Cannot compute joint embedding on RNA and ATAC object

I am following this tutorial to jointly embed my single-nucelus multiome capture from 10X. I have already filtered both RNA and ATAC objects to the same cells, and made sure that the barcodes are identical, then I process with the embedding:

embedding = snap.tl.multi_spectral(
    adatas = [rna, atac],
    features = 'selected'
)[1]

Once running the above cell I am met with this error:

2024-10-09 15:44:29 - INFO - Compute normalized views...
thread '<unnamed>' panicked at src/embedding.rs:411:13:
called `Result::unwrap()` on an `Err` value: Cannot convert Array(F32) to f64 CsrMatrix

---------------------------------------------------------------------------
PanicException                            Traceback (most recent call last)
Cell In[35], line 1
----> 1 embedding = snap.tl.multi_spectral(
      2     adatas = [rna, atac],
      3     features = 'selected'
      4 )[1]

File /home/.conda/envs/scSnapATAC2_v2024.08/lib/python3.12/site-packages/snapatac2/tools/_embedding.py:533, in multi_spectral(adatas, n_comps, features, weights, random_state, weighted_by_sd)
    530 if weights is None:
    531     weights = [1.0 for _ in adatas]
--> 533 evals, evecs = internal.multi_spectral_embedding(adatas, features, weights, n_comps, random_state)
    535 if weighted_by_sd:
    536     idx = [i for i in range(evals.shape[0]) if evals[i] > 0]

PanicException: called `Result::unwrap()` on an `Err` value: Cannot convert Array(F32) to f64 CsrMatrix

It seems like the issue is that atac.X is a PyArray Element. Keep in mind atac is an AnnDataSet object that I processed with SnapATAC2 and then converted to an AnnData object in order to co-embed these ATAC data with the RNA data from the same capture.

I tried many ways to convert the PyArray to a Numpy array in order to have it be compatible, but I was unsuccessful. It seems as if I could do this if I went into Rust to modulate this PyArray, but I have no experience in Rust. Any help with this would be greatly appreciate it. Thank you!

scanpy==1.10.3
snapatac2==2.7.0
polars==1.6.0
numpy==1.26.4
pandas==2.1.1
h5py==3.11.0
anndata==0.10.8
matplotlib==3.9.2
seaborn==0.13.2
scipy==1.14.1

Start with tutorials/modality.html and the multi_spectral entry point in snapatac2/tools/_embedding.py at line 533, then inspect src/embedding.rs:411 where the panic is reported. Reproduce the call with the listed package versions and the AnnDataSet-to-AnnData conversion; done means the joint embedding completes without the F32-to-f64 CsrMatrix error.