script_diffusion_umap problem

[chengfengke]

excuse me,i would like to know why I got ---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
Cell In[8], line 6
3 adata_w.var_names = gene_names
5 autoencoder = load_VAE()
----> 6 cell_w = autoencoder(torch.tensor(adata_w.X).cuda(),return_latent=True).detach().cpu().numpy()
8 # concat this cell_w with cell_gen and send them to the autoencoder
9 cell_gen_all = autoencoder(torch.tensor(np.concatenate((cell_gen,cell_w),axis=0)).cuda(),return_decoded=True).cpu().detach().numpy()

File ~/anaconda3/envs/scDiffusion/lib/python3.9/site-packages/scipy/sparse/_base.py:404, in _spbase.len(self)
403 def len(self):
--> 404 raise TypeError("sparse array length is ambiguous; use getnnz()"
405 " or shape[0]")

TypeError: sparse array length is ambiguous; use getnnz() or shape[0] when I executing

if not generated all type of cells, use the real cell to balance the batchnorm in the scimilarity

adata_w = adata.copy()[::5]
adata_w.var_names = gene_names

autoencoder = load_VAE()
cell_w = autoencoder(torch.tensor(adata_w.X).cuda(),return_latent=True).detach().cpu().numpy()

concat this cell_w with cell_gen and send them to the autoencoder

cell_gen_all = autoencoder(torch.tensor(np.concatenate((cell_gen,cell_w),axis=0)).cuda(),return_decoded=True).cpu().detach().numpy()
cell_w.shape
this part of code, and I got different output with your previous results,cell_gen.shape=(600, 18996),not (12000,18996)

[chengfengke]

is this a problem of the version of scipy？

[EperLuo]

Hi! Thank you for raising this issue. We did find a small problem with the code here. The adata_w.X is a spare matrix, not a numpy array, so the right way is:

# if not generated all type of cells, use the real cell to balance the batchnorm in the scimilarity
adata_w = adata.copy()[::5]
adata_w.var_names = gene_names

autoencoder = load_VAE()
cell_w = autoencoder(torch.tensor(adata_w.X.toarray()).cuda(),return_latent=True).detach().cpu().numpy()

We will update this part of the code after we finish the revision.

[chengfengke]

thank for your quick reply, I will try it.

[chengfengke]

I am not sure if it's right? and data.X is an array so it can not use to array(),and the two metrics can not alignment.

[JeanRadig]

Hi chengfengke,

In case you still have this issue, you can look at the fork I have created to reproduce the results of the paper: https://github.com/JeanRadig/scDiffusion.git.

Under Jean_Instructions you will find two notebooks that should be able to yield the plotting you are trying to get. Hope this helps :)