so much need to commit

Joseph Redmon

2016-05-06 c7b10ceadb1a78e7480d281444a31ae2a7dc1b05

 src/network_kernels.cu

@@ -1,3 +1,7 @@
#include "cuda_runtime.h"
#include "curand.h"
#include "cublas_v2.h"

extern "C" {
#include <stdio.h>
#include <time.h>
@@ -7,22 +11,27 @@
#include "image.h"
#include "data.h"
#include "utils.h"
#include "params.h"
#include "parser.h"

#include "crop_layer.h"
#include "connected_layer.h"
#include "rnn_layer.h"
#include "gru_layer.h"
#include "crnn_layer.h"
#include "detection_layer.h"
#include "region_layer.h"
#include "convolutional_layer.h"
#include "activation_layer.h"
#include "deconvolutional_layer.h"
#include "maxpool_layer.h"
#include "avgpool_layer.h"
#include "normalization_layer.h"
#include "batchnorm_layer.h"
#include "cost_layer.h"
#include "local_layer.h"
#include "softmax_layer.h"
#include "dropout_layer.h"
#include "route_layer.h"
#include "shortcut_layer.h"
#include "blas.h"
}

@@ -34,20 +43,29 @@
{
    int i;
    for(i = 0; i < net.n; ++i){
        state.index = i;
        layer l = net.layers[i];
        if(l.delta_gpu){
            scal_ongpu(l.outputs * l.batch, 0, l.delta_gpu, 1);
            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){
@@ -56,6 +74,8 @@
            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 == AVGPOOL){
@@ -64,6 +84,8 @@
            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);
        }
        state.input = l.output_gpu;
    }
@@ -75,6 +97,7 @@
    float * original_input = state.input;
    float * original_delta = state.delta;
    for(i = net.n-1; i >= 0; --i){
        state.index = i;
        layer l = net.layers[i];
        if(i == 0){
            state.input = original_input;
@@ -88,6 +111,10 @@
            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 == AVGPOOL){
@@ -96,18 +123,26 @@
            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);
        }
    }
}
@@ -125,6 +160,14 @@
            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);
        }
    }
}
@@ -132,8 +175,11 @@
float train_network_datum_gpu(network net, float *x, float *y)
{
    network_state state;
    state.index = 0;
    state.net = net;
    int x_size = get_network_input_size(net)*net.batch;
    int y_size = get_network_output_size(net)*net.batch;
    if(net.layers[net.n-1].type == DETECTION) y_size = net.layers[net.n-1].truths*net.batch;
    if(!*net.input_gpu){
        *net.input_gpu = cuda_make_array(x, x_size);
        *net.truth_gpu = cuda_make_array(y, y_size);
@@ -171,6 +217,8 @@
{
    int size = get_network_input_size(net) * net.batch;
    network_state state;
    state.index = 0;
    state.net = net;
    state.input = cuda_make_array(input, size);
    state.truth = 0;
    state.train = 0;