Leiden, Louvain with Katana Graph
In this contribution , we have introduced a new flavour 'katana' in the already existing leiden and louvain implementation. The implementation of this new flavour is based on the KatanaGraph library , which comes out to be a faster alternative to the already existing flavours, providing more than 2x speedup to the leiden and louvain execution.
Codecov Report
Merging #2409 (686e6a9) into master (9969734) will decrease coverage by
0.30%. The diff coverage is33.76%.
Additional details and impacted files
@@ Coverage Diff @@
## master #2409 +/- ##
==========================================
- Coverage 71.87% 71.58% -0.30%
==========================================
Files 98 98
Lines 11496 11544 +48
==========================================
+ Hits 8263 8264 +1
- Misses 3233 3280 +47
| Impacted Files | Coverage Δ | |
|---|---|---|
| scanpy/tools/_louvain.py | 15.51% <0.00%> (-3.84%) |
:arrow_down: |
| scanpy/tools/_leiden.py | 57.53% <49.05%> (-27.89%) |
:arrow_down: |
Hi @ivirshup, I am part of Intel Labs and we are trying to accelerate the genomics pipeline. We are trying to push some changes into scanpy details about which are mentioned in the blog : https://community.intel.com/t5/Blogs/Tech-Innovation/Artificial-Intelligence-AI/Intel-Labs-Accelerates-Single-cell-RNA-Seq-Analysis/post/1390715#:~:text=Intel%20Labs%20has%20accelerated%20a,of%20a%20single%20A100%20GPU. We are facing some issues while pushing some changes in the leiden and louvain. The error states some issues with pca in the scanpy/tests/external/test_scrublet.py::test_scrublet_params when we have not made any changes for the same. Can you please help us to resolve this issue.
Please ignore the scrublet issue. It's unrelated and we'll fix it asap.
Thank you very much for the (planned) contribution!
Hi @Zethson, Thank you so much for the response. Please let us know if any updates/changes are required to be done from our side in the above contribution, we will do them at the earliest. Also, please let us know if we need to do anything else for fixing the issues in the PR at the moment.
@Sakshi-2797 please have some patience since we're all busy at the moment. It might take us a bit of time to get back to you and your teams PRs