-
Notifications
You must be signed in to change notification settings - Fork 9
Open
Description
Hi,
I have already installed PyTorch, and I verified that CUDA is available on my system using the following code:
import torch
print(torch.cuda.is_available())
The above code returns True. But when I run Scube.align() function, it runs very slowly
And I checked the GPU usage through nvidia-smi, I found that Scube didn't use GPU.
Here's the core code I'm running:
adata_list = []
for sample_id in sample_ids:
adata = data_reader.read_dataset(dataset,sample_id)
adata_list.append(adata)
from SPACEL import Scube
Scube.align(adata_list,
cluster_key='ground_truth',
n_neighbors = 4,
p=1,
write_loc_path = f'{output_dir}/aligned_coordinates.csv'
)
It took 20 minutes to align two anndata
AnnData object with n_obs × n_vars = 2378 × 254
AnnData object with n_obs × n_vars = 2423 × 254
Is this normal? Is there something I'm doing wrong during the usage, or are there some configuration or compatibility issues with Scube.align itself? I hope to get your help. Thank you!
Best regards.
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels