calculate similarities differently depending on return_loss in Clip f…

…orward

dalle_pytorch/dalle_pytorch.py

@@ -65,7 +65,7 @@ def generate_images(
 
     if exists(clipper):
         scores = clipper(text_seq, img_seq, return_loss = False)
-        return images, scores.diag()
+        return images, scores
 
     return images
 
@@ -202,11 +202,13 @@ def forward(
 
         text_latents, image_latents = map(lambda t: F.normalize(t, p = 2, dim = -1), (text_latents, image_latents))
 
-        sim = einsum('i d, j d -> i j', text_latents, image_latents) * self.temperature.exp()
+        temp = self.temperature.exp()
 
         if not return_loss:
+            sim = einsum('n d, n d -> n', text_latents, image_latents) * temp
             return sim
 
+        sim = einsum('i d, j d -> i j', text_latents, image_latents) * temp
         labels = torch.arange(b, device = device)
         loss = F.cross_entropy(sim, labels)
         return loss