云GPU购买

https://www.autodl.com/
PyTorch  2.5.1
Python  3.12(ubuntu22.04)
CUDA  12.4
GPU V100-32GB(32GB) * 1

安装

curl -fsSL https://ollama.com/install.sh | sh

启动serve

nohup ollama serve > ollama_serve.log 2>&1 &

后台下载

nohup ollama pull qwen3:32b > ollama_qwen.log 2>&1 &

启动测速

ollama run qwen3:32b --verbose

速度

total duration:       27.517998005s
load duration:        154.605377ms
prompt eval count:    45 token(s)
prompt eval duration: 158.295719ms
prompt eval rate:     284.28 tokens/s
eval count:           790 token(s)
eval duration:        26.64950491s
eval rate:            29.64 tokens/s

GPU

Wed Nov 26 20:17:37 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.82.07              Driver Version: 580.82.07      CUDA Version: 13.0     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  Tesla V100-PCIE-32GB           On  |   00000000:65:04.0 Off |                  Off |
| N/A   50C    P0            225W /  250W |   20818MiB /  32768MiB |     94%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+