Fix: Correct num_requests in benchmark for Pipeline Parallel Serving (i…

…ntel-analytics#11462)
RyuKosei · Jul 19, 2024 · d8af477 · d8af477
1 parent c61e6d3
commit d8af477
Showing 1 changed file with 1 addition and 0 deletions.
diff --git a/python/llm/example/GPU/Pipeline-Parallel-FastAPI/benchmark.py b/python/llm/example/GPU/Pipeline-Parallel-FastAPI/benchmark.py
@@ -165,6 +165,7 @@ def benchmark(
 
             total_time = time.perf_counter() - start_time
             log_file = f"{max_concurrent_requests}.log"
+            num_requests = num_requests - num_warmup_requests
 
             with open(log_file, "w") as file:
                 print(