探索和构建 LLaMA 3 架构：深入探讨组件、编码和推理技术（七）前馈神经网络

在Transformer架构中，前馈层扮演着至关重要的角色，通常位于注意力层和标准化处理之后。前馈层由三个线性变换组成。

python 复制代码

class FeedForward(nn.Module):
    def __init__(self, args: ModelArgs):
        super().__init__()
        # Assuming 'hidden_dim' is calculated as per your specifications
        hidden_dim = 4 * args.dim
        hidden_dim = int(2 * hidden_dim / 3)  # Applying your specific transformation
        if args.ffn_dim_multiplier is not None:
            hidden_dim = int(args.ffn_dim_multiplier * hidden_dim)
        #hidden_dim = int(2 * hidden_dim / 3)  # Applying your specific transformation
        hidden_dim = args.multiple_of * ((hidden_dim + args.multiple_of - 1) // args.multiple_of)

        self.w1 = nn.Linear(args.dim, hidden_dim, bias=False)
        self.w2 = nn.Linear(hidden_dim, args.dim, bias=False)  # This layer seems to be missing in your original setup
        self.w3 = nn.Linear(args.dim, hidden_dim, bias=False)  # Corrected to match checkpoint

    def forward(self, x: torch.Tensor):
        swish = F.silu(self.w1(x))  # Apply first transformation
        x_V = self.w3(x) 
        x = swish * x_V        # Apply contraction to original dimension
        x = self.w2(x)  # Apply optional additional transformation
        return x

在前向传递过程中，输入张量x经历多层线性变换。第一次转换后应用的SwiGLU激活函数增强了模型的表达能力。最终的变换将张量映射回其原始维度。 SwiGLU 激活和多个前馈层的这种独特组合增强了模型的性能。

系列博客

探索和构建 LLaMA 3 架构：深入探讨组件、编码和推理技术（一）Llama3 模型架构
https://duanzhihua.blog.csdn.net/article/details/138208650

探索和构建 LLaMA 3 架构：深入探讨组件、编码和推理技术（二）RoPE位置编码
https://duanzhihua.blog.csdn.net/article/details/138212328

探索和构建 LLaMA 3 架构：深入探讨组件、编码和推理技术（三）KV缓存
https://duanzhihua.blog.csdn.net/article/details/138213306

探索和构建 LLaMA 3 架构：深入探讨组件、编码和推理技术（四）分组多查询注意力
https://duanzhihua.blog.csdn.net/article/details/138216050

探索和构建 LLaMA 3 架构：深入探讨组件、编码和推理技术（五）RMS 均方根归一化
https://duanzhihua.blog.csdn.net/article/details/138216630

探索和构建 LLaMA 3 架构：深入探讨组件、编码和推理技术（六）SwiGLU 激活函数
https://duanzhihua.blog.csdn.net/article/details/138217261