void layer::randomize_orthogonal_data(
layer_data::ptr data,
layer_data_custom::ptr data_custom,
random_generator& generator) const
{
randomize_data(
data,
data_custom,
generator);
}
int main(int argc, char* argv[]) {
if(argc != 2) {
std::cerr << "Usage : " << argv[0] << "<0,1>" <<std::endl;
std::cerr << "with : " << std::endl;
std::cerr << "0 : quadratic loss" << std::endl;
std::cerr << "1 : cross entropy loss" << std::endl;
return -1;
}
bool quadratic_loss = (atoi(argv[1]) == 0);
srand(time(NULL));
// We compare our computation of the gradient to
// a finite difference approximation
// The loss is also involved
std::cout << "---------------------------------" << std::endl;
std::cout << "Comparing the analytical gradient and numerical approximation " << std::endl;
auto input = gaml::mlp::input<X>(INPUT_DIM, fillInput);
auto l1 = gaml::mlp::layer(input, HIDDEN_LAYER_SIZE, gaml::mlp::mlp_sigmoid(), gaml::mlp::mlp_dsigmoid());
auto l2 = gaml::mlp::layer(l1, HIDDEN_LAYER_SIZE, gaml::mlp::mlp_identity(), gaml::mlp::mlp_didentity());
auto l3 = gaml::mlp::layer(l2, HIDDEN_LAYER_SIZE, gaml::mlp::mlp_tanh(), gaml::mlp::mlp_dtanh());
auto l4 = gaml::mlp::layer(l3, OUTPUT_DIM, gaml::mlp::mlp_sigmoid(), gaml::mlp::mlp_dsigmoid());
auto mlp = gaml::mlp::perceptron(l4, output_of);
std::cout << "We use the following architecture : " << std::endl;
std::cout << mlp << std::endl;
std::cout << "which has a total of " << mlp.psize() << " parameters"<< std::endl;
gaml::mlp::parameters_type params(mlp.psize());
gaml::mlp::parameters_type paramsph(mlp.psize());
gaml::mlp::values_type derivatives(mlp.psize());
gaml::mlp::values_type forward_sweep(mlp.size());
X x;
auto loss_ce = gaml::mlp::loss::CrossEntropy();
auto loss_quadratic = gaml::mlp::loss::Quadratic();
auto f = [&mlp, ¶ms] (const typename decltype(mlp)::input_type& x) -> gaml::mlp::values_type {
auto output = mlp(x, params);
gaml::mlp::values_type voutput(mlp.output_size());
fillOutput(voutput.begin(), output);
return voutput;
};
auto df = [&mlp, &forward_sweep, ¶ms] (const typename decltype(mlp)::input_type& x, unsigned int parameter_dim) -> gaml::mlp::values_type {
return mlp.deriv(x, params, forward_sweep, parameter_dim);
};
unsigned int nbtrials = 100;
unsigned int nbfails = 0;
std::cout << "I will compare " << nbtrials << " times a numerical approximation and the analytical gradient we compute" << std::endl;
for(unsigned int t = 0 ; t < nbtrials ; ++t) {
randomize_data(params, -1.0, 1.0);
randomize_data(x, -1.0, 1.0);
// Compute the output at params
auto output = mlp(x, params);
gaml::mlp::values_type raw_output(OUTPUT_DIM);
fillOutput(raw_output.begin(), output);
gaml::mlp::values_type raw_outputph(OUTPUT_DIM);
// For computing the loss, we need a target
gaml::mlp::values_type raw_target(OUTPUT_DIM);
randomize_data(raw_target);
double norm_dh = 0.0;
for(unsigned int i = 0 ; i < mlp.psize() ; ++i) {
// Let us compute params + h*[0 0 0 0 0 0 1 0 0 0 0 0], the 1 at the ith position
std::copy(params.begin(), params.end(), paramsph.begin());
double dh = (sqrt(DBL_EPSILON) * paramsph[i]);
paramsph[i] += dh;
norm_dh += dh*dh;
// Compute the output at params + h
auto outputph = mlp(x, paramsph);
fillOutput(raw_outputph.begin(), outputph);
// We now compute the approximation of the derivative
if(quadratic_loss)
derivatives[i] = (loss_quadratic(raw_target, raw_outputph) - loss_quadratic(raw_target, raw_output))/dh;
else
derivatives[i] = (loss_ce(raw_target, raw_outputph) - loss_ce(raw_target, raw_output))/dh;
}
// We now compute the analytical derivatives
mlp(x, params);
std::copy(mlp.begin(), mlp.end(), forward_sweep.begin());
gaml::mlp::values_type our_derivatives(mlp.psize());
for(unsigned int i = 0 ; i < mlp.psize() ; ++i) {
if(quadratic_loss)
our_derivatives[i] = loss_quadratic.deriv(x, raw_target, forward_sweep, f, df, i);
else
our_derivatives[i] = loss_ce.deriv(x, raw_target, forward_sweep, f, df, i);
}
// We finally compute the norm of the difference
double error = 0.0;
auto diter = derivatives.begin();
for(auto& ourdi : our_derivatives) {
error = (ourdi - *diter) * (ourdi - *diter);
diter++;
}
error = sqrt(error);
std::cout << "Error between the analytical and numerical gradients " << error << " with a step size of " << sqrt(norm_dh) << " in norm" << std::endl;
if(error > 1e-7)
++nbfails;
/*
std::cout << "numerical " << std::endl;
for(auto & di : derivatives)
std::cout << di << " ";
std::cout << std::endl;
std::cout << "our :" << std::endl;
for(auto& di : our_derivatives)
std::cout << di << " ";
std::cout << std::endl;
*/
}
std::cout << nbfails << " / " << nbtrials << " with an error higher than 1e-7" << std::endl;
}
void train_lsd2(char *cfgfile, char *weightfile, char *acfgfile, char *aweightfile, int clear)
{
#ifdef GPU
char *train_images = "/home/pjreddie/data/coco/trainvalno5k.txt";
char *backup_directory = "/home/pjreddie/backup/";
srand(time(0));
char *base = basecfg(cfgfile);
printf("%s\n", base);
network net = parse_network_cfg(cfgfile);
if(weightfile){
load_weights(&net, weightfile);
}
if(clear) *net.seen = 0;
char *abase = basecfg(acfgfile);
network anet = parse_network_cfg(acfgfile);
if(aweightfile){
load_weights(&anet, aweightfile);
}
if(clear) *anet.seen = 0;
int i, j, k;
layer imlayer = {};
for (i = 0; i < net.n; ++i) {
if (net.layers[i].out_c == 3) {
imlayer = net.layers[i];
break;
}
}
printf("Learning Rate: %g, Momentum: %g, Decay: %g\n", net.learning_rate, net.momentum, net.decay);
int imgs = net.batch*net.subdivisions;
i = *net.seen/imgs;
data train, buffer;
list *plist = get_paths(train_images);
//int N = plist->size;
char **paths = (char **)list_to_array(plist);
load_args args = {};
args.w = net.w;
args.h = net.h;
args.paths = paths;
args.n = imgs;
args.m = plist->size;
args.d = &buffer;
args.min = net.min_crop;
args.max = net.max_crop;
args.angle = net.angle;
args.aspect = net.aspect;
args.exposure = net.exposure;
args.saturation = net.saturation;
args.hue = net.hue;
args.size = net.w;
args.type = CLASSIFICATION_DATA;
args.classes = 1;
char *ls[1] = {"coco"};
args.labels = ls;
pthread_t load_thread = load_data_in_thread(args);
clock_t time;
network_state gstate = {};
gstate.index = 0;
gstate.net = net;
int x_size = get_network_input_size(net)*net.batch;
int y_size = 1*net.batch;
gstate.input = cuda_make_array(0, x_size);
gstate.truth = 0;
gstate.delta = 0;
gstate.train = 1;
float *X = (float*)calloc(x_size, sizeof(float));
float *y = (float*)calloc(y_size, sizeof(float));
network_state astate = {};
astate.index = 0;
astate.net = anet;
int ay_size = get_network_output_size(anet)*anet.batch;
astate.input = 0;
astate.truth = 0;
astate.delta = 0;
astate.train = 1;
float *imerror = cuda_make_array(0, imlayer.outputs);
float *ones_gpu = cuda_make_array(0, ay_size);
fill_ongpu(ay_size, 1, ones_gpu, 1);
float aloss_avg = -1;
float gloss_avg = -1;
//data generated = copy_data(train);
while (get_current_batch(net) < net.max_batches) {
i += 1;
time=clock();
pthread_join(load_thread, 0);
train = buffer;
load_thread = load_data_in_thread(args);
printf("Loaded: %lf seconds\n", sec(clock()-time));
data generated = copy_data(train);
time=clock();
float gloss = 0;
for(j = 0; j < net.subdivisions; ++j){
get_next_batch(train, net.batch, j*net.batch, X, y);
cuda_push_array(gstate.input, X, x_size);
*net.seen += net.batch;
forward_network_gpu(net, gstate);
fill_ongpu(imlayer.outputs, 0, imerror, 1);
astate.input = imlayer.output_gpu;
astate.delta = imerror;
astate.truth = ones_gpu;
forward_network_gpu(anet, astate);
backward_network_gpu(anet, astate);
scal_ongpu(imlayer.outputs, 1, imerror, 1);
axpy_ongpu(imlayer.outputs, 1, imerror, 1, imlayer.delta_gpu, 1);
backward_network_gpu(net, gstate);
printf("features %f\n", cuda_mag_array(imlayer.delta_gpu, imlayer.outputs));
printf("realness %f\n", cuda_mag_array(imerror, imlayer.outputs));
gloss += get_network_cost(net) /(net.subdivisions*net.batch);
cuda_pull_array(imlayer.output_gpu, imlayer.output, x_size);
for(k = 0; k < net.batch; ++k){
int index = j*net.batch + k;
copy_cpu(imlayer.outputs, imlayer.output + k*imlayer.outputs, 1, generated.X.vals[index], 1);
generated.y.vals[index][0] = 0;
}
}
harmless_update_network_gpu(anet);
data merge = concat_data(train, generated);
randomize_data(merge);
float aloss = train_network(anet, merge);
update_network_gpu(net);
update_network_gpu(anet);
free_data(merge);
free_data(train);
free_data(generated);
if (aloss_avg < 0) aloss_avg = aloss;
aloss_avg = aloss_avg*.9 + aloss*.1;
gloss_avg = gloss_avg*.9 + gloss*.1;
printf("%d: gen: %f, adv: %f | gen_avg: %f, adv_avg: %f, %f rate, %lf seconds, %d images\n", i, gloss, aloss, gloss_avg, aloss_avg, get_current_rate(net), sec(clock()-time), i*imgs);
if(i%1000==0){
char buff[256];
sprintf(buff, "%s/%s_%d.weights", backup_directory, base, i);
save_weights(net, buff);
sprintf(buff, "%s/%s_%d.weights", backup_directory, abase, i);
save_weights(anet, buff);
}
if(i%100==0){
char buff[256];
sprintf(buff, "%s/%s.backup", backup_directory, base);
save_weights(net, buff);
sprintf(buff, "%s/%s.backup", backup_directory, abase);
save_weights(anet, buff);
}
}
char buff[256];
sprintf(buff, "%s/%s_final.weights", backup_directory, base);
save_weights(net, buff);
#endif
}
void train_lsd3(char *fcfg, char *fweight, char *gcfg, char *gweight, char *acfg, char *aweight, int clear)
{
#ifdef GPU
//char *train_images = "/home/pjreddie/data/coco/trainvalno5k.txt";
char *train_images = "/home/pjreddie/data/imagenet/imagenet1k.train.list";
//char *style_images = "/home/pjreddie/data/coco/trainvalno5k.txt";
char *style_images = "/home/pjreddie/zelda.txt";
char *backup_directory = "/home/pjreddie/backup/";
srand(time(0));
network fnet = load_network(fcfg, fweight, clear);
network gnet = load_network(gcfg, gweight, clear);
network anet = load_network(acfg, aweight, clear);
char *gbase = basecfg(gcfg);
char *abase = basecfg(acfg);
printf("Learning Rate: %g, Momentum: %g, Decay: %g\n", gnet.learning_rate, gnet.momentum, gnet.decay);
int imgs = gnet.batch*gnet.subdivisions;
int i = *gnet.seen/imgs;
data train, tbuffer;
data style, sbuffer;
list *slist = get_paths(style_images);
char **spaths = (char **)list_to_array(slist);
list *tlist = get_paths(train_images);
char **tpaths = (char **)list_to_array(tlist);
load_args targs= get_base_args(gnet);
targs.paths = tpaths;
targs.n = imgs;
targs.m = tlist->size;
targs.d = &tbuffer;
targs.type = CLASSIFICATION_DATA;
targs.classes = 1;
char *ls[1] = {"zelda"};
targs.labels = ls;
load_args sargs = get_base_args(gnet);
sargs.paths = spaths;
sargs.n = imgs;
sargs.m = slist->size;
sargs.d = &sbuffer;
sargs.type = CLASSIFICATION_DATA;
sargs.classes = 1;
sargs.labels = ls;
pthread_t tload_thread = load_data_in_thread(targs);
pthread_t sload_thread = load_data_in_thread(sargs);
clock_t time;
float aloss_avg = -1;
float floss_avg = -1;
network_state fstate = {};
fstate.index = 0;
fstate.net = fnet;
int x_size = get_network_input_size(fnet)*fnet.batch;
int y_size = get_network_output_size(fnet)*fnet.batch;
fstate.input = cuda_make_array(0, x_size);
fstate.truth = cuda_make_array(0, y_size);
fstate.delta = cuda_make_array(0, x_size);
fstate.train = 1;
float *X = (float*)calloc(x_size, sizeof(float));
float *y = (float*)calloc(y_size, sizeof(float));
float *ones = cuda_make_array(0, anet.batch);
float *zeros = cuda_make_array(0, anet.batch);
fill_ongpu(anet.batch, .99, ones, 1);
fill_ongpu(anet.batch, .01, zeros, 1);
network_state astate = {};
astate.index = 0;
astate.net = anet;
int ax_size = get_network_input_size(anet)*anet.batch;
int ay_size = get_network_output_size(anet)*anet.batch;
astate.input = 0;
astate.truth = ones;
astate.delta = cuda_make_array(0, ax_size);
astate.train = 1;
network_state gstate = {};
gstate.index = 0;
gstate.net = gnet;
int gx_size = get_network_input_size(gnet)*gnet.batch;
int gy_size = get_network_output_size(gnet)*gnet.batch;
gstate.input = cuda_make_array(0, gx_size);
gstate.truth = 0;
gstate.delta = 0;
gstate.train = 1;
while (get_current_batch(gnet) < gnet.max_batches) {
i += 1;
time=clock();
pthread_join(tload_thread, 0);
pthread_join(sload_thread, 0);
train = tbuffer;
style = sbuffer;
tload_thread = load_data_in_thread(targs);
sload_thread = load_data_in_thread(sargs);
printf("Loaded: %lf seconds\n", sec(clock()-time));
data generated = copy_data(train);
time=clock();
int j, k;
float floss = 0;
for(j = 0; j < fnet.subdivisions; ++j){
layer imlayer = gnet.layers[gnet.n - 1];
get_next_batch(train, fnet.batch, j*fnet.batch, X, y);
cuda_push_array(fstate.input, X, x_size);
cuda_push_array(gstate.input, X, gx_size);
*gnet.seen += gnet.batch;
forward_network_gpu(fnet, fstate);
float *feats = fnet.layers[fnet.n - 2].output_gpu;
copy_ongpu(y_size, feats, 1, fstate.truth, 1);
forward_network_gpu(gnet, gstate);
float *gen = gnet.layers[gnet.n-1].output_gpu;
copy_ongpu(x_size, gen, 1, fstate.input, 1);
fill_ongpu(x_size, 0, fstate.delta, 1);
forward_network_gpu(fnet, fstate);
backward_network_gpu(fnet, fstate);
//HERE
astate.input = gen;
fill_ongpu(ax_size, 0, astate.delta, 1);
forward_network_gpu(anet, astate);
backward_network_gpu(anet, astate);
float *delta = imlayer.delta_gpu;
fill_ongpu(x_size, 0, delta, 1);
scal_ongpu(x_size, 100, astate.delta, 1);
scal_ongpu(x_size, .00001, fstate.delta, 1);
axpy_ongpu(x_size, 1, fstate.delta, 1, delta, 1);
axpy_ongpu(x_size, 1, astate.delta, 1, delta, 1);
//fill_ongpu(x_size, 0, delta, 1);
//cuda_push_array(delta, X, x_size);
//axpy_ongpu(x_size, -1, imlayer.output_gpu, 1, delta, 1);
//printf("pix error: %f\n", cuda_mag_array(delta, x_size));
printf("fea error: %f\n", cuda_mag_array(fstate.delta, x_size));
printf("adv error: %f\n", cuda_mag_array(astate.delta, x_size));
//axpy_ongpu(x_size, 1, astate.delta, 1, delta, 1);
backward_network_gpu(gnet, gstate);
floss += get_network_cost(fnet) /(fnet.subdivisions*fnet.batch);
cuda_pull_array(imlayer.output_gpu, imlayer.output, x_size);
for(k = 0; k < gnet.batch; ++k){
int index = j*gnet.batch + k;
copy_cpu(imlayer.outputs, imlayer.output + k*imlayer.outputs, 1, generated.X.vals[index], 1);
generated.y.vals[index][0] = .01;
}
}
/*
image sim = float_to_image(anet.w, anet.h, anet.c, style.X.vals[j]);
show_image(sim, "style");
cvWaitKey(0);
*/
harmless_update_network_gpu(anet);
data merge = concat_data(style, generated);
randomize_data(merge);
float aloss = train_network(anet, merge);
update_network_gpu(gnet);
free_data(merge);
free_data(train);
free_data(generated);
free_data(style);
if (aloss_avg < 0) aloss_avg = aloss;
if (floss_avg < 0) floss_avg = floss;
aloss_avg = aloss_avg*.9 + aloss*.1;
floss_avg = floss_avg*.9 + floss*.1;
printf("%d: gen: %f, adv: %f | gen_avg: %f, adv_avg: %f, %f rate, %lf seconds, %d images\n", i, floss, aloss, floss_avg, aloss_avg, get_current_rate(gnet), sec(clock()-time), i*imgs);
if(i%1000==0){
char buff[256];
sprintf(buff, "%s/%s_%d.weights", backup_directory, gbase, i);
save_weights(gnet, buff);
sprintf(buff, "%s/%s_%d.weights", backup_directory, abase, i);
save_weights(anet, buff);
}
if(i%100==0){
char buff[256];
sprintf(buff, "%s/%s.backup", backup_directory, gbase);
save_weights(gnet, buff);
sprintf(buff, "%s/%s.backup", backup_directory, abase);
save_weights(anet, buff);
}
}
#endif
}
