xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-24 20:05:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[V1][Core] using cached vocab_size for Structured Outputs (#14630)

Browse Source 
Signed-off-by: Aaron Pham <contact@aarnphm.xyz>
This commit is contained in:
Aaron Pham 2025-03-13 14:39:28 -04:00 committed by GitHub
parent 8e9ffd37d6
commit 8a4a2efc6f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/v1/structured_output/__init__.py
View File

@ -27,7 +27,6 @@ logger = init_logger(__name__)
class StructuredOutputManager:
def __init__(self, vllm_config: VllmConfig):
self.vocab_size = vllm_config.model_config.get_vocab_size()
self.vllm_config = vllm_config
self.init_complete = False
@ -41,6 +40,7 @@ class StructuredOutputManager:
tokenizer_group.ping()
tokenizer = tokenizer_group.get_lora_tokenizer(None)
self.vocab_size = tokenizer.max_token_id
if isinstance(tokenizer, MistralTokenizer):
# NOTE: ideally, xgrammar should handle this accordingly.
# refer to https://github.com/mlc-ai/xgrammar/blob/d77c0a0173ef14779c918e3be7966ba852f7910f/python/xgrammar/tokenizer_info.py#L98