Learn Before
Concept

Computational Challenges of LLM Inference

During the inference stage of large language models, significant computational challenges arise from two primary operations: efficiently calculating the conditional probability Pr(yx)\Pr(\mathbf{y}|\mathbf{x}) of potential output sequences given an input, and performing the arg max\argmax search operation to find the optimal sequence y^\hat{\mathbf{y}} that maximizes this probability.

0

1

Updated 2026-05-03

Contributors are:

Who are from:

Tags

Foundations of Large Language Models

Ch.5 Inference - Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences