You can fix the problem on Mac OS X Mavericks and Yosemite by doing the following:
cd ~/.matplotlib/
fc-list # Should take a couple minutes.
After fc-list finishes running you will be able to import the package using import matplotlib.pyplot as plt.
In your example if you use PCA to initialize your t-SNE you get widely spaced centroids; if you use random initialization you'll get tiny centroids and an uninteresting picture.
for i, txt in enumerate(n):
ax.annotate(txt, (z i » ,y i » ))