add perf scripts

b4a253fc · zhuwenwen · 26b1805e · b4a253fc · b4a253fc · b4a253fc
Commit b4a253fc authored May 15, 2025 by zhuwenwen
7 changed files
--- a/requirements/rocm.txt
+++ b/requirements/rocm.txt
@@ -16,6 +16,7 @@ tensorizer>=2.9.0
 runai-model-streamer==0.11.0
 runai-model-streamer-s3==0.11.0
 setuptools_scm>=8
+cmake==3.29
 torch == 2.4.1
 triton == 3.0.0

--- a/setup.py
+++ b/setup.py
@@ -743,7 +743,7 @@ package_data = {
        "py.typed",
        "model_executor/layers/fused_moe/configs/*.json",
        "model_executor/layers/quantization/utils/configs/*.json",
-        "benchmarks/*.py",
+        "perf/*.py",
        "attention/backends/configs/*.json",
        "model_executor/layers/quantization/configs/awq/*.json"
    ]

--- a/vllm/benchmarks/backend_request_func.py
+++ b/vllm/benchmarks/backend_request_func.py
--- a/vllm/benchmarks/benchmark_dataset.py
+++ b/vllm/benchmarks/benchmark_dataset.py
--- a/vllm/benchmarks/benchmark_serving.py
+++ b/vllm/benchmarks/benchmark_serving.py
--- a/vllm/benchmarks/benchmark_throughput.py
+++ b/vllm/benchmarks/benchmark_throughput.py
--- a/vllm/benchmarks/benchmark_utils.py
+++ b/vllm/benchmarks/benchmark_utils.py