Revision history [back]

in short, (and sad, as it is) : no. there is no cuda optimisation for this.

however, IF you have a decent gpu, try to enable the opencl target for your network, like this , and this