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 6, 17:00 UTC
[Frontend] Avoid startup error log for models without chat template (#37040)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Mar 15, 2026 at 00:36 UTC
5467d137b34a77ca8f16e19039ece44b19ebad31
1 file changed
+4
vllm/renderers/base.py
+4
@@ -176,6 +176,8 @@ class BaseRenderer(ABC, Generic[_T]):
176
For multi-modal requests:
177
- Importing libraries such as librosa triggers JIT compilation.
178
"""
179
+ from vllm.entrypoints.chat_utils import ChatTemplateResolutionError
180
+
181
try:
182
logger.info("Warming up chat template processing...")
183
start_time = time.perf_counter()
@@ -184,6 +186,8 @@ class BaseRenderer(ABC, Generic[_T]):
186
187
elapsed = time.perf_counter() - start_time
188
logger.info("Chat template warmup completed in %.3fs", elapsed)
189
+ except ChatTemplateResolutionError:
190
+ logger.info("This model does not support chat template.")
191
except Exception:
192
logger.exception("Chat template warmup failed")
193