# How is the prompt + answer handled during training

**URL:** <https://discuss.huggingface.co/t/how-is-the-prompt-answer-handled-during-training/78308>\
**Category:** Beginners\
**Created:** [March 20, 2024, 5:28pm UTC](https://discuss.huggingface.co/t/how-is-the-prompt-answer-handled-during-training/78308 "2024-03-20T17:28:55Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![AdL8](https://avatars.discourse-cdn.com/v4/letter/a/3be4f8/32.png) [@AdL8](https://discuss.huggingface.co/u/AdL8)\
**Post date:** [March 20, 2024, 5:28pm UTC](https://discuss.huggingface.co/t/how-is-the-prompt-answer-handled-during-training/78308/1 "2024-03-20T17:28:55Z")

</div>

Hello,

I am really confused on how the model train on the prompt and the answer of the prompt. Does the model use the whole prompt as a context and tries to predict each token of the answer, sliding the context window. Or does it start the training by trying to predict also each token in the prompt ?