skip partitioning ops with bool args/outputs (pytorch#3679)

Summary:
Pull Request resolved: pytorch#3679

bool is not currently supported in operators or in shader generation code

Reviewed By: jorgep31415

Differential Revision: D57509811

fbshipit-source-id: 1f9979611993745f056dbb1d1abb25dc8a02fb1a

backends/vulkan/partitioner/vulkan_partitioner.py

@@ -50,6 +50,10 @@ def node_val_is_compatible(self, node_val: Any) -> bool:
             if len(node_val.shape) > 4:
                 return False
 
+            # bool dtype not currently supported
+            if node_val.dtype == torch.bool:
+                return False
+
         if isinstance(node_val, (list, tuple)):
             for item in node_val:
                 if not self.node_val_is_compatible(item):