add swarm-resource配置

This commit is contained in:
Your Name
2024-07-02 06:56:33 +00:00
parent bb625454ba
commit 7f295d8f08

10
finetune/configure_gpu.sh Normal file → Executable file
View File

@@ -10,13 +10,13 @@ NODE_RESOURCES=${NODE_RESOURCES%,} # 移除最后一个逗号
# 生成完整的daemon.json内容 # 生成完整的daemon.json内容
DAEMON_JSON=$(cat <<EOF DAEMON_JSON=$(cat <<EOF
{ {
"default-runtime": "nvidia",
"runtimes": { "runtimes": {
"nvidia": { "nvidia": {
"path": "nvidia-container-runtime", "path": "/usr/bin/nvidia-container-runtime",
"runtimeArgs": [] "runtimeArgs": []
} }
}, },
"default-runtime": "nvidia",
"node-generic-resources": [ "node-generic-resources": [
$NODE_RESOURCES $NODE_RESOURCES
] ]
@@ -32,8 +32,12 @@ fi
# 写入新的daemon.json内容 # 写入新的daemon.json内容
echo "$DAEMON_JSON" | sudo tee /etc/docker/daemon.json > /dev/null echo "$DAEMON_JSON" | sudo tee /etc/docker/daemon.json > /dev/null
# 添加swarm-resource配置
sudo sed -i '/^#.*swarm-resource/s/^#//' /etc/nvidia-container-runtime/config.toml
sudo sed -i '/swarm-resource =/s/=.*/= "DOCKER_RESOURCE_GPU"/' /etc/nvidia-container-runtime/config.toml
# 重启Docker服务 # 重启Docker服务
sudo systemctl restart docker sudo systemctl restart docker.service
# 验证配置 # 验证配置
docker info | grep -i 'nvidia' docker info | grep -i 'nvidia'