parser.add_argument('--k', type=int, default=500, help="Number mixture components in MoG prior") parser.add_argument('--iter_max', type=int, default=20000, help="Number of training iterations") parser.add_argument('--iter_save', type=int, default=10000, help="Save model every n iterations") parser.add_argument('--run', type=int, default=0, help="Run ID. In case you want to run replicates") args = parser.parse_args() layout = [ ('model={:s}', 'gmvae'), ('z={:02d}', args.z), ('k={:03d}', args.k), ('run={:04d}', args.run) ] model_name = '_'.join([t.format(v) for (t, v) in layout]) pprint(vars(args)) print('Model name:', model_name) device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') train_loader, labeled_subset, _ = ut.get_mnist_data(device, use_test_subset=True) gmvae = GMVAE(z_dim=args.z, k=args.k, name=model_name).to(device) ut.load_model_by_name(gmvae, global_step=args.iter_max) ut.evaluate_lower_bound(gmvae, labeled_subset, run_iwae=False) samples = torch.reshape(gmvae.sample_x(200), (10, 20, 28, 28)) f, axarr = plt.subplots(10,20) for i in range(samples.shape[0]): for j in range(samples.shape[1]): axarr[i,j].imshow(samples[i,j].detach().numpy()) axarr[i,j].axis('off') plt.show()
('model={:s}', 'vae'), ('z={:02d}', args.z), ('run={:04d}', args.run) ] model_name = '_'.join([t.format(v) for (t, v) in layout]) pprint(vars(args)) print('Model name:', model_name) device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') train_loader, labeled_subset, _ = ut.get_mnist_data(device, use_test_subset=True) vae = VAE(z_dim=args.z, name=model_name).to(device) if args.train: writer = ut.prepare_writer(model_name, overwrite_existing=True) train(model=vae, train_loader=train_loader, labeled_subset=labeled_subset, device=device, tqdm=tqdm.tqdm, writer=writer, iter_max=args.iter_max, iter_save=args.iter_save) ut.evaluate_lower_bound(vae, labeled_subset, run_iwae=args.train == 2) x = vae.sample_x(100).view(100, 1, 28, 28) db.printTensor(x) ImUtil.showBatch(x, show=True) input('Press key to exit') else: ut.load_model_by_name(vae, global_step=args.iter_max) ut.evaluate_lower_bound(vae, labeled_subset, run_iwae=True)
parser.add_argument('--train', type=int, default=1, help="Flag for training") args = parser.parse_args() layout = [('model={:s}', 'gvae'), ('x-dim={:03d}', args.x_dim), ('z-dim={:02d}', args.z_dim), ('z-num={:02d}', args.z_num), ('run={:04d}', args.run)] model_name = '_'.join([t.format(v) for (t, v) in layout]) pprint(vars(args)) print('Model name:', model_name) device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') train_loader, labeled_subset, _ = ut.get_mnist_data(device, use_test_subset=True) gvae = GVAE(x_dim=args.x_dim, z_dim=args.z_dim, z_num=args.z_num, name=model_name).to(device) if args.train: writer = ut.prepare_writer(model_name, overwrite_existing=True) train(model=gvae, train_loader=train_loader, labeled_subset=labeled_subset, device=device, tqdm=tqdm.tqdm, writer=writer, iter_max=args.iter_max, iter_save=args.iter_save) ut.evaluate_lower_bound(gvae, labeled_subset) else: ut.load_model_by_name(gvae, global_step=args.iter_max) ut.evaluate_lower_bound(gvae, labeled_subset)
z_dim=args.z, k=args.k, name=model_name).to(device) if args.train: writer = ut.prepare_writer(model_name, overwrite_existing=True) train(model=gmvae, train_loader=train_loader, labeled_subset=None, device=device, tqdm=tqdm.tqdm, writer=writer, iter_max=args.iter_max, iter_save=args.iter_save) ut.evaluate_lower_bound(gmvae, None, run_iwae=args.train == 2, ox=torch.tensor(testgen)) evaluate(model=gmvae, test_loader=test_loader, labeled_subset=None, device=device, tqdm=tqdm.tqdm, iter_max=len(testgen), iter_save=args.iter_save, samples=testsamples, meta=meta, group_by=color_by, group_on=group_on) else: #writer = ut.prepare_writer(model_name, overwrite_existing=True)
device=device, tqdm=tqdm.tqdm, writer=writer, lr=args.lr, lr_gamma=args.lr_gamma, lr_milestones=lr_milestones, iw=args.iw, iter_max=args.iter_max, iter_save=args.iter_save) model.set_to_eval() val_set = ut.get_load_data(device, split='val', in_memory=True, log_normal=True, shift_scale=shift_scale) ut.evaluate_lower_bound(model, val_set, run_iwae=(args.iw > 1)) else: ut.load_model_by_name(model, global_step=args.iter_max) if args.mode in ['val', 'test']: model.set_to_eval() val_set = ut.get_load_data(device, split=args.mode, in_memory=True, log_normal=True, shift_scale=shift_scale) ut.evaluate_lower_bound(model, val_set, run_iwae=(args.iw > 1)) make_image_load(model, log_normal=True, shift_scale=shift_scale["use"]) make_image_load_z(model, log_normal=True, shift_scale=shift_scale["use"])