diff --git a/.DS_Store b/.DS_Store index 21c200a..7c2bc08 100644 Binary files a/.DS_Store and b/.DS_Store differ diff --git a/image-20240430194452839.png b/assets/image-20240430194452839.png similarity index 100% rename from image-20240430194452839.png rename to assets/image-20240430194452839.png diff --git a/人人都能看懂的Transformer/.DS_Store b/人人都能看懂的Transformer/.DS_Store index 81b81e9..db0abb0 100644 Binary files a/人人都能看懂的Transformer/.DS_Store and b/人人都能看懂的Transformer/.DS_Store differ diff --git a/人人都能看懂的Transformer/第四章——多头注意力机制——QK矩阵相乘.md b/人人都能看懂的Transformer/第四章——多头注意力机制——QK矩阵相乘.md index a8ad718..b90b4ff 100644 --- a/人人都能看懂的Transformer/第四章——多头注意力机制——QK矩阵相乘.md +++ b/人人都能看懂的Transformer/第四章——多头注意力机制——QK矩阵相乘.md @@ -85,7 +85,7 @@ $$ 现在我们知道矩阵相乘能代表相似度的高低,回到实际中,过程图如下 -image-20240430194452839 +image-20240430194452839 上面我放的文字,实际传给机器的时候是数值。