Remove mask if use fusion mask (NVIDIA#9723)

* Remove mask if use fusion mask

Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com>

* Apply isort and black reformatting

Signed-off-by: hsiehjackson <hsiehjackson@users.noreply.github.com>

---------

Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com>
Signed-off-by: hsiehjackson <hsiehjackson@users.noreply.github.com>
Co-authored-by: hsiehjackson <hsiehjackson@users.noreply.github.com>
Signed-off-by: Malay Nagda <malayn@malayn-mlt.client.nvidia.com>

nemo/collections/nlp/modules/common/text_generation_utils.py

@@ -794,6 +794,9 @@ def generate(
     if random_seed is not None:
         seed_everything(random_seed)
 
+    if hasattr(model, 'get_attention_mask_from_fusion') and model.get_attention_mask_from_fusion:
+        compute_attention_mask = False
+
     output = synced_generate(
         model,
         inference_strategy,