def create_model(num_classes, ema=False):
model = MobileNet(num_classes)
#model = WideResNet(num_classes)
model.cuda()
if ema:
for param in model.parameters():
param.detach_()
return model
# import model
from dnn121 import DenseNet121, MobileNet
import wisenet as models
# import dataset
from mixmatch_dataset import train_val_split, NIH_CXR_BASE, CxrDataset, CXR_unlabeled
from utils import Bar, Logger, AverageMeter, accuracy, mkdir_p, savefig
from tensorboardX import SummaryWriter
test_set = CxrDataset(NIH_CXR_BASE, "~/cxr-jingyi/Age/NIH_test_2500.csv")
test_loader = data.DataLoader(test_set, batch_size=32, shuffle=False, num_workers=32)
model = MobileNet(16)
model = model.cuda()
#checkpoint = torch.load('/home/jingyi/cxr-jingyi/Age/result/supervised/model_best.pth.tar')
checkpoint = torch.load('/home/jingyi/cxr-jingyi/Age/checkpoint/cifar10-semi/exp/ckpt.pth.tar')
#model.load_state_dict(checkpoint['state_dict'])
model.load_state_dict(checkpoint['net'])
def validate(val_loader, model, mode = 'valid'):
top1 = AverageMeter()
top5 = AverageMeter()
predict = []
# switch to evaluate mode
model.eval()
with torch.no_grad():