To preprocess wikipedia-vectors.csv into the format in which you used it in the exercises, you have to take its transpose:

import pandas as pd
from scipy.sparse import csr_matrix

df = pd.read_csv('wikipedia-vectors.csv', index_col=0)
articles = csr_matrix(df.transpose())
titles = list(df.columns)

The reason for taking this transpose is that without it, there would be 13,000 columns (corresponding to the 13,000 words in the file), which is a lot of columns for a CSV to have.