以下代码实现了LoRA（低秩适配）微调层，请阅读代码并填写空缺部分。[1] 和 [2] 处应分别填入：

import torch

import torch.nn as nn

import math

class LoRALinear(nn.Module):

   """在原始线性层旁添加低秩分解矩阵实现LoRA微调"""

   def __init__(self, original_linear: nn.Linear, rank: int = 8, alpha: float = 16.0):

       super().__init__()

       self.original_linear = original_linear

       self.rank = rank

       self.alpha = alpha

       self.scaling = alpha / rank

       in_features = original_linear.in_features

       out_features = original_linear.out_features

       # [1] 初始化低秩矩阵 A 和 B

       # LoRA 的核心：ΔW = A @ B，其中 A: (in_features, rank)，B: (rank, out_features)

          [1]_____________________________________________________________________

          [2]_____________________________________________________________________

       nn.init.kaiming_uniform_(self.lora_A, a=math.sqrt(5))

       nn.init.zeros_(self.lora_B)

       # 冻结原始权重

       self.original_linear.weight.requires_grad = False

       if self.original_linear.bias is not None:

           self.original_linear.bias.requires_grad = False

   def forward(self, x):

       # 原始线性层的输出

       original_output = self.original_linear(x)

       # [2] 计算 LoRA 的增量输出：x 先经过 A 再经过 B，并乘以缩放因子