fix bug with @jihoonerd

README.md

@@ -325,6 +325,7 @@ Offer training wrappers
 - [ ] train on a toy task, offer in colab
 - [ ] add attention to unet - apply some personal tricks with efficient attention
 - [ ] figure out the big idea behind latent diffusion and what can be ported over
+- [ ] consider U2-net for decoder https://arxiv.org/abs/2005.09007
 
 ## Citations
 

dalle2_pytorch/dalle2_pytorch.py

@@ -450,7 +450,7 @@ class DiffusionPrior(nn.Module):
 
         alphas = 1. - betas
         alphas_cumprod = torch.cumprod(alphas, axis=0)
-        alphas_cumprod_prev = F.pad(alphas_cumprod[:-1], (0, 1), value = 1.)
+        alphas_cumprod_prev = F.pad(alphas_cumprod[:-1], (1, 0), value = 1.)
 
         timesteps, = betas.shape
         self.num_timesteps = int(timesteps)
@@ -941,7 +941,7 @@ class Decoder(nn.Module):
 
         alphas = 1. - betas
         alphas_cumprod = torch.cumprod(alphas, axis=0)
-        alphas_cumprod_prev = F.pad(alphas_cumprod[:-1], (0, 1), value = 1.)
+        alphas_cumprod_prev = F.pad(alphas_cumprod[:-1], (1, 0), value = 1.)
 
         timesteps, = betas.shape
         self.num_timesteps = int(timesteps)

setup.py

@@ -10,7 +10,7 @@ setup(
       'dream = dalle2_pytorch.cli:dream'
     ],
   },
-  version = '0.0.15',
+  version = '0.0.16',
   license='MIT',
   description = 'DALL-E 2',
   author = 'Phil Wang',