Marco Bertini
Marco Bertini
The latest 0.7 branch has a VGG-D model. Did you create a script to convert those Caffe Zoo models ?
I've tried to get some GPU info with a small test program (changing one of the NVIDIA CUDA SDK 6.5 samples - i.e. samples/1_Utilities/bandwidthTest/bandwidthTest.cu) and I get the following: size_t...
I've added the following function to cnnclassify.c: void CUDAMem() { int currentDevice = 0; struct cudaDeviceProp deviceProp; cudaError_t error_id = cudaGetDeviceProperties(&deviceProp, currentDevice); printf(" Device %d: %s\n", currentDevice, deviceProp.name); size_t cudaFree,...
I've just tried using the new CUDA SDK 7.0 RC but there's still the same behavior.
It's a NVIDIA GeForce GT 750M with 2048 MB
This is true if you execute the model.compile with function objects instead of the model.compile with strings. Probably adding some text that says to update the keys in the plot...
I have to confirm what reported by @TimNie: tested using Keras 2.1.3 on Google Colab (Tensorflow 1.4.1) and on my workstation with Tensorflow 1.5.0: the curves are like those reported...