Kernel Restarting

[zysmu]

Hi,

Thanks for providing such a usful tool SCimilarity.

I run SCimilarity in a server with 386G, and I don't use the docker.

I run the "individual cell research" as the Tutorials step by step. I have successfully run the code:
" from scimilarity.utils import lognorm_counts, align_dataset
from scimilarity import CellQuery"

But I get "Kernel Restarting", when I run the code:
"model_path = "/models/model_v1.1"
cq = CellQuery(model_path)"

Best,
zhouY

[javh]

Greetings @zysmu,

That much RAM should be more than sufficient (60-80GB should be enough). The first thing
I suggest trying is updating hnswlib if it's an older version. We've had memory issues with older versions of hnswlib.

[lidanwu]

Jupyter notebook as a configuration set the max memory usage allowed and you can change it by launch it with higher memory limit. for example,
jupyter notebook --ServerApp.max_buffer_size=64000000000

With that being said, I did encountered similar issue even when using python console to run the code directly. I was running the cell type annotation tutorial on a 64-bit windows with 128GB RAM. I have the latest version of hnswlib (0.8.0) and have verified the available RAM using psutil. But as soon as I did the ca.get_predictions_knn() step, the python session just crashed.