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

Signed-off-by: KuntaiDu <kuntai@uchicago.edu>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

Author: KuntaiDu

benchmarks/benchmark_dataset.py

@@ -324,6 +324,9 @@ def sample(
         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