# Quantizing Facebook's segment anything model

**URL:** <https://discuss.huggingface.co/t/quantizing-facebooks-segment-anything-model/42674>\
**Category:** Beginners\
**Created:** [June 9, 2023, 4:02pm UTC](https://discuss.huggingface.co/t/quantizing-facebooks-segment-anything-model/42674 "2023-06-09T16:02:08Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![shaivalp](https://avatars.discourse-cdn.com/v4/letter/s/eada6e/32.png) [@shaivalp](https://discuss.huggingface.co/u/shaivalp)\
**Post date:** [June 9, 2023, 4:02pm UTC](https://discuss.huggingface.co/t/quantizing-facebooks-segment-anything-model/42674/1 "2023-06-09T16:02:08Z")

</div>

Hello,  
I am trying to quantize Facebook’s segment anything transformer model “facebook/sam-vit-huge”, especially the encoder side so that I can get a lower inference time. I have an NVIDIA RTX 3060 12GB RAM GPU. Can anyone help me get started with this?

---

<div class="post-metadata">

**Author:** ![1elsa1](https://avatars.discourse-cdn.com/v4/letter/1/7ab992/32.png) [@1elsa1](https://discuss.huggingface.co/u/1elsa1)\
**Post date:** [June 28, 2023, 10:06am UTC](https://discuss.huggingface.co/t/quantizing-facebooks-segment-anything-model/42674/2 "2023-06-28T10:06:37Z")

</div>

I concern too