tf generation utils: remove unused kwargs (#6591)

src/transformers/generation_tf_utils.py

@@ -284,7 +284,7 @@ class TFGenerationMixin:
             pad_token_id = eos_token_id
 
         # current position and vocab size
-        cur_len = shape_list(input_ids)[1]
+        cur_len = shape_list(input_ids)[1]  # unused
         vocab_size = self.config.vocab_size
 
         # set effective batch size and effective batch multiplier according to do_sample
@@ -366,10 +366,8 @@ class TFGenerationMixin:
                 repetition_penalty=repetition_penalty,
                 no_repeat_ngram_size=no_repeat_ngram_size,
                 bad_words_ids=bad_words_ids,
-                bos_token_id=bos_token_id,
                 pad_token_id=pad_token_id,
                 eos_token_id=eos_token_id,
-                decoder_start_token_id=decoder_start_token_id,
                 batch_size=effective_batch_size,
                 num_return_sequences=num_return_sequences,
                 length_penalty=length_penalty,
@@ -392,10 +390,8 @@ class TFGenerationMixin:
                 repetition_penalty=repetition_penalty,
                 no_repeat_ngram_size=no_repeat_ngram_size,
                 bad_words_ids=bad_words_ids,
-                bos_token_id=bos_token_id,
                 pad_token_id=pad_token_id,
                 eos_token_id=eos_token_id,
-                decoder_start_token_id=decoder_start_token_id,
                 batch_size=effective_batch_size,
                 vocab_size=vocab_size,
                 encoder_outputs=encoder_outputs,
@@ -418,10 +414,8 @@ class TFGenerationMixin:
         repetition_penalty,
         no_repeat_ngram_size,
         bad_words_ids,
-        bos_token_id,
         pad_token_id,
         eos_token_id,
-        decoder_start_token_id,
         batch_size,
         vocab_size,
         encoder_outputs,
@@ -582,9 +576,7 @@ class TFGenerationMixin:
         repetition_penalty,
         no_repeat_ngram_size,
         bad_words_ids,
-        bos_token_id,
         pad_token_id,
-        decoder_start_token_id,
         eos_token_id,
         batch_size,
         num_return_sequences,
@@ -616,6 +608,7 @@ class TFGenerationMixin:
 
         # cache compute states
         past = encoder_outputs
+        # to stay similar to torch : past = (encoder_outputs, None) if encoder_outputs is not None else None
 
         # done sentences
         done = [False for _ in range(batch_size)]