跨国深谈← 返回←蒸馏 (distill)

集里怎么说它

• 《LLM 重造 DoorDash 搜索与推荐:四个基础组件》(06:10起):本集反复强调的模式:离线用一次昂贵的推理,然后蒸馏进可以廉价、快速服务的模型;不是用 LLM 替换检索排序系统,而是把推理蒸馏进生产级架构。

① 提到它的金句

5 条

我实际上发现先生成代码会容易得多，让那段代码有点像是一个关于世界可能样子的草案，作为团队与之互动来完善它，然后从中提炼出规范。
I’ve actually found it is much easier to produce the code first, to have that code sort of like present a straw man for what the world could look like, engage with that as a team to refine it, and then distill the spec out of that.
—— Ryan Lopopolo · [28:20]

指向原始笔记的链接

离线使用一次昂贵的推理,然后把它蒸馏进可以廉价、快速地提供服务的模型中。
Use expensive reasoning once offline, then distill it into models that can be served cheaply and quickly.
—— Raghav Saboo · [06:10]

指向原始笔记的链接

我认为我们会同时做出快的和慢的模型，然后把它们蒸馏成一个。
I think we are going to simultaneously make the fast and the slow models and then distill them to one.
—— Kirthana Gopalakrishnan · [30:00]

指向原始笔记的链接

让小模型变得非常聪明的最好方法是蒸馏。你训练最大的模型，然后蒸馏。
The best way to make a small model very smart is by doing distillation. You train the biggest model, then you distill.
—— Jason Ma · [19:00]

指向原始笔记的链接

人们每天都在蒸馏我的产品。他们拿走它，把它拆解成原始部件，这就叫竞争。
People distill my products every single day. They take it, they strip it down to its raw parts, and that’s called competition.
—— 嘉宾 · [00:21]

指向原始笔记的链接

② 出现在这些集

1 集

• 《LLM 重造 DoorDash 搜索与推荐:四个基础组件》 — 作为概念

③ 关联

点进去有真内容 —— 本页主要出口

Raghav Saboo · DoorDash · LLM · 语义理解 · 分级相关性 · 语义 ID · 记忆 · 嵌入 · 检索 · 排名