Closed richardrasu closed 7 years ago
Have you already try https://github.com/fmassa/optimize-net? "OptNet - Reducing memory usage in torch neural nets"
@rave78 I have not yet, but I shall. Thanks for the pointer. That looks useful.
OK, so I was being thick and labouring under the illusion that I had unlimited everything on this machine, but apparently I was wrong. Apologies to all who wasted their time reading this.
Hi,
I'm fiddling about with various neural-style repos which depend on torch. Torch has, up until now, been problem free, however on a well-endowed machine I ran a transfer for a very large end image size, and after a while received this message:
I wouldn't mind so much, but I've got approximately 224GB RAM free (it's a server with 256GB in it). I'm not seeing any obvious signs of memory exhaustion while it's running. I'm on Ubuntu 16.04.2 LTS. Any quick fixes to this, or am I just expecting too much out of my output resolution (9000 px)?
Thanks for any assistance you can offer.
Richard