Force use_cache to be False in PyTorch (#15385)

* use_cache = False for PT models if labels is passed

* Fix for BigBirdPegasusForConditionalGeneration

* add warning if users specify use_cache=True

* Use logger.warning instead of warnings.warn
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/bart/modeling_bart.py

@@ -1318,6 +1318,9 @@ class BartForConditionalGeneration(BartPretrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:
+            if use_cache:
+                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
+            use_cache = False
             if decoder_input_ids is None and decoder_inputs_embeds is None:
                 decoder_input_ids = shift_tokens_right(
                     labels, self.config.pad_token_id, self.config.decoder_start_token_id

src/transformers/models/bigbird_pegasus/modeling_bigbird_pegasus.py

@@ -2513,6 +2513,9 @@ class BigBirdPegasusForConditionalGeneration(BigBirdPegasusPreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:
+            if use_cache:
+                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
+            use_cache = False
             if decoder_input_ids is None and decoder_inputs_embeds is None:
                 decoder_input_ids = shift_tokens_right(
                     labels, self.config.pad_token_id, self.config.decoder_start_token_id

src/transformers/models/blenderbot/modeling_blenderbot.py

@@ -1287,6 +1287,9 @@ class BlenderbotForConditionalGeneration(BlenderbotPreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:
+            if use_cache:
+                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
+            use_cache = False
             if decoder_input_ids is None:
                 decoder_input_ids = shift_tokens_right(
                     labels, self.config.pad_token_id, self.config.decoder_start_token_id

src/transformers/models/blenderbot_small/modeling_blenderbot_small.py

@@ -1258,6 +1258,9 @@ class BlenderbotSmallForConditionalGeneration(BlenderbotSmallPreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:
+            if use_cache:
+                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
+            use_cache = False
             if decoder_input_ids is None:
                 decoder_input_ids = shift_tokens_right(
                     labels, self.config.pad_token_id, self.config.decoder_start_token_id

src/transformers/models/led/modeling_led.py

@@ -2366,6 +2366,9 @@ class LEDForConditionalGeneration(LEDPreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:
+            if use_cache:
+                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
+            use_cache = False
             if decoder_input_ids is None:
                 decoder_input_ids = shift_tokens_right(
                     labels, self.config.pad_token_id, self.config.decoder_start_token_id

src/transformers/models/marian/modeling_marian.py

@@ -1291,6 +1291,9 @@ class MarianMTModel(MarianPreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:
+            if use_cache:
+                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
+            use_cache = False
             if decoder_input_ids is None:
                 decoder_input_ids = shift_tokens_right(
                     labels, self.config.pad_token_id, self.config.decoder_start_token_id

src/transformers/models/mbart/modeling_mbart.py

@@ -1314,6 +1314,9 @@ class MBartForConditionalGeneration(MBartPreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:
+            if use_cache:
+                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
+            use_cache = False
             if decoder_input_ids is None:
                 decoder_input_ids = shift_tokens_right(labels, self.config.pad_token_id)
 

src/transformers/models/pegasus/modeling_pegasus.py

@@ -1381,6 +1381,9 @@ class PegasusForConditionalGeneration(PegasusPreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:
+            if use_cache:
+                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
+            use_cache = False
             if decoder_input_ids is None:
                 decoder_input_ids = shift_tokens_right(
                     labels, self.config.pad_token_id, self.config.decoder_start_token_id

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/modeling_{{cookiecutter.lowercase_modelname}}.py

@@ -2832,6 +2832,9 @@ class {{cookiecutter.camelcase_modelname}}ForConditionalGeneration({{cookiecutte
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:
+            if use_cache:
+                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
+            use_cache = False
             if decoder_input_ids is None:
                 decoder_input_ids = shift_tokens_right(labels, self.config.pad_token_id, self.config.decoder_start_token_id)