt


#7

Thank you )) now I can work with FastText set. Sorry, maybe you can give me advice, how can I upload data with Windows?


#8

Unfortunately you cannot use neptune data upload on Windows at the moment. We are aware of the problem and we are working on fixing it.

There is alternative solution here.
You can download your data directly from your notebook using curl.
For example using this command in your notebook:
!curl -O -X GET https://www.kaggle.com/c/8076/download/train.csv.zip
will download train data to your experiment source directory for Kaggle Toxic Comment Classification Challenge.


#9

Thank you ) i will try