Open wariobrega opened 11 months ago
Today I also encountered the same bug. Any fix now?
It seems that a little bit different from the above, this is the traceback info
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In[675], line 1
----> 1 mu.pp.neighbors(mdata_cc, key_added='wnn', n_neighbors=15)
2 mu.tl.umap(mdata_cc, neighbors_key='wnn', random_state=10)
File [/home/to/envs/scverse/lib/python3.10/site-packages/muon/_core/preproc.py:491](http://localhost:8989/home/to/envs/scverse/lib/python3.10/site-packages/muon/_core/preproc.py#line=490), in neighbors(mdata, n_neighbors, n_bandwidth_neighbors, n_multineighbors, neighbor_keys, metric, low_memory, key_added, weight_key, add_weights_to_modalities, eps, copy, random_state)
489 currtheta = theta
490 else:
--> 491 thetas[:, lasti] = theta
492 lasti += 1
493 ratios[ratioidx, i1] = currtheta [/](http://localhost:8989/) (np.max(thetas, axis=1) + eps)
ValueError: could not broadcast input array from shape (0,) into shape (13498,)
Hey @jefferyUstc, thanks for reporting this.
So far I wasn't able to reproduce this error. Is there any data that we could reproduce it on? Does it also happen with the latest version of muon from the main branch?
Hello,
I have a dataset made of unpaired RNA/ATAC seq data.
I performed PCA / nearest neighbors on each modality using a standard approach and I was going for the joint neighbors embedding.
I obtain however the following error:
This usually happens after a couple of minutes after running the command, so way along the line
The data is made of 5106 Cells from ATAC and 4215 Cells from RNA.
I am using the latest stable version of muon on a conda environment running python 3.10. The conda env contains only muon and its minimal dependencies + ipython kernel/jupyter specs.
Can you help me understand this issue and how to fix it?
Thanks,
Daniele