# CPU usage during batched training

**URL:** <https://discuss.huggingface.co/t/cpu-usage-during-batched-training/34291>\
**Category:** 🤗Transformers\
**Created:** [March 21, 2023, 10:40am UTC](https://discuss.huggingface.co/t/cpu-usage-during-batched-training/34291 "2023-03-21T10:40:32Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![XDoubleU](https://avatars.discourse-cdn.com/v4/letter/x/6f9a4e/32.png) [@XDoubleU](https://discuss.huggingface.co/u/XDoubleU)\
**Post date:** [March 21, 2023, 10:40am UTC](https://discuss.huggingface.co/t/cpu-usage-during-batched-training/34291/1 "2023-03-21T10:40:32Z")

</div>

I noticed that during training the CPU only uses 1 thread. Since my batches are quite small and I’m swapping often I’m wondering if parallelization is possible and if it actually would increase training speed?