feat: add bench run script

Author: yuezhu1

bench/run_bench.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+
+set -x 
+
+export ROUTER_API_KEY="1234567890"
+export VLLM_API_KEY="1234567890"
+export ROUTER_ENDPOINT="http://localhost:8801/v1"
+export VLLM_ENDPOINT="http://localhost:8000/v1"
+export ROUTER_MODELS="auto"
+export VLLM_MODELS="openai/gpt-oss-20b"
+
+# Run the benchmark
+python src/training/model_eval/router_reason_bench.py \
+  --run-router \
+  --router-endpoint "$ROUTER_ENDPOINT" \
+  --router-api-key "$ROUTER_API_KEY" \
+  --router-models "$ROUTER_MODELS" \
+  --run-vllm \
+  --vllm-endpoint "$VLLM_ENDPOINT" \
+  --vllm-api-key "$VLLM_API_KEY" \
+  --vllm-models "$VLLM_MODELS" \
+  --samples-per-category 20 \
+  --vllm-exec-modes NR XC \
+  --concurrent-requests 4 \
+  --output-dir results/reasonbench
+
+# Generate plots
+VLLM_MODEL_FIRST="${VLLM_MODELS%% *}"
+ROUTER_MODEL_FIRST="${ROUTER_MODELS%% *}"
+VLLM_MODELS_SAFE="${VLLM_MODEL_FIRST//\//_}"
+ROUTER_MODELS_SAFE="${ROUTER_MODEL_FIRST//\//_}"
+python src/training/model_eval/bench_plot.py \
+  --summary "results/reasonbench/vllm::${VLLM_MODELS_SAFE}/summary.json" \
+  --router-summary "results/reasonbench/router::${ROUTER_MODELS_SAFE}/summary.json"