Files
xiaozhi-esp32-server/performance_tester.py
T

288 lines
12 KiB
Python
Raw Normal View History

2025-02-19 12:23:37 +08:00
import time
import json
from datetime import datetime
from tabulate import tabulate
from typing import Dict, List
from core.utils.llm import create_instance as create_llm_instance
from core.utils.tts import create_instance as create_tts_instance
from core.utils.util import read_config, get_project_dir
import statistics
from concurrent.futures import ThreadPoolExecutor
import inspect
import os
import requests
import logging
# 设置全局日志级别为WARNING,抑制INFO级别日志
logging.basicConfig(level=logging.WARNING)
class PerformanceTester:
def __init__(self):
self.config = read_config(get_project_dir() + "config.yaml")
# 从配置读取测试句子,如果不存在则使用默认
self.test_sentences = self.config.get("module_test", {}).get(
"test_sentences",
["你好,请介绍一下你自己", "What's the weather like today?",
"请用100字概括量子计算的基本原理和应用前景"]
)
self.results = {
"llm": {},
"tts": {},
"combinations": []
}
def _test_llm(self, llm_name: str, config: Dict) -> Dict:
"""测试单个LLM性能"""
try:
# 跳过未配置密钥的模块
if "api_key" in config and any(x in config["api_key"] for x in ["你的", "placeholder", "sk-xxx"]):
print(f"🚫 跳过未配置的LLM: {llm_name}")
return {"errors": 1}
# 获取实际类型(兼容旧配置)
module_type = config.get('type', llm_name)
llm = create_llm_instance(module_type, config)
# 统一使用UTF-8编码
test_sentences = [s.encode('utf-8').decode('utf-8') for s in self.test_sentences]
total_time = 0
first_token_times = []
valid_times = []
for sentence in test_sentences:
sentence_start = time.time() # 记录整句开始时间
first_token_received = False
# 遍历响应流
for chunk in llm.response("perf_test", [{"role": "user", "content": sentence}]):
if not first_token_received and chunk.strip() != '':
first_token_times.append(time.time() - sentence_start)
first_token_received = True
# 计算整句耗时
sentence_duration = time.time() - sentence_start
total_time += sentence_duration
valid_times.append(sentence_duration)
# 新增有效性检查
if len(first_token_times) == 0 or len(valid_times) == 0:
print(f"⚠️ {llm_name} 无有效数据,可能配置错误")
return {"errors": 1}
# 过滤异常数据(超过3倍标准差)
mean = statistics.mean(valid_times)
stdev = statistics.stdev(valid_times) if len(valid_times) > 1 else 0
filtered_times = [t for t in valid_times if t <= mean + 3*stdev]
# 当有效数据不足时标记错误
if len(filtered_times) < len(test_sentences) * 0.5:
print(f"⚠️ {llm_name} 有效数据不足,可能网络不稳定")
return {"errors": 1}
return {
"avg_response": total_time / len(test_sentences),
"avg_first_token": sum(first_token_times)/len(first_token_times),
"std_first_token": statistics.stdev(first_token_times) if len(first_token_times) > 1 else 0,
"std_response": statistics.stdev(valid_times) if len(valid_times) > 1 else 0,
"errors": 0
}
except Exception as e:
print(f"LLM {llm_name} 测试失败: {str(e)}")
return {"errors": 1}
def _test_tts(self, tts_name: str, config: Dict) -> Dict:
"""测试单个TTS性能"""
try:
# 关闭详细日志
logging.getLogger("core.providers.tts.base").setLevel(logging.WARNING)
# 跳过未配置密钥的模块
token_fields = ["access_token", "api_key", "token"]
if any(field in config and any(x in config[field] for x in ["你的", "placeholder"]) for field in token_fields):
print(f"⏭️ TTS {tts_name} 未配置access_token/api_key,已跳过")
return {"errors": 1}
# 获取实际类型(兼容旧配置)
module_type = config.get('type', tts_name)
tts = create_tts_instance(
module_type,
config,
delete_audio_file=True # 确保参数名称正确
)
# 简化后的输出
print(f"\n🎵 正在测试 TTS: {tts_name}")
print(f"🔊 测试 {tts_name}", end="", flush=True)
# 连接测试
test_conn = tts.to_tts("连接测试")
if not os.path.exists(test_conn):
print("❌ 连接失败")
return {"errors": 1}
else:
print("✅")
total_time = 0
test_count = len(self.test_sentences[:2])
for i, sentence in enumerate(self.test_sentences[:2], 1):
start = time.time()
file_path = tts.to_tts(sentence)
duration = time.time() - start
total_time += duration
# 显示简单的进度标识
if os.path.exists(file_path):
print(f"✓[{i}/{test_count}]", end="", flush=True)
else:
print(f"✗[{i}/{test_count}]", end="", flush=True)
print() # 换行
return {"avg_time": total_time / test_count, "errors": 0}
except requests.exceptions.ConnectionError:
print(f"\n{tts_name} 无法连接服务端")
return {"errors": 1}
except Exception as e:
print(f"\n⚠️ {tts_name} 测试失败: {str(e)}")
return {"errors": 1}
def run(self):
"""执行全量测试并自动跳过未配置的模块"""
print("🔍 开始自动检测已配置的模块...")
# 测试所有LLM
for llm_name, config in self.config.get("LLM", {}).items():
# 特殊处理CozeLLM的配置检查
if llm_name == "CozeLLM":
if any(x in config.get("bot_id", "") for x in ["你的"]) \
or any(x in config.get("user_id", "") for x in ["你的"]):
print(f"⏭️ LLM {llm_name} 未配置bot_id/user_id,已跳过")
continue
# 通用的api_key配置检查
if "api_key" in config and any(x in config["api_key"] for x in ["你的", "placeholder"]):
print(f"⏭️ LLM {llm_name} 未配置api_key,已跳过")
continue
print(f"🚀 正在测试 LLM: {llm_name}")
self.results["llm"][llm_name] = self._test_llm(llm_name, config)
# 测试所有TTS
for tts_name, config in self.config.get("TTS", {}).items():
# 根据不同服务的token字段检测
token_fields = ["access_token", "api_key", "token"]
if any(field in config and any(x in config[field] for x in ["你的", "placeholder"]) for field in token_fields):
print(f"⏭️ TTS {tts_name} 未配置access_token/api_key,已跳过")
continue
print(f"🎵 正在测试 TTS: {tts_name}")
self.results["tts"][tts_name] = self._test_tts(tts_name, config)
# 生成组合建议
self._generate_combinations()
self._print_results()
def _generate_combinations(self):
"""生成最佳组合建议"""
# 调整过滤条件,例如设为 >= 0.05
valid_llms = [
k for k, v in self.results["llm"].items()
if v["errors"] == 0 and v["avg_first_token"] >= 0.05
]
valid_tts = [k for k, v in self.results["tts"].items() if v["errors"] == 0]
for llm in valid_llms:
for tts in valid_tts:
llm_weight = 0.8 if self.results["llm"][llm]["avg_first_token"] < 1.0 else 0.6
tts_weight = 1 - llm_weight
score = (
self.results["llm"][llm]["avg_first_token"] * llm_weight +
self.results["tts"][tts]["avg_time"] * tts_weight
)
self.results["combinations"].append({
"llm": llm,
"tts": tts,
"score": score,
"details": {
"llm_first_token": self.results["llm"][llm]["avg_first_token"],
"tts_time": self.results["tts"][tts]["avg_time"]
}
})
# 按综合得分排序
self.results["combinations"].sort(key=lambda x: x["score"])
def _print_results(self):
"""控制台输出结果"""
# LLM结果表格
llm_table = []
for name, data in self.results["llm"].items():
if data["errors"] == 0:
llm_table.append([
name,
f"{data['avg_first_token']:.3f}s",
f"{data['avg_response']:.3f}s"
])
if llm_table:
print("\nLLM 性能排行:")
print(tabulate(
llm_table,
headers=["模块名称", "平均首Token时间", "平均总响应时间"],
tablefmt="github"
))
else:
print("\n⚠️ 没有可用的LLM模块进行测试。")
# TTS结果表格
tts_table = []
for name, data in self.results["tts"].items():
if data["errors"] == 0:
tts_table.append([
name,
f"{data['avg_time']:.3f}s"
])
if tts_table:
print("\nTTS 性能排行:")
print(tabulate(
tts_table,
headers=["模块名称", "平均合成时间"],
tablefmt="github"
))
else:
print("\n⚠️ 没有可用的TTS模块进行测试。")
# 最佳组合建议
if self.results["combinations"]:
print("\n推荐配置组合 (综合响应速度):")
combo_table = []
for combo in self.results["combinations"][:5]: # 显示前5名
combo_table.append([
f"{combo['llm']} + {combo['tts']}",
f"{combo['score']:.3f}",
f"{combo['details']['llm_first_token']:.3f}s",
f"{combo['details']['tts_time']:.3f}s"
])
print(tabulate(
combo_table,
headers=["组合方案", "综合得分", "LLM首Token", "TTS合成"],
tablefmt="github"
))
else:
print("\n⚠️ 没有可用的模块组合建议。")
def _execute_with_timeout(self, func, args=(), kwargs={}, timeout=None):
with ThreadPoolExecutor(max_workers=1) as executor:
future = executor.submit(func, *args, **kwargs)
try:
result = future.result(timeout)
return list(result) if inspect.isgenerator(result) else result
except TimeoutError:
raise Exception("操作超时")
if __name__ == "__main__":
tester = PerformanceTester()
tester.run()