From 6c403519d10cf37cfa1f7c508b7514a97c0291d6 Mon Sep 17 00:00:00 2001 From: Mark Saroufim Date: Tue, 13 Jan 2026 15:24:06 -0800 Subject: [PATCH 1/2] Concurrency on custom nvidia runners --- .github/workflows/nvidia_workflow.yml | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) diff --git a/.github/workflows/nvidia_workflow.yml b/.github/workflows/nvidia_workflow.yml index 59a5e4d4..26bf2f22 100644 --- a/.github/workflows/nvidia_workflow.yml +++ b/.github/workflows/nvidia_workflow.yml @@ -15,11 +15,27 @@ on: required: false type: string +concurrency: + group: nvidia-workflow-global + cancel-in-progress: false + run-name: 'NVIDIA Job - ${{ github.event.inputs.run_id }}' jobs: + select-runner: + runs-on: ubuntu-latest + outputs: + runner: ${{ steps.pick.outputs.runner }} + steps: + - id: pick + run: | + runners=("b200-02-gpu0" "b200-02-gpu1" "b200-02-gpu2" "b200-02-gpu3" "b200-02-gpu4" "b200-02-gpu5" "b200-02-gpu6" "b200-02-gpu7") + index=$(( ${{ github.run_number }} % 8 )) + echo "runner=${runners[$index]}" >> $GITHUB_OUTPUT + run: - runs-on: [nvidia-docker-b200-8-x86-64] + needs: select-runner + runs-on: [self-hosted, Linux, X64, nvidia-docker-b200-8-x86-64, "${{ needs.select-runner.outputs.runner }}"] timeout-minutes: 10 steps: - uses: actions/checkout@v3 From 27e348333c188d6aa455a74bdda9cd2bf877ccc8 Mon Sep 17 00:00:00 2001 From: Mark Saroufim Date: Tue, 13 Jan 2026 15:26:05 -0800 Subject: [PATCH 2/2] update --- .github/workflows/nvidia_workflow.yml | 14 +------------- 1 file changed, 1 insertion(+), 13 deletions(-) diff --git a/.github/workflows/nvidia_workflow.yml b/.github/workflows/nvidia_workflow.yml index 26bf2f22..c64924b1 100644 --- a/.github/workflows/nvidia_workflow.yml +++ b/.github/workflows/nvidia_workflow.yml @@ -22,20 +22,8 @@ concurrency: run-name: 'NVIDIA Job - ${{ github.event.inputs.run_id }}' jobs: - select-runner: - runs-on: ubuntu-latest - outputs: - runner: ${{ steps.pick.outputs.runner }} - steps: - - id: pick - run: | - runners=("b200-02-gpu0" "b200-02-gpu1" "b200-02-gpu2" "b200-02-gpu3" "b200-02-gpu4" "b200-02-gpu5" "b200-02-gpu6" "b200-02-gpu7") - index=$(( ${{ github.run_number }} % 8 )) - echo "runner=${runners[$index]}" >> $GITHUB_OUTPUT - run: - needs: select-runner - runs-on: [self-hosted, Linux, X64, nvidia-docker-b200-8-x86-64, "${{ needs.select-runner.outputs.runner }}"] + runs-on: nvidia-docker-b200-8-x86-64 timeout-minutes: 10 steps: - uses: actions/checkout@v3