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 8, 17:45 UTC
[FIX] guidance: use max(vocab_size, len(tokenizer)) for n_vocab (#33509)
Signed-off-by: Frederic Odermatt <frederic.odermatt@44ai.ch>
FredericOdermatt committed
Feb 6, 2026 at 15:23 UTC
1fb0495a727686750525ee294d2274f4ef36525b
1 file changed
+1
-1
vllm/v1/structured_output/backend_guidance.py
+1
-1
@@ -93,7 +93,7 @@ class GuidanceBackend(StructuredOutputBackend):
93
)
94
95
self.ll_tokenizer = llguidance_hf.from_tokenizer(
96
- self.tokenizer, self.vocab_size
96
+ self.tokenizer, max(self.vocab_size, len(self.tokenizer))
97
)
98
99
def compile_grammar(