sgd.py

# Copyright (c) Facebook, Inc. and its affiliates.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
#

import torch
from torch.optim import Optimizer


class SGD(Optimizer):
    def __init__(self, params, lr, momentum=0, weight_decay=0, nesterov=False):
        if lr < 0.0:
            raise ValueError(f"Invalid learning rate: {lr}")
        if weight_decay < 0.0:
            raise ValueError(f"Invalid weight_decay value: {weight_decay}")

        defaults = dict(lr=lr, momentum=momentum, weight_decay=weight_decay, nesterov=nesterov)
        super(SGD, self).__init__(params, defaults)

    @torch.no_grad()
    def step(self):
        for group in self.param_groups:
            weight_decay = group["weight_decay"]
            momentum = group["momentum"]
            nesterov = group["nesterov"]

            for p in group["params"]:
                if p.grad is None:
                    continue

                d_p = p.grad
                if weight_decay != 0:
                    d_p = d_p.add(p, alpha=weight_decay)
                d_p.mul_(-group["lr"])

                if momentum != 0:
                    param_state = self.state[p]
                    if "momentum_buffer" not in param_state:
                        buf = param_state["momentum_buffer"] = d_p.clone().detach()
                    else:
                        buf = param_state["momentum_buffer"]
                        buf.mul_(momentum).add_(d_p)

                    if nesterov:
                        d_p.add_(buf, alpha=momentum)
                    else:
                        d_p = buf

                p.add_(d_p)

        return None