# How to train RAG-models?

**URL:** <https://community.deeplearning.ai/t/how-to-train-rag-models/651360>\
**Category:** Generative AI with Large Language Models\
**Tags:** week-module-3\
**Created:** [June 21, 2024, 5:52pm UTC](https://community.deeplearning.ai/t/how-to-train-rag-models/651360 "2024-06-21T17:52:22Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![someone555777](https://avatars.discourse-cdn.com/v4/letter/s/67e7ee/32.png) [@someone555777](https://community.deeplearning.ai/u/someone555777)\
**Post date:** [June 21, 2024, 5:52pm UTC](https://community.deeplearning.ai/t/how-to-train-rag-models/651360/1 "2024-06-21T17:52:22Z")

</div>

So, rag-data is external. Should I take this in accounting when I try to train a model? Should I try to retrieve potential data of RAG on the moment when prediction was done by human and pass it as simple prompt?

Should it be like mock-concept in testing for example?

---

<div class="post-metadata">

**Author:** ![roma2023](https://avatars.discourse-cdn.com/v4/letter/r/d6d6ee/32.png) [@roma2023](https://community.deeplearning.ai/u/roma2023)\
**Post date:** [August 10, 2024, 3:17pm UTC](https://community.deeplearning.ai/t/how-to-train-rag-models/651360/2 "2024-08-10T15:17:54Z")

</div>

Yes, It could indeed be similar to a mock-concept in testing, where you simulate real-world conditions to evaluate the model’s performance.