Learn Before
Formula for the Expansion of the Search Space at Each Step
In sequence generation, the set of sequences generated at step , denoted , can be obtained by expanding each sequence in the previous set with all possible next tokens from the vocabulary . This is expressed recursively as the Cartesian product: . This operation denotes that each sequence in is concatenated with each token in . However, if a sequence in is already complete (for instance, ending with an token), it will not be expanded any further.
0
1
Tags
Ch.5 Inference - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Related
Formula for the Search Space as a Union of Complete Sequences
Formula for the Expansion of the Search Space at Each Step
A simplified language model has a vocabulary consisting of only three unique tokens: 'cat', 'sat', and 'on'. The model is configured to generate an output sequence with a fixed length of exactly two tokens. Which of the following options correctly represents the complete set of all possible output sequences the model can generate?
Analyzing Search Space Dimensions
Growth of the Generative Search Space
Mathematical Formulation of the Search Problem in LLM Inference
Learn After
A language model with a vocabulary of 5,000 unique tokens is generating text. If, at a certain stage, the model is considering 20 different unique sequences of tokens, each 4 tokens long, how many total unique sequences of 5 tokens can possibly be generated in the very next step?
Search Space Growth Analysis
Comparing Computational Load in Language Models