optim.py

import torch.optim as optim


class Optim(object):

    def set_parameters(self, params):
        self.params = list(params)  # careful: params may be a generator
        if self.method == 'sgd':
            self.optimizer = optim.SGD(self.params, lr=self.lr)
        elif self.method == 'adagrad':
            self.optimizer = optim.Adagrad(self.params, lr=self.lr)
        elif self.method == 'adadelta':
            self.optimizer = optim.Adadelta(self.params, lr=self.lr)
        elif self.method == 'adam':
            self.optimizer = optim.Adam(self.params, lr=self.lr)
        else:
            raise RuntimeError("Invalid optim method: " + self.method)

    def __init__(self, method, lr, max_grad_norm, lr_decay=1, start_decay_at=None):
        self.last_ppl = None
        self.lr = lr
        self.max_grad_norm = max_grad_norm
        self.method = method
        self.lr_decay = lr_decay
        self.start_decay_at = start_decay_at
        self.start_decay = False

    def step(self):
        self.optimizer.step()

    def updateLearningRate(self, ppl, epoch):
        if self.start_decay_at is not None and epoch >= self.start_decay_at:
            self.start_decay = True
        if self.last_ppl is not None and ppl > self.last_ppl:
            self.start_decay = True

        if self.start_decay:
            self.lr = self.lr * self.lr_decay
            print("Decaying learning rate to %g" % self.lr)

        self.last_ppl = ppl
        self.optimizer.param_groups[0]['lr'] = self.lr