def main():
from train_infogan import parse_args
opt = parse_args()
train_set, test_set = get_merged_common_dataset(opt)
size = train_set.features.shape[1]
discriminator = Discriminator(opt, size, train_set.binary)
discriminator.load_state_dict(torch.load(opt.disc_path))
discriminator.eval()
xgboost_test(discriminator.extract_features, opt)
with open(args.config, 'r') as f:
y = yaml.load(f, Loader=yaml.SafeLoader)
opt = argparse.Namespace(**y)
# Loss functions
adversarial_loss = torch.nn.MSELoss()
categorical_loss = torch.nn.CrossEntropyLoss()
continuous_loss = torch.nn.MSELoss()
# Loss weights
lambda_cat = 1
lambda_con = 0.1
# Initialize generator and discriminator
generator = Generator(opt)
discriminator = Discriminator(opt)
if cuda:
generator.cuda()
discriminator.cuda()
adversarial_loss.cuda()
categorical_loss.cuda()
continuous_loss.cuda()
# Initialize weights
generator.apply(weights_init_normal)
discriminator.apply(weights_init_normal)
# Configure data loader
data = CuratedBreastCancerData(opt.batch_size, test_split=0.0) # train this unsupervised model on all data
categorical_size = train_config.n_classes if train_config.n_classes else train_dataset.n_studies
load_checkpoints = 's{0:02d}_c{1:02d}_{2}'.format(
train_config.code_dim, categorical_size,
os.path.split(train_config.gexs_csv)[-1].split('.')[0])
load_path = os.path.join('checkpoints', load_checkpoints)
# No need to load generator
# generator = Generator(
# train_config.latent_dim,
# categorical_size,
# train_config.code_dim,
# train_dataset.n_genes
# )
# generator_checkpoint = os.path.join(load_path, 'generator.pth')
# generator.load_state_dict(torch.load(generator_checkpoint))
discriminator = Discriminator(categorical_size, train_config.code_dim,
train_dataset.n_genes)
discriminator_checkpoint = os.path.join(load_path, 'discriminator.pth')
discriminator.load_state_dict(torch.load(discriminator_checkpoint))
if cuda:
#generator = generator.cuda().eval()
discriminator = discriminator.cuda().eval()
print('CUDA is here')
else:
print('CPU using')
codes, hidden, clusters, tags, valids = [], [], [], [], []
for gex in tqdm(test_dataset.traverse_gexs()):
with torch.no_grad():
batch_gex = FloatTensor(gex)
valid, cluster, pred_code = discriminator(batch_gex)
import matplotlib.pyplot as plt
#from mpl_toolkits.axes_grid1 import ImageGrid
#from sklearn.decomposition import PCA
#from sklearn.linear_model import LogisticRegression
#from scipy import stats
# load config
with open('config1.yml', 'r') as f:
y = yaml.load(f, Loader=yaml.SafeLoader)
opt = argparse.Namespace(**y)
data = CuratedBreastCancerData(opt.batch_size, test_split=0.0)
generator = Generator(opt)
generator.load_state_dict(torch.load('generator.pth'))
discriminator = Discriminator(opt)
discriminator.load_state_dict(torch.load('discriminator.pth'))
if cuda:
generator = generator.cuda().eval()
discriminator = discriminator.cuda().eval()
print('CUDA is here')
else:
print('CPU using')
codes, hidden, clusters, tags, valids = [], [], [], [], []
for gex in data.traverse_gex_study():
with torch.no_grad():
batch_gex = FloatTensor(gex)
valid, cluster, pred_code = discriminator(batch_gex)
hi_z = discriminator.hidden(batch_gex)
with open(args.config, 'r') as f:
y = yaml.load(f, Loader=yaml.SafeLoader)
opt = argparse.Namespace(**y)
# Loss functions
adversarial_loss = torch.nn.MSELoss()
categorical_loss = torch.nn.CrossEntropyLoss()
continuous_loss = torch.nn.MSELoss()
# Loss weights
lambda_cat = 1
lambda_con = 0.1
# Initialize generator and discriminator
generator = Generator(opt)
discriminator = Discriminator(opt)
if cuda:
generator.cuda()
discriminator.cuda()
adversarial_loss.cuda()
categorical_loss.cuda()
continuous_loss.cuda()
# Initialize weights
generator.apply(weights_init_normal)
discriminator.apply(weights_init_normal)
# Configure data loader
data = CuratedBreastCancerData(
opt.batch_size,
adversarial_loss = torch.nn.MSELoss()
categorical_loss = torch.nn.CrossEntropyLoss()
continuous_loss = torch.nn.MSELoss()
# Loss weights
lambda_cat = 1
lambda_con = 0.1
# Initialize generator and discriminator
categorical_size = config.n_classes if config.n_classes else dataset.n_studies
print('categorical_size:', categorical_size)
#import ipdb; ipdb.set_trace()
generator = Generator(config.latent_dim, categorical_size, config.code_dim,
dataset.n_genes)
discriminator = Discriminator(categorical_size, config.code_dim,
dataset.n_genes)
if cuda:
generator.cuda()
discriminator.cuda()
adversarial_loss.cuda()
categorical_loss.cuda()
continuous_loss.cuda()
# Initialize weights
generator.apply(weights_init_normal)
discriminator.apply(weights_init_normal)
# Optimizers
optimizer_G = torch.optim.Adam(generator.parameters(),
lr=config.lr,
encoder = Encoder(c_size, z_size)
def weights_init(m):
classname = m.__class__.__name__
if classname.find('Conv') != -1:
m.weight.data.normal_(0.0, 0.02)
elif classname.find('BatchNorm') != -1:
m.weight.data.normal_(1.0, 0.02)
m.bias.data.fill_(0)
g = Generator(c_size + z_size)
g.apply(weights_init)
d = Discriminator()
d.apply(weights_init)
q = Q(c_size)
q.apply(weights_init)
dq = D_Q_commonlayer()
dq.apply(weights_init)
for i in [dq, d, q, g, encoder]:
i.cuda()
#i.apply(weights_init)
trainer = Trainer(g, dq, d, q, encoder, batch_size, img_size, c_size, z_size,
dataloader, version, c_loss_weight, RF_loss_weight,
generator_loss_weight, epoch)
trainer.train()