Deep Learning Academy · 课时

自注意力：查询、键与值

让每个词元都能关注其他词元

自注意力：查询、键与值 是 CoddyKit 上的免费 Deep Learning Academy 课时。这是第 1 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 Deep Learning Academy 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 Deep Learning Academy 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

Tokens That Talk

In a sequence, the meaning of one word depends on others. Self-attention lets every token look at every other token to gather the context it needs.

Three Roles per Token

Each token plays three roles: a query that asks, a key that answers, and a value that carries content. These come from the same word, used three ways.

The Query

A token's query describes what it is looking for. Think of it as the question this word is asking about the rest of the sentence.

The Key

Every token also exposes a key, a label advertising what it offers. A query is compared against all keys to find good matches.

The Value

Once a match is found, the value is the actual information that gets passed along. Keys decide how much, values decide what.

Make Q, K, V

You build queries, keys, and values by projecting the input through three learned linear layers. Same input, three different weight matrices.

q = self.W_q(x)
k = self.W_k(x)
v = self.W_v(x)

Score by Similarity

To see how well a query matches a key, you take their dot product. A bigger score means the two tokens are more relevant to each other.

scores = q @ k.transpose(-2, -1)

Scores to Weights

Raw scores become attention weights with softmax, so each query's weights are positive and sum to one across all keys.

weights = scores.softmax(dim=-1)

Blend the Values

The output for each token is a weighted sum of all values, mixed by the attention weights. Relevant tokens contribute more.

out = weights @ v

Why It Beats RNNs

Self-attention connects any two tokens in one step, so distance does not matter. This parallel view is why transformers handle long context so well.

Learned, Not Fixed

The Q, K, V projections are trained by gradient descent. The network learns what to ask, what to advertise, and what to share, all from data.

Quick Check

Let's test how attention combines its pieces.

Recap

You learned that self-attention turns each token into a query, key, and value, scores query-key matches, and blends values by softmax weights. Nice work!

常见问题解答

「自注意力：查询、键与值」课时是免费的吗？

是的 — 「自注意力：查询、键与值」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 Deep Learning Academy 课程的其余内容，请升级到 CoddyKit PRO。 Deep Learning Academy 课程共包含 4 节课。

「自注意力：查询、键与值」这节课中我会学到什么？

让每个词元都能关注其他词元你通过在浏览器中直接运行的动手代码来练习 Deep Learning Academy，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 Deep Learning Academy 需要有经验吗？

无需任何先前经验。CoddyKit 上的 Deep Learning Academy 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 1 节课，共 4 节。

「自注意力：查询、键与值」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 Deep Learning Academy 课中编写并运行代码吗？

能。每节 Deep Learning Academy 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 Deep Learning Academy