One training step updates the table from the current gradients and then
evaluates the loss and the next gradients.
295def bobTrainingStep =
296 (lambda unrestricted scalars : Bytes .
297 (bobThen
298 (bobLaunch b"bob-adamw" (bobKernel bobAdamPlace 256) (bobAdamBlock scalars))
299 (bobLossLaunch bobTrainingLossAddress)))The compiler supplied declaration spans and resolved links from this source snapshot. This page does not assert that this file belongs to a checked closure.