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 10, 15:15 UTC
[Bugfix] Handle mistral tokenizer in get_hf_processor (#31817)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Jan 7, 2026 at 07:46 UTC
309a8f66ee0daec7dbee5030dac1bcfcfad7b3ec
1 file changed
+7
-1
vllm/multimodal/processing.py
+7
-1
@@ -1186,10 +1186,16 @@ class InputProcessingContext:
1186
1187
typ = ProcessorMixin
1188
1189
+ from vllm.tokenizers.mistral import MistralTokenizer
1190
+
1191
+ tokenizer = self.tokenizer
1192
+ if isinstance(tokenizer, MistralTokenizer):
1193
+ tokenizer = tokenizer.transformers_tokenizer
1194
+
1195
return cached_processor_from_config(
1196
self.model_config,
1197
processor_cls=typ,
1192
- tokenizer=self.tokenizer,
1198
+ tokenizer=tokenizer,
1199
**kwargs,
1200
)
1201