search

scverse / scanpy

Single-cell analysis in Python. Scales to >1M cells.
https://scanpy.readthedocs.io
BSD 3-Clause "New" or "Revised" License
1.92k stars 599 forks source link

very slow when I run sc.tl.draw_graph #1242

Closed icml-compbio closed 1 year ago

ivirshup commented 4 years ago

@icml-compbio The draw_graph function calls out to forceatlas2 if you have it installed. This does seem slower than using UMAP.

@YubinXie I see some multithreading being used on my machine when I run neighbors. Is there none on yours? One thing I'd check first is to make sure UMAP is up to date and install pynndescent.

ivirshup commented 4 years ago

Python 3.8 should work. I think that warning just has to do with numba and pytorch.

ivirshup commented 4 years ago

Ah, my bad, I read the path wrong on my phone.

In general, the most recent numba release cycle had a lot of deprecations, so many packages are throwing numba warnings.

fidelram commented 4 years ago

sc.tl.draw_graph is much slower than UMAP. 20 mins for 67k cells does not surprise me.

draw_graph uses by default https://github.com/bhargavchippada/forceatlas2 You may want to search in that package if computations can be speed up.

eroell commented 1 year ago

We will close the issue for now, as based on the provided information and the discussion so far, it seems that the question has been addressed and hopefully useful insights obtained :)

However, please don't hesitate to reopen this issue or create a new one if you have any more questions or run into any related problems in the future.

Thanks for being a part of our community! :)