remove duplicate

Signed-off-by: Maanu Grover <maanug@nvidia.com>

Author: maanug-nv

nemo/tron/config.py

@@ -115,9 +115,6 @@ class GPTDatasetConfig(MCoreGPTDatasetConfig):
     data_sharding: bool = True
     """Disable data sharding."""
 
-    create_attention_mask_in_dataloader: bool = True
-    """If set, do not create attention_masks in dataloader."""
-
     def __post_init__(self) -> None:
         super(MCoreGPTDatasetConfig, self).__post_init__()
 

nemo/tron/llm/utils.py

@@ -76,7 +76,7 @@ def _broadcast(item):
             dtype=torch.float32,
             device=torch.cuda.current_device(),
         )
-        if cfg.dataset_config.create_attention_mask_in_dataloader:
+        if cfg.dataset_config.create_attention_mask:
             attention_mask = torch.empty(
                 (
                     mbs,