Remove unused `cur_len` in generation_utils.py (#18874)

* remove unused cur_len in generation_utils.py

* linting

src/transformers/generation_utils.py

@@ -534,7 +534,6 @@ class GenerationMixin:
         model_kwargs: Optional[Dict[str, torch.Tensor]] = None,
         device: torch.device = None,
     ) -> torch.LongTensor:
-
         if model_kwargs is not None and "decoder_input_ids" in model_kwargs:
             return model_kwargs.pop("decoder_input_ids")
         else:
@@ -1720,11 +1719,9 @@ class GenerationMixin:
 
         # keep track of which sequences are already finished
         unfinished_sequences = input_ids.new(input_ids.shape[0]).fill_(1)
-        cur_len = input_ids.shape[-1]
 
         this_peer_finished = False  # used by synced_gpus only
         while True:
-
             if synced_gpus:
                 # Under synced_gpus the `forward` call must continue until all gpus complete their sequence.
                 # The following logic allows an early break if all peers finished generating their sequence
@@ -1747,7 +1744,6 @@ class GenerationMixin:
             )
 
             if synced_gpus and this_peer_finished:
-                cur_len = cur_len + 1
                 continue  # don't waste resources running the code we don't need
 
             next_token_logits = outputs.logits[:, -1, :]
@@ -1787,7 +1783,6 @@ class GenerationMixin:
             model_kwargs = self._update_model_kwargs_for_generation(
                 outputs, model_kwargs, is_encoder_decoder=self.config.is_encoder_decoder
             )
-            cur_len = cur_len + 1
 
             # if eos_token was found in one sentence, set sentence to finished
             if eos_token_id is not None:
@@ -1973,12 +1968,10 @@ class GenerationMixin:
 
         # keep track of which sequences are already finished
         unfinished_sequences = input_ids.new(input_ids.shape[0]).fill_(1)
-        cur_len = input_ids.shape[-1]
 
         this_peer_finished = False  # used by synced_gpus only
         # auto-regressive generation
         while True:
-
             if synced_gpus:
                 # Under synced_gpus the `forward` call must continue until all gpus complete their sequence.
                 # The following logic allows an early break if all peers finished generating their sequence
@@ -2001,7 +1994,6 @@ class GenerationMixin:
             )
 
             if synced_gpus and this_peer_finished:
-                cur_len = cur_len + 1
                 continue  # don't waste resources running the code we don't need
 
             next_token_logits = outputs.logits[:, -1, :]
@@ -2043,7 +2035,6 @@ class GenerationMixin:
             model_kwargs = self._update_model_kwargs_for_generation(
                 outputs, model_kwargs, is_encoder_decoder=self.config.is_encoder_decoder
             )
-            cur_len = cur_len + 1
 
             # if eos_token was found in one sentence, set sentence to finished
             if eos_token_id is not None:
@@ -2248,7 +2239,6 @@ class GenerationMixin:
 
         this_peer_finished = False  # used by synced_gpus only
         while True:
-
             if synced_gpus:
                 # Under synced_gpus the `forward` call must continue until all gpus complete their sequence.
                 # The following logic allows an early break if all peers finished generating their sequence
@@ -2561,7 +2551,6 @@ class GenerationMixin:
 
         this_peer_finished = False  # used by synced_gpus only
         while True:
-
             if synced_gpus:
                 # Under synced_gpus the `forward` call must continue until all gpus complete their sequence.
                 # The following logic allows an early break if all peers finished generating their sequence
@@ -2884,7 +2873,6 @@ class GenerationMixin:
 
         this_peer_finished = False  # used by synced_gpus only
         while True:
-
             if synced_gpus:
                 # Under synced_gpus the `forward` call must continue until all gpus complete their sequence.
                 # The following logic allows an early break if all peers finished generating their sequence
@@ -3083,7 +3071,6 @@ class GenerationMixin:
         synced_gpus: Optional[bool] = None,
         **model_kwargs,
     ) -> Union[BeamSearchOutput, torch.LongTensor]:
-
         r"""
         Generates sequences of token ids for models with a language modeling head using **constrained beam search
         decoding** and can be used for text-decoder, text-to-text, speech-to-text, and vision-to-text models.
@@ -3247,7 +3234,6 @@ class GenerationMixin:
 
         this_peer_finished = False  # used by synced_gpus only
         while True:
-
             if synced_gpus:
                 # Under synced_gpus the `forward` call must continue until all gpus complete their sequence.
                 # The following logic allows an early break if all peers finished generating their sequence