def main():
rootdir = "/mnt/xxx/dataset/office_caltech_10/"
tr.manual_seed(1)
domain_str = ['webcam', 'dslr']
X_s = data_loader.load_train(root_dir=rootdir,
domain=domain_str[0],
batch_size=BATCH_SIZE[0])
X_t = data_loader.load_test(root_dir=rootdir,
domain=domain_str[1],
batch_size=BATCH_SIZE[1])
# train and test
start = time.time()
mmd_type = ['mmd', 'jmmd', 'jpmmd', 'djpmmd']
for mt in mmd_type:
print('-' * 10 + domain_str[0] + ' --> ' + domain_str[1] + '-' * 10)
print('MMD loss type: ' + mt + '\n')
acc, loss = {}, {}
train_acc = []
test_acc = []
train_loss = []
test_loss = []
y_pse = tr.zeros(14, 64).long().cuda()
mdl = DaNN.DaNN(n_input=28 * 28, n_hidden=256, n_class=10)
mdl = mdl.to(DEVICE)
# optimization
opt_Adam = optim.Adam(mdl.parameters(), lr=LEARNING_RATE)
for ep in tqdm(range(1, N_EPOCH + 1)):
tmp_train_acc, tmp_train_loss, mdl = \
model_train(model=mdl, optimizer=opt_Adam, epoch=ep, data_src=X_s, data_tar=X_t, y_pse=y_pse,
mmd_type=mt)
tmp_test_acc, tmp_test_loss = model_test(mdl, X_t, ep)
train_acc.append(tmp_train_acc)
test_acc.append(tmp_test_acc)
train_loss.append(tmp_train_loss)
test_loss.append(tmp_test_loss)
acc['train'], acc['test'] = train_acc, test_acc
loss['train'], loss['test'] = train_loss, test_loss
# visualize
plt.plot(acc['train'], label='train-' + mt)
plt.plot(acc['test'], label='test-' + mt, ls='--')
plt.title(domain_str[0] + ' to ' + domain_str[1])
plt.xticks(np.linspace(1, N_EPOCH, num=5, dtype=np.int8))
plt.xlim(1, N_EPOCH)
plt.ylim(0, 100)
plt.legend(loc='upper right')
plt.xlabel("epochs")
plt.ylabel("accuracy")
plt.savefig(domain_str[0] + '_' + domain_str[1] + "_acc.jpg")
plt.close()
# time and save model
end = time.time()
print("Total run time: %.2f" % float(end - start))
accuracy = correct * 100. / len(data_tar.dataset)
res = 'Test: total loss: {:.6f}, correct: [{}/{}], testing accuracy: {:.4f}%'.format(
total_loss_test, correct, len(data_tar.dataset), accuracy)
tqdm.write(res)
RESULT_TEST.append([e, total_loss_test, accuracy])
log_test.write(res + '\n')
if __name__ == '__main__':
rootdir = '../../../data/office_caltech_10/'
torch.manual_seed(1)
data_src = data_loader.load_data(root_dir=rootdir,
domain='amazon',
batch_size=BATCH_SIZE[0])
data_tar = data_loader.load_test(root_dir=rootdir,
domain='webcam',
batch_size=BATCH_SIZE[1])
model = DaNN.DaNN(n_input=28 * 28, n_hidden=256, n_class=10)
model = model.to(DEVICE)
optimizer = optim.SGD(model.parameters(),
lr=LEARNING_RATE,
momentum=MOMEMTUN,
weight_decay=L2_WEIGHT)
for e in tqdm(range(1, N_EPOCH + 1)):
model = train(model=model,
optimizer=optimizer,
epoch=e,
data_src=data_src,
data_tar=data_tar)
test(model, data_tar, e)
torch.save(model, 'model_dann.pkl')
2.样本种类一致
另外，每类样本数据个数可以不一样 ，也就是两个数据集的样本个数可以不一样
'''
# src_dir = './source_datasets.mat'
# tar_dir = './target_datasets.mat'
src_dir = './source_datasets_Eq_shuffle.mat'
tar_dir = './target_datasets_Eq_shuffle.mat'
torch.manual_seed(1)
data_src = data_loader.load_data(root_dir=src_dir,
domain='source_data_train',
batch_size=BATCH_SIZE[0])
# root_dir=src_dir, domain='source_data', batch_size=BATCH_SIZE[0])
# root_dir=src_dir, domain='source_data_device', batch_size=BATCH_SIZE[0])
data_tar = data_loader.load_test(root_dir=tar_dir,
domain='target_data_train',
batch_size=BATCH_SIZE[1])
# root_dir=tar_dir, domain='target_data', batch_size=BATCH_SIZE[1])
# root_dir=tar_dir, domain='target_data_device', batch_size=BATCH_SIZE[1])
# print(data_src.size())
model = CNN1d.CNN1d(n_hidden=100, n_class=Classes)
# 打印输出模型结构
print(model)
model = model.to(DEVICE)
# 定义优化器
optimizer = optim.Adamax(
model.parameters(),
lr=LEARNING_RATE,
# momentum=MOMEMTUN,
# f = gzip.open(filename, 'rb')
# training_data, validation_data, test_data = cPickle.load(f)
# f.close()
from __future__ import print_function
import data_loader
tmp_data = data_loader.load_test()
training_data = tmp_data
print(training_data[0][0])
print(training_data[1][1])
#### Libraries
# Standard library
import cPickle
import gzip
import os.path
import random
# Third-party libraries
import numpy as np
print("Expanding the MNIST training set")
if False:
print("The expanded training set already exists. Exiting.")
else:
# f = gzip.open("../data/mnist.pkl.gz", 'rb')
# training_data, validation_data, test_data = cPickle.load(f)
# f.close()
# f = gzip.open(filename, 'rb')
# training_data, validation_data, test_data = cPickle.load(f)
# f.close()
from __future__ import print_function
import data_loader
tmp_data = data_loader.load_test()
training_data = tmp_data
print (training_data[0][0])
print (training_data[1][1])
#### Libraries
# Standard library
import cPickle
import gzip
import os.path
import random
# Third-party libraries
import numpy as np
print("Expanding the MNIST training set")
if False:
print("The expanded training set already exists. Exiting.")
else:
# f = gzip.open("../data/mnist.pkl.gz", 'rb')
# training_data, validation_data, test_data = cPickle.load(f)