torch.optim.rmsprop 的源代码

```html

import torch
from torch import Tensor
from .optimizer import (Optimizer, _default_to_fused_or_foreach, _use_grad_for_differentiable,
                        _differentiable_doc, _foreach_doc, _maximize_doc, _view_as_real)
from typing import List, Optional

__all__ = ["RMSprop", "rmsprop"]


class RMSprop(Optimizer):
    def __init__(
        self,
        params,
        lr=1e-2,
        alpha=0.99,
        eps=1e-8,
        weight_decay=0,
        momentum=0,
        centered=False,
        foreach: Optional[bool] = None,
        maximize: bool = False,
        differentiable: bool = False,
    ):
        if not 0.0 <= lr:
            raise ValueError(f"Invalid learning rate: {lr}")
        if not 0.0 <= eps:
            raise ValueError(f"Invalid epsilon value: {eps}")
        if not 0.0 <= momentum:
            raise ValueError(f"Invalid momentum value: {momentum}")
        if not 0.0 <= weight_decay:
            raise ValueError(f"Invalid weight_decay value: {weight_decay}")
        if not 0.0 <= alpha:
            raise ValueError(f"Invalid alpha value: {alpha}")

        defaults = dict(
            lr=lr,
            momentum=momentum,
            alpha=alpha,
            eps=eps,
            centered=centered,
            weight_decay=weight_decay,
            foreach=foreach,
            maximize=maximize,
            differentiable=differentiable,
        )
        super().__init__(params, defaults)

    def __setstate__(self, state):
        super().__setstate__(state)
        for group in self.param_groups:
            group.setdefault("momentum", 0)
            group.setdefault("centered", False)
            group.setdefault("foreach", None)
            group.setdefault("maximize", False)
            group.setdefault("differentiable", False)

    def _init_group(self, group, params_with_grad, grads, square_avgs, momentum_buffer_list, grad_avgs):
        has_complex = False
        for p in group["params"]:
            if p.grad is None:
                continue
            has_complex |= torch.is_complex(p)
            params_with_grad.append(p)

            if p.grad.is_sparse:
                raise RuntimeError("RMSprop does not support sparse gradients")
            grads.append(p.grad)

            state = self.state[p]

            # State initialization
            if len(state) == 0:
                state["step"] = 0
                state["square_avg"] = torch.zeros_like(
                    p, memory_format=torch.preserve_format
                )
                if group["momentum"] > 0:
                    state["momentum_buffer"] = torch.zeros_like(
                        p, memory_format=torch.preserve_format
                    )
                if group["centered"]:
                    state[<span class="s2