[Mne_analysis] Error nproj = np.sum((S / S[0]) > 1e-2) raised when applying projectors {Disarmed}

[Eric Larson]

        External Email - Use Caution        

Are you on NumPy 1.19.4? A Windows update introduced a FPU bug a couple of
months ago that they're planning on patching in January I think:

https://secure-web.cisco.com/1NjLlYU0SqEFbhBTrJZ2O5foARTRqIBEgpBntWBp2mjVzbU9h6asq-_9ex0SFksZSGn0wQzcuHI3AqUwfC5sLqGKKwyAreNozndtHF3WFtdv7kQEQLuhq6QBo9m3vPPxWaoNas5OvbIQXecGj6XyDC0r-UOSKSoxOCl2SdATSE8WteWwCzli89gP1mp34UKm-0_8R_n-ni7pMIhFeijLMkagyngKiaRLXSNeoBLp03rEpUmw2W6fUNxjp7E_LxxDZXd8_gnTsQEvtzyWxC-3Z0Q/https%3A%2F%2Fdevelopercommunity2.visualstudio.com%2Ft%2Ffmod-after-an-update-to-windows-2004-is-causing-a%2F1207405

In the meantime, a workaround is to install NumPy 1.19.3 on Windows. There
is also a NumPy 1.20 release coming soon that should fix it.

Eric


On Tue, Dec 8, 2020 at 4:06 AM Scheltienne Mathieu <
mathieu.scheltienne at epfl.ch> wrote:

>         External Email - Use Caution
>
> Dear all,
>
>
>
> I am using MNE for EEG analysis on my MacBook Pro on both mac OS and
> Windows (double boot).
>
> Today, I could not apply projectors on Windows but I could on mac OS. The
> error raised on Windows is:
>
>
>
> *C:\Program Files\Python\Python38\lib\site-packages\mne\io\proj.py:641:
> RuntimeWarning: invalid value encountered in greater*
>
> *    nproj = np.sum((S / S[0]) > 1e-2)*
>
>
>
> The error was raised when doing:
>
>
>
> *raw = mne.io.read_raw_fif(fname, preload=True)*
>
> *raw.set_eeg_reference(ref_channels='average', projection=True) # CAR *
>
> *eog_projs, _ = mne.preprocessing.compute_proj_eog(raw, n_grad=0, n_mag=0,
> n_eeg=1, reject=None,*
>
> *
>          no_proj=True, n_jobs=1)*
>
> *raw.add_proj(eog_projs)*
>
> *raw.plot(scalings=DEFAULT_SCALINGS)*
>
>
>
> Note that using only one of the 2 projectors raises the same error; and
> using raw.apply_proj() does it as well.
>
> Same file with same script on mac OS is working fine. Any idea for a
> solution?
>
>
>
> Have a nice day,
>
> Mathieu Scheltienne
> _______________________________________________
> Mne_analysis mailing list
> Mne_analysis at nmr.mgh.harvard.edu
> https://secure-web.cisco.com/14fb4NLvrBcSD9s03BqypslRkU6TnMV0kIaEwie9oY0sk9pXr0veZBzYBBKmVKcBaYmg-5PSf3qOvh2PXGBwld7o0D7gQJvPLfTw7CDod1b_CzzhKtVKAnqHOgjOvpSRiG68HZmkr-VM78jN0o5GmEIx16vfOK-mBuiET0cokp5S0Jy2IR1Hqy-OfzTw_L9irpFR1gCzt85hi6_KGYg4EuMeAOEnfCj6bwBkEpGJXLK8iw0sI_iR3MqI8hRZ40yPpDqRt5-j-gCVojrnEGY32uA/https%3A%2F%2Fmail.nmr.mgh.harvard.edu%2Fmailman%2Flistinfo%2Fmne_analysis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20201208/547c9f41/attachment.html