Allow for multiple example inputs when creating summary (#543)

Author: VSJMilewski

pytorch_lightning/core/memory.py

@@ -50,20 +50,31 @@ def get_variable_sizes(self):
         input_ = self.model.example_input_array
 
         if self.model.on_gpu:
-            input_ = input_.cuda(0)
+            device = next(self.model.parameters()).get_device()
+            # test if input is a list or a tuple
+            if isinstance(input_, (list, tuple)):
+                input_ = [input_i.cuda(device) if torch.is_tensor(input_i) else input_i
+                          for input_i in input_]
+            else:
+                input_ = input_.cuda(device)
 
         if self.model.trainer.use_amp:
-            input_ = input_.half()
+            # test if it is not a list or a tuple
+            if isinstance(input_, (list, tuple)):
+                input_ = [input_i.half() if torch.is_tensor(input_i) else input_i
+                          for input_i in input_]
+            else:
+                input_ = input_.half()
 
         with torch.no_grad():
 
             for _, m in mods:
-                if type(input_) is list or type(input_) is tuple:  # pragma: no cover
+                if isinstance(input_, (list, tuple)):  # pragma: no cover
                     out = m(*input_)
                 else:
                     out = m(input_)
 
-                if type(input_) is tuple or type(input_) is list:  # pragma: no cover
+                if isinstance(input_, (list, tuple)):  # pragma: no cover
                     in_size = []
                     for x in input_:
                         if type(x) is list:
@@ -75,7 +86,7 @@ def get_variable_sizes(self):
 
                 in_sizes.append(in_size)
 
-                if type(out) is tuple or type(out) is list:  # pragma: no cover
+                if isinstance(out, (list, tuple)):  # pragma: no cover
                     out_size = np.asarray([x.size() for x in out])
                 else:
                     out_size = np.array(out.size())