diff --git a/src/transformers/models/codegen/modeling_codegen.py b/src/transformers/models/codegen/modeling_codegen.py
index 6ddd9d2addb2..0412b87b83a9 100644
--- a/src/transformers/models/codegen/modeling_codegen.py
+++ b/src/transformers/models/codegen/modeling_codegen.py
@@ -539,6 +539,14 @@ def forward(
 
         output_shape = input_shape + (hidden_states.size(-1),)
 
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with `config.gradient_checkpointing=True`. Setting "
+                    "`use_cache=False`..."
+                )
+                use_cache = False
+
         presents = () if use_cache else None
         all_self_attentions = () if output_attentions else None
         all_hidden_states = () if output_hidden_states else None
@@ -547,12 +555,6 @@ def forward(
                 all_hidden_states = all_hidden_states + (hidden_states,)
 
             if self.gradient_checkpointing and self.training:
-                if use_cache:
-                    logger.warning_once(
-                        "`use_cache=True` is incompatible with `config.gradient_checkpointing=True`. Setting "
-                        "`use_cache=False`..."
-                    )
-                    use_cache = False
 
                 def create_custom_forward(module):
                     def custom_forward(*inputs):