# Using low precision matrix multiplication to boost performance

**URL:** <https://discourse.scverse.org/t/using-low-precision-matrix-multiplication-to-boost-performance/2522>\
**Category:** scvi-tools\
**Created:** [September 19, 2024, 2:23am UTC](https://discourse.scverse.org/t/using-low-precision-matrix-multiplication-to-boost-performance/2522 "2024-09-19T02:23:08Z")\
**Posts on this page:** 1\
**Showing post:** 5

<div class="post-metadata">

**Author:** ![abs51295](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.scverse.org/abs51295/32/994_2.png) [@abs51295](https://discourse.scverse.org/u/abs51295)\
**Post date:** [September 19, 2024, 5:43am UTC](https://discourse.scverse.org/t/using-low-precision-matrix-multiplication-to-boost-performance/2522/5 "2024-09-19T05:43:04Z")

</div>

How do I track the training so that I know that changing to low precision did/didn’t work? What would you look for?

Also in [Increase scVI integration speed - #2 by Valentine\_Svensson](https://discourse.scverse.org/t/increase-scvi-integration-speed/1772/2) @Valentine_Svensson didn’t see any significant increase by changing batch\_size. Is it still the case?

---

_[View the full topic](https://discourse.scverse.org/t/using-low-precision-matrix-multiplication-to-boost-performance/2522)._