Severin Dicks

Dear Rilango, thank you for the quick reply. On our institute's sever we use Debian 10. On my local machine I use Ubuntu 20.04 LTS.

Here is the `conda list`output: ``` # packages in environment at /home/sdicks-local/conda/rapids-0.18: # # Name Version Build Channel _libgcc_mutex 0.1 conda_forge conda-forge _openmp_mutex 4.5 1_gnu conda-forge abseil-cpp 20200225.2 he1b5a44_2 conda-forge...

@rilango I ran it with 70000 cells today and it ran perfectly.

You seem to have the same error that I had. @rilango pointed out that if you oversubscribe the VRAM more than 2x notebooks tend to crash. `This example relies heavily...

@m0nib what kind of dataset do you have with almost 70000 features? I would suggest to restrict that featurespace and cleanup the vram as much as possible

I was just able to confirm that this version is a lot faster than the version in the newest release. On my A100 80GB It runs in 1m 4 sec...

diffusion maps and TSNE also work with cupy and rapids. Diffusion Maps don't see a massive speedup maybe 2X on modern GPUs. TSNE sees a massive speedup. For larger anndata...

To me having the meta data (obs and var) in VRAM only makes sense for large GPUs like your RTX8000 or A100. I wrote a small anndata like object (https://github.com/Intron7/rapids_singlecell)...

Using RMM works but only to a certain extend. As far as I understand it you can oversubscribe VRAM to a maximum of 2X. If you go above that you’ll...

I created a PR to this branch to add GPU support for : *`tl.rank_gene_groups` with method='logreg' *`tl.embedding_density` *`correlation_matrix` *`diffmap` I added `.layers` support for `pp.pca`. This helps with the "Pearson...