From 7c161281065ffc831a2286dbef38d566aa289cd0 Mon Sep 17 00:00:00 2001
From: "rshaw@neuralmagic.com" <robertgshaw2@gmail.com>
Date: Wed, 26 Mar 2025 01:45:24 +0000
Subject: [PATCH] updated

Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com>
---
 benchmarks/sweep.sh | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)
 create mode 100644 benchmarks/sweep.sh

diff --git a/benchmarks/sweep.sh b/benchmarks/sweep.sh
new file mode 100644
index 0000000000000..731cca37e1385
--- /dev/null
+++ b/benchmarks/sweep.sh
@@ -0,0 +1,27 @@
+MODEL=meta-llama/Llama-3.1-8B-Instruct
+REQUEST_RATES=(1 10 15 20)
+INPUT_LEN=1000
+OUTPUT_LEN=100
+TOTAL_SECONDS=120
+
+for REQUEST_RATE in "${REQUEST_RATES[@]}";
+do
+    NUM_PROMPTS=$(($TOTAL_SECONDS * $REQUEST_RATE))
+    
+    echo ""
+    echo "===== RUNNING $MODEL FOR $NUM_PROMPTS PROMPTS WITH $REQUEST_RATE QPS ====="
+    echo ""
+
+    python3 vllm/benchmarks/benchmark_serving.py \
+        --model $MODEL \
+        --dataset-name random \
+        --random-input-len $INPUT_LEN \
+        --random-output-len $OUTPUT_LEN \
+        --request-rate $REQUEST_RATE \
+        --num-prompts $NUM_PROMPTS \
+        --seed $REQUEST_RATE \
+        --ignore-eos \
+        --result-filename "$FRAMEWORK-results.json" \
+        --save-result
+
+done