# Decoder only fine-tuning enough for UMT5

**URL:** <https://discuss.huggingface.co/t/decoder-only-fine-tuning-enough-for-umt5/64022>\
**Category:** Models\
**Created:** [November 29, 2023, 5:03pm UTC](https://discuss.huggingface.co/t/decoder-only-fine-tuning-enough-for-umt5/64022 "2023-11-29T17:03:12Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![nitinsurya](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/nitinsurya/32/21687_2.png) [@nitinsurya](https://discuss.huggingface.co/u/nitinsurya)\
**Post date:** [November 29, 2023, 5:03pm UTC](https://discuss.huggingface.co/t/decoder-only-fine-tuning-enough-for-umt5/64022/1 "2023-11-29T17:03:12Z")

</div>

Looking into fine-tuning UMT5 (encoder-decoder) model for a translation task.

Has anyone explored end task’s performance difference when finetuning such (unsupervised trained encoder-decoder) model in

- whole model finetuning setting
- decoder only finetuning