For convenience and speed, I preload all images (roughly 20000) before training and draw loss curve using gnu at the end of every 10 epochs (so there are total 10 points to draw). Then gnu gives me this error:
cannot open </usr/bin/gnuplot -persist > /dev/null 2>&1> in mode w. This might be because eg the executable doesn't exist, but it could also be because you are out of memory. at /tmp/luarocks_torch-scm-1-5129/torch7/lib/TH/THDiskFile.c:751.
The error is caused by this line:
ff = gnuplot.pngfigure('Training_aug_'..loader.N..'.png')
But if I load less images, say 13000, then gnu works fine. Is there a way to avoid this error while loading 20000 images? THX
For convenience and speed, I preload all images (roughly 20000) before training and draw loss curve using gnu at the end of every 10 epochs (so there are total 10 points to draw). Then gnu gives me this error:
cannot open </usr/bin/gnuplot -persist > /dev/null 2>&1> in mode w. This might be because eg the executable doesn't exist, but it could also be because you are out of memory. at /tmp/luarocks_torch-scm-1-5129/torch7/lib/TH/THDiskFile.c:751.The error is caused by this line:
ff = gnuplot.pngfigure('Training_aug_'..loader.N..'.png')But if I load less images, say 13000, then gnu works fine. Is there a way to avoid this error while loading 20000 images? THX