# Rapids-singlecell

**URL:** https://discourse.scverse.org/c/help/rsc/48.md

[Latest](https://discourse.scverse.org/latest.md) · [Categories](https://discourse.scverse.org/categories.md) · [Tags](https://discourse.scverse.org/tags.md)

---

## [About the Rapids-singlecell category](https://discourse.scverse.org/t/about-the-rapids-singlecell-category/3865)

<div class="topic-metadata">

**Author:** [@Zethson](https://discourse.scverse.org/u/Zethson)\
**Replies:** 0

</div>

---

## [Client.restart errors for dask cuda clusters (pooled mode rapids)](https://discourse.scverse.org/t/client-restart-errors-for-dask-cuda-clusters-pooled-mode-rapids/3958)

<div class="topic-metadata">

**Author:** [@mkarikom](https://discourse.scverse.org/u/mkarikom)\
**Replies:** 0\
**Last updated:** [April 10, 2026, 1:45pm UTC](https://discourse.scverse.org/t/client-restart-errors-for-dask-cuda-clusters-pooled-mode-rapids/3958 "2026-04-10T13:45:41Z")

</div>

I need to process several mdata modalities serially (updating shared mdata embeddings in between as needed for shared embeddings) in rapids but my total cluster VRAM is only ~350gb, so my strategy is to re-initialize the…