Relaxation History

View the loss convergence from generate() or shell_relax():

cell.plot_shell_relax()

Returns a matplotlib figure showing:

  • Total loss (sum of component losses)

  • Best loss (running minimum)

  • Bond loss (mean squared deviation from target bond distance)

  • Angle loss (mean squared deviation from target bond angle)

  • Repulsion loss (overlap and non-bonded clearance violations per atom)

Parameters

cell.plot_shell_relax(log_y=False)

  • log_y: if True, use a log scale for the y-axis (useful when loss spans many orders of magnitude).

Accessing raw history

The history arrays are stored on the supercell:

hist = cell.shell_relax_history
hist["step"]              # step index
hist["loss"]              # total loss per step
hist["best_loss"]
hist["bond_loss"]
hist["angle_loss"]
hist["repulsion_loss"]