diff options
Diffstat (limited to 'main.py')
| -rw-r--r-- | main.py | 244 |
1 files changed, 109 insertions, 135 deletions
@@ -31,24 +31,18 @@ class UNet(nn.Module): # with the exception of the last block which does not include a max-pooling layer. # ------- # input: 28x28x1 - self.e11 = nn.Conv2d(1, 64, kernel_size=3, - padding=1) # output: 28x28x64 - self.e12 = nn.Conv2d(64, 64, kernel_size=3, - padding=1) # output: 28x28x64 + self.e11 = nn.Conv2d(1, 64, kernel_size=3, padding=1) # output: 28x28x64 + self.e12 = nn.Conv2d(64, 64, kernel_size=3, padding=1) # output: 28x28x64 self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2) # output: 14x14x64 # input: 14x14x64 - self.e21 = nn.Conv2d(64, 128, kernel_size=3, - padding=1) # output: 14x14x128 - self.e22 = nn.Conv2d(128, 128, kernel_size=3, - padding=1) # output: 14x14x128 + self.e21 = nn.Conv2d(64, 128, kernel_size=3, padding=1) # output: 14x14x128 + self.e22 = nn.Conv2d(128, 128, kernel_size=3, padding=1) # output: 14x14x128 self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2) # output: 7x7x128 # input: 7x7x128 - self.e31 = nn.Conv2d(129, 256, kernel_size=3, - padding=1) # output: 7x7x256 - self.e32 = nn.Conv2d(257, 256, kernel_size=3, - padding=1) # output: 7x7x256 + self.e31 = nn.Conv2d(129, 256, kernel_size=3, padding=1) # output: 7x7x256 + self.e32 = nn.Conv2d(257, 256, kernel_size=3, padding=1) # output: 7x7x256 # Decoder # In the decoder, the output of the encoder is upsampled using the ConvTranspose2d function. @@ -57,19 +51,19 @@ class UNet(nn.Module): # ------- # input: 7x7x256 self.upconv1 = nn.ConvTranspose2d( - 256, 128, kernel_size=2, stride=2) # output: 14x14x128 - self.d11 = nn.Conv2d(256, 128, kernel_size=3, - padding=1) # output: 14x14x(128x2) - self.d12 = nn.Conv2d(128, 128, kernel_size=3, - padding=1) # output: 14x14x128 + 256, 128, kernel_size=2, stride=2 + ) # output: 14x14x128 + self.d11 = nn.Conv2d( + 256, 128, kernel_size=3, padding=1 + ) # output: 14x14x(128x2) + self.d12 = nn.Conv2d(128, 128, kernel_size=3, padding=1) # output: 14x14x128 # input: 14x14x128 self.upconv2 = nn.ConvTranspose2d( - 128, 64, kernel_size=2, stride=2) # output: 28x28x64 - self.d21 = nn.Conv2d(128, 64, kernel_size=3, - padding=1) # output: 28x28x(64x2) - self.d22 = nn.Conv2d(64, 64, kernel_size=3, - padding=1) # output: 28x28x64 + 128, 64, kernel_size=2, stride=2 + ) # output: 28x28x64 + self.d21 = nn.Conv2d(128, 64, kernel_size=3, padding=1) # output: 28x28x(64x2) + self.d22 = nn.Conv2d(64, 64, kernel_size=3, padding=1) # output: 28x28x64 # Output # The output of the decoder is passed through a convolutional layer with the Conv2d function to obtain the final output. @@ -77,58 +71,27 @@ class UNet(nn.Module): # input: 28x28x64 self.outconv = nn.Conv2d(64, 1, kernel_size=1) # output: 28x28x2 - def forward(self, x, y): - # Input - y = self.inconv3(y) - y = y.view(-1, 1, 7, 7) - - # Encoder - x = F.relu(self.e11(x)) - x1 = F.relu(self.e12(x)) - x = self.pool1(x1) - - x = F.relu(self.e21(x)) - x2 = F.relu(self.e22(x)) - x = self.pool2(x2) - - x = torch.cat([x, y], dim=1) - x = F.relu(self.e31(x)) - x = torch.cat([x, y], dim=1) - x = F.relu(self.e32(x)) - - # Decoder - x = self.upconv1(x) - x = torch.cat([x, x], dim=1) - x = F.relu(self.d11(x)) - x = F.relu(self.d12(x)) + def forward(self, xt, t, vec): + # Encode t and vec + t = F.relu(self.encodet(t)) + vec = F.relu(self.encodevec(vec)) - x = self.upconv2(x) - x = torch.cat([x, x1], dim=1) - x = F.relu(self.d21(x)) - x = F.relu(self.d22(x)) - - # Output - x = self.outconv(x) + # Concat all 3 + x = torch.cat((xt, t, vec), ) + # TODO return x -DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') +DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu") -DIFFU_STEPS = 10 -BETA = torch.linspace(0.0001, 0.2, DIFFU_STEPS+1, device=DEVICE) +DIFFU_STEPS = 100 +BETA = torch.linspace(0.0001, 0.2, DIFFU_STEPS, device=DEVICE) ALPHA = 1 - BETA ALPHA_BAR = torch.cumprod(ALPHA, dim=0) SIGMA2 = BETA -def q_xt_x0(x0, t): - alpha_bar = ALPHA_BAR[t] - mean = x0 * torch.sqrt(alpha_bar) - std = 1 - alpha_bar - return torch.distributions.Normal(mean, std) - - def q_xt_xt_1(xt_1, t): beta = BETA[t] mean = torch.sqrt(1 - beta) * xt_1 @@ -136,27 +99,36 @@ def q_xt_xt_1(xt_1, t): return torch.distributions.Normal(mean, std) -def p_xt_1_xt(model, xt, vec): - t = vec[..., -1].to(dtype=torch.long) - eps_theta = model(xt, vec) - alpha_bar = ALPHA_BAR[t].unsqueeze( - -1).unsqueeze(-1).unsqueeze(-1).repeat(1, 1, 28, 28) - alpha = ALPHA[t].unsqueeze(-1).unsqueeze(-1).unsqueeze(-1).repeat(1, 1, 28, 28) - beta = BETA[t].unsqueeze(-1).unsqueeze(-1).unsqueeze(-1).repeat(1, 1, 28, 28) - eps_coef = (1 - alpha) / torch.sqrt(1 - alpha_bar) - # mean = 1 / torch.sqrt(alpha) * (xt - eps_coef * eps_theta) - mean = (xt - beta * eps_theta) - # std = torch.sqrt( - # SIGMA2[t]).unsqueeze(-1).unsqueeze(-1).unsqueeze(-1).repeat(1, 1, 28, 28) - std = beta +def q_xt_x0(x0, t): + alpha_bar = ALPHA_BAR[t] + mean = torch.sqrt(alpha_bar) * x0 + std = 1 - alpha_bar return torch.distributions.Normal(mean, std) +def p_xt_1_xt(model, xt, t, vec): + if not 0 < t <= DIFFU_STEPS: + raise Exception('Steps out of range') + alpha_bar_t = ALPHA_BAR[t] + alpha_bar_t_1 = ALPHA_BAR[t-1] + alpha_t = ALPHA[t] + beta_t = BETA[t] + beta_tilde = (1 - alpha_bar_t_1) / (1 - alpha_bar_t) * beta_t + + epsilon_theta = model(xt, t, vec) + + # sigma_theta = torch.exp(nu * torch.log(beta_t) + (1 - nu) * torch.log(beta_tilde)) + sigma_theta = beta_tilde + mu_theta = (xt - beta_t / (torch.sqrt(1 - alpha_bar_t) * epsilon_theta)) / torch.sqrt(alpha_t) + + return torch.distributions.Normal(mu_theta, sigma_theta) + + class MNISTDiffusionDataset(Dataset): def __init__(self, train=True): super().__init__() self.mnist_data = datasets.MNIST( - root='./data', + root="./data", train=train, download=True, transform=transforms.ToTensor(), @@ -168,24 +140,20 @@ class MNISTDiffusionDataset(Dataset): img = img.to(DEVICE) # Add noise to the image. - t_1 = torch.randint(0, DIFFU_STEPS, (1,), device=DEVICE) - t = t_1 + 1 - xt_1 = q_xt_x0(img, t_1).sample() - eps = torch.distributions.Normal(0, 1).sample(img.shape).to(DEVICE) - xt = xt_1 * torch.sqrt(ALPHA[t]) + BETA[t] * eps + t = torch.randint(1, DIFFU_STEPS, (1,), device=DEVICE) + xt = q_xt_x0(img, t).sample() + eps = xt - img # Convert the label to a one-hot vector. - vector = torch.nn.functional.one_hot( + vec = torch.nn.functional.one_hot( torch.tensor(label), num_classes=10, ) - # Add contextual information (t) to the label. - vector = torch.cat([vector, torch.tensor([t_1+1])]) - return ( xt.clone().detach().to(dtype=torch.float32, device=DEVICE), - vector.clone().detach().to(dtype=torch.float32, device=DEVICE), + t.clone().detach().to(dtype=torch.float32, device=DEVICE), + vec.clone().detach().to(dtype=torch.float32, device=DEVICE), ( img.clone().detach().to(dtype=torch.float32, device=DEVICE), eps, @@ -199,11 +167,12 @@ class MNISTDiffusionDataset(Dataset): def loss_fn(y_pred, y_true): x0, eps, t = y_true - return nn.MSELoss()(y_pred, eps) + eps_theta = y_pred + return nn.MSELoss()(eps_theta, eps) mnist_data = datasets.MNIST( - root='./data', + root="./data", train=True, download=True, transform=transforms.ToTensor(), @@ -213,22 +182,22 @@ img = img.to(DEVICE) fig = plt.figure(figsize=(DIFFU_STEPS, 2)) for t_1 in range(0, DIFFU_STEPS): xt_1 = q_xt_x0(img, t_1).sample() - x_t = q_xt_xt_1(xt_1, t_1+1).sample() + x_t = q_xt_xt_1(xt_1, t_1 + 1).sample() ax = fig.add_subplot(2, DIFFU_STEPS, t_1 + 1) - ax.imshow(xt_1[0].cpu(), cmap='gray') - ax.axis('off') + ax.imshow(xt_1[0].cpu(), cmap="gray") + ax.axis("off") ax = fig.add_subplot(2, DIFFU_STEPS, DIFFU_STEPS + t_1 + 1) - ax.imshow(x_t[0].cpu(), cmap='gray') - ax.axis('off') + ax.imshow(x_t[0].cpu(), cmap="gray") + ax.axis("off") fig.tight_layout() -fig.savefig('img.tmp.png') +fig.savefig("img.tmp.png") ############ # Training # ############ -torch.multiprocessing.set_start_method('spawn') +torch.multiprocessing.set_start_method("spawn") # Load the model. model = UNet().to(DEVICE) @@ -248,7 +217,7 @@ epochs = 1 trainer.train(model, train_loader, epochs, optimizer, criterion) # Save the model. -torch.save(model.state_dict(), 'model.pth') +torch.save(model.state_dict(), "model.pth") ############## @@ -280,19 +249,19 @@ torch.save(model.state_dict(), 'model.pth') # Load the model. model = UNet().to(DEVICE) -model.load_state_dict(torch.load('model.pth')) +model.load_state_dict(torch.load("model.pth")) # Set the model to evaluation mode. model.eval() n = 10 fig = plt.figure(figsize=(2 * 2 * n, 3 * 2)) -gs = plt.GridSpec(nrows=3, ncols=2*2*n) +gs = plt.GridSpec(nrows=3, ncols=2 * 2 * n) for i in range(n): # Get the i-th input and its label. input, vector, (x0, eps, t) = train_dataset[i] - xt_1 = q_xt_x0(input, t-1).sample() + xt_1 = q_xt_x0(input, t - 1).sample() xt = q_xt_xt_1(xt_1, t).sample() xt_1_pred = p_xt_1_xt( model, @@ -301,36 +270,34 @@ for i in range(n): ).sample() # Plot the input. - ax = fig.add_subplot(gs[0:1, 1 + 4 * i:3 + 4 * i]) - ax.imshow(xt[0].cpu(), cmap='gray') - ax.axis('off') + ax = fig.add_subplot(gs[0:1, 1 + 4 * i : 3 + 4 * i]) + ax.imshow(xt[0].cpu(), cmap="gray") + ax.axis("off") # Plot the label. - ax = fig.add_subplot(gs[1:2, 4 * i:2 + 4 * i]) - ax.imshow(xt_1[0].cpu(), cmap='gray') - ax.axis('off') + ax = fig.add_subplot(gs[1:2, 4 * i : 2 + 4 * i]) + ax.imshow(xt_1[0].cpu(), cmap="gray") + ax.axis("off") # Plot the model output. - ax = fig.add_subplot(gs[1:2, 2 + 4 * i:4 + 4 * i]) - ax.imshow(xt_1_pred[0, 0].cpu().detach(), cmap='gray') - ax.axis('off') + ax = fig.add_subplot(gs[1:2, 2 + 4 * i : 4 + 4 * i]) + ax.imshow(xt_1_pred[0, 0].cpu().detach(), cmap="gray") + ax.axis("off") # Plot the difference between the label and the model output. - ax = fig.add_subplot(gs[2:3, 1 + 4 * i:3 + 4 * i]) - ax.imshow( - (xt_1.cpu() - xt_1_pred[0, 0].cpu().detach())[0], cmap='coolwarm') - ax.axis('off') + ax = fig.add_subplot(gs[2:3, 1 + 4 * i : 3 + 4 * i]) + ax.imshow((xt_1.cpu() - xt_1_pred[0, 0].cpu().detach())[0], cmap="coolwarm") + ax.axis("off") fig.tight_layout() -fig.savefig('diff.tmp.png') +fig.savefig("diff.tmp.png") # Plot the evolution of the noise. fig = plt.figure(figsize=(n, DIFFU_STEPS)) noises = np.random.normal(0, 1, (n, 1, 28, 28)) noises = torch.Tensor(noises).to(DEVICE) -vector = torch.nn.functional.one_hot( - torch.tensor(range(n)), num_classes=10).to(DEVICE) +vector = torch.nn.functional.one_hot(torch.tensor(range(n)), num_classes=10).to(DEVICE) vector = vector.clone().detach().to(dtype=torch.float32) # Apply the model multiple times. @@ -339,44 +306,51 @@ for i in range(DIFFU_STEPS): noises = p_xt_1_xt( model, noises, - torch.cat([ - vector, - torch.tensor([t] * n) - .unsqueeze(-1) - .to(device=DEVICE) - .to(dtype=torch.long), - ], dim=-1), + torch.cat( + [ + vector, + torch.tensor([t] * n) + .unsqueeze(-1) + .to(device=DEVICE) + .to(dtype=torch.long), + ], + dim=-1, + ), ).sample() for j in range(n): ax = fig.add_subplot(DIFFU_STEPS, n, i * n + j + 1) - ax.imshow(noises[j, 0].cpu().detach(), cmap='gray') - ax.axis('off') + ax.imshow(noises[j, 0].cpu().detach(), cmap="gray") + ax.axis("off") fig.tight_layout() -fig.savefig('diffu.tmp.png') +fig.savefig("diffu.tmp.png") # Plot bench of generated images. fig = plt.figure(figsize=(n, n)) noises = np.random.normal(0, 1, (n * n, 1, 28, 28)) noises = torch.Tensor(noises).to(DEVICE) -vector = torch.nn.functional.one_hot( - torch.tensor([range(n)] * n), num_classes=10).to(DEVICE) +vector = torch.nn.functional.one_hot(torch.tensor([range(n)] * n), num_classes=10).to( + DEVICE +) vector = vector.clone().detach().to(dtype=torch.float32) # Apply the model multiple times. for i in range(DIFFU_STEPS): t = DIFFU_STEPS - i - 1 - noises = model(noises, torch.cat( - [vector, torch.tensor([[t] * n] * n).unsqueeze(-1) - .to(DEVICE)], dim=-1)) + noises = model( + noises, + torch.cat( + [vector, torch.tensor([[t] * n] * n).unsqueeze(-1).to(DEVICE)], dim=-1 + ), + ) for i in range(n * n): ax = fig.add_subplot(n, n, i + 1) - ax.imshow(noises[i, 0].cpu().detach(), cmap='gray') - ax.axis('off') + ax.imshow(noises[i, 0].cpu().detach(), cmap="gray") + ax.axis("off") fig.tight_layout() -fig.savefig('bench.tmp.png') +fig.savefig("bench.tmp.png") -plt.close('all') +plt.close("all") |