litelotus/.nix
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

01-server: ollama (Vulkan) on the Tesla M10s #21

Merged
litelotus merged 1 commit from feat/ollama-vulkan-01 into main 2026-09-14 09:36:29 +00:00
Conversation 0 Commits 1 Files changed 1 +12
litelotus commented 2026-09-14 09:13:45 +00:00
Owner
Copy link

Enables ollama on the k3s master node with the Vulkan backend:

  • services.ollama with pkgs.ollama-vulkan — Maxwell (cc 5.0) is removed from CUDA 13 runtimes, so CUDA builds can't target the M10; Vulkan compiles shaders at runtime and the pinned legacy_580 driver (R580, last Maxwell branch) still provides the Vulkan ICD
  • host 0.0.0.0:11434 + openFirewall so the k8s headroom pod can reach it
  • loadModels pulls qwen3:8b automatically

Chain being built: hermes-agent (k8s) -> headroom (k8s) -> ollama:11434 on the k3s master node.

NOTE: do NOT upgrade the nvidia driver past legacy_580 — R580 is the last branch supporting Maxwell.

Enables ollama on the k3s master node with the Vulkan backend: - `services.ollama` with `pkgs.ollama-vulkan` — Maxwell (cc 5.0) is removed from CUDA 13 runtimes, so CUDA builds can't target the M10; Vulkan compiles shaders at runtime and the pinned legacy_580 driver (R580, last Maxwell branch) still provides the Vulkan ICD - host 0.0.0.0:11434 + openFirewall so the k8s headroom pod can reach it - loadModels pulls qwen3:8b automatically Chain being built: hermes-agent (k8s) -> headroom (k8s) -> ollama:11434 on the k3s master node. NOTE: do NOT upgrade the nvidia driver past legacy_580 — R580 is the last branch supporting Maxwell.
01-server: enable ollama with Vulkan backend on the Tesla M10s
Some checks failed
ci/crow/push/check Pipeline failed
Details
6f648c9f0d 
- services.ollama with package pkgs.ollama-vulkan (Maxwell is dropped from
  CUDA 13 runtimes, so CUDA builds can't target the M10; Vulkan has no
  prebuilt-arch cutoff and R580 driver still supports it)
- listen on 0.0.0.0:11434 + openFirewall so the k8s headroom pod can reach it
- loadModels: qwen3:8b (pulled automatically by ollama-model-loader)
- part of the local LLM chain: hermes (k8s) -> headroom (k8s) -> ollama:11434
litelotus deleted branch feat/ollama-vulkan-01 2026-09-14 09:36:36 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
litelotus/.nix!21
No description provided.