Open SamDuffield opened 5 months ago
Would be very useful to be able to support lr = lambda i: 1e-3 * i ** -0.5 to support Robbins-Monro style scheduling (and others)
lr = lambda i: 1e-3 * i ** -0.5
Up for debate but could be useful to add this support for the temperature parameter in the spirit of simulated annealing
temperature
Would be very useful to be able to support
lr = lambda i: 1e-3 * i ** -0.5
to support Robbins-Monro style scheduling (and others)Up for debate but could be useful to add this support for the
temperature
parameter in the spirit of simulated annealing