It is not common to save word-vectors to CSV.
And, that file you've linked isn't typical dense high-dimensional word-vector embedding data.
Rather, there are a few named, meaningful scalar values from Spotify's analysis. (I see things like danceability, energy, speechiness, etc.)
An actual word2vec model doesn't usually label the values with such easily-interpretable names.
This might still be interesting multi-dimensional data for clustering/plotting, but you wouldn't at any point convert it to a "word2vec object" to do that.
(And, if you did either train some word-vectors from the artist/title info alone, or use external word-vectors to convert the artist/title to word-vector dimensions, the results might be disappointing - those words may not characterize the underlying patterns very well, except in some really coarse ways that are obvious from shared words (like love-songs having 'love' in the title, or those with 'remix' in the title being a little more dancey, etc).
What is your real ultimate goal in working with this data?