pytc.vmc.optimizer.create_optimizer

pytc.vmc.optimizer.create_optimizer(optimizer_type, learning_rate, opt_kwargs=None)

Create an optimizer based on specified type and parameters.

Parameters:

• optimizer_type – “adam”, “sgd”, “rmsprop”, “lion”, or “newton”

• learning_rate – Initial learning rate (float) or an optax schedule (callable).

• opt_kwargs – Additional optimizer parameters. For optax optimizers: can include ‘decay_rate’ and ‘transition_steps’ to customize the default 1/(1+t) schedule. For Newton: can include ‘min_learning_rate’ (default 0.01).