If I have code like this :

import matplotlib.pyplot as plt
import numpy as np

time_axis = np.arange(0, 30 + 1 / 50, 1 / 50)
another_axis = np.arange(0, 10, 1)

grid = np.zeros((len(another_axis), len(time_axis)))
fig, ax = plt.subplots(12, 1, figsize=(10, 15), constrained_layout=True)
cbar = None


for i in range(12):
    grid = np.random.random((len(another_axis), len(time_axis)))
    pcm = ax[i].pcolormesh(time_axis, another_axis, grid, vmin=0, vmax=1)
    cbar = fig.colorbar(pcm, ax=ax[i])

plt.show()

I have a problem when not all 12 grids can fit into memory at once. At the point that plt.show() is called, is matplotlib holding memory proportional to 12*grid.size? Is there anything I can do to reduce this or is this just the way plotting and pcolormesh work?