# Calculate tokens per second while fine-tuning llm?

**URL:** <https://discuss.huggingface.co/t/calculate-tokens-per-second-while-fine-tuning-llm/107098>\
**Category:** DeepSpeed\
**Created:** [September 17, 2024, 4:41pm UTC](https://discuss.huggingface.co/t/calculate-tokens-per-second-while-fine-tuning-llm/107098 "2024-09-17T16:41:03Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![superbigtree](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/superbigtree/32/20868_2.png) [@superbigtree](https://discuss.huggingface.co/u/superbigtree)\
**Post date:** [September 17, 2024, 4:41pm UTC](https://discuss.huggingface.co/t/calculate-tokens-per-second-while-fine-tuning-llm/107098/1 "2024-09-17T16:41:03Z")

</div>

Hi everyone, how is the tokens per second calculated during training? And how different is it compared to the inference?