# How to best version a model after retraining?

**URL:** <https://discuss.huggingface.co/t/how-to-best-version-a-model-after-retraining/23705>\
**Category:** Models\
**Created:** [September 28, 2022, 6:32am UTC](https://discuss.huggingface.co/t/how-to-best-version-a-model-after-retraining/23705 "2022-09-28T06:32:29Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![michaelwechner](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/michaelwechner/32/5514_2.png) [@michaelwechner](https://discuss.huggingface.co/u/michaelwechner)\
**Post date:** [September 28, 2022, 6:32am UTC](https://discuss.huggingface.co/t/how-to-best-version-a-model-after-retraining/23705/1 "2022-09-28T06:32:29Z")

</div>

We recently published a model [https://huggingface.co/Wyona/message-classification-question-other-smalltalk-modified](https://huggingface.co/Wyona/message-classification-question-other-smalltalk-modified) and we would like to retrain it continuously with an updated dataset to improve it. How do we best version the model after retraining, such that people can use a particular version, instead of using the latest version, which might not perform as a previous version?

---

<div class="post-metadata">

**Author:** ![michaelwechner](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/michaelwechner/32/5514_2.png) [@michaelwechner](https://discuss.huggingface.co/u/michaelwechner)\
**Post date:** [September 28, 2022, 7:23am UTC](https://discuss.huggingface.co/t/how-to-best-version-a-model-after-retraining/23705/2 "2022-09-28T07:23:49Z")

</div>

Just got answered on [Discord](https://discord.com/channels/879548962464493619/1024567195830861895/1024572991901814854) resp. the answer is:

You can use git tags and just use the same repo if you don’t want to create multiple ones

 ![Bildschirmfoto 2022-09-28 um 09.12.02](https://us1.discourse-cdn.com/hellohellohello/original/2X/5/55a81d63677df9f1a9b44d0299690e3d8d61b644.jpeg)

You can also add tags programmatically using the `huggingface_hub` library

`Repository.add_tag`