Stable, needs to be way faster

Joseph Redmon

2015-01-14 08b757a0bf76efe8c76b453063a1bb19315bcaa6

 src/network_gpu.c

@@ -24,6 +24,7 @@
{
    int i;
    for(i = 0; i < net.n; ++i){
        //clock_t time = clock();
        if(net.types[i] == CONVOLUTIONAL){
            convolutional_layer layer = *(convolutional_layer *)net.layers[i];
            forward_convolutional_layer_gpu(layer, input);
@@ -59,6 +60,8 @@
            forward_crop_layer_gpu(layer, input);
            input = layer.output_cl;
        }
        check_error(cl);
        //printf("Forward %d %s %f\n", i, get_layer_string(net.types[i]), sec(clock() - time));
    }
}

@@ -68,6 +71,7 @@
    cl_mem prev_input;
    cl_mem prev_delta;
    for(i = net.n-1; i >= 0; --i){
        //clock_t time = clock();
        if(i == 0){
            prev_input = input;
            prev_delta = 0;
@@ -99,6 +103,8 @@
            softmax_layer layer = *(softmax_layer *)net.layers[i];
            backward_softmax_layer_gpu(layer, prev_delta);
        }
        check_error(cl);
        //printf("Backward %d %s %f\n", i, get_layer_string(net.types[i]), sec(clock() - time));
    }
}