def rbm(opytimizer):
# Gathers params
lr = opytimizer[0][0]
momentum = opytimizer[1][0]
decay = opytimizer[2][0]
# Creates an RBM
model = RBM(n_visible=784,
n_hidden=128,
steps=1,
learning_rate=lr,
momentum=momentum,
decay=decay,
temperature=1,
use_gpu=False)
# Training an RBM
error, _ = model.fit(train, batch_size=128, epochs=5)
return error
T = args.temp
gpu = args.gpu
batch_size = args.batch_size
epochs = args.epochs
seed = args.seed
# Loads the training data
train, _, _ = s.load_dataset(name=dataset, val_split=split)
# Defining the torch seed
torch.manual_seed(seed)
# Creates an RBM
model = RBM(n_visible=n_visible,
n_hidden=n_hidden,
steps=steps,
learning_rate=lr,
momentum=momentum,
decay=decay,
temperature=T,
use_gpu=True)
# Fits an RBM
mse, pl = model.fit(train, batch_size=batch_size, epochs=epochs)
# Saves the RBM
torch.save(model, f'models/{output_model}.pth')
# Outputting extracted weights to a numpy file
s.save_torch_as_numpy(model.W, output_file=f'weights/{output_weight}.npy')
download=True,
transform=torchvision.transforms.ToTensor(),
)
test = torchvision.datasets.MNIST(
root="./data",
train=False,
download=True,
transform=torchvision.transforms.ToTensor(),
)
# Creating an RBM
model = RBM(
n_visible=784,
n_hidden=128,
steps=1,
learning_rate=0.1,
momentum=0,
decay=0,
temperature=1,
use_gpu=True,
)
# Training an RBM
mse, pl = model.fit(train, batch_size=128, epochs=5)
# Reconstructing test set
rec_mse, v = model.reconstruct(test)
# Saving model
torch.save(model, "model.pth")
# Checking the model's history
import torchvision
import learnergy.visual.image as im
from learnergy.models.bernoulli import RBM
# Creating training dataset
train = torchvision.datasets.MNIST(root='./data',
train=True,
download=True,
transform=torchvision.transforms.ToTensor())
# Creating an RBM
model = RBM(n_visible=784,
n_hidden=128,
steps=1,
learning_rate=0.1,
momentum=0,
decay=0,
temperature=1,
use_gpu=True)
# Training an RBM
model.fit(train, epochs=5)
# Creating weights' mosaic
im.create_mosaic(model.W)
from learnergy.models.bernoulli import RBM
# Creates an RBM-based class
model = RBM(
n_visible=784,
n_hidden=128,
steps=1,
learning_rate=0.1,
momentum=0,
decay=0,
temperature=1,
use_gpu=False,
)
# Creating training and validation/testing dataset
train = torchvision.datasets.MNIST(root='./data',
train=True,
download=True,
transform=torchvision.transforms.ToTensor())
test = torchvision.datasets.MNIST(root='./data',
train=False,
download=True,
transform=torchvision.transforms.ToTensor())
# Creating an RBM
model = RBM(n_visible=784,
n_hidden=128,
steps=1,
learning_rate=0.1,
momentum=0,
decay=0,
temperature=1,
use_gpu=True)
# Training an RBM
model.fit(train, batch_size=batch_size, epochs=1)
# Creating the Fully Connected layer to append on top of RBM
fc = nn.Linear(model.n_hidden, n_classes)
# Check if model uses GPU
if model.device == 'cuda':
# If yes, put fully-connected on GPU
fc = fc.cuda()