Memory error on a 6 hour 64 channel recording

[KaiserArndt2]

When trying to run KS2 I can't get past learnAndSolve8b. I have tried modifying the ops.NT in config file (using values ranging from 16 X 1024 to 128 X 1024), all attempts have ended in memory errors at the learnAndSolve8b step. I have seen thread 135 about possibly cutting the recordings or template matching and neither of those options work for us. I was hoping you would have a solution to this memory error.

Thanks,
Kaiser

[JoseGuzman]

I'm not sure that it's related to issue #135. It happens also to me with smaller data. It may be also related to issue #218?

I think it happens when some batches are not having spikes at all - e.g. recordings with very spurious activity or when setting ops.minFR very low (I tend to use 1/600, 1 spike in 600 sampling points). Check #44 (comment) for details.

[marius10p]

What Jose said, try increasing your batch size much more. Also, remember to add ops.ntbuff:

ops.NT = 64*1024+ ops.ntbuff; % must be multiple of 32 + ntbuff. This is the batch size (try decreasing if out of memory).