Processing 3rd Frame and get cudnn Argument Error about deepdreamvideo HOT 5 CLOSED

I figure I would throw some information into this, since I'm getting the same issue and have done some testing.

I can run this on 720p and above, without issue. 360p and 480p crash occasionally with the same error. I'm using cuDNNv2, cuda7.0 on a 780TI on lubuntu 15.04. I can process images individually using Dhar/image-dreamer dreamify.py at any resolution. The issue seems to be related the inception_5a or b output when using default iteration and octave parameters. Removing those from the loop yields flawless processing at 480p and 360p.

from deepdreamvideo.

Using the "Fear and Loathing in Las Vegas" demo video downloaded from Mega, I get the following error:

LayerParameter
GPU mode [device id: 0]
using GPU, but you'd still better make a cup of coffee
Processing frame #1
/home/daniel/anaconda/lib/python2.7/site-packages/scipy/ndimage/interpolation.py:549: UserWarning: From scipy 0.13.0, the output shape of zoom() is calculated with round() instead of int() - for these inputs the size of the returned array has changed.
  "the returned array has changed.", UserWarning)
F0711 15:55:59.168756 16791 cudnn_conv_layer.cu:53] Check failed: status == CUDNN_STATUS_SUCCESS (3 vs. 0)  CUDNN_STATUS_BAD_PARAM
*** Check failure stack trace: ***
Aborted (core dumped)

from deepdreamvideo.

Building caffe without cudnn appears to resolve the issue. This is however not an optimal configuration.

from deepdreamvideo.

yes this seems to be an issue in the newer caffe release and also discussed at: BVLC/caffe#2235 & BVLC/caffe#2682
As this is an issue in caffe, I'll close this issue for now, lets hope this issue will be fixed soon.

from deepdreamvideo.

Thanks for the info.

from deepdreamvideo.