Added support for Tensor Cores CC >= 7.0 (V100). For FP16/32 (mixed precisi...

AlexeyAB

2018-02-25 cad4d1618fee74471d335314cb77070fee951a42

 src/network_kernels.cu

@@ -117,7 +117,7 @@
   int i;
   for (i = 0; i < net.n; ++i) {
      layer l = net.layers[i];
      cuda_convert_f32_to_f16(l.weights_gpu, l.c*l.n*l.size*l.size, (half *)l.weights_gpu16);
      cuda_convert_f32_to_f16(l.weights_gpu, l.c*l.n*l.size*l.size, l.weights_gpu16);
   }
#endif
    forward_network_gpu(net, state);