some fixes

Joseph Redmon

2015-12-18 9802287b5890d9b2cc250adba1b9810657a95c9c

 src/network.c

@@ -8,10 +8,10 @@

#include "crop_layer.h"
#include "connected_layer.h"
#include "local_layer.h"
#include "convolutional_layer.h"
#include "deconvolutional_layer.h"
#include "detection_layer.h"
#include "region_layer.h"
#include "normalization_layer.h"
#include "maxpool_layer.h"
#include "avgpool_layer.h"
@@ -19,6 +19,7 @@
#include "softmax_layer.h"
#include "dropout_layer.h"
#include "route_layer.h"
#include "shortcut_layer.h"

int get_current_batch(network net)
{
@@ -26,6 +27,17 @@
    return batch_num;
}

void reset_momentum(network net)
{
    if (net.momentum == 0) return;
    net.learning_rate = 0;
    net.momentum = 0;
    net.decay = 0;
    #ifdef GPU
        if(gpu_index >= 0) update_network_gpu(net);
    #endif
}

float get_current_rate(network net)
{
    int batch_num = get_current_batch(net);
@@ -41,6 +53,7 @@
            for(i = 0; i < net.num_steps; ++i){
                if(net.steps[i] > batch_num) return rate;
                rate *= net.scales[i];
                if(net.steps[i] > batch_num - 1) reset_momentum(net);
            }
            return rate;
        case EXP:
@@ -60,6 +73,8 @@
    switch(a){
        case CONVOLUTIONAL:
            return "convolutional";
        case LOCAL:
            return "local";
        case DECONVOLUTIONAL:
            return "deconvolutional";
        case CONNECTED:
@@ -72,8 +87,6 @@
            return "softmax";
        case DETECTION:
            return "detection";
        case REGION:
            return "region";
        case DROPOUT:
            return "dropout";
        case CROP:
@@ -82,6 +95,8 @@
            return "cost";
        case ROUTE:
            return "route";
        case SHORTCUT:
            return "shortcut";
        case NORMALIZATION:
            return "normalization";
        default:
@@ -107,6 +122,7 @@
{
    int i;
    for(i = 0; i < net.n; ++i){
        state.index = i;
        layer l = net.layers[i];
        if(l.delta){
            scal_cpu(l.outputs * l.batch, 0, l.delta, 1);
@@ -115,12 +131,12 @@
            forward_convolutional_layer(l, state);
        } else if(l.type == DECONVOLUTIONAL){
            forward_deconvolutional_layer(l, state);
        } else if(l.type == LOCAL){
            forward_local_layer(l, state);
        } else if(l.type == NORMALIZATION){
            forward_normalization_layer(l, state);
        } else if(l.type == DETECTION){
            forward_detection_layer(l, state);
        } else if(l.type == REGION){
            forward_region_layer(l, state);
        } else if(l.type == CONNECTED){
            forward_connected_layer(l, state);
        } else if(l.type == CROP){
@@ -137,6 +153,8 @@
            forward_dropout_layer(l, state);
        } else if(l.type == ROUTE){
            forward_route_layer(l, net);
        } else if(l.type == SHORTCUT){
            forward_shortcut_layer(l, state);
        }
        state.input = l.output;
    }
@@ -155,6 +173,8 @@
            update_deconvolutional_layer(l, rate, net.momentum, net.decay);
        } else if(l.type == CONNECTED){
            update_connected_layer(l, update_batch, rate, net.momentum, net.decay);
        } else if(l.type == LOCAL){
            update_local_layer(l, update_batch, rate, net.momentum, net.decay);
        }
    }
}
@@ -180,10 +200,6 @@
            sum += net.layers[i].cost[0];
            ++count;
        }
        if(net.layers[i].type == REGION){
            sum += net.layers[i].cost[0];
            ++count;
        }
    }
    return sum/count;
}
@@ -201,6 +217,7 @@
    float *original_input = state.input;
    float *original_delta = state.delta;
    for(i = net.n-1; i >= 0; --i){
        state.index = i;
        if(i == 0){
            state.input = original_input;
            state.delta = original_delta;
@@ -224,16 +241,18 @@
            backward_dropout_layer(l, state);
        } else if(l.type == DETECTION){
            backward_detection_layer(l, state);
        } else if(l.type == REGION){
            backward_region_layer(l, state);
        } else if(l.type == SOFTMAX){
            if(i != 0) backward_softmax_layer(l, state);
        } else if(l.type == CONNECTED){
            backward_connected_layer(l, state);
        } else if(l.type == LOCAL){
            backward_local_layer(l, state);
        } else if(l.type == COST){
            backward_cost_layer(l, state);
        } else if(l.type == ROUTE){
            backward_route_layer(l, net);
        } else if(l.type == SHORTCUT){
            backward_shortcut_layer(l, state);
        }
    }
}
@@ -245,6 +264,8 @@
    if(gpu_index >= 0) return train_network_datum_gpu(net, x, y);
#endif
    network_state state;
    state.index = 0;
    state.net = net;
    state.input = x;
    state.delta = 0;
    state.truth = y;
@@ -297,6 +318,8 @@
{
    int i,j;
    network_state state;
    state.index = 0;
    state.net = net;
    state.train = 1;
    state.delta = 0;
    float sum = 0;
@@ -433,6 +456,8 @@
#endif

    network_state state;
    state.net = net;
    state.index = 0;
    state.input = input;
    state.truth = 0;
    state.train = 0;
@@ -540,12 +565,12 @@
    return acc;
}

float *network_accuracies(network net, data d)
float *network_accuracies(network net, data d, int n)
{
    static float acc[2];
    matrix guess = network_predict_data(net, d);
    acc[0] = matrix_topk_accuracy(d.y, guess,1);
    acc[1] = matrix_topk_accuracy(d.y, guess,5);
    acc[0] = matrix_topk_accuracy(d.y, guess, 1);
    acc[1] = matrix_topk_accuracy(d.y, guess, n);
    free_matrix(guess);
    return acc;
}