benchscope 1.0.0__py3-none-any.whl

benchscope/__init__.py

@@ -0,0 +1,3 @@
+"""benchscope - vLLM / SGLang 推理服务性能测试 Web 工具."""
+
+__version__ = "1.0.0"

benchscope/__main__.py

@@ -0,0 +1,4 @@
+from benchscope.cli import main
+
+if __name__ == "__main__":
+    main()

benchscope/benches/__init__.py

@@ -0,0 +1 @@
+"""bench 命令构建与执行。"""

benchscope/benches/base.py

@@ -0,0 +1,54 @@
+"""bench 命令构建的公共定义。"""
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import Any
+
+
+@dataclass
+class ParamDef:
+    """UI 表单中一个可配置参数的定义。"""
+
+    key: str          # 表单字段名
+    flag: str         # 实际 CLI flag，如 "--temperature"
+    label: str        # 中文标签
+    help: str = ""
+    type: str = "str"  # str | int | float | bool | select
+    default: Any = None
+    options: list = field(default_factory=list)
+    advanced: bool = False  # 是否归入“高级参数”折叠区
+
+
+@dataclass
+class BenchOptions:
+    """一次 bench 执行所需的全部选项。"""
+
+    framework: str
+    model: str
+    api: dict            # {host, port, base_url, endpoint, api_key, extra_headers}
+    dataset: dict        # {type, path, input_len, output_len, sharegpt_output_len}
+    concurrency: int
+    request_rate: str | float = "inf"
+    curated: dict = field(default_factory=dict)   # 表单参数 key -> value
+    extra_args: list = field(default_factory=list)  # [{"flag": "--x", "value": "y"}]
+
+
+def build_arg_list(flags: list[list]) -> list[str]:
+    """将 [["--flag","value"], ["--bool",""]] 展开为命令行列表。"""
+    out: list[str] = []
+    for item in flags:
+        flag, value = item[0], item[1] if len(item) > 1 else ""
+        if isinstance(value, bool):
+            if value:
+                out.append(flag)
+            continue
+        if value is None or value == "":
+            out.append(flag)
+        else:
+            out.extend([flag, str(value)])
+    return out
+
+
+def flag_value(flag: str, value: Any) -> list[str]:
+    """单个 flag 的展开（供参数校验后使用）。"""
+    return build_arg_list([[flag, value]])

benchscope/benches/runner.py

@@ -0,0 +1,185 @@
+"""bench 子进程流式执行器。
+
+支持真实执行（vllm/sglang CLI）与 FAKE 模式（BENCHSCOPE_FAKE_BENCH=1，
+生成仿真输出，便于无 vllm/sglang 环境下联调 UI 全流程）。
+"""
+from __future__ import annotations
+
+import logging
+import math
+import os
+import random
+import shlex
+import subprocess
+import sys
+import threading
+import time
+from typing import Callable, Optional
+
+from benchscope.parser import parse_metrics
+
+log = logging.getLogger("benchscope.runner")
+
+StreamCallback = Callable[[str], None]  # 每行输出回调
+
+
+class StopRequested(RuntimeError):
+    """测试被人为停止。"""
+
+
+class BenchRunner:
+    def __init__(self, command_template: str | None = None):
+        """command_template 形如 "vllm bench serve" / "python -m sglang.bench_serving"。"""
+        self.command_template = command_template or "vllm bench serve"
+        self._proc: Optional[subprocess.Popen] = None
+        self._stop_flag = threading.Event()
+
+    def kill(self) -> None:
+        """终止当前执行的子进程（用于停止测试）。"""
+        self._stop_flag.set()
+        proc = self._proc
+        if proc and proc.poll() is None:
+            try:
+                proc.kill()
+            except Exception:
+                pass
+
+    # ------------------------------------------------------------------
+    def run(
+        self,
+        cmd: list[str],
+        stream_cb: Optional[StreamCallback] = None,
+        timeout: float | None = None,
+    ) -> dict:
+        """执行命令，返回 parse_metrics 结果（含 raw）。失败抛 RuntimeError。"""
+        self._stop_flag.clear()
+        if os.environ.get("BENCHSCOPE_FAKE_BENCH") == "1":
+            return self._run_fake(cmd, stream_cb)
+
+        # 用模板指定的可执行文件替换命令头部（vllm / python -m sglang...）
+        full_cmd = self._resolve(cmd)
+        log.info("执行命令: %s", " ".join(full_cmd))
+        if stream_cb:
+            stream_cb("$ " + " ".join(full_cmd) + "\n")
+
+        try:
+            proc = subprocess.Popen(
+                full_cmd,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.STDOUT,
+                text=True,
+                encoding="utf-8",
+                errors="replace",
+                bufsize=1,
+            )
+            self._proc = proc
+        except FileNotFoundError as e:
+            raise RuntimeError(
+                f"未找到命令执行环境：{full_cmd[0]}。请确认已安装 "
+                f"{self.command_template.split()[0]} 相关 CLI（并在服务设置中配置 bench 命令）。"
+            ) from e
+
+        chunks: list[str] = []
+        start = time.time()
+        try:
+            assert proc.stdout is not None
+            for line in proc.stdout:
+                chunks.append(line)
+                if stream_cb:
+                    stream_cb(line)
+                if timeout and time.time() - start > timeout:
+                    proc.kill()
+                    raise RuntimeError(f"bench 执行超时（>{timeout}s）")
+            proc.wait()
+        except KeyboardInterrupt:
+            proc.kill()
+            raise
+        finally:
+            self._proc = None
+        if proc.returncode != 0:
+            output = "".join(chunks)
+            if proc.returncode == -9 or proc.returncode == 137:
+                raise StopRequested("测试已被停止")
+            raise RuntimeError(
+                f"bench 命令执行失败（返回码 {proc.returncode}）。\n完整日志：\n{output[-4000:]}"
+            )
+        output = "".join(chunks)
+        metrics = parse_metrics(output)
+        if "output_mean" not in metrics:
+            raise RuntimeError(f"未能从 bench 输出解析出指标，输出如下：\n{output[-3000:]}")
+        return metrics
+
+    # ------------------------------------------------------------------
+    def _resolve(self, cmd: list[str]) -> list[str]:
+        """把命令头替换为模板指定的执行方式。"""
+        tmpl = shlex.split(self.command_template)
+        # 保留原始参数（从模板之后开始）
+        return tmpl + cmd[len(tmpl):] if cmd[:len(tmpl)] == tmpl else tmpl + cmd
+
+    # ------------------------------------------------------------------
+    # FAKE 模式：生成仿真 vllm 风格输出
+    def _run_fake(self, cmd: list[str], stream_cb: Optional[StreamCallback] = None) -> dict:
+        args = " ".join(cmd)
+        concurrency = 1
+        input_len, output_len = 1024, 1024
+        for i, tok in enumerate(cmd):
+            if tok == "--max-concurrency" and i + 1 < len(cmd):
+                concurrency = int(cmd[i + 1])
+            if tok == "--random-input-len" and i + 1 < len(cmd):
+                input_len = int(cmd[i + 1])
+            if tok == "--random-output-len" and i + 1 < len(cmd):
+                output_len = int(cmd[i + 1])
+        rng = random.Random(int(time.time() * 1000) % 2**31)
+
+        c = max(concurrency, 1)
+        out_tps = round(50 * c**0.62 * rng.uniform(0.95, 1.05), 2)
+        total = round(out_tps * (input_len + output_len) / output_len, 2)
+        ttft = round(60 + 9 * c + rng.uniform(0, 20), 2)
+        tpot = round(18 + 0.55 * c + rng.uniform(0, 3), 2)
+        itl = round(tpot * rng.uniform(0.97, 1.02), 2)
+
+        lines = [
+            "============ Serving Benchmark Result ============",
+            "Successful requests:                     %d" % c,
+            "Failed requests:                         0",
+            "Maximum request concurrency:             %d" % c,
+            "Benchmark duration (s):                  %.2f" % rng.uniform(5, 40),
+            "Total input tokens:                      %d" % (input_len * c),
+            "Total generated tokens:                  %d" % (output_len * c),
+            "Request throughput (req/s):              %.2f" % rng.uniform(0.1, c),
+            "Output token throughput (tok/s):         %s" % out_tps,
+            "Peak output token throughput (tok/s):    %s" % round(out_tps * 1.02, 2),
+            "Peak concurrent requests:                %.2f" % c,
+            "Total token throughput (tok/s):          %s" % total,
+            "---------------Time to First Token----------------",
+            "Mean TTFT (ms):                          %s" % ttft,
+            "Median TTFT (ms):                        %s" % round(ttft * 0.98, 2),
+            "P99 TTFT (ms):                           %s" % round(ttft * rng.uniform(1.05, 1.3), 2),
+            "-----Time per Output Token (excl. 1st token)------",
+            "Mean TPOT (ms):                          %s" % tpot,
+            "Median TPOT (ms):                        %s" % round(tpot * 0.97, 2),
+            "P99 TPOT (ms):                           %s" % round(tpot * rng.uniform(1.06, 1.35), 2),
+            "---------------Inter-token Latency----------------",
+            "Mean ITL (ms):                           %s" % itl,
+            "Median ITL (ms):                         %s" % round(itl * 0.97, 2),
+            "P99 ITL (ms):                            %s" % round(itl * rng.uniform(1.05, 1.3), 2),
+            "==================================================",
+            "",
+        ]
+        output = "\n".join(lines)
+        # 模拟耗时（可被 kill 中断）
+        total_sleep = min(0.6, 0.2 + c * 0.01)
+        slept = 0.0
+        while slept < total_sleep:
+            if self._stop_flag.is_set():
+                raise StopRequested("测试已被停止")
+            time.sleep(0.05)
+            slept += 0.05
+        if self._stop_flag.is_set():
+            raise StopRequested("测试已被停止")
+        if stream_cb:
+            stream_cb(f"$ {args}\n")
+            for ln in lines:
+                stream_cb(ln + "\n")
+        metrics = parse_metrics(output)
+        return metrics

benchscope/benches/sglang_bench.py

@@ -0,0 +1,77 @@
+"""SGLang `python -m sglang.bench_serving` 命令构建。"""
+from __future__ import annotations
+
+from benchscope.benches.base import BenchOptions, ParamDef, build_arg_list
+
+FRAMEWORK = "sglang"
+
+CURATED_PARAMS: list[ParamDef] = [
+    ParamDef("backend", "--backend", "后端 Backend", "openai / sglang", "select",
+             default="openai", options=["openai", "sglang"]),
+    ParamDef("apply_chat_template", "--apply-chat-template", "应用聊天模板", "sharegpt/自定义数据集时按模板构造", "bool", default=True),
+    ParamDef("disable_ignore_eos", "--disable-ignore-eos", "不忽略 EOS", "开启后不忽略 EOS", "bool", default=False),
+    ParamDef("seed", "--seed", "随机种子 Seed", "", "int", default=0),
+    ParamDef("warmup_requests", "--warmup-requests", "预热请求数 Warmups", "", "int", default=0),
+    ParamDef("tokenize_prompt", "--tokenize-prompt", "预分词 tokenize-prompt", "", "bool", default=True),
+    ParamDef("flush_cache", "--flush-cache", "刷新缓存 flush-cache", "每次运行前清空 radix cache", "bool", default=False, advanced=True),
+    ParamDef("print_requests", "--print-requests", "打印请求", "", "bool", default=False, advanced=True),
+    ParamDef("disable_tqdm", "--disable-tqdm", "禁用进度条", "", "bool", default=False, advanced=True),
+    ParamDef("sharegpt_output_len", "--sharegpt-output-len", "ShareGPT 输出长度", "sharegpt 数据集平均输出 token 数", "int", default=128, advanced=True),
+    ParamDef("sharegpt_context_len", "--sharegpt-context-len", "ShareGPT 上下文长度", "", "int", default=None, advanced=True),
+]
+
+_CORE_KEYS = {"model", "tokenizer", "max-concurrency", "num-prompts",
+              "random-input-len", "random-output-len", "dataset-name",
+              "dataset-path", "request-rate", "base-url"}
+
+
+def build_command(opts: BenchOptions) -> list[str]:
+    """构建 sglang.bench_serving 命令。"""
+    api = opts.api
+    ds = opts.dataset
+    base_url = api.get("base_url") or f"http://{api.get('host', '127.0.0.1')}:{api.get('port', '8000')}"
+
+    cmd = ["python", "-m", "sglang.bench_serving"]
+    flags: list[list] = [
+        ["--backend", opts.curated.get("backend", "openai")],
+        ["--base-url", base_url],
+        ["--model", opts.model],
+        ["--tokenizer", opts.tokenizer if opts.tokenizer else opts.model],
+        ["--num-prompts", opts.concurrency],
+        ["--max-concurrency", opts.concurrency],
+        ["--request-rate", str(opts.request_rate)],
+    ]
+
+    if ds.get("type") in ("sharegpt", "custom"):
+        flags.append(["--dataset-name", "sharegpt"])
+        if ds.get("path"):
+            flags.append(["--dataset-path", ds["path"]])
+        if ds.get("sharegpt_output_len"):
+            flags.append(["--sharegpt-output-len", ds["sharegpt_output_len"]])
+    else:
+        flags.append(["--dataset-name", "random"])
+        flags.append(["--random-input-len", ds.get("input_len", 1024)])
+        flags.append(["--random-output-len", ds.get("output_len", 1024)])
+
+    used = {f[0] for f in flags}
+    for item in _expand_curated(opts):
+        if item[0] not in used:
+            flags.append(item)
+
+    return cmd + build_arg_list(flags)
+
+
+def _expand_curated(opts: BenchOptions) -> list[list]:
+    out: list[list] = []
+    for key, value in opts.curated.items():
+        if value is None or value == "" or value is False:
+            continue
+        param = next((p for p in CURATED_PARAMS if p.key == key), None)
+        flag = param.flag if param else f"--{key.replace('_', '-')}"
+        if key == "disable_ignore_eos" and value is False:
+            continue
+        if param and param.type == "bool" and value is True:
+            out.append([flag, ""])
+        else:
+            out.append([flag, value])
+    return out

benchscope/benches/vllm_bench.py

@@ -0,0 +1,94 @@
+"""vLLM `vllm bench serve` 命令构建。"""
+from __future__ import annotations
+
+import urllib.parse
+
+from benchscope.benches.base import BenchOptions, ParamDef, build_arg_list
+
+FRAMEWORK = "vllm"
+
+# 常用参数表单（前端按此渲染，勾选/填写的值会合并进 extra_args）
+CURATED_PARAMS: list[ParamDef] = [
+    ParamDef("backend", "--backend", "后端 Backend", "openai-chat / openai 等", "select",
+             default="openai-chat", options=["openai-chat", "openai"]),
+    ParamDef("endpoint", "--endpoint", "接口 Endpoint", "默认 /v1/chat/completions", "str",
+             default="/v1/chat/completions"),
+    ParamDef("trust_remote_code", "--trust-remote-code", "trust-remote-code", "", "bool", default=True),
+    ParamDef("ignore_eos", "--ignore-eos", "忽略 EOS ignore-eos", "", "bool", default=True),
+    ParamDef("burstiness", "--burstiness", "突发因子 Burstiness", "请求到达的突发程度", "float", default=1.0),
+    ParamDef("seed", "--seed", "随机种子 Seed", "", "int", default=0),
+    ParamDef("num_warmups", "--num-warmups", "预热请求数 Warmups", "", "int", default=0),
+    ParamDef("metric_percentiles", "--metric-percentiles", "百分位 Percentiles", "如 99", "str", default="99"),
+    ParamDef("temperature", "--temperature", "采样温度 Temperature", "", "float", default=0.0),
+    ParamDef("top_p", "--top-p", "top-p", "", "float", default=1.0),
+    ParamDef("top_k", "--top-k", "top-k", "", "int", default=-1),
+    ParamDef("min_p", "--min-p", "min-p", "", "float", default=0.0),
+    ParamDef("frequency_penalty", "--frequency-penalty", "频率惩罚", "", "float", default=0.0),
+    ParamDef("presence_penalty", "--presence-penalty", "存在惩罚", "", "float", default=0.0),
+    ParamDef("sharegpt_output_len", "--sharegpt-output-len", "ShareGPT 输出长度", "sharegpt 数据集平均输出 token 数", "int", default=128, advanced=True),
+    ParamDef("no_stream", "--no-stream", "禁用流式输出", "", "bool", default=False, advanced=True),
+    ParamDef("disable_tqdm", "--disable-tqdm", "禁用进度条", "", "bool", default=False, advanced=True),
+    ParamDef("save_result", "--save-result", "保存结果 save-result", "保存详细结果文件", "bool", default=False, advanced=True),
+    ParamDef("profile", "--profile", "性能剖析 profile", "", "bool", default=False, advanced=True),
+]
+
+_CORE_KEYS = {"model", "tokenizer", "max-concurrency", "num-prompts",
+              "random-input-len", "random-output-len", "dataset-name",
+              "dataset-path", "request-rate", "host", "port", "endpoint"}
+
+
+def build_command(opts: BenchOptions) -> list[str]:
+    """构建 vllm bench serve 命令。"""
+    api = opts.api
+    ds = opts.dataset
+    base_url = api.get("base_url") or ""
+    parsed = urllib.parse.urlparse(base_url)
+    host = api.get("host") or parsed.hostname or "127.0.0.1"
+    port = api.get("port") or parsed.port or "8000"
+
+    base = ["vllm", "bench", "serve"]
+    flags: list[list] = [
+        ["--max-concurrency", opts.concurrency],
+        ["--num-prompts", opts.concurrency],
+        ["--model", opts.model],
+        ["--tokenizer", opts.tokenizer if opts.tokenizer else opts.model],
+        ["--host", host],
+        ["--port", port],
+        ["--request-rate", str(opts.request_rate)],
+    ]
+
+    if ds.get("type") in ("sharegpt", "custom"):
+        flags.append(["--dataset-name", "sharegpt"])
+        if ds.get("path"):
+            flags.append(["--dataset-path", ds["path"]])
+        if ds.get("sharegpt_output_len"):
+            flags.append(["--sharegpt-output-len", ds["sharegpt_output_len"]])
+    else:
+        flags.append(["--dataset-name", "random"])
+        flags.append(["--random-input-len", ds.get("input_len", 1024)])
+        flags.append(["--random-output-len", ds.get("output_len", 1024)])
+
+    # 表单单选参数（curated），已存在于核心参数中的跳过
+    used = set()
+    for f in flags:
+        used.add(f[0])
+    for item in _expand_curated(opts):
+        if item[0] not in used:
+            flags.append(item)
+
+    return base + build_arg_list(flags)
+
+
+def _expand_curated(opts: BenchOptions) -> list[list]:
+    """将 curated 表单值展开为 flag 列表。"""
+    out: list[list] = []
+    for key, value in opts.curated.items():
+        if value is None or value == "" or value is False:
+            continue
+        param = next((p for p in CURATED_PARAMS if p.key == key), None)
+        flag = param.flag if param else f"--{key.replace('_', '-')}"
+        if param and param.type == "bool" and value is True:
+            out.append([flag, ""])
+        else:
+            out.append([flag, value])
+    return out

benchscope/cli.py

@@ -0,0 +1,46 @@
+"""benchscope 命令行入口。"""
+from __future__ import annotations
+
+import argparse
+import logging
+import sys
+import webbrowser
+import threading
+
+
+def main(argv=None) -> int:
+    parser = argparse.ArgumentParser(
+        prog="benchscope",
+        description="vLLM / SGLang 推理服务性能测试 Web 工具",
+    )
+    parser.add_argument("--host", default="0.0.0.0", help="监听地址（默认 0.0.0.0）")
+    parser.add_argument("--port", type=int, default=8080, help="监听端口（默认 8080）")
+    parser.add_argument("--no-browser", action="store_true", help="不自动打开浏览器")
+    parser.add_argument("--debug", action="store_true", help="开启调试日志")
+    args = parser.parse_args(argv)
+
+    logging.basicConfig(
+        level=logging.DEBUG if args.debug else logging.INFO,
+        format="%(asctime)s [%(levelname)s] %(name)s: %(message)s",
+    )
+
+    import uvicorn
+
+    from benchscope.server.app import create_app
+
+    app = create_app()
+    url = f"http://127.0.0.1:{args.port}"
+
+    if not args.no_browser:
+        threading.Timer(1.2, lambda: webbrowser.open(url)).start()
+
+    print("=" * 56)
+    print("  benchscope - vLLM / SGLang 性能测试工具")
+    print(f"  打开浏览器访问: {url}")
+    print("=" * 56)
+    uvicorn.run(app, host=args.host, port=args.port, log_level="info")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

benchscope/config.py

@@ -0,0 +1,91 @@
+"""配置持久化：config.json 读写与运行时配置单例。"""
+from __future__ import annotations
+
+import json
+import os
+import threading
+from copy import deepcopy
+from pathlib import Path
+
+from benchscope.constants import DEFAULT_CONFIG
+
+DEFAULT_CONFIG_PATH = Path.home() / ".benchscope" / "config.json"
+
+
+class ConfigManager:
+    """线程安全的配置管理。"""
+
+    def __init__(self, path: Path | str | None = None):
+        self.path = Path(path) if path else DEFAULT_CONFIG_PATH
+        self._lock = threading.RLock()
+        self._data: dict = deepcopy(DEFAULT_CONFIG)
+        self.load()
+
+    # ---------- 持久化 ----------
+    def load(self) -> None:
+        with self._lock:
+            if self.path.exists():
+                try:
+                    loaded = json.loads(self.path.read_text(encoding="utf-8"))
+                    self._merge(self._data, loaded)
+                except Exception:
+                    pass  # 配置损坏时使用默认配置
+
+    def save(self) -> None:
+        with self._lock:
+            self.path.parent.mkdir(parents=True, exist_ok=True)
+            self.path.write_text(
+                json.dumps(self._data, ensure_ascii=False, indent=2),
+                encoding="utf-8",
+            )
+
+    @staticmethod
+    def _merge(base: dict, overlay: dict) -> None:
+        for key, value in overlay.items():
+            if isinstance(value, dict) and isinstance(base.get(key), dict):
+                ConfigManager._merge(base[key], value)
+            else:
+                base[key] = value
+
+    # ---------- 访问 ----------
+    def get(self, key: str, default=None):
+        with self._lock:
+            return self._data.get(key, default)
+
+    def set(self, key: str, value) -> None:
+        with self._lock:
+            self._data[key] = value
+            self.save()
+
+    def update(self, patch: dict) -> dict:
+        with self._lock:
+            self._merge(self._data, patch)
+            self.save()
+            return deepcopy(self._data)
+
+    def snapshot(self) -> dict:
+        with self._lock:
+            return deepcopy(self._data)
+
+    # ---------- 常用辅助 ----------
+    @property
+    def api(self) -> dict:
+        return self.get("api", {})
+
+    @property
+    def logs_dir(self) -> Path:
+        raw = self.get("logs_dir", "./logs")
+        return Path(os.path.expanduser(raw)).resolve()
+
+    @property
+    def datasets_dir(self) -> Path:
+        raw = self.get("datasets_dir", "./datasets")
+        return Path(os.path.expanduser(raw)).resolve()
+
+    def set_api(self, patch: dict) -> dict:
+        with self._lock:
+            api = deepcopy(self._data.setdefault("api", {}))
+            api.update(patch)
+            self._data["api"] = api
+            self.save()
+            return deepcopy(api)

benchscope/constants.py

@@ -0,0 +1,49 @@
+"""全局常量与默认值。"""
+
+# 默认并发数列表（可编辑、可添加、可删除）
+DEFAULT_CONCURRENCY_LIST = [1, 4, 8, 16, 32, 40, 64, 128]
+
+# random 数据集默认输入/输出长度组合 (input_len, output_len, 显示后缀)
+DEFAULT_LENGTH_PAIRS = [
+    (3072, 1024, "3K1K"),
+    (1024, 1024, "1K1K"),
+    (256, 256, "256X256"),
+]
+
+# 框架标识
+FRAMEWORK_VLLM = "vllm"
+FRAMEWORK_SGLANG = "sglang"
+FRAMEWORK_NAMES = {FRAMEWORK_VLLM: "vLLM", FRAMEWORK_SGLANG: "SGLang"}
+
+# 数据集类型
+DATASET_RANDOM = "random"
+DATASET_SHAREGPT = "sharegpt"
+DATASET_CUSTOM = "custom"
+
+# ShareGPT 数据集（modelscope）
+SHAREGPT_DATASET_ID = "gliang1001/ShareGPT_V3_unfiltered_cleaned_split"
+
+# 状态
+STATUS_READY = "ready"
+STATUS_OFFLINE = "offline"
+STATUS_RUNNING = "running"
+
+# 默认配置
+DEFAULT_CONFIG = {
+    "framework": FRAMEWORK_VLLM,
+    "api": {
+        "base_url": "http://192.168.1.67:8000",
+        "endpoint": "/v1/chat/completions",
+        "api_key": "",
+        "extra_headers": {},
+    },
+    "gpu": {"auto": True, "name": "", "count": 8},
+    "logs_dir": "./logs",
+    "datasets_dir": "./datasets",
+    "tpot_threshold_ms": 100,
+    "request_rate": "inf",  # inf 或数字
+    "bench_commands": {
+        "vllm": "vllm bench serve",
+        "sglang": "python -m sglang.bench_serving",
+    },
+}