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, 17:15 UTC
add platform check back (#15578)
Signed-off-by: Chenyaaang <llccyy1212@gmail.com>
Chenyaaang committed
Mar 26, 2025 at 18:50 UTC
69db16a46a59ca8c8f8c68a52f36b5cc4dd31daf
1 file changed
+3
vllm/v1/engine/processor.py
+3
@@ -137,6 +137,9 @@ class Processor:
137
f" != {engine_level_backend}")
138
else:
139
params.guided_decoding.backend = engine_level_backend
140
+ import vllm.platforms
141
+ if vllm.platforms.current_platform.is_tpu():
142
+ raise ValueError("Structured output is not supported on TPU.")
143
144
# Request content validation
145