diff --git a/openspeech/models/contextnet/model.py b/openspeech/models/contextnet/model.py
index 2a73d9f..6de8f3d 100644
--- a/openspeech/models/contextnet/model.py
+++ b/openspeech/models/contextnet/model.py
@@ -56,7 +56,13 @@ class ContextNetModel(OpenspeechCTCModel):
     """
     def __init__(self, configs: DictConfig, tokenizer: Tokenizer) -> None:
         super(ContextNetModel, self).__init__(configs, tokenizer)
-        self.fc = Linear(self.configs.model.encoder_dim, self.num_classes, bias=False)
+        supported_models = {
+            'small': 0.5,
+            'medium': 1,
+            'large': 2,
+        }
+        alpha = supported_models[self.configs.model.model_size]
+        self.fc = Linear(int(self.configs.model.encoder_dim * alpha), self.num_classes, bias=False)
 
     def build_model(self):
         self.encoder = ContextNetEncoder(