add deallocate pipeline output optimization

Signed-off-by: Jimmy Zhang <jiemingz@nvidia.com>
JimmyZhang12 · Jan 31, 2024 · 492a1df · 492a1df
1 parent 618ff06
commit 492a1df
Showing 1 changed file with 1 addition and 0 deletions.
diff --git a/nemo/collections/nlp/models/language_modeling/megatron_base_model.py b/nemo/collections/nlp/models/language_modeling/megatron_base_model.py
@@ -475,6 +475,7 @@ def build_transformer_config(self) -> TransformerConfig:
             'recompute_num_layers': recompute_num_layers,
             'distribute_saved_activations': False,  # not currently used in NeMo
             'fp8': None,
+            'deallocate_pipeline_outputs': True
         }
 
         # populate the transformer config dict