iframe-proxy | Sunbelt Computer Software

Repository navigation

History

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

$ErrorActionPreference = "Stop"

# Check for 64-bit Windows installation

if (-not [Environment]::Is64BitOperatingSystem) {

Write-Error "diffusers-workflow requires a 64-bit Windows installation"

Exit 1

}

# Check for Visual C++ Runtime DLLs

$vcRuntime = Get-Item -ErrorAction SilentlyContinue "$env:windir\System32\msvcp140.dll"

if (-not $vcRuntime.Exists) {

$vcRuntimeUrl = "https://visualstudio.microsoft.com/downloads/#microsoft-visual-c-redistributable-for-visual-studio-2019"

Write-Error "Unable to find Visual C++ Runtime DLLs"

Write-Output "Download and install the Visual C++ Redistributable for Visual Studio 2019 package from: $vcRuntimeUrl"

Exit 1

}

# Check for Python

try {

$pythonVersion = (python --version).split(" ")[1]

}

catch {

Write-Error "Unable to find python"

$pythonUrl = "https://docs.python.org/3/using/windows.html#installation-steps"

Write-Output "Note the check box during installation of Python to install the Python Launcher for Windows."

Write-Output "Install Python from: $pythonUrl"

Exit 1

}

# Check for supported Python version (3.10 - 3.14)

$supportedPythonVersions = "3.14", "3.13", "3.12", "3.11", "3.10"

if ($env:INSTALL_PYTHON_VERSION) {

$pythonVersion = $env:INSTALL_PYTHON_VERSION

}

else {

$pythonVersion = $null

foreach ($version in $supportedPythonVersions) {

try {

$pver = (python --version).split(" ")[1]

$result = $pver.StartsWith($version)

}

catch {

$result = $false

}

if ($result) {

$pythonVersion = $version

break

}

if (-not $pythonVersion) {

$supportedPythonVersions = ($supportedPythonVersions | ForEach-Object { "Python $_" }) -join ", "

Write-Error "No usable Python version found, supported versions are: $supportedPythonVersions"

Write-Output "diffusers-workflow requires Python version >= 3.10 and <= 3.14"

Exit 1

}

# Print Python version

Write-Output "Python version is: $pythonVersion"

# Remove the venv if it exists

if (Test-Path -Path ".\venv" -PathType Container) {

Remove-Item -LiteralPath ".\venv" -Recurse -Force

}

# Create virtual environment

python -m venv venv

# Activate virtual environment

.\venv\scripts\activate

# Upgrade pip

python.exe -m pip install --upgrade pip

if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

# Install PyTorch first - GPU-aware, so the resolver below sees it satisfied

# and leaves it alone. torchaudio comes with it, from the same index so its

# build matches: dw does not import it, but diffusers' MiniMax H3 blocks use it

# to resample an audio reference that is not already at the audio VAE's rate,

# and without it those pipelines fail at the first setup block.

# PyPI's Windows wheel bundles a CUDA 12.x runtime; pytorch.org's cu130 index

# carries a CUDA 13 build that needs a newer driver (>= 580). So we probe for a

# working NVIDIA driver via nvidia-smi and, if found, read the CUDA version it

# supports: a CUDA 13-capable driver gets the cu130 build, an older one stays on

# PyPI's CUDA 12 wheel rather than a build its driver cannot load (which would

# silently fall back to the CPU).

$driverCuda = $null

if (Get-Command nvidia-smi -ErrorAction SilentlyContinue) {

# Finding the binary only proves it exists; a stale or broken driver can

# leave nvidia-smi installed but failing, so require it to run too.

try {

$smiOutput = (& nvidia-smi 2>$null) -join "`n"

if ($LASTEXITCODE -eq 0) {

$cudaMatch = [regex]::Match($smiOutput, 'CUDA Version:\s*(\d+)')

if ($cudaMatch.Success) {

$driverCuda = [int]$cudaMatch.Groups[1].Value

}

else {

$driverCuda = 12

}

catch {

$driverCuda = $null

}

if ($null -eq $driverCuda) {

Write-Output "No CUDA GPU detected - installing CPU-only torch"

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

}

elseif ($driverCuda -ge 13) {

Write-Output "CUDA $driverCuda driver detected - installing GPU-enabled torch (cu130)"

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu130

}

else {

Write-Output "CUDA $driverCuda driver detected - installing PyPI torch (bundled CUDA 12 runtime)"

pip install torch torchvision torchaudio

}

if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

# Everything else resolves from pyproject.toml - the single source of the

# dependency list. Editable install with the server + dev extras.

pip install -e ".[server,dev]"

if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

# Diffusers from GitHub (releases lag the newest model pipelines) - after

# the resolver, so it isn't replaced by the released version. Forced, because

# main keeps one dev version string for a release cycle and --upgrade would

# keep an older commit; --no-deps so torch is not reinstalled, then a plain

# install adds any dependency the new commit needs (as dw/server/updater.py)

pip install --force-reinstall --no-deps git+https://github.com/huggingface/diffusers

if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

pip install diffusers

if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

# Windows/CUDA-specific: bitsandbytes' pyproject marker is linux-only

pip install bitsandbytes kernels

if ($LASTEXITCODE -ne 0) { exit $LASTEXITCODE }

# Web UI dependencies - optional, only needed to build/run the SPA served by dw-serve

if (Get-Command npm -ErrorAction SilentlyContinue) {

Write-Output ""

Write-Output "Installing web UI dependencies..."

Push-Location ui

npm install

$npmExitCode = $LASTEXITCODE

Pop-Location

if ($npmExitCode -ne 0) { exit $npmExitCode }

}

else {

Write-Output ""

Write-Output "npm was not found - skipping web UI dependency install."

Write-Output "Install Node.js/npm and run 'npm install' in the ui\ folder to build the web UI."

}

Write-Output ""

Write-Output "Installation complete!"

Write-Output ""

Write-Output "To activate the virtual environment, run:"

Write-Output " .\venv\scripts\activate"

Write-Output ""

Write-Output "Note: Some LLM workflows (e.g., Phi mini-instruct) require flash_attn,"

Write-Output "which requires the CUDA Toolkit: https://developer.nvidia.com/cuda-toolkit"

Sunbelt Computer Software

PL/B Language Development and Support

Navigation Menu

Repository navigation

install.ps1

install.ps1

Sunbelt Computer Software

PL/B Language Development and Support

FilesExpand file tree

install.ps1

Latest commit

History

install.ps1

File metadata and controls