# Default distributed strategy used in single-node multi-GPU env

**URL:** <https://discuss.huggingface.co/t/default-distributed-strategy-used-in-single-node-multi-gpu-env/54610>\
**Category:** 🤗Transformers\
**Created:** [September 12, 2023, 6:42am UTC](https://discuss.huggingface.co/t/default-distributed-strategy-used-in-single-node-multi-gpu-env/54610 "2023-09-12T06:42:03Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![mk6](https://avatars.discourse-cdn.com/v4/letter/m/7993a0/32.png) [@mk6](https://discuss.huggingface.co/u/mk6)\
**Post date:** [September 12, 2023, 6:42am UTC](https://discuss.huggingface.co/t/default-distributed-strategy-used-in-single-node-multi-gpu-env/54610/1 "2023-09-12T06:42:03Z")

</div>

Hi,

It seems that in a single-node multi-GPU env, all GPUs are automatically used. I’m wondering what distributed strategy is used? DP? DDP? I didn’t find any documentation describing related behaviors.

Thanks