Skip to content

feat: add fluidstack-gpu DIB element for GPU qualification IPA image#11

Merged
nacef3005 merged 1 commit into
mainfrom
nacef/gpu-support
Jun 10, 2026
Merged

feat: add fluidstack-gpu DIB element for GPU qualification IPA image#11
nacef3005 merged 1 commit into
mainfrom
nacef/gpu-support

Conversation

@nacef3005

Copy link
Copy Markdown
Collaborator
  • Adds a new fluidstack-gpu DIB element that builds a dedicated Ubuntu Noble 24.04 IPA ramdisk for GPU qualification workloads

  • The image bundles NVIDIA driver 570 + CUDA 12.8 + NCCL (all from the NVIDIA CUDA apt repo at consistent versions) and pre-compiled nccl-tests binaries at /usr/local/bin/nccl-tests/

  • Adds a gpu-noble matrix entry to the CI workflow, producing ipa-ubuntu-noble-gpu-<branch>-fs in parallel with the existing fish and ironwood builds

  • cuda-drivers-570 — full driver stack (kernel modules, nvidia-smi, compute libs) from NVIDIA repo

  • libnccl2 + cuda-cudart-12-8 — NCCL and CUDA runtime

  • nccl-tests *_perf binaries compiled at build time against the installed toolkit; dev packages stripped afterwards to keep image size down

  • nvidia, nvidia_uvm, nvidia_drm loaded at boot via modules-load.d

@nacef3005 nacef3005 self-assigned this Jun 9, 2026
@nacef3005 nacef3005 marked this pull request as ready for review June 9, 2026 14:21

nacef3005 commented Jun 9, 2026

Copy link
Copy Markdown
Collaborator Author

nacef3005 commented Jun 10, 2026

Copy link
Copy Markdown
Collaborator Author

Merge activity

  • Jun 10, 1:29 PM UTC: A user started a stack merge that includes this pull request via Graphite.
  • Jun 10, 1:30 PM UTC: Graphite rebased this pull request as part of a merge.
  • Jun 10, 1:30 PM UTC: @nacef3005 merged this pull request with Graphite.

@nacef3005 nacef3005 changed the base branch from nacef/tpu-ironwood-support to graphite-base/11 June 10, 2026 13:29
@nacef3005 nacef3005 changed the base branch from graphite-base/11 to main June 10, 2026 13:29
- Adds a new `fluidstack-gpu` DIB element that builds a dedicated Ubuntu Noble 24.04 IPA ramdisk for GPU qualification workloads
- The image bundles NVIDIA driver 570 + CUDA 12.8 + NCCL (all from the NVIDIA CUDA apt repo at consistent versions) and pre-compiled `nccl-tests` binaries at `/usr/local/bin/nccl-tests/`
- Adds a `gpu-noble` matrix entry to the CI workflow, producing `ipa-ubuntu-noble-gpu-<branch>-fs` in parallel with the existing fish and ironwood builds

- `cuda-drivers-570` — full driver stack (kernel modules, `nvidia-smi`, compute libs) from NVIDIA repo
- `libnccl2` + `cuda-cudart-12-8` — NCCL and CUDA runtime
- `nccl-tests` `*_perf` binaries compiled at build time against the installed toolkit; dev packages stripped afterwards to keep image size down
- `nvidia`, `nvidia_uvm`, `nvidia_drm` loaded at boot via `modules-load.d`
@nacef3005 nacef3005 force-pushed the nacef/gpu-support branch from b3efa25 to ee1132f Compare June 10, 2026 13:29
@nacef3005 nacef3005 merged commit b6eafe1 into main Jun 10, 2026
1 check passed
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.

2 participants