Files
llm/scripts/check_gpu_host.ps1
T

45 lines
1.1 KiB
PowerShell

param(
[string]$SshTarget = "docker-gpu.cin.su",
[int]$ConnectTimeoutSeconds = 5
)
$ErrorActionPreference = "Stop"
function Invoke-Remote {
param(
[Parameter(Mandatory = $true)]
[string]$Command
)
ssh `
-o BatchMode=yes `
-o ConnectTimeout=$ConnectTimeoutSeconds `
$SshTarget `
$Command
if ($LASTEXITCODE -ne 0) {
throw "Remote command failed on ${SshTarget} with exit code ${LASTEXITCODE}: ${Command}"
}
}
Write-Host "Checking SSH access to $SshTarget..."
Invoke-Remote "hostname"
Write-Host ""
Write-Host "Checking NVIDIA GPU..."
Invoke-Remote "nvidia-smi --query-gpu=name,memory.total,driver_version --format=csv,noheader"
Write-Host ""
Write-Host "Checking Docker..."
Invoke-Remote "docker version --format '{{.Server.Version}}'"
Write-Host ""
Write-Host "Checking Docker Compose..."
Invoke-Remote "docker compose version"
Write-Host ""
Write-Host "Checking Docker GPU runtime..."
Invoke-Remote "docker run --rm --gpus all nvidia/cuda:12.4.1-base-ubuntu22.04 nvidia-smi --query-gpu=name,memory.total --format=csv,noheader"
Write-Host ""
Write-Host "GPU host preflight completed."