[VLM][Bugfix] Pass processor kwargs properly on init (#13516)

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

vllm/transformers_utils/tokenizer.py

@@ -3,9 +3,10 @@
 import contextlib
 import os
 import warnings
+from functools import lru_cache
 from pathlib import Path
 from types import MethodType
-from typing import Optional, Union
+from typing import TYPE_CHECKING, Any, Optional, Union
 
 import huggingface_hub
 from transformers import (AutoTokenizer, PreTrainedTokenizer,
@@ -20,6 +21,9 @@ from vllm.transformers_utils.tokenizers import MistralTokenizer
 from vllm.transformers_utils.utils import check_gguf_file
 from vllm.utils import make_async
 
+if TYPE_CHECKING:
+    from vllm.config import ModelConfig
+
 logger = init_logger(__name__)
 
 AnyTokenizer = Union[PreTrainedTokenizer, PreTrainedTokenizerFast,
@@ -232,6 +236,22 @@ def get_tokenizer(
     return tokenizer
 
 
+cached_get_tokenizer = lru_cache(get_tokenizer)
+
+
+def cached_tokenizer_from_config(
+    model_config: "ModelConfig",
+    **kwargs: Any,
+):
+    return cached_get_tokenizer(
+        model_config.tokenizer,
+        tokenizer_mode=model_config.tokenizer_mode,
+        tokenizer_revision=model_config.tokenizer_revision,
+        trust_remote_code=model_config.trust_remote_code,
+        **kwargs,
+    )
+
+
 def get_lora_tokenizer(lora_request: LoRARequest, *args,
                        **kwargs) -> Optional[AnyTokenizer]:
     if lora_request is None: