Plotting
hippynn.plotting
is only available if matplotlib is installed.
By default, hippynn will plot loss metrics over time when training ends.
On top of this, hippynn can make diagnostic plots during its evaluation phase.
For example, Let’s assume you have a molecule_energy
node that you are training to.
A simple plot maker would look like this:
from hippynn import plotting
plot_maker = hippynn.plotting.PlotMaker(
plotting.Hist2D.compare(molecule_energy, saved=True,shown=False),
plot_every=10
)
from hippynn.experiment import assemble_for_training
training_modules,db_info = assemble_for_training(train_loss, validation_losses, plot_maker=plot_maker)
The plot maker is thus passed to assemble_for_training()
and attached to the model evaluator.