train_loader = torch.utils.data.DataLoader(dataset,
batch_size=args.batches,
shuffle=True,
drop_last=True)
test_loader = torch.utils.data.DataLoader(datasets.MNIST(
'./DATA/MNIST_data',
train=False,
transform=transforms.Compose([transforms.ToTensor()])),
batch_size=args.batches)
#############################
elif args.model == 'ising':
# For the Ising Model data set
model_size = args.ising_size * args.ising_size
image_size = args.ising_size
train_loader = torch.utils.data.DataLoader(rbm_pytorch.CSV_Ising_dataset(
args.training_data, size=model_size),
shuffle=True,
batch_size=args.batches,
drop_last=True)
# Read the model, example
rbm = rbm_pytorch.RBM(k=args.kCD, n_vis=model_size, n_hid=hidden_layers)
# load the model, if the file is present
if args.ckpoint is not None:
print("Loading saved network state from file", args.ckpoint)
rbm.load_state_dict(torch.load(args.ckpoint))
##############################
# Training parameters
image_size = 28
if parameters['initialize_with_training']:
print("Loading MNIST training set...")
train_loader = datasets.MNIST('./DATA/MNIST_data',
train=True,
download=True,
transform=transforms.Compose(
[transforms.ToTensor()]))
#############################
elif parameters['model'] == 'ising':
# For the Ising Model data set
image_size = parameters['ising']['size']
model_size = image_size * image_size
if parameters['initialize_with_training']:
print("Loading Ising training set...")
train_loader = rbm_pytorch.CSV_Ising_dataset(
parameters['ising']['train_data'], size=model_size, skiprows=1)
# Compute magnetization and susceptibility
train_mag = []
for i in xrange(len(train_loader)):
data = train_loader[i][0].view(-1, model_size)
train_mag.append(
ising_magnetization(get_ising_variables(data.numpy())))
tr_magarray = np.asarray(train_mag)
ising_averages(tr_magarray, model_size, "training_set")
# Read the model, example
rbm = rbm_pytorch.RBM(n_vis=model_size, n_hid=hidden_layers)
# load the model, if the file is present
try:
print(args)
pprint(parameters)
model_size = parameters['ising']['size'] * parameters['ising']['size']
rbm = rbm_pytorch.RBM(n_vis=model_size, n_hid=parameters['hidden_layers'])
if parameters['do_convergence_analysis']:
###################################################################
#### Convergence analysis: observables/loglikelihood vs epochs ####
###################################################################
print("Convergence analysis L=" + str(parameters['ising']['size']))
print("Loading Ising training set...")
train_loader = torch.utils.data.DataLoader(rbm_pytorch.CSV_Ising_dataset(parameters['ising']['train_data'], size=model_size), shuffle=True,
batch_size=parameters['batch_size'], drop_last=True)
analysis_file = open(parameters['output_dir'] + "analysis_" + str(parameters['temperature']) + "_L" + str(parameters['ising']['size']) + "/data.dat", 'w')
analysis_file.write("trained rbms from " + str(parameters['checkpoint'])+ "\n")
pbar = tqdm(range(parameters['start_epoch'], parameters['final_epoch']))
n = 0
print ("Range of epochs: ", parameters['start_epoch'], " to ", parameters['final_epoch'])
npoints = int((parameters['final_epoch']-parameters['start_epoch'])/parameters['step'])
print("npoints",npoints)
epochs = np.zeros(npoints)
mag = np.zeros(npoints)
transform=transforms.Compose([
transforms.ToTensor()
]))
train_loader = torch.utils.data.DataLoader(dataset, batch_size=args.batches, shuffle=True, drop_last=True)
test_loader = torch.utils.data.DataLoader(
datasets.MNIST('./DATA/MNIST_data', train=False, transform=transforms.Compose([
transforms.ToTensor()
])),
batch_size=args.batches)
#############################
elif args.model == 'ising':
# For the Ising Model data set
model_size = args.ising_size * args.ising_size
image_size = args.ising_size
train_loader = torch.utils.data.DataLoader(rbm_pytorch.CSV_Ising_dataset(args.training_data, size=args.ising_size), shuffle=True,
batch_size=args.batches, drop_last=True)
# Read the model, example
rbm = rbm_pytorch.RBM(k=args.kCD, n_vis=model_size, n_hid=hidden_layers)
# load the model, if the file is present
if args.ckpoint is not None:
print("Loading saved network state from file", args.ckpoint)
rbm.load_state_dict(torch.load(args.ckpoint))
##############################
# Training parameters
learning_rate = 0.01 # Mnist 0.3
mom = 0.0 # momentum
shuffle=True,
drop_last=True)
test_loader = torch.utils.data.DataLoader(
datasets.MNIST('./DATA/MNIST_data',
train=False,
transform=transforms.Compose([transforms.ToTensor()])),
batch_size=parameters['batch_size'])
#############################
elif parameters['model'] == 'ising':
# For the Ising Model data set
model_size = parameters[
'ising_size'] # note that you have to give the full dimension LxL if you are in 2 dim
image_size = parameters['ising_size']
train_loader = torch.utils.data.DataLoader(
rbm_pytorch.CSV_Ising_dataset(parameters['training_data'],
size=model_size),
shuffle=True,
batch_size=parameters['batch_size'],
drop_last=True)
# Read the model, example
rbm = rbm_pytorch.RBM(k=parameters['kCD'],
n_vis=model_size,
n_hid=hidden_layers)
# load the model, if the file is present
if parameters['ckpoint'] is not None:
print("Loading saved network state from file", parameters['ckpoint'])
rbm.load_state_dict(torch.load(parameters['ckpoint']))
##############################
model_size = MNIST_SIZE
image_size = 28
if parameters['initialize_with_training']:
print("Loading MNIST training set...")
train_loader = datasets.MNIST('./DATA/MNIST_data', train=True, download=True,
transform=transforms.Compose([
transforms.ToTensor()
]))
#############################
elif parameters['model'] == 'ising':
# For the Ising Model data set
image_size = parameters['ising']['size']
model_size = image_size * image_size
if parameters['initialize_with_training']:
print("Loading Ising training set...")
train_loader = rbm_pytorch.CSV_Ising_dataset(
parameters['ising']['train_data'], size=image_size)
# Read the model, example
rbm = rbm_pytorch.RBM(n_vis=model_size, n_hid=hidden_layers)
# load the model, if the file is present
try:
print("Loading saved RBM network state from file",
parameters['checkpoint'])
rbm.load_state_dict(torch.load(parameters['checkpoint']))
except IOError as e:
print("I/O error({0}): {1}".format(e.errno, e.strerror))
except:
print("Unexpected error:", sys.exc_info()[0])
raise
