From a736216a0503171c83d6f4bfb57e2be4ac082804 Mon Sep 17 00:00:00 2001 From: Andrew Beveridge Date: Mon, 18 May 2026 00:34:20 -0400 Subject: [PATCH] ci: verify new ephemeral GPU runner image + dispatcher MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit No-op touch on audio_separator/separator/__init__.py to fire the run-integration-tests workflow's three self-hosted GPU jobs (ensemble-presets, core-models, stems-and-quality). This is the end-to-end verification step for karaoke-gen PR #781 (GPU image NVIDIA driver-load fix + secure-boot disabled for GPU VMs). All three jobs should land on fresh ephemeral GPU VMs created by the updated dispatcher, the gha-gpu-modprobe.service should rebuild the NVIDIA module against the running kernel on first boot, and nvidia-smi should report Tesla T4 at the "Verify GPU availability" step. Smoke-test of the new image (`gha-runner-gpu-20260518-035713`) on 2026-05-18 already showed end-to-end success — this PR exercises it through the real dispatcher path. Co-Authored-By: Claude Opus 4.7 (1M context) --- audio_separator/separator/__init__.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/audio_separator/separator/__init__.py b/audio_separator/separator/__init__.py index 2d36e8f2..b00f88f9 100644 --- a/audio_separator/separator/__init__.py +++ b/audio_separator/separator/__init__.py @@ -1 +1,5 @@ from .separator import Separator + +# No-op touch (2026-05-18): trigger the run-integration-tests workflow to +# verify the new gha-runner-gpu image + secure-boot-off ephemeral dispatcher +# load the NVIDIA driver cleanly at boot. See karaoke-gen PR #781.