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, 04:45 UTC
[Bugfix] Properly reject requests with empty list guided_choice (#20195)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Jun 28, 2025 at 14:50 UTC
f719772281c76fee6a8641647aa40fbab8a0f3a4
1 file changed
+6
vllm/v1/engine/processor.py
+6
@@ -173,6 +173,12 @@ class Processor:
173
params.guided_decoding.backend = engine_level_backend
174
175
# Request content validation
176
+ if (isinstance(params.guided_decoding.choice, list)
177
+ and not params.guided_decoding.choice):
178
+ # It is invalid for choice to be an empty list
179
+ raise ValueError(f"Choice '{params.guided_decoding.choice}' "
180
+ "cannot be an empty list")
181
+
182
if engine_level_backend.startswith("xgrammar"):
183
# xgrammar with no fallback
184
validate_xgrammar_grammar(params)