-
Notifications
You must be signed in to change notification settings - Fork 69
Description
This my version

Hello cell2location developers,
First of all, thank you for creating such a powerful and impressive tool. I have been following the official tutorial (https://cell2location.readthedocs.io/en/latest/notebooks/cell2location_tutorial.html
) to test the pipeline, and I encountered a few issues that I would like to ask for your advice on:
use_gpu=True not recognized
In many functions shown in the tutorial, setting use_gpu=True raises an error like “got an unexpected keyword argument 'use_gpu'”. Has this parameter been removed in recent versions? If so, what is the recommended way to control GPU usage?
mod.samples does not exist
When trying to compute the expected expression per cell type, the tutorial uses:
expected_dict = mod.module.model.compute_expected_per_cell_type(
mod.samples["post_sample_q05"],
mod.adata_manager
)
But in my installation, Cell2location objects do not have the attribute samples. I already ran mod.sample_posterior(...), but mod.samples is still missing. What is the correct way in the current version to access posterior samples (e.g. q05/q50/q95) for compute_expected_per_cell_type?
QC plotting extremely slow
Running
fig = mod.plot_spatial_QC_across_batches()
takes a very long time (I waited >6 hours on GPU without finishing). Is this expected for large datasets, or am I possibly using the function incorrectly? Is there a recommended way to speed this up (for example, subsetting data, lowering resolution, or using exported posterior instead of raw sampling)?
I want to emphasize again that I find this package very impressive and useful, and I ask these questions with genuine respect and appreciation for your work. Thank you very much for your time and support!