Decoding Strategies in LLM Greedy Beam Search Top K Top P Temperature Explained
In this video, we understand how Large Language Models generate text using different decoding strategies. This is one of the most important concepts to control output quality, creativity, and accuracy in models like GPT.
Here is the GitHub repo link
https://github.com/switch2ai
You can download all the code, scripts, and documents from the above GitHub repository.
We start by understanding how decoder works.
Given an input
Once upon a time
The model generates probability for each token in vocabulary
token1 → probability
token2 → probability
token3 → probability
The next token is select…
Watch on YouTube ↗
(saves to browser)
DeepCamp AI