Seaborn heatmap, format of annotations in non-diagonal cells

Question

I generated a heatmap using seaborn version 10.1. I would like to format the annotations of each cell with fmt='.2g'

However, this seems to only affect the cells on the diagonal.

import seaborn as sn

x = np.array([[0.99082, 0.00102, 0.0, 0.0],
              [0.0, 0.98767, 0.00529, 0.00088],
              [0.01744, 0.00097, 0.94961, 0.00291],
              [0.00990, 0.00099, 0.00594, 0.94356]])

sn.heatmap(x, annot=True, fmt='.2g', cmap=plt.cm.Blues)

What I get is this:

I have been through the Seaborn documentation, but I couldn't find any setting that applies the formatting to the annotations of the non-diagonal cells. Does anybody have an idea how to do that?

Edit: For example, I would expect 0.017 be formatted as 0.02 and 0.0099 be formatted as 0.01, but 0.00 should be 0.

Answer 1

• Use fmt='0.2f', which comes from str.format() notation, to set the format for all annotations
  • Real Python: A Guide to the Newer Python String Format Techniques
  • mkaz.blog: Python String Format Cookbook
• To change values of '0.00' to '0', use the matplotlib.text API.

import seaborn as sns

x = np.array([[0.99082, 0.00102, 0.0, 0.0],
              [0.0, 0.98767, 0.00529, 0.00088],
              [0.01744, 0.00097, 0.94961, 0.00291],
              [0.00990, 0.00099, 0.00594, 0.94356]])

sns.heatmap(x, annot=True, fmt='0.2f', cmap=plt.cm.Blues)

# update the desired text annotations
for text in ax.texts:
    if text.get_text() == '0.00':
        text.set_text('0')

Reference:

change certain squares in a seaborn heatmap