Fix reduce_across_processes inconsistent return type (#4733)

Author: NicolasHug

references/classification/utils.py

@@ -403,7 +403,9 @@ def store_model_weights(model, checkpoint_path, checkpoint_key="model", strict=T
 
 def reduce_across_processes(val):
     if not is_dist_avail_and_initialized():
-        return val
+        # nothing to sync, but we still convert to tensor for consistency with the distributed case.
+        return torch.tensor(val)
+
     t = torch.tensor(val, device="cuda")
     dist.barrier()
     dist.all_reduce(t)