Skip to content

feat(runners): enable GPU passthrough for ai-lab lane #661

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We鈥檒l occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
POWERFULMOVES merged 2 commits into from
Feb 19, 2026
Merged

feat(runners): enable GPU passthrough for ai-lab lane #661

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
9376e31
feat(runners): enable GPU passthrough for ai-lab lane
hunnibear Feb 19, 2026
e8d8265
fix(pr-661): lane_hosts.json schema + min-runner enforcement
hunnibear Feb 19, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 27 additions & 1 deletion pmoves/integrations/github-runners/compose/lane_hosts.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,29 @@
{
"lane_hosts": {}
"lane_hosts": {
"self-hosted,Linux,X64": {
"host": "vps / ai-lab",
"runner_name": "pmoves-vps-runner",
"registration_script": "integrations/github-runners/compose/register-runner.sh"
},
"self-hosted,ai-lab,gpu": {
"host": "ai-lab",
"runner_name": "pmoves-ai-lab-runner",
"registration_script": "integrations/github-runners/compose/register-runner.sh"
},
"self-hosted,ai-lab": {
"host": "ai-lab",
"runner_name": "pmoves-ai-lab-runner",
"registration_script": "integrations/github-runners/compose/register-runner.sh"
},
"self-hosted,vps,Linux,X64": {
"host": "vps",
"runner_name": "pmoves-vps-runner",
"registration_script": "integrations/github-runners/compose/register-runner.sh"
},
"self-hosted,gpu": {
"host": "ai-lab",
"runner_name": "pmoves-ai-lab-runner",
"registration_script": "integrations/github-runners/compose/register-runner.sh"
}
}
}
11 changes: 9 additions & 2 deletions pmoves/integrations/github-runners/compose/runner_phase_policy.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,11 @@
{
"default_phase": "",
"phases": {}
"default_phase": "local-certification",
"phases": {
"local-certification": {
"description": "Both runners on local Docker containers for development/CI",
"required_online": ["self-hosted,Linux,X64", "self-hosted,ai-lab,gpu"],

@chatgpt-codex-connector chatgpt-codex-connector Bot Feb 19, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

P2 Badge Require VPS lane explicitly in local-cert policy

This phase is described as "Both runners on local Docker containers", but required_online currently uses self-hosted,Linux,X64 plus self-hosted,ai-lab,gpu; because lane checks are subset-based in runner_lane_map.py, the ai-lab runner (self-hosted,ai-lab,gpu,Linux,X64) can satisfy both requirements by itself, so the policy can pass while the VPS runner is offline and fail to enforce the stated two-runner requirement.

Useful? React with 馃憤聽/ 馃憥.

"required_offline": [],
"min_runners": 2
}
}
}
83 changes: 59 additions & 24 deletions pmoves/tools/local_cert_runners.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,31 +83,66 @@ def docker_rm(container_name: str) -> None:
run_cmd(["docker", "rm", "-f", container_name], check=False)


LOKI_URL = os.getenv("LOKI_URL", "http://localhost:3100/loki/api/v1/push")

# Default resource limits per lane (overridable via env)
LANE_RESOURCES: dict[str, dict[str, str]] = {
"ai-lab": {
"cpus": os.getenv("RUNNER_AI_LAB_CPUS", "4"),
"memory": os.getenv("RUNNER_AI_LAB_MEMORY", "8g"),
"gpus": "all",
},
"vps": {
"cpus": os.getenv("RUNNER_VPS_CPUS", "2"),
"memory": os.getenv("RUNNER_VPS_MEMORY", "4g"),
"gpus": "",
},
}


def docker_run(repo: str, image: str, lane: RunnerLane, token: str) -> None:
run_cmd(
[
"docker",
"run",
"-d",
"--name",
lane.container_name,
"--restart",
"unless-stopped",
"-e",
f"REPO_URL=https://github.com/{repo}",
"-e",
f"RUNNER_NAME={lane.runner_name}",
"-e",
f"RUNNER_TOKEN={token}",
"-e",
f"LABELS={lane.labels}",
"-e",
"RUNNER_WORKDIR=/tmp/runner/_work",
"-v",
"/var/run/docker.sock:/var/run/docker.sock",
image,
]
)
resources = LANE_RESOURCES.get(lane.lane, {"cpus": "4", "memory": "8g"})
cmd = [
"docker",
"run",
"-d",
"--name",
lane.container_name,
"--restart",
"unless-stopped",
"--cpus",
resources["cpus"],
"--memory",
resources["memory"],
"--log-driver",
"loki",
"--log-opt",
f"loki-url={LOKI_URL}",
"--log-opt",
f"loki-external-labels=job=gha-runner,lane={lane.lane},runner={lane.runner_name}",
]
# GPU passthrough for ai-lab lane only
gpus = resources.get("gpus", "")
if gpus:
cmd.extend(["--gpus", gpus])
cmd.extend(["-e", "NVIDIA_VISIBLE_DEVICES=all"])
cmd.extend(["-e", "NVIDIA_DRIVER_CAPABILITIES=compute,utility"])
cmd.extend([
"-e",
f"REPO_URL=https://github.com/{repo}",
"-e",
f"RUNNER_NAME={lane.runner_name}",
"-e",
f"RUNNER_TOKEN={token}",
"-e",
f"LABELS={lane.labels}",
"-e",
"RUNNER_WORKDIR=/tmp/runner/_work",
"-v",
"/var/run/docker.sock:/var/run/docker.sock",
image,
])
run_cmd(cmd)


def cmd_up(repo: str, image: str) -> int:
Expand Down
9 changes: 9 additions & 0 deletions pmoves/tools/runner_lane_map.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -208,6 +208,15 @@ def evaluate_phase(
else:
notes.append(f"[{phase_name}] cannot validate offline state without --check-gh: {lane}")

min_runners = int(phase_cfg.get("min_runners", 0))
if min_runners > 0 and check_gh:
distinct_names = {r.name for r in runners if r.online}
if len(distinct_names) < min_runners:
failures.append(
f"[{phase_name}] min_runners={min_runners} requires {min_runners} distinct "
f"online runners, found {len(distinct_names)}: {sorted(distinct_names)}"
)

return failures, notes


Expand Down
Loading