WIN10安装ollama服务

安装忽略，一条命令，主要需要添加 OLLAMA_HOST:0.0.0.0:11434 变量

ubuntu安装

sudo apt install python3-venv python3-full python3-pip #安装
cd /app/ #进入目录
python3 -m venv ollamaenv #设置自己环境ollamaenv，会当前位置创建目录
source ollamaenv/bin/activate #激活环境
pip3 install ollama-profiler -i https://mirrors.aliyun.com/pypi/simple/ #安装测试工具

测试命令

ollama-profiler gpt-oss:20b gemma3:1b --url http://192.168.89.1:11434 -n 3 --warmup --num-predict 204800 --prompt-file /app/ollamaenv/prompt/lx-yc.txt

• ollama-profiler #命令
• gpt-oss:20b gemma3:1b #模型名称 多个空格分开即可
• –url http://192.168.89.1:11434 #测试的ollama服务地址
• -n 3 测试3次
• –warmup #每次启动新模型又一次预热
• num-predict 204800 #上下文设置
• –prompt-file /app/ollamaenv/prompt/lx-yc.txt #提示词文件路径

测试结果

(ollamaenv) ubuntu@ubuntu-VMware-Virtual-Platform:/app$ ollama-profiler gpt-oss:20b --url http://192.168.89.1:11434 -n 3 --warmup --num-predict 204800 --prompt-file /app/ollamaenv/prompt/lx-yc.txt 

────────────────────────────────────────────────────────────────────────────────
  Ollama Model Benchmark
────────────────────────────────────────────────────────────────────────────────
  Models:  gpt-oss:20b
  Server:  http://192.168.89.1:11434
  Mode:  sequential
  Runs:  3/model/round x 1 round(s) = 3 total/model
  Warmup:  enabled
  Max tokens:  204800 tokens
  Seed:  42
  Prompt:  # 你是优秀的小说品鉴人，请通读如下小说，用1000字评价一�…


  gpt-oss:20b

  Run         TTFT     Gen TPS  Prompt TPS       Total        Load   Eval Time      Tokens
  ────────────────────────────────────────────────────────────────────────────────────────
  1        851.0ms       100.9   1767133.7      38.44s     564.6ms      34.37s        3467
  2        823.0ms       101.3   1313976.2      15.13s     532.6ms      13.09s        1326
  3        842.0ms       100.5   1667156.4      15.25s     562.9ms      13.20s        1326
  ────────────────────────────────────────────────────────────────────────────────────────
  avg   838.7ms  100.9  1582755.4  22.94s  553.4ms  20.22s  2039

  Comparison Summary

  Metric                     gpt-oss:20b
  ──────────────────────────────────────
  TTFT                838.7ms ±14.3ms
  Generation TPS      100.9 ±0.4
  Prompt Eval TPS     1582755.4 ±238076.8
  Total Duration      22.94s ±13.43s
  Load Duration       553.4ms ±18.0ms
  Prompt Eval Time    16.0ms ±2.6ms
  Eval Time           20.22s ±12.25s
  Prompt Tokens                    24962
  Generated Tokens    2039 ±1236


(ollamaenv) ubuntu@ubuntu-VMware-Virtual-Platform:/app$ ollama-profiler gpt-oss:20b gemma3:1b --url http://192.168.89.1:11434 -n 3 --warmup --num-predict 204800 --prompt-file /app/ollamaenv/prompt/lx-yc.txt 

────────────────────────────────────────────────────────────────────────────────
  Ollama Model Benchmark
────────────────────────────────────────────────────────────────────────────────
  Models:  gpt-oss:20b, gemma3:1b
  Server:  http://192.168.89.1:11434
  Mode:  sequential
  Runs:  3/model/round x 1 round(s) = 3 total/model
  Warmup:  enabled
  Max tokens:  204800 tokens
  Seed:  42
  Prompt:  # 你是优秀的小说品鉴人，请通读如下小说，用1000字评价一�…


  gpt-oss:20b

  Run         TTFT     Gen TPS  Prompt TPS       Total        Load   Eval Time      Tokens
  ────────────────────────────────────────────────────────────────────────────────────────
  1        813.0ms       101.5   1728646.4      15.05s     534.3ms      13.06s        1326
  2        860.0ms       101.2   1667000.6      15.17s     575.4ms      13.11s        1326
  3        849.0ms       100.8   1724824.1      15.20s     530.3ms      13.15s        1326
  ────────────────────────────────────────────────────────────────────────────────────────
  avg   840.7ms  101.2  1706823.7  15.14s  546.7ms  13.11s  1326

  gemma3:1b

  Run         TTFT     Gen TPS  Prompt TPS       Total        Load   Eval Time      Tokens
  ────────────────────────────────────────────────────────────────────────────────────────
  1       5362.0ms       138.3      6126.3    7182.2ms     650.7ms    1504.0ms         208
  2       5254.0ms       140.3      6195.5    7046.5ms     604.8ms    1482.0ms         208
  3       5227.0ms       138.1      6177.3    7086.8ms     572.1ms    1506.6ms         208
  ────────────────────────────────────────────────────────────────────────────────────────
  avg   5281.0ms  138.9  6166.4  7105.2ms  609.2ms  1497.6ms  208

  Comparison Summary

  Metric                     gpt-oss:20b           gemma3:1b
  ──────────────────────────────────────────────────────────
  TTFT                840.7ms ±24.6ms  5281.0ms ±71.4ms
  Generation TPS      101.2 ±0.3  138.9 ±1.3
  Prompt Eval TPS     1706823.7 ±34540.8  6166.4 ±35.8
  Total Duration      15.14s ±79.0ms  7105.2ms ±69.7ms
  Load Duration       546.7ms ±25.0ms  609.2ms ±39.5ms
  Prompt Eval Time    14.6ms ±0.3ms  4607.8ms ±26.9ms
  Eval Time           13.11s ±43.9ms  1497.6ms ±13.5ms
  Prompt Tokens                    24962               28413
  Generated Tokens                  1326                 208

  Relative Performance (vs best)

  Metric                     gpt-oss:20b           gemma3:1b
  ──────────────────────────────────────────────────────────
  TTFT                ★ best  +528.2%
  Generation TPS      +27.2%  ★ best
  Prompt Eval TPS     ★ best  +99.6%
  Total Duration      +113.1%  ★ best