add self conditioning for elucidated ddpm

denoising_diffusion_pytorch/elucidated_diffusion.py

@@ -1,4 +1,5 @@
 from math import sqrt
+from random import random
 import torch
 from torch import nn, einsum
 import torch.nn.functional as F
@@ -52,7 +53,7 @@ def __init__(
     ):
         super().__init__()
         assert net.learned_sinusoidal_cond
-        assert not net.self_condition, 'not supported yet'
+        self.self_condition = net.self_condition
 
         self.net = net
 
@@ -100,7 +101,7 @@ def c_noise(self, sigma):
     # preconditioned network output
     # equation (7) in the paper
 
-    def preconditioned_network_forward(self, noised_images, sigma, clamp = False):
+    def preconditioned_network_forward(self, noised_images, sigma, self_cond = None, clamp = False):
         batch, device = noised_images.shape[0], noised_images.device
 
         if isinstance(sigma, float):
@@ -110,7 +111,8 @@ def preconditioned_network_forward(self, noised_images, sigma, clamp = False):
 
         net_out = self.net(
             self.c_in(padded_sigma) * noised_images,
-            self.c_noise(sigma)
+            self.c_noise(sigma),
+            self_cond
         )
 
         out = self.c_skip(padded_sigma) * noised_images +  self.c_out(padded_sigma) * net_out
@@ -161,6 +163,10 @@ def sample(self, batch_size = 16, num_sample_steps = None, clamp = True):
 
         images = init_sigma * torch.randn(shape, device = self.device)
 
+        # for self conditioning
+
+        x_start = None
+
         # gradually denoise
 
         for sigma, sigma_next, gamma in tqdm(sigmas_and_gammas, desc = 'sampling time step'):
@@ -171,19 +177,24 @@ def sample(self, batch_size = 16, num_sample_steps = None, clamp = True):
             sigma_hat = sigma + gamma * sigma
             images_hat = images + sqrt(sigma_hat ** 2 - sigma ** 2) * eps
 
-            model_output = self.preconditioned_network_forward(images_hat, sigma_hat, clamp = clamp)
+            self_cond = x_start if self.self_condition else None
+
+            model_output = self.preconditioned_network_forward(images_hat, sigma_hat, self_cond, clamp = clamp)
             denoised_over_sigma = (images_hat - model_output) / sigma_hat
 
             images_next = images_hat + (sigma_next - sigma_hat) * denoised_over_sigma
 
             # second order correction, if not the last timestep
 
             if sigma_next != 0:
-                model_output_next = self.preconditioned_network_forward(images_next, sigma_next, clamp = clamp)
+                self_cond = model_output if self.self_condition else None
+
+                model_output_next = self.preconditioned_network_forward(images_next, sigma_next, self_cond, clamp = clamp)
                 denoised_prime_over_sigma = (images_next - model_output_next) / sigma_next
                 images_next = images_hat + 0.5 * (sigma_next - sigma_hat) * (denoised_over_sigma + denoised_prime_over_sigma)
 
             images = images_next
+            x_start = model_output
 
         images = images.clamp(-1., 1.)
         return unnormalize_to_zero_to_one(images)
@@ -211,7 +222,15 @@ def forward(self, images):
 
         noised_images = images + padded_sigmas * noise  # alphas are 1. in the paper
 
-        denoised = self.preconditioned_network_forward(noised_images, sigmas)
+        self_cond = None
+
+        if self.self_condition and random() < 0.5:
+            # from hinton's group's bit diffusion paper
+            with torch.no_grad():
+                self_cond = self.preconditioned_network_forward(noised_images, sigmas)
+                self_cond.detach_()
+
+        denoised = self.preconditioned_network_forward(noised_images, sigmas, self_cond)
 
         losses = F.mse_loss(denoised, images, reduction = 'none')
         losses = reduce(losses, 'b ... -> b', 'mean')

setup.py

@@ -3,7 +3,7 @@
 setup(
   name = 'denoising-diffusion-pytorch',
   packages = find_packages(),
-  version = '0.27.1',
+  version = '0.27.2',
   license='MIT',
   description = 'Denoising Diffusion Probabilistic Models - Pytorch',
   author = 'Phil Wang',