class TestVAE(unittest.TestCase):
def setUp(self) -> None:
# self.model2 = VAE(3, 10)
self.model = VanillaVAE(3, 10)
def test_summary(self):
print(summary(self.model, (3, 64, 64), device='cpu'))
# print(summary(self.model2, (3, 64, 64), device='cpu'))
def test_forward(self):
x = torch.randn(16, 3, 64, 64)
y = self.model(x)
print("Model Output size:", y[0].size())
# print("Model2 Output size:", self.model2(x)[0].size())
def test_loss(self):
x = torch.randn(16, 3, 64, 64)
result = self.model(x)
loss = self.model.loss_function(*result, M_N=0.005)
print(loss)
plt.savefig('../plots/{}.png'.format(save_as),bbox_inches='tight', transparent="True", pad_inches=0)
plt.show()
if __name__ == '__main__':
# plotting with old VAE
#images, rec_images = random_sample(5)
#subplot(images, rec_images, save_as='testing_old')
# plotting with Vanillia - delicious
model_path = '../models/Vanilla_128_lr5e-4stable_cont_lr1-4-19.pth'
data_dir = '../data/128x128CACD2000'
model = VanillaVAE(layer_count=4, in_channels=3, latent_dim=100, size=128)
#images, rec_images = random_sample(5, model, model_path=model_path, data_dir=data_dir)
#subplot(images, rec_images, save_as='testing_vanillia_layer4_size64_stableLR_20epochs_final')
# here interpolation is done
interpolations1 = random_interpolate(5, model, model_path, subset=None, test_split=0.2)
interpolations2 = random_interpolate(5, model, model_path, subset=None, test_split=0.2)
# here interpolation between age is done
file1 = 'm_Matt_Damon_0001.jpg'
file2 = '42_Matt_Damon_0010.jpg'
file3 = '15_Daniel_Radcliffe_0001.jpg'
file4 = '24_Daniel_Radcliffe_0005.jpg'
interpolations_age1 = age_interpolate(5, model, model_path, file1, file2)
interpolations_age2 = age_interpolate(5, model, model_path, file3, file4)
def setUp(self) -> None:
# self.model2 = VAE(3, 10)
self.model = VanillaVAE(3, 10)
to_tensor = transforms.ToTensor()
# define transformations
trafo = transforms.Compose([PIL, to_tensor, normalize])
# set up Model
#model = standard_vae()
if torch.cuda.is_available():
device = 'cuda'
else:
device = 'cpu'
model = VanillaVAE(layer_count=4,
in_channels=3,
latent_dim=100,
size=128,
name='Vanilla_128_lr5e-4stable_cont_lr1-4')
model.to(device)
summary(model, (3, 128, 128))
model_path = '../models/Vanilla_128_lr5e-4stable-9.pth'
train(model,
epochs,
batch,
trafo,
subset_size=None,
test_split=0.2,
lrs=lrs,
args = parser.parse_args()
# Housekeeping
torch.manual_seed(0)
device = torch.device("cuda" if args.device=="cuda" else "cpu")
train_loader_kwargs = {'num_workers': 1, 'pin_memory': True} if args.device=="cuda" else {}
# Initalise VAE model
model_kwargs = {
"input_size": args.input_size,
"latent_size" : args.latent_size
}
if args.model_type == "vanilla":
model = VanillaVAE(**model_kwargs)
elif args.model_type == "conditional":
model_kwargs.update({"n_classes": 10})
model = ConditionalVAE(**model_kwargs)
# Load MNIST dataset
train_loader = torch.utils.data.DataLoader(
datasets.MNIST('../data', train=True, download=True,
transform=transforms.ToTensor()),
batch_size=args.batch_size, shuffle=True, **train_loader_kwargs
)
# Train model
train_kwargs = {
"learning_rate": args.learning_rate,
"epochs" : args.epochs,
from datetime import datetime
from numpy.random import normal
import numpy as np
import torch
from models import VAE, VanillaVAE
from torchsummary import summary
from utils import standard_vae, newest, set_split
import glob
from PIL import Image
from models import VanillaEncoder
if torch.cuda.is_available():
device = 'cuda'
else:
device = 'cpu'
model = VanillaVAE(layer_count=4, in_channels=3, latent_dim=100, size=128, name='trying_shit')
model.to(device)
summary(model, (3, 128, 128))
file_names =
matt_damon_young = '34_Matt_Damon_0001.jpg'
matt_damon_old1 = '42_Matt_Damon_0010.jpg'
matt_damon_old2 = '42_Matt_Damon_0011.jpg'
daniel_radcliff_young = '15_Daniel_Radcliffe_0001.jpg'
daniel_radcliff_old1 = '24_Daniel_Radcliffe_0017.jpg'
daniel_radcliff_old2 = '24_Daniel_Radcliffe_0005.jpg'