# GPT\_J custom dataset

**URL:** <https://discuss.huggingface.co/t/gpt-j-custom-dataset/25780>\
**Category:** Beginners\
**Created:** [November 9, 2022, 4:31am UTC](https://discuss.huggingface.co/t/gpt-j-custom-dataset/25780 "2022-11-09T04:31:13Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![vmhf22](https://avatars.discourse-cdn.com/v4/letter/v/3ec8ea/32.png) [@vmhf22](https://discuss.huggingface.co/u/vmhf22)\
**Post date:** [November 9, 2022, 4:31am UTC](https://discuss.huggingface.co/t/gpt-j-custom-dataset/25780/1 "2022-11-09T04:31:13Z")

</div>

Hi,

I am a CS student trying to use GPTJ for my custom dataset.  
I have successfully tried generate() API and that works.  
To fine-tune it further, is there a way for me to train with my custom dataset.  
Any pointers would be appreciated.

Thank you!

---

<div class="post-metadata">

**Author:** ![vmhf22](https://avatars.discourse-cdn.com/v4/letter/v/3ec8ea/32.png) [@vmhf22](https://discuss.huggingface.co/u/vmhf22)\
**Post date:** [November 9, 2022, 7:14pm UTC](https://discuss.huggingface.co/t/gpt-j-custom-dataset/25780/2 "2022-11-09T19:14:37Z")

</div>

I checked documentation at [GPT-J](https://huggingface.co/docs/transformers/model_doc/gptj),  
but don’t see any way to train GPTJ with my own custom dataset.  
Could someone please confirm if its possible to do so and any pointer to API to use?

Thanks so much!