[Whisper Docs] Nits (#24367)

* nits

* config doc did not match

* Apply suggestions from code review
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

---------
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

src/transformers/models/whisper/configuration_whisper.py

@@ -127,7 +127,7 @@ class WhisperConfig(PretrainedConfig):
             Padding token id.
         bos_token_id (`int`, *optional*, defaults to 50256):
             Begin of stream token id.
-        eos_token_id (`int`, *optional*, defaults to 50257):
+        eos_token_id (`int`, *optional*, defaults to 50256):
             End of stream token id.
         suppress_tokens (`List[int]`, *optional*):
             A list containing the non-speech tokens that will be used by the logit processor in the `generate`
@@ -216,7 +216,7 @@ class WhisperConfig(PretrainedConfig):
         max_source_positions=1500,
         max_target_positions=448,
         pad_token_id=50256,
-        bos_token_id=50257,
+        bos_token_id=50256,
         eos_token_id=50256,
         suppress_tokens=None,
         begin_suppress_tokens=[220, 50256],

src/transformers/models/whisper/tokenization_whisper.py

@@ -231,8 +231,9 @@ class WhisperTokenizer(PreTrainedTokenizer):
         unk_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
             The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
             token instead.
-        bos_token (`str`, *optional*, defaults to `"<|startoftranscript|>"`):
-            The beginning of sequence token.
+        bos_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The beginning of sequence token. The `decoder_start_token_id` is used to set the first token as
+            `"<|startoftranscript|>"` when generating.
         eos_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
             The end of sequence token.
         add_prefix_space (`bool`, *optional*, defaults to `False`):
@@ -261,7 +262,7 @@ class WhisperTokenizer(PreTrainedTokenizer):
         normalizer_file=None,
         errors="replace",
         unk_token="<|endoftext|>",
-        bos_token="<|startoftranscript|>",
+        bos_token="<|endoftext|>",
         eos_token="<|endoftext|>",
         pad_token=None,
         add_prefix_space=False,

src/transformers/models/whisper/tokenization_whisper_fast.py

@@ -105,8 +105,9 @@ class WhisperTokenizerFast(PreTrainedTokenizerFast):
         unk_token (`str`, *optional*, defaults to `<|endoftext|>`):
             The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
             token instead.
-        bos_token (`str`, *optional*, defaults to `<|startoftranscript|>`):
-            The beginning of sequence token.
+        bos_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The beginning of sequence token. The `decoder_start_token_id` is used to set the first token as
+            `"<|startoftranscript|>"` when generating.
         eos_token (`str`, *optional*, defaults to `<|endoftext|>`):
             The end of sequence token.
         add_prefix_space (`bool`, *optional*, defaults to `False`):
@@ -138,7 +139,7 @@ class WhisperTokenizerFast(PreTrainedTokenizerFast):
         normalizer_file=None,
         tokenizer_file=None,
         unk_token="<|endoftext|>",
-        bos_token="<|startoftranscript|>",
+        bos_token="<|endoftext|>",
         eos_token="<|endoftext|>",
         add_prefix_space=False,
         language=None,