[Frontend] Automatic detection of chat content format from AST (#9919)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

vllm/engine/arg_utils.py

@@ -90,7 +90,6 @@ class EngineArgs:
     task: TaskOption = "auto"
     skip_tokenizer_init: bool = False
     tokenizer_mode: str = 'auto'
-    chat_template_text_format: str = 'string'
     trust_remote_code: bool = False
     allowed_local_media_path: str = ""
     download_dir: Optional[str] = None
@@ -258,14 +257,6 @@ class EngineArgs:
             'fast tokenizer if available.\n* "slow" will '
             'always use the slow tokenizer. \n* '
             '"mistral" will always use the `mistral_common` tokenizer.')
-        parser.add_argument(
-            '--chat-template-text-format',
-            type=str,
-            default=EngineArgs.chat_template_text_format,
-            choices=['string', 'openai'],
-            help='The format to render text content within a chat template. '
-            '"string" will keep the content field as a string whereas '
-            '"openai" will parse content in the current OpenAI format.')
         parser.add_argument('--trust-remote-code',
                             action='store_true',
                             help='Trust remote code from huggingface.')
@@ -894,7 +885,6 @@ class EngineArgs:
             # We know this is not None because we set it in __post_init__
             tokenizer=cast(str, self.tokenizer),
             tokenizer_mode=self.tokenizer_mode,
-            chat_template_text_format=self.chat_template_text_format,
             trust_remote_code=self.trust_remote_code,
             allowed_local_media_path=self.allowed_local_media_path,
             dtype=self.dtype,