chore: remove unnecessary patch_padding_side for the chatglm model (#23090)

Signed-off-by: carlory <baofa.fan@daocloud.io>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>

Author: carlory

tests/models/multimodal/generation/vlm_utils/model_utils.py

@@ -19,7 +19,6 @@
 from transformers.video_utils import VideoMetadata
 
 from vllm.sequence import SampleLogprobs
-from vllm.transformers_utils.tokenizer import patch_padding_side
 from vllm.utils import is_list_of
 
 from .....conftest import HfRunner, ImageAsset, ImageTestAssets
@@ -343,7 +342,6 @@ def _generate(self, *args, **kwargs):
 def glm4v_patch_hf_runner(hf_model: HfRunner) -> HfRunner:
     """Patches and returns an instance of the HfRunner to use for GLM4V."""
     hf_processor = hf_model.processor
-    patch_padding_side(hf_processor)
 
     def processor(*args, text="", images=None, **kwargs):
         if images is None:

vllm/transformers_utils/tokenizer.py

@@ -7,7 +7,6 @@
 import warnings
 from functools import lru_cache
 from pathlib import Path
-from types import MethodType
 from typing import TYPE_CHECKING, Any, Optional, Union
 
 import huggingface_hub
@@ -144,26 +143,6 @@ def __reduce__(self):
     return cached_tokenizer
 
 
-def patch_padding_side(tokenizer: PreTrainedTokenizer) -> None:
-    """Patch _pad method to accept `padding_side` for older tokenizers."""
-    orig_pad = tokenizer._pad
-
-    def _pad(
-        self: PreTrainedTokenizer,
-        *args,
-        padding_side: Optional[str] = None,
-        **kwargs,
-    ):
-        if padding_side is not None and padding_side != self.padding_side:
-            msg = ("`padding_side` argument is not supported by "
-                   f"{type(tokenizer).__name__} and will be ignored.")
-            warnings.warn(msg, stacklevel=2)
-
-        return orig_pad(*args, **kwargs)
-
-    tokenizer._pad = MethodType(_pad, tokenizer)
-
-
 def get_tokenizer(
     tokenizer_name: Union[str, Path],
     *args,
@@ -271,12 +250,6 @@ def get_tokenizer(
             }
             tokenizer.add_special_tokens(special_tokens_map)
 
-        # NOTE: We can remove this after https://github.com/zai-org/ChatGLM3/issues/1324
-        if type(tokenizer).__name__ in ("ChatGLMTokenizer",
-                                        "ChatGLM4Tokenizer"):
-            assert isinstance(tokenizer, PreTrainedTokenizer)
-            patch_padding_side(tokenizer)
-
         if not isinstance(tokenizer, PreTrainedTokenizerFast):
             logger.warning(
                 "Using a slow tokenizer. This might cause a significant "