EEGlab ICAlabel not working with specific data

I have 64 channel EEG data run through preprocessing steps before running ICA. I have run 18 datasets using the same process with no issues. I have 2 datasets where the ICA takes hours and then once the ICA is finished, this error occurs:

Error using ICL_feature_extractor (line 26)

Your ICA decomposition must be real to use ICLabel

Error in iclabel (line 52)

features = ICL_feature_extractor(EEG, flag_autocorr);

The data doesn't look any different than my other datasets. Any suggestions?

[KSSV]

Which error?

[Walter Roberson]

iclabel appears to be a third-party toolbox.

You could get that message if your ica generated nan or inf or complex values.

[Rebecca Kenny]

Can you expand on what you mean by nan or inf or complex values? And how do I correct for that? Even the function in eeglab to "inspect/label components by map" doesn't work.

Still learning matlab and eeglab and I haven't seen this error before.

[Walter Roberson]

What shows up for

nnz(~isfinite(EEG))

if the result is not 0 then you have samples that are infinite or Not A Number (nan). nan is often used to indicate missing data.

[Moiz Ashfaque]

Hi @Rebecca Kenny, did you get around to solve this error? I seem to be facing the same error with my 9 channels EEG data.