seaborn cycle through colours with matplotlib scatter

To build on Carsten's answer, if you have a large number of categories to assign colours to, you might wish to zip the colours to a very large seaborn palette, for example the xkcd_palette or crayon_palette.. Note that this practice is usually a chartjunk anti-pattern: using more than 5-6 colours is usually overkill, and you might need to consider changing your chart type.

import matplotlib.pyplot as plt
import seaborn as sns

palette = zip(df['category'].unique(), sns.crayons.values())

You have to tell matplotlib which color to use. To Use, for example, seaborn's default color palette:

import matplotlib.pyplot as plt
import seaborn as sns
import itertools
ax=fig.add_subplot(111)

palette = itertools.cycle(sns.color_palette())

for f in files:
    ax.scatter(args, color=next(palette))

The itertools.cycle makes sure we don't run out of colors and start with the first one again after using the last one.

Update:

As per @Iceflower's comment, creating a custom color palette via

palette = sns.color_palette(None, len(files))

might be a better solution. The difference is that my original answer at the top iterates through the default colors as often as it has to, whereas this solution creates a palette with as much hues as there are files. That means that no color is repeated, but the difference between colors might be very subtle.