def train(self):
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
self.tools.logging("The model is set.")
self.tools.logging("The {} device is selected.".format(device))
self.model = learning.AliNet()
self.model = self.model.to(device)
self.dataloaders = self.dataloaders.to(device)
summary(self.model, input_size=(3, 25, 25))
self.optimizer_ft = optim.Adam(self.model.parameters(), lr=1e-4)
self.exp_lr_scheduler = lr_scheduler.StepLR(self.optimizer_ft,
step_size=25,
gamma=0.1)
self.model = learning.train_model(self.model,
self.dataloaders,
self.optimizer_ft,
self.exp_lr_scheduler,
num_epochs=10)
def fake_outfits():
return {
# '234': ['1', '2', '3'],
# '235': ['1', '2', '4'],
# '236': ['1', '5', '6'],
# '334': ['7', '2', '3'],
# '335': ['7', '2', '4'],
# '336': ['7', '5', '6'],
'431': ['1', '2', '6'],
'432': ['8', '5', '3']
}
if __name__ == "__main__":
if False:
df = generate_matrix(fake_ratings())
model = train_model(df)
print(make_item_prediction('123', '7', model))
print(make_item_prediction('126', '1', model))
make_outfit_prediction('123', fake_outfits(), model)
make_outfit_prediction('126', fake_outfits(), model)
if False:
response = requests.get("localhost:4000/items?limit=10")
print(response.json())
if False:
c = pymongo.MongoClient()
db = c.cloutfits
items = db.items
for i in items.find():
print(i['id'])
suffix = '/train'
model = models.vgg16(pretrained=False)
# modify the last Fc layer into size 2 for positive sample and negative sample
model.classifier[6] = nn.Linear(
in_features=4096, out_features=2,
bias=True) #### what does this "6" of classifier[6] ??????
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(),
lr=0.01,
weight_decay=0.0001)
for epoch in range(100):
# print('\nepoch %d:' % epoch)
# train the model
train_model(model, tr_data_loader, criterion, optimizer, gpu_id=1)
if epoch % 1 == 0:
print('\nepoch %d:' % epoch)
# validate the trained model for every 1 epoch
val = validate_model(model,
tr_noaug_data_loader,
criterion,
epoch,
prediction_folder='../results/train_data/',
gpu_id=1)
cvs_path = 'ep_' + str(epoch) + '_predictions.csv'
arr = cvs_to_arr(cvs_file='../results/train_data/' + cvs_path)
metrics = get_metrics(arr)
print('train_accuracy', metrics[0])
# validate the test model for every 1 epoch
os.makedirs(model_dir+'/test_data/')
os.makedirs(model_dir+'/val_data/')
os.makedirs(model_dir+'/saved_models/')
suffix = '/train'
model = models.vgg16(pretrained=True)
# modify the last Fc layer into size 2 for positive sample and negative sample
model.classifier[6] = nn.Linear(in_features=4096, out_features=2,
bias=True) #### what does this "6" of classifier[6] ??????
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
for epoch in range(100):
# print('\nepoch %d:' % epoch)
# train the model
train_model(model, tr_data_loader, criterion, optimizer, gpu_id=DEVICE_ID)
if epoch % 1 == 0:
print('\nepoch %d:' % epoch)
# validate the trained model for every 1 epoch
val = validate_model(model, tr_noaug_data_loader, criterion, epoch,
prediction_folder = model_dir+'/train_data/', gpu_id=DEVICE_ID)
cvs_path = 'ep_' + str(epoch) + '_predictions.csv'
arr = cvs_to_arr(cvs_file = model_dir+'/train_data/' + cvs_path)
metrics = get_metrics(arr)
print('train_accuracy', metrics[0])
# validate the test model for every 1 epoch
val = validate_model(model, ts_noaug_data_loader, criterion, epoch,
prediction_folder = model_dir + '/test_data/', gpu_id=DEVICE_ID)
cvs_path = 'ep_' + str(epoch) + '_predictions.csv'
arr = cvs_to_arr(cvs_file = model_dir + '/test_data/' + cvs_path)
num_workers=1)
test_im = 'data/test/'
test_img_database = datasets.ImageFolder(root=test_im, transform=transform)
# batch_size, num_workers = 1 always and shuffle=False. Bcz we are loading images individually from the folder
test_dataset_loader = torch.utils.data.DataLoader(test_img_database,
batch_size=1,
shuffle=False,
num_workers=1)
# Assign classes
classes = (0, 1, 2, 3, 4, 5, 6, 7, 8, 9)
if __name__ == '__main__':
print(net)
train_model(train_dataset_loader, num_iteration, learn_rate, mom)
correct = 0
total = 0
for j, data in enumerate(test_dataset_loader, 0):
test_x = (data)[0] # test image
test_y = (data)[1]
outputs = net(test_x)
_, predicted = torch.max(outputs.data, 1)
total += test_y.size(0)
correct += (predicted == test_y).sum().item()
print('Accuracy of the network on the 10 * 80 test images: %d %%' %
(100 * correct / total))
class_correct = list(0. for i in range(10))