optax.scale_by_adamax

optax.scale_by_adamax#

optax.scale_by_adamax(
b1: jax.typing.ArrayLike = 0.9,
b2: jax.typing.ArrayLike = 0.999,
eps: jax.typing.ArrayLike = 1e-08,
) optax.GradientTransformation[source]#

Rescale updates according to the Adamax algorithm.

See optax.adamax() for more details.

Parameters:
  • b1 โ€“ Decay rate for the exponentially weighted average of grads.

  • b2 โ€“ Decay rate for the exponentially weighted maximum of grads.

  • eps โ€“ Term added to the denominator to improve numerical stability.

Returns:

A optax.GradientTransformation object.