| | |
| | | printf("Learning Rate: %g, Momentum: %g, Decay: %g\n", net.learning_rate, net.momentum, net.decay); |
| | | //net.seen=0; |
| | | int imgs = 1024; |
| | | int i = net.seen/imgs; |
| | | char **labels = get_labels("data/inet.labels.list"); |
| | | list *plist = get_paths("/data/imagenet/cls.train.list"); |
| | | char **paths = (char **)list_to_array(plist); |
| | | printf("%d\n", plist->size); |
| | | int N = plist->size; |
| | | clock_t time; |
| | | pthread_t load_thread; |
| | | data train; |
| | |
| | | args.paths = paths; |
| | | args.classes = 1000; |
| | | args.n = imgs; |
| | | args.m = plist->size; |
| | | args.m = N; |
| | | args.labels = labels; |
| | | args.d = &buffer; |
| | | args.type = CLASSIFICATION_DATA; |
| | | |
| | | load_thread = load_data_in_thread(args); |
| | | int epoch = net.seen/N; |
| | | while(1){ |
| | | ++i; |
| | | time=clock(); |
| | | pthread_join(load_thread, 0); |
| | | train = buffer; |
| | |
| | | net.seen += imgs; |
| | | if(avg_loss == -1) avg_loss = loss; |
| | | avg_loss = avg_loss*.9 + loss*.1; |
| | | printf("%d: %f, %f avg, %lf seconds, %d images\n", i, loss, avg_loss, sec(clock()-time), net.seen); |
| | | printf("%.3f: %f, %f avg, %lf seconds, %d images\n", (float)net.seen/N, loss, avg_loss, sec(clock()-time), net.seen); |
| | | free_data(train); |
| | | if((i % 30000) == 0) net.learning_rate *= .1; |
| | | if(i%1000==0){ |
| | | if(net.seen/N > epoch){ |
| | | epoch = net.seen/N; |
| | | char buff[256]; |
| | | sprintf(buff, "%s/%s_%d.weights",backup_directory,base, i); |
| | | sprintf(buff, "%s/%s_%d.weights",backup_directory,base, epoch); |
| | | save_weights(net, buff); |
| | | if(epoch%22 == 0) net.learning_rate *= .1; |
| | | } |
| | | } |
| | | pthread_join(load_thread, 0); |
| | | free_data(buffer); |
| | | free_network(net); |
| | | free_ptrs((void**)labels, 1000); |
| | | free_ptrs((void**)paths, plist->size); |
| | | free_list(plist); |
| | | free(base); |
| | | } |
| | | |
| | | void validate_imagenet(char *filename, char *weightfile) |