Saving a GPy plot as an image file

How can I save a GPy plot/visualization to an image file? I am on a Linux system with no attached display, so this is my only option for viewing visualizations.

This is my current attempt:

import GPy
import numpy as np

# sample inputs and outputs
X = np.random.uniform(-3.,3.,(50,2))
Y = np.sin(X[:,0:1]) * np.sin(X[:,1:2])+np.random.randn(50,1)*0.05

# define kernel
ker = GPy.kern.Matern52(2,ARD=True) + GPy.kern.White(2)

# create simple GP model
m = GPy.models.GPRegression(X,Y,ker)

# optimize and plot
m.optimize(messages=True,max_f_eval = 1000)

GPy.plotting.change_plotting_library("matplotlib")
fig = m.plot()
fig.savefig("model.png")

I get an error on the fig = m.plot() line:

Unable to access the X Display, is $DISPLAY set properly?

It seems like m.plot() is trying to show the figure immediately. How can I avoid this and only write to an image file?

[timothy-shields]

I have resolved the issue. Here is the code in case anyone else encounters the same problem:

import matplotlib
import GPy
import numpy as np

# Configure GPy -> matplotlib -> Agg
matplotlib.use("Agg")
GPy.plotting.change_plotting_library("matplotlib")

X = np.random.uniform(-3., 3., (50, 2))
Y = np.sin(X[:, 0:1]) * np.sin(X[:, 1:2]) + np.random.randn(50, 1) * 0.05
kernel = GPy.kern.Matern52(2, ARD=True) + GPy.kern.White(2)
model = GPy.models.GPRegression(X, Y, kernel)
model.optimize(messages=True, max_f_eval = 1000)

ax = model.plot()
ax.figure.savefig("model.pdf")

where the key lines are

# Configure GPy -> matplotlib -> Agg
matplotlib.use("Agg")
GPy.plotting.change_plotting_library("matplotlib")

[sgugler]

I found a workaround: If you print out the dictionary from ax = m.plot(plot_limits=[-15, 15], samples=5), so print(ax), we see that it contains a dictionary,

{'dataplot': [<matplotlib.collections.PathCollection object at 0x7f3d9f6f2f60>], 'gpmean': [[<matplotlib.lines.Line2D object at 0x7f3d9f6f2eb8>]], 'gpconfidence': [<matplotlib.collections.PolyCollection object at 0x7f3d9f6f27b8>]}

the first entry of which is the matplotlib plot we want. So the following worked for me:

dataplot = ax['dataplot'][0] 
dataplot.figure.savefig("gp-test.pdf")