LLM-Papers.md

Papers

llya 30

llya 30u30

多模态

具身智能经典：PaLM-E，基于图文VLM的具身VQA任务推理模型

PaLM-E: An Embodied Multimodal Language Model

Tip

具身智能经典：PaLM-E，基于图文VLM的具身VQA任务推理模型

Spectral Torical Band Inequalities and Generalizations of the Schoen-Yau Black Hole Existence Theorem

https://arxiv.org/abs/2301.08270

Language Is Not All You Need: Aligning Perception with Language Models

https://arxiv.org/abs/2302.14045

Attention is All You Need

quickstart

concepts

• attention mechanism
• self-attention
• multi-headed sel-attention(多头自注意力机制)
• encoder-dcoder 多头输入, 逐个输出
• auto-regressive(自回归): 过去时刻的输出可以作为当前时刻的输入
• positional encoding(位置编码)
• $LayerNorm(x + Sublayer(x))$

Architecture

encoder-decoder architecture

二维矩阵(batch, feature)

三维矩阵(batch, sequence, feature)

$X=[x1,x2,...,xn]$, 为 $ \mu = \frac{1}{n} \sum_{i=1}^n x_i $，标准化后的值变为 $ x_i' = x_i - \mu $。

通过除以数据的标准差（方差的平方根），将数据的分散程度缩放到1。这样可以消除不同数据尺度（单位或范围）的差异。

标准差 $ \sigma = \sqrt{\frac{1}{n} \sum_{i=1}^n (x_i - \mu)^2} $，标准化后的值变为 $ x_i'' = \frac{x_i'}{\sigma} = \frac{x_i - \mu}{\sigma} $。

Papers

llya 30

多模态

具身智能经典：PaLM-E，基于图文VLM的具身VQA任务推理模型

PaLM-E: An Embodied Multimodal Language Model

Tip

具身智能经典：PaLM-E，基于图文VLM的具身VQA任务推理模型

Spectral Torical Band Inequalities and Generalizations of the Schoen-Yau Black Hole Existence Theorem

https://arxiv.org/abs/2301.08270

Language Is Not All You Need: Aligning Perception with Language Models

https://arxiv.org/abs/2302.14045

Attention is All You Need

quickstart

concepts

• attention mechanism
• self-attention
• multi-headed sel-attention(多头自注意力机制)
• encoder-dcoder 多头输入, 逐个输出
• auto-regressive(自回归): 过去时刻的输出可以作为当前时刻的输入
• positional encoding(位置编码)
• $LayerNorm(x + Sublayer(x))$

Architecture

encoder-decoder architecture

二维矩阵(batch, feature)

三维矩阵(batch, sequence, feature)

$X=[x1,x2,...,xn]$, 为 $ \mu = \frac{1}{n} \sum_{i=1}^n x_i $，标准化后的值变为 $ x_i' = x_i - \mu $。

通过除以数据的标准差（方差的平方根），将数据的分散程度缩放到1。这样可以消除不同数据尺度（单位或范围）的差异。

标准差 $ \sigma = \sqrt{\frac{1}{n} \sum_{i=1}^n (x_i - \mu)^2} $，标准化后的值变为 $ x_i'' = \frac{x_i'}{\sigma} = \frac{x_i - \mu}{\sigma} $。