What operations are part of a standard Transformer block? (Select all that apply.)

Question

BlackTom AI · Accepted Answer

The question asks which operations are part of a standard Transformer block and specifies that it is a select-all-that-apply type.
Option 1: Layer normalization. In a typical Transformer block, after the multi-head self-attention and feed-forward sublayers, the outputs are usually passed through layer normalization (often implemented with a pre-norm or post-norm variant). This is a core component that stabilizes training and helps with gradient flow, so this option is correct.
Option 2: Convolutional layer. Standard Transformer blocks do not include a convolutional layer as a built-in component. The hallmark of the Transformer is self-attention, not convolution, so this choice is incorrect.
Option 3: Residual connection. Residual (or skip) connections are integral to Transformer blocks, allowing the sublayer input to be added to its output, which helps with optimization and information flow. This option is correct.
Option 4: Self-attention layer. The self-attention mechanism is the defining feature of the Transformer, enabling tokens to attend to other tokens within the same sequence. This is certainly part of a standard Transformer block, so this option is correct.
In summary, options 1, 3, and 4 correspond to standard components of a Transformer block, while option 2 does not. Each correct choice reflects a widely recognized aspect of the Transformer's architectural design, whereas the incorrect choice represents an element not used in the canonical block structure.

What operations are part of a standard Transformer block? (Select all that apply.)多项选择题

类似问题

Which of the following best describes a key property of Transformer architecture used in LLMs?

The transformer architecture only consists of a single encoder component and single decoder component.

What is the main architectural innovation in ChatGPT that allows it to handle complex language tasks?

A team of data scientists is experiencing poor performance with their Transformer model. After inspecting their implementation, you find several suspicious design choices. Which one would you identify as the MOST problematic based on the architectural principles discussed in the lecture?

The implementation and loading phase of the Database Life Cycle (DBLC) involves _____.

Which of the following parameters is used to determine the concentration of protein in a spectrophotometer?

更多留学生实用工具

智能学习助手

风格化写作助手

论文查重助手

文献引用助手

课堂转译助手

课堂笔记助手

Quiz搜索助手

学校历年真题

智能刷题助手

智能匹配练习

希望你的学习变得更简单