Skip to content

GPU VM - #64

Merged
wongcallum merged 10 commits into
masterfrom
draft/gpu-compute-vm
Aug 27, 2026
Merged

GPU VM#64
wongcallum merged 10 commits into
masterfrom
draft/gpu-compute-vm

Conversation

@wongcallum

Copy link
Copy Markdown
Owner

No description provided.

liz's declarative microvm guest already builds vm-gpu against
inputs.unstable, but the standalone checks.x86_64-linux.vm-gpu fell
back to the default stable nixpkgs, so CUDA llama-cpp hashed
differently and had to be compiled from source twice per CI run.
modules/ssh.nix contributes the openssh host keys to the persistence
aspect, not to the ssh aspect, so importing ssh alone left vm-gpu with
no impermanence at all. It regenerated its host keys on every restart,
tripping host key verification for anyone reaching it over ssh.

vm-gallery imports persistence directly and vm-coder picks it up via
the coder aspect; vm-gpu was the only guest missing it.
@wongcallum
wongcallum force-pushed the draft/gpu-compute-vm branch from dd3a0f6 to 13ca6e3 Compare August 27, 2026 11:40
@wongcallum
wongcallum merged commit 5ff1338 into master Aug 27, 2026
5 checks passed
@wongcallum
wongcallum deleted the draft/gpu-compute-vm branch August 27, 2026 13:35
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.

1 participant