align the ema model device back after sampling from the cascading ddp…

…m in the decoder

dalle2_pytorch/train.py

@@ -105,6 +105,10 @@ def __init__(
         self.register_buffer('initted', torch.Tensor([False]))
         self.register_buffer('step', torch.tensor([0.]))
 
+    def restore_ema_model_device(self):
+        device = self.initted.device
+        self.ema_model.to(device)
+
     def update(self):
         self.step += 1
 
@@ -305,6 +309,11 @@ def sample(self, *args, **kwargs):
 
         if self.use_ema:
             self.decoder.unets = trainable_unets             # restore original training unets
+
+        # cast the ema_model unets back to original device
+        for ema in self.ema_unets:
+            ema.restore_ema_model_device()
+
         return output
 
     def forward(

setup.py

@@ -10,7 +10,7 @@
       'dream = dalle2_pytorch.cli:dream'
     ],
   },
-  version = '0.2.11',
+  version = '0.2.12',
   license='MIT',
   description = 'DALL-E 2',
   author = 'Phil Wang',