AdamaxWStep¶
The AdamaxW optimizer step (adamax with weight decay).
Based on Loshchilov et al, 2019 (but see the note about the weight decay parameter), AdamaxW is a variant of the Adamax optimizer that additionally regularizes weights to have small l2 norm, which helps with generalization, especially when the amount of training data is low relative to the number of parameters. This is analogous to l2 regularization terms on the weights, but works better with adaptive gradient algorithms like Adam. Like all step nodes, this node only processes gradients, and the resulting updates must be applied manually to the weights (this can be accomplished using the Add node). However, you can also pass it to the StepSolver node which implements the full optimization loop. The learning rate can instead be given as a schedule, by wiring one of the Schedule nodes into the learning_rate_schedule port. The weight decay can be used in conjunction with a mask data structure that has the same nested structure as the weights being optimized, but which contains booleans indicating which weights should be decayed. More Info... Version 0.2.0
Ports/Properties¶
gradients¶
Gradients to be transformed.
weights¶
Optional current weights.
state¶
Explicit state of the node.
learning_rate_schedule¶
Optional learning rate schedule.
weight_decay_mask¶
Mask structure for the weight decay.
learning_rate¶
Learning rate. A typical choice may be 0.001 here, but this is problem dependent. If a learning rate schedule is provided, this value should be left unspecified.
beta1¶
Exponential decay rate for the first moment estimates.
beta2¶
Exponential decay rate for the second moment estimates.
epsilon¶
Small value applied to the denominator outside the square root to avoid dividing by zero when rescaling. Note that larger epsilon values have been explored in the literature.
epsilon_inroot¶
Small value applied to the denominator inside the square root to avoid dividing by zero when rescaling. A case where this is needed is when differentiating the optimizer itself, eg for bilevel optimization.
weight_decay¶
Strength of the weight decay. This is multiplied by the learning rate as in e.g., PyTorch and Optax, but differs from the paper, where it is only multiplied by the schedule multiplier but not the base learning rate.
mu_precision¶
Numeric precision for the first-order accumulator. Keep resolves to the precision of the inputs.
set_breakpoint¶
Set a breakpoint on this node. If this is enabled, your debugger (if one is attached) will trigger a breakpoint.
metadata¶
User-definable meta-data associated with the node. Usually reserved for technical purposes.