Open lyciansarpedon opened 4 years ago
Seen this recently exactly on a windows laptop. Not sure but sound like something messed up with the environment, are you working on the base env? Try creating a fresh conda environment and installing scanpy there.
Seen this recently exactly on a windows laptop. Not sure but sound like something messed up with the environment, are you working on the base env? Try creating a fresh conda environment and installing scanpy there.
Thanks for the suggestion, @giovp. Was doing it in the base env earlier. Made a new env and tried it again, but ran into the same exact error.
What's your version of numba
in this environment?
What's your version of
numba
in this environment?
It's version 0.48.0, @ivirshup. For what it's worth, the build is py38he350917_0 and the source is conda-forge. (In the base environment where I was getting the same error, the version is the same but the build is py37h47e9c7a_0.)
Could you show the result of print(adata_10x.X)
?
I'm not sure how much I can help with this, since I don't have a windows machine to try this on. I think this might be best raised an issue over on the numba
repository. @giovp, maybe you could look into more about how came across this?
Here's what I got, @ivirshup:
print(adata_10x.X)
(0, 70) 1.0
(0, 166) 1.0
(0, 178) 2.0
(0, 326) 1.0
(0, 363) 1.0
(0, 410) 1.0
(0, 412) 1.0
(0, 492) 41.0
(0, 494) 1.0
(0, 495) 1.0
(0, 496) 1.0
(0, 525) 1.0
(0, 556) 2.0
(0, 558) 6.0
(0, 671) 1.0
(0, 684) 1.0
(0, 735) 1.0
(0, 770) 1.0
(0, 793) 1.0
(0, 820) 1.0
(0, 859) 2.0
(0, 871) 1.0
(0, 908) 15.0
(0, 926) 1.0
(0, 941) 1.0
: :
(2699, 31849) 1.0
(2699, 31855) 1.0
(2699, 31887) 1.0
(2699, 31949) 2.0
(2699, 31970) 2.0
(2699, 32022) 17.0
(2699, 32044) 1.0
(2699, 32047) 2.0
(2699, 32059) 1.0
(2699, 32065) 1.0
(2699, 32066) 1.0
(2699, 32082) 1.0
(2699, 32186) 1.0
(2699, 32193) 1.0
(2699, 32322) 1.0
(2699, 32442) 1.0
(2699, 32543) 1.0
(2699, 32581) 1.0
(2699, 32641) 1.0
(2699, 32696) 3.0
(2699, 32697) 1.0
(2699, 32698) 7.0
(2699, 32702) 1.0
(2699, 32705) 1.0
(2699, 32708) 3.0
I took a look over at the numba repository, any chance this is related to https://github.com/numba/numba/issues/4529 (i.e. #843 , which I somehow missed when I was looking through issues in this repository before I posted)?
Ah, sorry, looks like I meant to say print(repr(adata.X))
.
I think the specific bug in that previous issue was fixed, but this seems kinda related in that it's windows and 32 bit numbers.
Overall, I think you should open an issue on numba for this. Please tag me if you do!
(Python & GitHub novice here, apologies in advance.)
Running through a tutorial using the 10xGenomics 3K PBMC dataset in Jupyter Notebook on Windows 10, caught an error at sc.pp.calculate_qc_metrics. Based on a quick look with my untrained eyes, this may not be a scanpy issue per se so much as an underlying data structure conflict issue in numba and/or llvmlite?
Trimmed down code I used to reach that point (the skipped steps, in ellipses, don't seem to be necessary, but I may still have a few extras there):
That spat out:
Unlikely to be related, but this was after I had issues installing scanpy from conda (as in #1142), which I got around by installing through pip.
Versions: