# Eval\_batch\_size VS per\_device\_eval\_batch\_size

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Feval-batch-size-vs-per-device-eval-batch-size%2F49490%3E%5C
**Category:** DeepSpeed\
**Created:** [August 4, 2023, 1:51pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Feval-batch-size-vs-per-device-eval-batch-size%2F49490 "2023-08-04T13:51:21Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![charles0120](https://avatars.discourse-cdn.com/v4/letter/c/eb8c5e/32.png) [@charles0120](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fcharles0120)%5C
**Post date:** [August 4, 2023, 1:51pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Feval-batch-size-vs-per-device-eval-batch-size%2F49490%2F1 "2023-08-04T13:51:21Z")

</div>

I noticed that there are two batch size args for evaluation in training args. What’s the difference between the two? It seems eval\_batch\_size is still used as per-device evaluation batch size in the trainer source codes?