def test_autoencode_vgg16_celeba():
args = config.config([
'--config', '../configs/autoencoder/vgg16/celeba.yaml', '--epochs',
'5', '--dataroot', '../data', '--dataset_test_len', '256',
'--dataset_train_len', '256'
])
best_loss = train_autoencoder.main(args)
assert best_loss < 0.04
def test_autoencode_fc_mnist():
args = config.config([
'--config', '../configs/autoencoder/fc/mnist.yaml', '--epochs', '80',
'--dataroot', '../data', '--dataset_test_len', '256',
'--dataset_train_len', '256'
])
best_loss = train_autoencoder.main(args)
assert best_loss < 1.3
def test_autoencode_resnet_cifar10():
args = config.config([
'--config', '../configs/autoencoder/resnet/cifar10.yaml', '--epochs',
'20', '--optim_lr', '1e-5', '--dataroot', '../data',
'--dataset_test_len', '256', '--dataset_train_len', '256', '--seed',
'0'
])
best_loss = train_autoencoder.main(args)
assert best_loss < 1.2
def test_mnist_fc():
args = config.config(['--config', '../configs/classify/fc/mnist.yaml',
'--epochs', '3',
'--dataroot', '../data',
'--dataset_test_len', '256',
'--dataset_train_len', '256',
'--seed', '0',
'--run_id', '3'
])
ave_precision, best_precision, train_accuracy, test_accuracy = train_classifier.main(args)
assert ave_precision > 0.2
assert best_precision > 0.2
def test_cifar10_vgg16():
args = config.config(['--config', '../configs/classify/vgg16/cifar10.yaml',
'--optim_lr', '0.05',
'--epochs', '6',
'--dataroot', '../data',
'--dataset_test_len', '256',
'--dataset_train_len', '256',
'--seed', '0',
'--run_id', '1'
])
ave_precision, best_precision, train_accuracy, test_accuracy = train_classifier.main(args)
assert ave_precision > 0.2
assert best_precision > 0.2
def test_cifar10_resnet():
args = config.config(['--config', '../configs/classify/resnet/cifar10-batchnorm.yaml',
'--epochs', '80',
'--optim_lr', '0.01',
'--dataroot', '../data',
'--dataset_test_len', '256',
'--dataset_train_len', '256',
'--seed', '0',
'--run_id', '4'
])
ave_precision, best_precision, train_accuracy, test_accuracy = train_classifier.main(args)
""" WARNING this model does not run reliably due to the shortcut containing convnets"""
assert best_precision > 0.13
assert train_accuracy > 20.0
'optimizer': optim.state_dict(),
'amp': amp.state_dict()
}
torch.save(best, run_dir + '/best_amp.pt')
for event in pygame.event.get():
if event.type == pygame.QUIT:
print('PYGAME QUIT')
pygame.quit()
elif event.type == pygame.KEYDOWN:
if event.key == pygame.K_c and pygame.key.get_mods(
) & pygame.KMOD_CTRL:
print("pressed CTRL-C as an event")
pygame.quit()
return ave_precision, best_precision, train_accuracy, test_accuracy
if __name__ == '__main__':
""" configuration """
args = config.config()
pygame.init()
wandb.init(project='bald-classification', name=args.name)
wandb.config.update(args)
Digo.init(api_key='qpqD6cHYvGG2XNHeGlHog8360T8uWHya',
project_name='HumanSegmantation',
workspace_name='Yunsang')
torch.cuda.set_device(args.device)
main(args)