Update modeling_quiet.py

modeling_quiet.py

@@ -1097,8 +1097,6 @@ class QuietForCausalLM(QuietPreTrainedModel, GenerationMixin):
         logits = self.lm_head(mixed_hidden_states)
         return logits
 
-        
-    from .generate import custom_generate
     @torch.no_grad()
     def generate(
         self,
@@ -1140,7 +1138,7 @@ class QuietForCausalLM(QuietPreTrainedModel, GenerationMixin):
             input_ids, past_key_values=None, attention_mask=attention_mask, use_cache=use_cache, **model_kwargs
         )
         
-        # Call the external custom generation function, ensuring it's integrated properly
+        from .generate import custom_generate
         return custom_generate(
             self,
             input_ids=input_ids,