Peyton-D

Hello, I tried to use your last commit, but I'm still getting a CUDA OOM error in the final clustering phase. How much dedicated GPU memory do you have? I...

Thanks for the quick response. Yes, kmeans_plus_plus inside of clustering_qr seems to be the cause of each crash every time. My recording duration is 90 min. Here's the problematic line...

Just ran another attempt with normal version. Here's the problem line: File "C:\Users\ColginLab\miniconda3\envs\kilosort\lib\site-packages\kilosort\clustering_qr.py", line 167, in kmeans_plusplus vtot = (Xg**2).sum(1) [kilosort4_normal_version.log](https://github.com/user-attachments/files/17166365/kilosort4_normal_version.log)

Alright, I'll look into getting more GPU memory. Thanks for the help!

@EmmettJT Yes, my issue was fixed by updating my nvidia driver to version 561.09 and cuda version to 12.6, specifically. I did not end up using this commit. My recordings...

I'm having the same issue using a single NP2.0 in 2- and even 1-shank configurations. The 2-shank sorting attempt got to 39% complete during the "kilosort.spikedetect: Re-computing universal templates from...