From d5b1eb081e193c54ac21390a0f6ba7013e4f3b11 Mon Sep 17 00:00:00 2001
From: Simon Mo <simon.mo@hey.com>
Date: Wed, 5 Jun 2024 11:42:08 -0500
Subject: [PATCH] [CI] Add nightly benchmarks (#5260)

---
 .../nightly-benchmarks/kickoff-pipeline.sh    | 26 +++++++++++++
 .buildkite/nightly-benchmarks/sample.yaml     | 39 +++++++++++++++++++
 2 files changed, 65 insertions(+)
 create mode 100755 .buildkite/nightly-benchmarks/kickoff-pipeline.sh
 create mode 100644 .buildkite/nightly-benchmarks/sample.yaml

diff --git a/.buildkite/nightly-benchmarks/kickoff-pipeline.sh b/.buildkite/nightly-benchmarks/kickoff-pipeline.sh
new file mode 100755
index 00000000..d3bf3b72
--- /dev/null
+++ b/.buildkite/nightly-benchmarks/kickoff-pipeline.sh
@@ -0,0 +1,26 @@
+#!/usr/bin/env bash
+
+set -euo pipefail
+
+# Install system packages
+apt update
+apt install -y curl jq
+
+# Install minijinja for templating
+curl -sSfL https://github.com/mitsuhiko/minijinja/releases/latest/download/minijinja-cli-installer.sh | sh
+source $HOME/.cargo/env
+
+# If BUILDKITE_PULL_REQUEST != "false", then we check the PR labels using curl and jq
+if [ "$BUILDKITE_PULL_REQUEST" != "false" ]; then
+  PR_LABELS=$(curl -s "https://api.github.com/repos/vllm-project/vllm/pulls/$BUILDKITE_PULL_REQUEST" | jq -r '.labels[].name')
+
+  if [[ $PR_LABELS == *"perf-benchmarks"* ]]; then
+    echo "This PR has the 'perf-benchmarks' label. Proceeding with the nightly benchmarks."
+  else
+    echo "This PR does not have the 'perf-benchmarks' label. Skipping the nightly benchmarks."
+    exit 0
+  fi
+fi
+
+# Upload sample.yaml
+buildkite-agent pipeline upload .buildkite/nightly-benchmarks/sample.yaml
diff --git a/.buildkite/nightly-benchmarks/sample.yaml b/.buildkite/nightly-benchmarks/sample.yaml
new file mode 100644
index 00000000..50e6e820
--- /dev/null
+++ b/.buildkite/nightly-benchmarks/sample.yaml
@@ -0,0 +1,39 @@
+steps:
+  # NOTE(simon): You can create separate blocks for different jobs
+  - label: "A100: NVIDIA SMI"
+    agents:
+      queue: A100
+    plugins:
+    - kubernetes:
+        podSpec:
+          containers:
+          # - image: us-central1-docker.pkg.dev/vllm-405802/vllm-ci-test-repo/vllm-test:$BUILDKITE_COMMIT
+          # TODO(simon): check latest main branch or use the PR image.
+          - image: us-central1-docker.pkg.dev/vllm-405802/vllm-ci-test-repo/vllm-test:45c35f0d58f4508bf43bd6af1d3d0d0ec0c915e6
+            command:
+            - bash -c 'nvidia-smi && nvidia-smi topo -m && pwd && ls'
+            resources:
+              limits:
+                nvidia.com/gpu: 8
+            volumeMounts:
+            - name: devshm
+              mountPath: /dev/shm
+          nodeSelector:
+            nvidia.com/gpu.product: NVIDIA-A100-SXM4-80GB
+          volumes:
+          - name: devshm
+            emptyDir:
+              medium: Memory
+  # TODO(simon): bring H100 online
+  # - label: "H100: NVIDIA SMI"
+  #   agents:
+  #     queue: H100
+  #   plugins:
+  #   - docker#v5.11.0:
+  #       image: us-central1-docker.pkg.dev/vllm-405802/vllm-ci-test-repo/vllm-test:45c35f0d58f4508bf43bd6af1d3d0d0ec0c915e6
+  #       command:
+  #       - bash -c 'nvidia-smi && nvidia-smi topo -m'
+  #       propagate-environment: true
+  #       ipc: host
+  #       gpus: all
+