cleanup conv-like attention

dalle_pytorch/attention.py

@@ -177,18 +177,18 @@ def forward(self, x, mask = None, rotary_pos_emb = None):
         dots_image = einsum('b i d, b i j d -> b i j', q_img, k_img)
         dots_image_to_text = einsum('b i d, b j d -> b i j', q_img, k_text)
 
-        # calculate causal attention for local convolution
+        # use padding of 0 on tensor of 1s and unfold for padding mask
 
         i, j = dots_image.shape[-2:]
-        img_seq = torch.arange(img_seq_len, device = device)
-        k_img_indices = rearrange(img_seq.float(), '(h w) -> () () h w', h = img_size)
-        k_img_indices = F.pad(k_img_indices, causal_padding, value = img_seq_len) # padding set to be max, so it is never attended to
-        k_img_indices = F.unfold(k_img_indices, kernel_size, dilation = dilation)
-        k_img_indices = rearrange(k_img_indices, 'b j i -> b i j')
+        ones = torch.ones((img_seq_len,), device = device)
+        ones = rearrange(ones, '(h w) -> () () h w', h = img_size)
+        ones = F.pad(ones, causal_padding, value = 0.)
+        ones = F.unfold(ones, kernel_size, dilation = dilation)
+        ones = rearrange(ones, 'b j i -> b i j')
 
         # mask image attention
 
-        padding_mask =  k_img_indices == img_seq_len
+        padding_mask = ones == 0.
 
         # concat text mask with image causal mask
 

setup.py

@@ -4,7 +4,7 @@
   name = 'dalle-pytorch',
   packages = find_packages(),
   include_package_data = True,
-  version = '1.5.0',
+  version = '1.5.1',
   license='MIT',
   description = 'DALL-E - Pytorch',
   author = 'Phil Wang',