# Fine-tuning Dataset Requirements
**URL:**
**Author:**  [@Buckeyes2019](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FBuckeyes2019)%5C
**Post date:** [September 4, 2020, 2:56pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Ffine-tuning-dataset-requirements%2F993%2F1 "2020-09-04T14:56:44Z")
I am looking to fine-tune a BART-large model for a summarization task and I am creating a dataset to tune on. How should I structure this dataset? Should it have a column of text blocks and another column with associated summaries? Or, will simply providing the raw text (the text blocks) without summaries suffice? Thanks!
---
**Author:**  [@valhalla](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fvalhalla)%5C
**Post date:** [September 6, 2020, 7:59am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Ffine-tuning-dataset-requirements%2F993%2F2 "2020-09-06T07:59:09Z")
Hi @Buckeyes2019
You can fine the seq2seq fine-tuning scripts here. The readme explains how the data should formatted and saved.
[https://github.com/huggingface/transformers/tree/master/examples/seq2seq](https://github.com/huggingface/transformers/tree/master/examples/seq2seq)
---
**Author:**  [@dsffs54654dasa](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fdsffs54654dasa)%5C
**Post date:** [February 23, 2026, 3:58pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Ffine-tuning-dataset-requirements%2F993%2F3 "2026-02-23T15:58:04Z")
