Source code for lmflow.optim.adagrad

#!/usr/bin/env python

import torch




class AdaGrad(torch.optim.Optimizer):
    def __init__(self, params, lr=0.001, eps=1e-8, weight_decay=0):
        defaults = dict(lr=lr, eps=eps, weight_decay=weight_decay)
        super().__init__(params, defaults)



    def step(self, closure=None):
        loss = None
        if closure is not None:
            loss = closure()

        for group in self.param_groups:
            for p in group["params"]:
                if p.grad is None:
                    continue
                grad = p.grad.data
                if group["weight_decay"] != 0:
                    grad = grad.add(group["weight_decay"], p.data)

                state = self.state[p]

                if len(state) == 0:
                    state["sum"] = torch.zeros_like(p.data)

                sum = state["sum"]
                sum.addcmul_(1, grad, grad)
                std = sum.sqrt().add_(group["eps"])
                p.data.addcdiv_(-group["lr"], grad, std)

        return loss