from configuration.train_configuration import TrainConfiguration from data_loading.sources.mnist_source import MNISTSource from data_loading.loaders.autoencoder_loader import AutoEncoderLoader from data_loading.loaders.multi_loader import MultiLoader import torch.optim as optim import torch.nn as nn from networks.autoencoder import AutoEncoder import random def get_loader(): source = MNISTSource('/home/ray/Data/MNIST') loader_params = {'crop_size': [28, 28]} return AutoEncoderLoader(source, loader_params) loader = (get_loader, dict()) # loader = (MultiLoader,dict(loader=get_loader,loader_args=dict(),num_procs=16)) model = (AutoEncoder, dict(inchans=1)) optimizer = (optim.Adam, dict(lr=1e-3)) loss = nn.BCELoss() train_config = TrainConfiguration(loader, optimizer, model, loss, False)
import random def get_loader(mode): path = '/home/ray/Data/KITTI/testing' if mode == 'test' else '/home/ray/Data/KITTI/training' source = KITTISource(path, max_frames=10000) return TripletDetectionLoader(source, crop_size=[255, 255], anchor_size=[127, 127], obj_types=['Car'], mode=mode) loader_test = (MultiLoader, dict(loader=get_loader, loader_args=dict(mode='test'), num_procs=1)) loader_train = (MultiLoader, dict(loader=get_loader, loader_args=dict(mode='train'), num_procs=1)) model = (TripletCorrelationalDetector, dict()) optimizer = (optim.Adam, dict(lr=1e-4)) loss = triplet_correlation_loss2 train_config = TrainConfiguration(loader_train, optimizer, model, loss, cuda=True) test_config = TestConfiguration(loader_test, model, loss, cuda=True)