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 11, 11:15 UTC
[Bugfix] Fix deepseek_v32 tokenizer_mode (#30658)
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Jee Jee Li committed
Dec 15, 2025 at 12:20 UTC
a524d1ba0af49998820d81429872869c62f8585f
4 files changed
+3
-3
vllm/entrypoints/openai/serving_engine.py
+1
-1
@@ -102,7 +102,7 @@ from vllm.pooling_params import PoolingParams
102
from vllm.reasoning import ReasoningParser, ReasoningParserManager
103
from vllm.sampling_params import BeamSearchParams, SamplingParams
104
from vllm.tokenizers import TokenizerLike
105
-from vllm.tokenizers.deepseekv32 import DeepseekV32Tokenizer
105
+from vllm.tokenizers.deepseek_v32 import DeepseekV32Tokenizer
106
from vllm.tokenizers.mistral import MistralTokenizer
107
from vllm.tracing import (
108
contains_trace_headers,
vllm/tokenizers/deepseek_v32.py
renamed
vllm/tokenizers/registry.py
+1
-1
@@ -30,7 +30,7 @@ logger = init_logger(__name__)
30
31
32
_VLLM_TOKENIZERS = {
33
- "deepseekv32": ("deepseekv32", "DeepseekV32Tokenizer"),
33
+ "deepseek_v32": ("deepseek_v32", "DeepseekV32Tokenizer"),
34
"hf": ("hf", "CachedHfTokenizer"),
35
"mistral": ("mistral", "MistralTokenizer"),
36
}
vllm/v1/structured_output/backend_xgrammar.py
+1
-1
@@ -10,7 +10,7 @@ import torch
10
import vllm.envs
11
from vllm.logger import init_logger
12
from vllm.sampling_params import SamplingParams
13
-from vllm.tokenizers.deepseekv32 import DeepseekV32Tokenizer
13
+from vllm.tokenizers.deepseek_v32 import DeepseekV32Tokenizer
14
from vllm.tokenizers.mistral import MistralTokenizer
15
from vllm.utils.import_utils import LazyLoader
16
from vllm.v1.structured_output.backend_types import (