diff --git a/nemo/lightning/megatron_parallel.py b/nemo/lightning/megatron_parallel.py
index 7d0b5ba31f0c..54c1c32d4a96 100644
--- a/nemo/lightning/megatron_parallel.py
+++ b/nemo/lightning/megatron_parallel.py
@@ -670,7 +670,6 @@ def init_ddp(self):
             model_chunk.buffers = (
                 dist_module.buffers
             )  # We need to do this explicitly since this is a attr pytorch uses
-            model_chunk.__class__.__getattr__ = getattr_proxy  # type: ignore
 
         # param_sync_func is set in nemo.lightning.pytorch.optim.megatron
         no_sync_func, grad_sync_func = extract_ddp_funcs(self.ddp_config, self)