So I have this new programming paradigm.......

Joseph Redmon

2016-09-25 481b57a96a9ef29b112caec1bb3e17ffb043ceae

 src/network_kernels.cu

@@ -22,7 +22,6 @@
#include "region_layer.h"
#include "convolutional_layer.h"
#include "activation_layer.h"
#include "deconvolutional_layer.h"
#include "maxpool_layer.h"
#include "reorg_layer.h"
#include "avgpool_layer.h"
@@ -51,49 +50,7 @@
        if(l.delta_gpu){
            fill_ongpu(l.outputs * l.batch, 0, l.delta_gpu, 1);
        }
        if(l.type == CONVOLUTIONAL){
            forward_convolutional_layer_gpu(l, state);
        } else if(l.type == DECONVOLUTIONAL){
            forward_deconvolutional_layer_gpu(l, state);
        } else if(l.type == ACTIVE){
            forward_activation_layer_gpu(l, state);
        } else if(l.type == LOCAL){
            forward_local_layer_gpu(l, state);
        } else if(l.type == DETECTION){
            forward_detection_layer_gpu(l, state);
        } else if(l.type == REGION){
            forward_region_layer_gpu(l, state);
        } else if(l.type == CONNECTED){
            forward_connected_layer_gpu(l, state);
        } else if(l.type == RNN){
            forward_rnn_layer_gpu(l, state);
        } else if(l.type == GRU){
            forward_gru_layer_gpu(l, state);
        } else if(l.type == CRNN){
            forward_crnn_layer_gpu(l, state);
        } else if(l.type == CROP){
            forward_crop_layer_gpu(l, state);
        } else if(l.type == COST){
            forward_cost_layer_gpu(l, state);
        } else if(l.type == SOFTMAX){
            forward_softmax_layer_gpu(l, state);
        } else if(l.type == NORMALIZATION){
            forward_normalization_layer_gpu(l, state);
        } else if(l.type == BATCHNORM){
            forward_batchnorm_layer_gpu(l, state);
        } else if(l.type == MAXPOOL){
            forward_maxpool_layer_gpu(l, state);
        } else if(l.type == REORG){
            forward_reorg_layer_gpu(l, state);
        } else if(l.type == AVGPOOL){
            forward_avgpool_layer_gpu(l, state);
        } else if(l.type == DROPOUT){
            forward_dropout_layer_gpu(l, state);
        } else if(l.type == ROUTE){
            forward_route_layer_gpu(l, net);
        } else if(l.type == SHORTCUT){
            forward_shortcut_layer_gpu(l, state);
        }
        l.forward_gpu(l, state);
        state.input = l.output_gpu;
    }
}
@@ -115,47 +72,7 @@
            state.input = prev.output_gpu;
            state.delta = prev.delta_gpu;
        }
        if(l.type == CONVOLUTIONAL){
            backward_convolutional_layer_gpu(l, state);
        } else if(l.type == DECONVOLUTIONAL){
            backward_deconvolutional_layer_gpu(l, state);
        } else if(l.type == ACTIVE){
            backward_activation_layer_gpu(l, state);
        } else if(l.type == LOCAL){
            backward_local_layer_gpu(l, state);
        } else if(l.type == MAXPOOL){
            if(i != 0) backward_maxpool_layer_gpu(l, state);
        } else if(l.type == REORG){
            backward_reorg_layer_gpu(l, state);
        } else if(l.type == AVGPOOL){
            if(i != 0) backward_avgpool_layer_gpu(l, state);
        } else if(l.type == DROPOUT){
            backward_dropout_layer_gpu(l, state);
        } else if(l.type == DETECTION){
            backward_detection_layer_gpu(l, state);
        } else if(l.type == REGION){
            backward_region_layer_gpu(l, state);
        } else if(l.type == NORMALIZATION){
            backward_normalization_layer_gpu(l, state);
        } else if(l.type == BATCHNORM){
            backward_batchnorm_layer_gpu(l, state);
        } else if(l.type == SOFTMAX){
            if(i != 0) backward_softmax_layer_gpu(l, state);
        } else if(l.type == CONNECTED){
            backward_connected_layer_gpu(l, state);
        } else if(l.type == RNN){
            backward_rnn_layer_gpu(l, state);
        } else if(l.type == GRU){
            backward_gru_layer_gpu(l, state);
        } else if(l.type == CRNN){
            backward_crnn_layer_gpu(l, state);
        } else if(l.type == COST){
            backward_cost_layer_gpu(l, state);
        } else if(l.type == ROUTE){
            backward_route_layer_gpu(l, net);
        } else if(l.type == SHORTCUT){
            backward_shortcut_layer_gpu(l, state);
        }
        l.backward_gpu(l, state);
    }
}

@@ -166,20 +83,8 @@
    float rate = get_current_rate(net);
    for(i = 0; i < net.n; ++i){
        layer l = net.layers[i];
        if(l.type == CONVOLUTIONAL){
            update_convolutional_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
        } else if(l.type == DECONVOLUTIONAL){
            update_deconvolutional_layer_gpu(l, rate, net.momentum, net.decay);
        } else if(l.type == CONNECTED){
            update_connected_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
        } else if(l.type == GRU){
            update_gru_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
        } else if(l.type == RNN){
            update_rnn_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
        } else if(l.type == CRNN){
            update_crnn_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
        } else if(l.type == LOCAL){
            update_local_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
        if(l.update_gpu){
            l.update_gpu(l, update_batch, rate, net.momentum, net.decay);
        }
    }
}
@@ -271,20 +176,8 @@
{
    int update_batch = net.batch*net.subdivisions;
    float rate = get_current_rate(net);
    if(l.type == CONVOLUTIONAL){
        update_convolutional_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
    } else if(l.type == DECONVOLUTIONAL){
        update_deconvolutional_layer_gpu(l, rate, net.momentum, net.decay);
    } else if(l.type == CONNECTED){
        update_connected_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
    } else if(l.type == RNN){
        update_rnn_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
    } else if(l.type == GRU){
        update_gru_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
    } else if(l.type == CRNN){
        update_crnn_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
    } else if(l.type == LOCAL){
        update_local_layer_gpu(l, update_batch, rate, net.momentum, net.decay);
    if(l.update_gpu){
        l.update_gpu(l, update_batch, rate, net.momentum, net.decay);
    }
}

@@ -463,7 +356,7 @@
    }
    for(i = 0; i < n; ++i){
        pthread_join(threads[i], 0);
        printf("%f\n", errors[i]);
        //printf("%f\n", errors[i]);
        sum += errors[i];
    }
    if (get_current_batch(nets[0]) % interval == 0) {
@@ -492,6 +385,7 @@

float *network_predict_gpu(network net, float *input)
{
    cuda_set_device(net.gpu_index);
    int size = get_network_input_size(net) * net.batch;
    network_state state;
    state.index = 0;