Create matplotlib figures that can be easily updated at runtime in Jupyter.
import numpy as np
from dyn_plot import Plot
# accepts all plt.figure() arguments
plot = Plot(figsize=(10, 5))
ts = np.linspace(0, 10, 200)
losses = []
for t in ts:
loss = -np.log(t + 0.001) + np.random.normal() * 0.4
losses.append(loss)
with plot: # clear and redraw
# use any methods from Axes
plot.set_title(f't = {t:.2f}')
plot.plot(losses)
# close to avoid an additional rendering of the inner figure
plot.close() 