def make_policy(epochs, params, lr, momentum, wd):
optimizer = SGD(params, lr, momentum=momentum, weight_decay=wd)
# this creates a scheduler which is used to adjust the learning rate, many of the scheduler are defined , choose one as per your need
# check if the policy is right
scheduler = PolyPolicy(optimizer, epochs, 1)
return optimizer, scheduler
def make_policy(epochs, network, lr, momentum):
optimizer = Adam([
{'params': network.parameters(), 'lr': lr},
], weight_decay=1e-4)
# this creates a scheduler which is used to adjust the learning rate, many of the scheduler are defined , choose one as per your need
# check if the policy is right
scheduler = PolyPolicy(optimizer, epochs, 1)
return optimizer, scheduler
def make_policy(epochs, network, lr, momentum, wd):
optimizer = SGD([
{
'params': network.parameters(),
'lr': lr
},
],
momentum=momentum,
weight_decay=wd)
scheduler = PolyPolicy(optimizer, epochs, 1)
return optimizer, scheduler
def make_policy(epochs, network, lr, momentum, wd):
optimizer = SGD(filter(lambda p: p.requires_grad, network.parameters()), lr = lr, momentum=momentum, weight_decay=wd)
scheduler = PolyPolicy(optimizer, epochs, 1)
return optimizer, scheduler