I am following the tutorial Simple_MultiGPU_GP_Regression and I noticed that during the training two options were established:
with gpytorch.beta_features.checkpoint_kernel(checkpoint_size), \
gpytorch.settings.max_preconditioner_size(preconditioner_size):
What are or refer to the variables checkpoint_kernel
and preconditioner_size
?
I have check the documentation preconditioner_size but it is not quite clear to me what it refers to.
I intuit that checkpoint_size
refers to something related to the number of training points associated with each GPU. But it is just an intuition.
Help is appreciated.
Thanks