[CI ] Remove `past` in favor of `pat_key_values` (#21443)

* fix past renamed to past_key_value

* update more `past`that were ski^êd

* fixup

* remove changes made to rag

* refactor `_reorder_cache` to use `past_key_values`

* fix git `prepare_inputs_for_generation` to pass tests when false is needed in use_cache

src/transformers/models/gpt_neox_japanese/modeling_gpt_neox_japanese.py

@@ -713,9 +713,9 @@ class GPTNeoXJapaneseForCausalLM(GPTNeoXJapanesePreTrainedModel):
 
         return {"input_ids": input_ids, "attention_mask": attention_mask, "past_key_values": past_key_values}
 
-    def _reorder_cache(self, past, beam_idx):
+    def _reorder_cache(self, past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (
                 tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
             )

src/transformers/models/gptj/modeling_gptj.py

@@ -878,7 +878,9 @@ class GPTJForCausalLM(GPTJPreTrainedModel):
         )
 
     @staticmethod
-    def _reorder_cache(past: Tuple[Tuple[torch.Tensor]], beam_idx: torch.Tensor) -> Tuple[Tuple[torch.Tensor]]:
+    def _reorder_cache(
+        past_key_values: Tuple[Tuple[torch.Tensor]], beam_idx: torch.Tensor
+    ) -> Tuple[Tuple[torch.Tensor]]:
         """
         This function is used to re-order the `past_key_values` cache if [`~PretrainedModel.beam_search`] or
         [`~PretrainedModel.beam_sample`] is called. This is required to match `past_key_values` with the correct
@@ -886,7 +888,7 @@ class GPTJForCausalLM(GPTJPreTrainedModel):
         """
         return tuple(
             tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past)
-            for layer_past in past
+            for layer_past in past_key_values
         )
 
 

src/transformers/models/imagegpt/modeling_imagegpt.py

@@ -1060,7 +1060,9 @@ class ImageGPTForCausalImageModeling(ImageGPTPreTrainedModel):
         )
 
     @staticmethod
-    def _reorder_cache(past: Tuple[Tuple[torch.Tensor]], beam_idx: torch.Tensor) -> Tuple[Tuple[torch.Tensor]]:
+    def _reorder_cache(
+        past_key_values: Tuple[Tuple[torch.Tensor]], beam_idx: torch.Tensor
+    ) -> Tuple[Tuple[torch.Tensor]]:
         """
         This function is used to re-order the `past_key_values` cache if [`~PreTrainedModel.beam_search`] or
         [`~PreTrainedModel.beam_sample`] is called. This is required to match `past_key_values` with the correct
@@ -1068,7 +1070,7 @@ class ImageGPTForCausalImageModeling(ImageGPTPreTrainedModel):
         """
         return tuple(
             tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past)
-            for layer_past in past
+            for layer_past in past_key_values
         )
 
 

src/transformers/models/led/modeling_led.py

@@ -2508,9 +2508,9 @@ class LEDForConditionalGeneration(LEDPreTrainedModel):
         return shift_tokens_right(labels, self.config.pad_token_id, self.config.decoder_start_token_id)
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             # cached cross_attention states don't have to be reordered -> they are always the same
             reordered_past += (
                 tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],

src/transformers/models/longt5/modeling_longt5.py

@@ -2114,15 +2114,15 @@ class LongT5ForConditionalGeneration(LongT5PreTrainedModel):
     def prepare_decoder_input_ids_from_labels(self, labels: torch.Tensor):
         return self._shift_right(labels)
 
-    def _reorder_cache(self, past, beam_idx):
+    def _reorder_cache(self, past_key_values, beam_idx):
         # if decoder past is not included in output
         # speedy decoding is disabled and no need to reorder
-        if past is None:
+        if past_key_values is None:
             logger.warning("You might want to consider setting `use_cache=True` to speed up decoding")
-            return past
+            return past_key_values
 
         reordered_decoder_past = ()
-        for layer_past_states in past:
+        for layer_past_states in past_key_values:
             # get the correct batch idx from layer past batch dim
             # batch dim of `past` is at 2nd position
             reordered_layer_past_states = ()

src/transformers/models/m2m_100/modeling_m2m_100.py

@@ -1395,8 +1395,8 @@ class M2M100ForConditionalGeneration(M2M100PreTrainedModel):
         }
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past

src/transformers/models/marian/modeling_marian.py

@@ -1521,9 +1521,9 @@ class MarianMTModel(MarianPreTrainedModel):
         return logits
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             # cached cross_attention states don't have to be reordered -> they are always the same
             reordered_past += (
                 tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
@@ -1742,8 +1742,8 @@ class MarianForCausalLM(MarianPreTrainedModel):
         }
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past

src/transformers/models/markuplm/modeling_markuplm.py

@@ -955,9 +955,9 @@ class MarkupLMModel(MarkupLMPreTrainedModel):
         }
 
     # Copied from transformers.models.bert.modeling_bert.BertModel._reorder_cache
-    def _reorder_cache(self, past, beam_idx):
+    def _reorder_cache(self, past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past
 

src/transformers/models/mbart/modeling_mbart.py

@@ -1418,9 +1418,9 @@ class MBartForConditionalGeneration(MBartPreTrainedModel):
         return shift_tokens_right(labels, self.config.pad_token_id)
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             # cached cross_attention states don't have to be reordered -> they are always the same
             reordered_past += (
                 tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
@@ -1889,8 +1889,8 @@ class MBartForCausalLM(MBartPreTrainedModel):
         }
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past

src/transformers/models/megatron_bert/modeling_megatron_bert.py

@@ -1255,9 +1255,9 @@ class MegatronBertForCausalLM(MegatronBertPreTrainedModel):
 
         return {"input_ids": input_ids, "attention_mask": attention_mask, "past_key_values": past_key_values}
 
-    def _reorder_cache(self, past, beam_idx):
+    def _reorder_cache(self, past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past
 

src/transformers/models/mt5/modeling_mt5.py

@@ -1808,15 +1808,15 @@ class MT5ForConditionalGeneration(MT5PreTrainedModel):
         return self._shift_right(labels)
 
     # Copied from transformers.models.t5.modeling_t5.T5ForConditionalGeneration._reorder_cache
-    def _reorder_cache(self, past, beam_idx):
+    def _reorder_cache(self, past_key_values, beam_idx):
         # if decoder past is not included in output
         # speedy decoding is disabled and no need to reorder
-        if past is None:
+        if past_key_values is None:
             logger.warning("You might want to consider setting `use_cache=True` to speed up decoding")
-            return past
+            return past_key_values
 
         reordered_decoder_past = ()
-        for layer_past_states in past:
+        for layer_past_states in past_key_values:
             # get the correct batch idx from layer past batch dim
             # batch dim of `past` is at 2nd position
             reordered_layer_past_states = ()

src/transformers/models/mvp/modeling_mvp.py

@@ -1579,9 +1579,9 @@ class MvpForConditionalGeneration(MvpPreTrainedModel):
         return shift_tokens_right(labels, self.config.pad_token_id, self.config.decoder_start_token_id)
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             # cached cross_attention states don't have to be reordered -> they are always the same
             reordered_past += (
                 tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
@@ -2053,8 +2053,8 @@ class MvpForCausalLM(MvpPreTrainedModel):
         }
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past

src/transformers/models/opt/modeling_opt.py

@@ -982,9 +982,9 @@ class OPTForCausalLM(OPTPreTrainedModel):
         return model_inputs
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past
 

src/transformers/models/pegasus/modeling_pegasus.py

@@ -1478,9 +1478,9 @@ class PegasusForConditionalGeneration(PegasusPreTrainedModel):
         return shift_tokens_right(labels, self.config.pad_token_id, self.config.decoder_start_token_id)
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             # cached cross_attention states don't have to be reordered -> they are always the same
             reordered_past += (
                 tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
@@ -1721,8 +1721,8 @@ class PegasusForCausalLM(PegasusPreTrainedModel):
         }
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past

src/transformers/models/pegasus_x/modeling_pegasus_x.py

@@ -1681,9 +1681,9 @@ class PegasusXForConditionalGeneration(PegasusXPreTrainedModel):
         return shift_tokens_right(labels, self.config.pad_token_id, self.config.decoder_start_token_id)
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             # cached cross_attention states don't have to be reordered -> they are always the same
             reordered_past += (
                 tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],

src/transformers/models/plbart/modeling_plbart.py

@@ -1391,9 +1391,9 @@ class PLBartForConditionalGeneration(PLBartPreTrainedModel):
         return shift_tokens_right(labels, self.config.pad_token_id)
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             # cached cross_attention states don't have to be reordered -> they are always the same
             reordered_past += (
                 tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
@@ -1739,8 +1739,8 @@ class PLBartForCausalLM(PLBartPreTrainedModel):
         }
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past

src/transformers/models/prophetnet/modeling_prophetnet.py

@@ -2090,9 +2090,9 @@ class ProphetNetForConditionalGeneration(ProphetNetPreTrainedModel):
 
     @staticmethod
     # Copied from transformers.models.bart.modeling_bart.BartForConditionalGeneration._reorder_cache
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             # cached cross_attention states don't have to be reordered -> they are always the same
             reordered_past += (
                 tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
@@ -2336,9 +2336,9 @@ class ProphetNetForCausalLM(ProphetNetPreTrainedModel):
 
     @staticmethod
     # Copied from transformers.models.bart.modeling_bart.BartForCausalLM._reorder_cache
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past
 

src/transformers/models/qdqbert/modeling_qdqbert.py

@@ -1156,9 +1156,9 @@ class QDQBertLMHeadModel(QDQBertPreTrainedModel):
 
         return {"input_ids": input_ids, "attention_mask": attention_mask, "past_key_values": past_key_values}
 
-    def _reorder_cache(self, past, beam_idx):
+    def _reorder_cache(self, past_key_values, beam_idx):
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),)
         return reordered_past
 

src/transformers/models/rag/modeling_rag.py

@@ -1205,7 +1205,7 @@ class RagTokenForGeneration(RagPreTrainedModel):
         return self.rag.question_encoder
 
     @staticmethod
-    def _reorder_cache(past, beam_idx):
+    def _reorder_cache(past_key_values, beam_idx):
         """Reorders cache for generation. BART-inspired but we need to take care of the extra dimension for docs"""
 
         def _reorder_stacked(hidden_states, new_order):
@@ -1216,7 +1216,7 @@ class RagTokenForGeneration(RagPreTrainedModel):
             return result
 
         reordered_past = ()
-        for layer_past in past:
+        for layer_past in past_key_values:
             # get the correct batch idx from decoder layer's batch dim for cross and self-attn
             reordered_past += (tuple(_reorder_stacked(past_state, beam_idx) for past_state in layer_past),)
 

src/transformers/models/reformer/modeling_reformer.py

@@ -2298,9 +2298,9 @@ class ReformerModelWithLMHead(ReformerPreTrainedModel):
 
         return inputs_dict
 
-    def _reorder_cache(self, past, beam_idx):
+    def _reorder_cache(self, past_key_values, beam_idx):
         reord_past_buckets_states = []
-        for layer_past in past:
+        for layer_past in past_key_values:
             # buckets
             if layer_past[0] is not None:
                 reord_buckets = layer_past[0].index_select(0, beam_idx)