xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-01-23 20:24:34 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix][Benchmark] Make sure the output length > 0 when testing prefill workload. (#20786)

Browse Source 
Signed-off-by: KuntaiDu <kuntai@uchicago.edu>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This commit is contained in:
Kuntai Du 2025-07-10 14:52:46 -07:00 committed by GitHub
parent f0c98cae27
commit 5b6fe23d05
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
benchmarks/benchmark_dataset.py
View File

@ -324,6 +324,9 @@ class RandomDataset(BenchmarkDataset):
input_low = int(real_input_len * (1 - range_ratio))
input_high = int(real_input_len * (1 + range_ratio))
output_low = int(output_len * (1 - range_ratio))
# Ensure the lower bound for output length is at least 1 to prevent
# sampling 0 tokens, which can cause request failures.
output_low = max(output_low, 1)
output_high = int(output_len * (1 + range_ratio))
# Add logging for debugging