Validate and recall always with batch=1

AlexeyAB

2017-10-31 accf3a59c8b17c70dcc39e8f7daede017190086d

 src/convolutional_kernels.cu

@@ -127,6 +127,7 @@
    activate_array_ongpu(l.output_gpu, l.outputs*l.batch, l.activation);
    //if(l.dot > 0) dot_error_gpu(l);
    if(l.binary || l.xnor) swap_binary(&l);
   //cudaDeviceSynchronize(); // for correct profiling of performance
}

void backward_convolutional_layer_gpu(convolutional_layer l, network_state state)