ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 2 dimensions. The detected shape was (1, 4) + inhomogeneous part.

[GaF123]

Thank you so much for your great tool!

I use seuratdisk convert rds to h5ad, and use velocyto convert bam to loom. And I merged these 2 files. and when I got to the: scv.tl.velocity_graph(adata)
It shows: ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 2 dimensions. The detected shape was (1, 4) + inhomogeneous part.

Here is my all codes:
import anndata
import scvelo as scv
import pandas as pd
import numpy as np
import scanpy as sc
import matplotlib as plt
file_path = '/Users/eaiscui/Desktop/temp/mouseforvelocyto_final.h5ad'
adata=sc.read_h5ad('mouseforvelocyto_final.h5ad')
adata.obs.seurat_clusters=adata.obs.seurat_clusters.astype('category')
ldata1 = anndata.read_loom('/Users/eaiscui/Desktop/loomfiles/Dminus_H2B.loom')
ldata2 = anndata.read_loom('/Users/eaiscui/Desktop/loomfiles/Dplus_H2B.loom')

rename barcodes in order to merge:

barcodes = [bc.split(':')[1] for bc in ldata1.obs.index.tolist()]
barcodes = [bc[0:len(bc)-1] + '_10' for bc in barcodes]
ldata1.obs.index = barcodes
barcodes = [bc.split(':')[1] for bc in ldata2.obs.index.tolist()]
barcodes = [bc[0:len(bc)-1] + '_11' for bc in barcodes]
ldata2.obs.index = barcodes

make variable names unique

ldata1.var_names_make_unique()
ldata2.var_names_make_unique()

concatenate the three loom

ldata = ldata1.concatenate([ldata2])
scv.utils.clean_obs_names(adata)
scv.utils.clean_obs_names(ldata)
adata = scv.utils.merge(adata, ldata)
scv.settings.verbosity = 3 # show errors(0), warnings(1), info(2), hints(3)
scv.settings.presenter_view = True # set max width size for presenter view
scv.set_figure_params('scvelo') # for beautified visualization
scv.pl.proportions(adata)
scv.pp.filter_genes(adata, min_shared_counts=20)
scv.pp.normalize_per_cell(adata)
scv.pp.filter_genes_dispersion(adata, n_top_genes=2000)
scv.pp.log1p(adata)
scv.pp.filter_and_normalize(adata, min_shared_counts=20, n_top_genes=2000)
scv.pp.moments(adata, n_pcs=30, n_neighbors=30)
scv.tl.velocity(adata)
scv.tl.velocity_graph(adata)

Then it shows:
ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 2 dimensions. The detected shape was (1, 4) + inhomogeneous part.

Please help! Thank you so much!

[jeffhsu3]

Similar issue but with alevin-fry data, concatenated with scanpy.

adatas = []
for sample in samples:
    adata = pyroe.load_fry(
        f"alevin_fry/{sample}/", 
        output_format="snRNA")
    adata.obs['sample'] = sample
    adatas.append(adata)

adata = sc.concat(adatas, index_unique='_')

[maartenciers]

It has to do with the numpy version they used. It works when you lower numpy to 1.21.1 but installing this is not possible in more recent python versions, so I switched my conda env to python 3.8 and then I could downgrade. It would be nice if they fixed this issue for more recent python/numpy versions.

[jeffhsu3]

Guessing this is the change in numpy: numpy/numpy#14794 . Would a try except at the np.array(res) be an accepted way to deal with handling versions of numpy?

[GaF123]

Thank you for your advice, it works for me!

[WeilerP]

The issue has been reported before and has been fixed in #1059.

[yamihn]

hi @WeilerP how did you solve the issue?

[ramadatta]

I used the following commands and created a new environment based on this #1040 (comment).

conda create -n scvelo_env python=3.9
conda activate scvelo_env
conda install numpy scipy cython numba matplotlib scikit-learn h5py click
pip install notebook
pip install scanpy
pip install leidenalg
pip install scvelo
pip install pandas==1.1.5 
pip install numpy==1.21.1

This error is resolved for me.

[brendamelano]

Hello, I have tried the tips above and in different threads including setting different environments with python=3.8 or 3.6. I have downgraded pandas and numpy with either numba compatability issues or kernels dying when generating umaps. Is there a way to fix this inhomogenous shape issue without downgrading libraries that the developers are working on?

[hernandezvargash]

Hi, same issue here.
I tried the downgrading solution, as explained here. I'm also working on a fresh install of Ubuntu 22.04. Only difference is I'm using mamba instead of conda.
Then I got a matplotlib error, so also downgraded to version 3.5.1 as suggested in this thread.
This took me back to the original "inhomogeneous" error.
So, I definitely agree with @brendamelano , we need a more stable solution with updated libraries.
Any suggestions?