[bugfix] Correct call to torch.no_grad (#5124)

Co-authored-by: Gregor Koporec <[email protected]>
Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: Sean Naren <[email protected]>

Author: 4 people

pytorch_lightning/utilities/distributed.py

@@ -206,6 +206,6 @@ def all_gather_ddp_if_available(
         if sync_grads:
             return AllGatherGrad.apply(tensor, group)
         else:
-            with torch.no_grad:
+            with torch.no_grad():
                 return AllGatherGrad.apply(tensor, group)
     return tensor