diff --git a/nemo/collections/multimodal/data/energon/config.py b/nemo/collections/multimodal/data/energon/config.py
index ab8ecf9fc06d3..45ca8e9db8006 100644
--- a/nemo/collections/multimodal/data/energon/config.py
+++ b/nemo/collections/multimodal/data/energon/config.py
@@ -64,7 +64,7 @@ class LLaVATemplateConfig(BaseConversationTemplateConfig):
 
 @dataclass
 class MultiModalSampleConfig:
-    image_token: ImageToken = ImageToken()
+    image_token: ImageToken = field(default_factory=ImageToken)
     ignore_place_holder: int = -100
-    conversation_template_config: LLaVATemplateConfig = LLaVATemplateConfig()
+    conversation_template_config: LLaVATemplateConfig = field(default_factory=LLaVATemplateConfig)
     image_following_text: bool = True