From 5f2ccc44bf555c027f97e9dae7ace7cef0e83c2d Mon Sep 17 00:00:00 2001
From: Logan Zou <74288839+logan-zou@users.noreply.github.com>
Date: Mon, 23 Jun 2025 10:53:45 +0800
Subject: [PATCH] =?UTF-8?q?Update=20=E7=AC=AC=E4=BA=8C=E7=AB=A0=20Transfor?=
 =?UTF-8?q?mer=E6=9E=B6=E6=9E=84.md?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

fix arg bug
---
 docs/chapter2/第二章 Transformer架构.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docs/chapter2/第二章 Transformer架构.md b/docs/chapter2/第二章 Transformer架构.md
index 1026216..ba0fd01 100644
--- a/docs/chapter2/第二章 Transformer架构.md	
+++ b/docs/chapter2/第二章 Transformer架构.md	
@@ -252,7 +252,7 @@ class MultiHeadAttention(nn.Module):
         # args: 配置对象
         super().__init__()
         # 隐藏层维度必须是头数的整数倍，因为后面我们会将输入拆成头数个矩阵
-        assert args.n_embd % args.n_heads == 0
+        assert args.dim % args.n_heads == 0
         # 模型并行处理大小，默认为1。
         model_parallel_size = 1
         # 本地计算头数，等于总头数除以模型并行处理大小。