cannot access local variable 'ad' where it is not associated with a value

[CollapsarSpectrum]

Parameter Passing Failure in anndata_to_train_val() During GEMS Construction.

When attempting to run the construct_gems_using_nsf() function with the example dataset (Mouse_embryo_svg_E9.5.h5ad), I encountered a bug. It appears that the anndata_to_train_val() function fails to correctly accept the ad argument passed from _gem_construction.py. This is strange because when I run the anndata_to_train_val() function separately with the adata from example dataset as the input, it works fine.

The code

fs.pp.construct_gems_using_nsf(adata,
                            n_gems = 20,
                            layer_key = 'count',
                            length_scale = 5.0)

Error massage

---------------------------------------------------------------------------
UnboundLocalError                         Traceback (most recent call last)
Cell In[5], line 1
----> 1 fs.pp.construct_gems_using_nsf(adata,
      2                             n_gems = 20,
      3                             layer_key = 'count',
      4                             n_inducing_pts = 500,
      5                             length_scale = 10)

File ~/anaconda3/envs/space_py311/lib/python3.11/site-packages/flowsig/preprocessing/_gem_construction.py:70, in construct_gems_using_nsf(adata, n_gems, layer_key, spatial_key, n_inducing_pts, length_scale)
     68 # Take raw count data for NSF
     69 training_fraction = 1.0
---> 70 D,Dval = anndata_to_train_val(ad,
     71                         layer=layer_key,
     72                         train_frac=training_fraction,
     73                         flip_yaxis=True)
     74 Ntr,J = D["Y"].shape
     75 Xtr = D["X"]

UnboundLocalError: cannot access local variable 'ad' where it is not associated with a value

pacakge version

spatial_factorization==0.0.1
FlowSig==0.1.2
commot==0.0.3
pyliger==0.2.3
tensorflow==2.18.0
numpy==1.26.4