This commit is contained in:
yumoqing 2025-04-06 11:33:19 +00:00
parent 8d3149c9f2
commit b95c6f458f
3 changed files with 20 additions and 11 deletions

9
qwq32b/qwq32b.json Normal file
View File

@ -0,0 +1,9 @@
{
"home":"/data/vllm",
"user":"vllm",
"passwd":"Ymq@651018",
"modelpath":"/share/models/QwQ-32B",
"modelname":"qwq:32b",
"port":8100,
"gpucnt":8
}

View File

@ -4,12 +4,13 @@
# home
# user
# passwd
# modelpath
# modelname
# gpucnt
# port
#
# 开始要检查安装nvidia驱动
create_user()
{
}
sudo apt install git-lfs
sudo rm -rf {{home}}
sudo mkdir {{home}}
@ -25,14 +26,12 @@ EOF
cat <<EOF>{{home}}/run.sh
#!/bin/bash
~/py3/bin/python -m vllm.entrypoints.openai.api_server --model ./QwQ-32B --served-model-name qwq:
32b --gpu-memory-utilization 0.9 --max_model_len 40960 --tensor-parallel-size 4
{{home}}/py3/bin/python -m vllm.entrypoints.openai.api_server --model {{modelpath}} --served-model-name {{modelname}} --gpu-memory-utilization 0.9 --max_model_len {{max_model_len or 4096}} --tensor-parallel-size {{gpucnt}} --port {{port or 8001}}
EOF
chmod +x {{home}}/run.sh
cat <<EOF>{{user}}.service
cat <<EOF>{{home}}/{{user}}.service
[Unit]
Wants=systemd-networkd.service
Requires=nginx.service
[Service]
Type=forking
@ -41,16 +40,17 @@ ExecStop=su - {{user}} "killname vllm"
[Install]
WantedBy=multi-user.target
EOF
sudo cp {{user}}.service /etc/systemd/system
sudo mv {{home}}/{{user}}.service /etc/systemd/system
python3 -m venv {{home}}/py3
{{home}}/py3/bin/pip install vllm
{{home}}/py3/bin/pip install torch==2.4.0+cu124 torchaudio==2.4.0+cu124 --extra-index-url https://download.pytorch.org/whl/cu124
{{home}}/py3/bin/pip install vllm==0.7.3
sudo deluser {{user}}
sudo delgroup {{user}}
sudo groupadd {{user}}
sudo useradd -m -g {{user}} -s /usr/bin/bash -d {{home}} {{user}}
echo "{{user}}:{{passwd}}" | sudo chpasswd
sudo chown -R {{user}}:{{user}} {{home}}
systemctl enable {{user}}.service
systemctl start {{user}}.service
sudo systemctl enable {{user}}.service
sudo systemctl start {{user}}.service
exit 0