Closed zbarry closed 5 years ago
FWIW, this is also happening on the dev branch. y.shape
-> (500, 540, 720)
.
@zbarry Possibly unrelated but your environment says you're using python 3.6 yet the log mentions python 3.7
@epnev that's actually because I realized I was working in 3.7 in the notebook when I started making the issue, then tried a 3.6 kernel and got the same error.
I see these lines pop up multiple times:
WARNING:root:Changing key init_batch in group online from 500 to 5000
INFO:root:(5000, 141, 141)
Not sure what that refers to yet, but those are also the two sizes in the broadcast mismatch.
That's strange indeed. Can you send the script you're trying to run? I'll try to take a look later tonight or tomorrow.
Errrrr.... you know what? This is where the 5000 came from:
filename = 'blarg_d1_540_d2_720_d3_1_order_C_frames_5000_.mmap'
yr, dims, t = cm.load_memmap(filename)
y = yr.T.reshape((t,) + dims, order='F')
# y = y[1000:3000] # perturbation at approximately frame 2000
y = y[2000:2500]
#y = y[2000:2200]
y_vis = y.transpose((1, 2, 0))
I was passing the time-cropped y
without even thinking about CaImAn pulling from y.filename
. Whoops. Don't usually run CaImAn from notebooks like this. Thanks, haha.
CNMF-E stage fails with an array broadcasting error after individual patch processing is finished and patches are trying to be integrated.
Environment
CentOS 7 / Jupyter notebook / Python 3.6
CNMF-E params
yields:
Error log
See: https://gist.github.com/zbarry/12e19f78775b915e9e05606d8b7bff28 for full log.