# Model size and VRAM requirements

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmodel-size-and-vram-requirements%2F167761%3E%5C
**Category:** Beginners\
**Created:** [August 25, 2025, 7:01pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmodel-size-and-vram-requirements%2F167761 "2025-08-25T19:01:27Z")\
**Posts on this page:** 1\
**Showing post:** 4

<div class="post-metadata">

**Author:** ![John6666](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/john6666/32/27664_2.png) [@John6666](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FJohn6666)%5C
**Post date:** [August 26, 2025, 1:42pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmodel-size-and-vram-requirements%2F167761%2F4 "2025-08-26T13:42:17Z")

</div>

> `facebook/bart-large` has roughly 0.4 billion parameters. Therefore, even with float32 is should fit in 14GB VRAM.

Yeah. I think so, too.

[When using `DataParallel` in a multi-GPU environment, unexpected out-of-memory errors can occur](https://discuss.pytorch.org/t/why-torch-nn-dataparallel-causes-unexpected-cuda-oom/135277). While somewhat complex, [using FSDP or similar methods might help avoid this](https://huggingface.co/docs/transformers/v4.48.0/en/perf_train_gpu_many).

---

_[View the full topic](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fmodel-size-and-vram-requirements%2F167761)._