diff --git a/src/transformers/models/speech_to_text/modeling_speech_to_text.py b/src/transformers/models/speech_to_text/modeling_speech_to_text.py
index a8c1a640f2ff..d08863f8353f 100755
--- a/src/transformers/models/speech_to_text/modeling_speech_to_text.py
+++ b/src/transformers/models/speech_to_text/modeling_speech_to_text.py
@@ -1027,7 +1027,7 @@ def forward(
         if self.gradient_checkpointing and self.training:
             if use_cache:
                 logger.warning_once(
-                    "`use_cache = True` is incompatible with gradient checkpointing. Setting `use_cache =" " False`..."
+                    "`use_cache = True` is incompatible with gradient checkpointing. Setting `use_cache = False`..."
                 )
                 use_cache = False
 
diff --git a/src/transformers/models/speecht5/modeling_speecht5.py b/src/transformers/models/speecht5/modeling_speecht5.py
index e67c55c23b4e..975f483395be 100644
--- a/src/transformers/models/speecht5/modeling_speecht5.py
+++ b/src/transformers/models/speecht5/modeling_speecht5.py
@@ -1662,6 +1662,13 @@ def forward(
 
         deepspeed_zero3_is_enabled = is_deepspeed_zero3_enabled()
 
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                )
+                use_cache = False
+
         # decoder layers
         all_hidden_states = () if output_hidden_states else None
         all_self_attentions = () if output_attentions else None
@@ -1691,11 +1698,6 @@ def forward(
             past_key_value = past_key_values[idx] if past_key_values is not None else None
 
             if self.gradient_checkpointing and self.training:
-                if use_cache:
-                    logger.warning_once(
-                        "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
-                    )
-                    use_cache = False
 
                 def create_custom_forward(module):
                     def custom_forward(*inputs):