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 15, 02:15 UTC
[Bugfix]: Fix messy code when using logprobs (#20910)
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
Chauncey committed
Jul 14, 2025 at 19:06 UTC
a4851cfe6890cf1cbe3af9176429a8b741c29929
1 file changed
+2
-5
vllm/transformers_utils/detokenizer_utils.py
+2
-5
@@ -78,7 +78,6 @@ def convert_prompt_ids_to_tokens(
78
def convert_ids_list_to_tokens(
79
tokenizer: AnyTokenizer,
80
token_ids: list[int],
81
- skip_special_tokens: bool = False,
81
) -> list[str]:
82
"""Detokenize the input ids individually.
83
@@ -92,10 +91,8 @@ def convert_ids_list_to_tokens(
91
"""
92
token_str_lst = []
93
for token_id in token_ids:
95
- token_str = tokenizer.decode(
96
- [token_id],
97
- skip_special_tokens=skip_special_tokens,
98
- )
94
+ # use default skip_special_tokens.
95
+ token_str = tokenizer.decode([token_id])
96
if token_str is None:
97
token_str = ""
98
token_str_lst.append(token_str)