Skip to content

Migrate Windows VM to CTK 13.3 - #98

Merged
trivialfis merged 10 commits into
dmlc:mainfrom
trivialfis:win-ctk13
Aug 3, 2026
Merged

trivialfis merged 10 commits into
dmlc:mainfrom
trivialfis:win-ctk13

Conversation

@trivialfis

@trivialfis trivialfis commented Aug 1, 2026 •

Copy link
Copy Markdown
Member

This PR updates the Windows AMI. In addition, it splits the bootstrap script to insert a reboot provisioned by Packer after Visual Studio is installed. The installation returns a non-zero code without a reboot.

TODOs:

@trivialfis
trivialfis marked this pull request as ready for review August 2, 2026 23:29

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR updates the Windows Server 2022 AMI build to install CUDA 13.3 (CTK 13.3) and restructures provisioning so Visual Studio installation is split into multiple steps with Packer-managed reboots, improving reliability of the image build.

Changes:

  • Split the previous monolithic bootstrap.ps1 install flow into dedicated scripts for Visual Studio install, workload install, verification, CUDA/R install, NVIDIA driver install, and driver verification.
  • Insert windows-restart provisioners between key install steps (VS, workload, NVIDIA driver) to handle installers that require reboot.
  • Add new verification scripts (verify_vs.ps1, verify_nvidia_driver.ps1) and ignore packer.log in the Windows VM image directory.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 2 comments.

Show a summary per file
File Description
vm_images/windows/windows.pkr.hcl Adds stepwise provisioning and multiple controlled reboots; updates sysprep provisioner exit codes.
vm_images/windows/bootstrap.ps1 Removes VS/CUDA/R installation from bootstrap to support the new split install flow.
vm_images/windows/install_vs.ps1 Installs VS Community (and vcredist) while tolerating “reboot required” exit codes.
vm_images/windows/install_vs_workload.ps1 Installs Native Desktop workload and tolerates “reboot required” exit codes.
vm_images/windows/verify_vs.ps1 Verifies VS workload presence and extracts VS display version via vswhere.
vm_images/windows/install_cuda_r.ps1 Installs CUDA 13.3 and R toolchain versions via Chocolatey.
vm_images/windows/install_nvidia_driver.ps1 Downloads, signature-checks, extracts, and installs pinned NVIDIA GRID driver from S3.
vm_images/windows/verify_nvidia_driver.ps1 Verifies driver install by running nvidia-smi.exe.
vm_images/windows/.gitignore Ignores packer.log in the Windows image directory.
Suppressed comments (1)

vm_images/windows/verify_vs.ps1:20

  • Same wildcard-expansion issue here: -products * should be quoted to avoid PowerShell expanding it to filenames before calling vswhere.
$displayVersion = & $vswhere `
    -latest `
    -products * `
    -property catalog_productDisplayVersion

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment thread vm_images/windows/verify_vs.ps1
Comment thread vm_images/windows/verify_nvidia_driver.ps1
@trivialfis
trivialfis merged commit ebb91c6 into dmlc:main Aug 3, 2026
5 of 6 checks passed
@trivialfis
trivialfis deleted the win-ctk13 branch August 3, 2026 06:37

This branch had an error being deployed

1 failed deployment
containers-staging — 1963022a Deployed Aug 3, 2026 by trivialfis via trigger-build / display-message #63
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants