Optimizers

The optimizer updates model parameters after each Trainer batch. It is selected by the trainer.optimizer field and is created by the trainer from the registered optimizer class.

Available implementations

adam (AdamOptimizer)
sgd (SGDOptimizer)
rmsprop (RMSpropOptimizer)
adamw (AdamWOptimizer)
adagrad (AdagradOptimizer)

Common configuration

learning_rate is configured beside optimizer in the trainer block.

Example

trainer:
  optimizer: AdamW
  learning_rate: 0.001

API reference

See xanesnet.components.optim for the XANESNET optimizer registry.