# Instruction tuning a pre-trained base model

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Finstruction-tuning-a-pre-trained-base-model%2F132076%3E%5C
**Category:** 🤗AutoTrain\
**Created:** [December 18, 2024, 10:15am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Finstruction-tuning-a-pre-trained-base-model%2F132076 "2024-12-18T10:15:21Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![anishkhandelwal06](https://avatars.discourse-cdn.com/v4/letter/a/dbc845/32.png) [@anishkhandelwal06](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fanishkhandelwal06)%5C
**Post date:** [December 18, 2024, 10:15am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Finstruction-tuning-a-pre-trained-base-model%2F132076%2F1 "2024-12-18T10:15:21Z")

</div>

how to pass instruction, context and answer in the training dataset?  
I am trying to instruction tune phi3.5 but unsure of how to pass above in training dataset.

My use-case is to create a planner agent for a chatbot which does the task decomposition.  
I generated the training dataset using gpt-4o and with a refined prompt and manually correction in the output wherever required.  
I maintain the entity context, current user input and last bot reply which was passed to the prompt along with instructions to generate the desired output.