karansag / phrase-analogies-large-vae Goto Github PK

I'm currently looking into using Dask to help parallelize and speed up our evaluation of analogies within the dataframe, since each calculation takes nearly a second to do. I'll keep my notes on possible ideas in here

In dataset_constructors/opposite_syntax_analogies.py, the links to files seem to be for files outside of the repository on your local device. Can we set this up with the files in the repo or as a way we can wget the files when we set up our tests, so people can easily replicate our findings?

In other words, where can we access the snli_1.0_train.csv file, and is there a way we can add it to the file structure of the repository? Also can we store the Opposites_Dataset.csv the same way?

If these files are too big for us to store on github, we can also add them to our .gitignore.