Find a interesting demo about transformer

This commit is contained in:
PinkR1ver 2024-10-10 11:25:05 +08:00
parent 6f6d21292c
commit 6428e7928b
2 changed files with 16 additions and 5 deletions

Binary file not shown.

After

Width:  |  Height:  |  Size: 1.1 MiB

View File

@ -1,9 +1,9 @@
---
title: "Transformer"
title: Transformer
tags:
- deep-learning
- attention
date: 2024-01-17
- deep-learning
- attention
date: 2024-10-09
---
> [!info]
@ -18,4 +18,15 @@ Transformer 是Seq2Seq model由Encoder和Decoder组成
这里贴的是原文Encoder的架构
![Pasted image 20230316162635](computer_sci/deep_learning_and_machine_learning/deep_learning/attachments/Pasted%20image%2020230316162635.png)
![Pasted image 20230316162642](computer_sci/deep_learning_and_machine_learning/deep_learning/attachments/Pasted%20image%2020230316162642.png)
![Pasted image 20230316162642](computer_sci/deep_learning_and_machine_learning/deep_learning/attachments/Pasted%20image%2020230316162642.png)
## Learn by Animation Demo
2024 VIS poster session 里有一个海报介绍的Transformer Explainer太清晰了包含word embeddingQKV到attention的计算dropout layer normalization ... ...
通过这个demo可以清晰地理解LLM中Transformer模块地应用
访问:[https://poloclub.github.io/transformer-explainer/](https://poloclub.github.io/transformer-explainer/)
![](computer_sci/deep_learning_and_machine_learning/deep_learning/attachments/Pasted%20image%2020241010112210.png)