# Autotrain ORPO Dataset format

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fautotrain-orpo-dataset-format%2F89057%3E%5C
**Category:** 🤗AutoTrain\
**Created:** [May 28, 2024, 10:37pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fautotrain-orpo-dataset-format%2F89057 "2024-05-28T22:37:10Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![amritap-ef](https://avatars.discourse-cdn.com/v4/letter/a/c57346/32.png) [@amritap-ef](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Famritap-ef)%5C
**Post date:** [May 28, 2024, 10:37pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fautotrain-orpo-dataset-format%2F89057%2F1 "2024-05-28T22:37:10Z")

</div>

Hi,

I’m wondering about what the correct format is for the “prompt” field for ORPO on autotrain? I see in this [example](https://huggingface.co/abhishek/autotrain-llama3-70b-orpo-v1) that the dataset (distilabel-capybara-dpo-7k-binarized) used that the prompt has been formatted using chatml format already - does that mean that I should similarly format my prompts? What about if I don’t want to use the chatml format - should I format the prompt accordingly?

Also, the [documentation](https://huggingface.co/docs/autotrain/llm_finetuning) for ORPO states that the columns needed for Reward/ORPO trainer is just the text and rejected text columns which is a bit confusing as I get an error if I try not to supply the “Prompt” field.

---

<div class="post-metadata">

**Author:** ![abhishek](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/abhishek/32/32360_2.png) [@abhishek](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fabhishek)%5C
**Post date:** [May 29, 2024, 5:53am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fautotrain-orpo-dataset-format%2F89057%2F2 "2024-05-29T05:53:04Z")

</div>

hi. it should be same as dpo instead. ill fix the docs asap.

---

<div class="post-metadata">

**Author:** ![amritap-ef](https://avatars.discourse-cdn.com/v4/letter/a/c57346/32.png) [@amritap-ef](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Famritap-ef)%5C
**Post date:** [May 29, 2024, 9:23am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fautotrain-orpo-dataset-format%2F89057%2F3 "2024-05-29T09:23:21Z")

</div>

Thanks, and RE the prompt format - should it be the raw text or with the chatml (etc) format?

---

<div class="post-metadata">

**Author:** ![QuestforAIEd](https://avatars.discourse-cdn.com/v4/letter/q/67e7ee/32.png) [@QuestforAIEd](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FQuestforAIEd)%5C
**Post date:** [October 8, 2024, 8:51pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fautotrain-orpo-dataset-format%2F89057%2F4 "2024-10-08T20:51:27Z")

</div>

Hi amritap,  
have you found any answer to your question?

---

<div class="post-metadata">

**Author:** ![abhishek](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/abhishek/32/32360_2.png) [@abhishek](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fabhishek)%5C
**Post date:** [October 8, 2024, 9:02pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fautotrain-orpo-dataset-format%2F89057%2F5 "2024-10-08T21:02:37Z")

</div>

we have all the dataset formats in docs: [LLM Finetuning](https://huggingface.co/docs/autotrain/tasks/llm_finetuning)