anishathalye / neural-style

Neural style in TensorFlow! 🎨
https://anishathalye.com/an-ai-that-can-mimic-any-artist/
GNU General Public License v3.0
5.54k stars 1.52k forks source link

Small images still run out memory #97

Closed andycui97 closed 6 years ago

andycui97 commented 7 years ago

Having read issues #21, #28, #60, and #61 I've assumed that a GPU with 16GB memory should be more than able to handle images of size 64x64.

Unfortunately I still experience crashes on the first iteration that have been described in #61.

I've tried having the three sessions I could find defined in stylize.py to grow their memory allocation using

config.gpu_options.allow_growth=True

but the problem persists.

Any ideas?

anishathalye commented 7 years ago

What GPU are you using?

On Jul 10, 2017, at 2:37 PM, Andrew Cui notifications@github.com wrote:

Having read issues #21, #28, #60, and #61 I've assumed that a GPU with 16GB memory should be more than able to handle images of size 64x64.

Unfortunately I still experience crashes on the first iteration that have been described in #61.

I've tried having the three sessions I could find defined in stylize.py to grow their memory allocation using

config.gpu_options.allow_growth=True

but the problem persists.

Any ideas?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

andycui97 commented 7 years ago

Tesla P100, I'm working on a DGX-1 machine.

anishathalye commented 7 years ago

Maybe something is wrong with your particular setup. I can do ~ 1000x1000 on my Titan X (Maxwell).