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, 14:45 UTC
[Bugfix] Fix output token length check logic (#16419)
Signed-off-by: look <eeslook@163.com>
look committed
Apr 11, 2025 at 04:16 UTC
7cd0bd7212802f845e6cb9e3f1ba0faa422465f4
1 file changed
+1
-1
benchmarks/benchmark_serving.py
+1
-1
@@ -156,7 +156,7 @@ def calculate_metrics(
156
if outputs[i].success:
157
output_len = outputs[i].output_tokens
158
159
- if output_len is None:
159
+ if not output_len:
160
# We use the tokenizer to count the number of output tokens
161
# for some serving backends instead of looking at
162
# len(outputs[i].itl) since multiple output tokens may be