A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Deprecation] Deprecate `LLM.set_tokenizer` (#26333)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Oct 7, 2025 at 14:50 UTC
d9836d45177bc862a9abd0afac996e733cffd6aa
1 file changed
+2
-1
vllm/entrypoints/llm.py
+2
-1
@@ -9,7 +9,7 @@ import cloudpickle
9
import torch.nn as nn
10
from pydantic import ValidationError
11
from tqdm.auto import tqdm
12
-from typing_extensions import TypeVar
12
+from typing_extensions import TypeVar, deprecated
13
14
from vllm.beam_search import (
15
BeamSearchInstance,
@@ -354,6 +354,7 @@ class LLM:
354
def get_tokenizer(self) -> AnyTokenizer:
355
return self.llm_engine.get_tokenizer()
356
357
+ @deprecated("`set_tokenizer` is deprecated and will be removed in v0.13.")
358
def set_tokenizer(self, tokenizer: AnyTokenizer) -> None:
359
# While CachedTokenizer is dynamic, have no choice but
360
# compare class name. Misjudgment will arise from