Sets the learning rate to the initial LR decayed by decay rate every steep step
(epoch, opt, optimizer)
source not stored for this graph (policy: none)
no outgoing calls
no test coverage detected