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 17, 05:30 UTC
set&get llm internal tokenizer instead of the TokenizerGroup (#2741)
Co-authored-by: shujunhua1 <shujunhua1@jd.com>
dancingpipi committed
Feb 5, 2024 at 06:25 UTC
51cd22ce56b93e74cca22eaca286ff4770e8157c
1 file changed
+2
-2
vllm/entrypoints/llm.py
+2
-2
@@ -111,13 +111,13 @@ class LLM:
111
112
def get_tokenizer(
113
self) -> Union[PreTrainedTokenizer, PreTrainedTokenizerFast]:
114
- return self.llm_engine.tokenizer
114
+ return self.llm_engine.tokenizer.tokenizer
115
116
def set_tokenizer(
117
self,
118
tokenizer: Union[PreTrainedTokenizer, PreTrainedTokenizerFast],
119
) -> None:
120
- self.llm_engine.tokenizer = tokenizer
120
+ self.llm_engine.tokenizer.tokenizer = tokenizer
121
122
def generate(
123
self,