From def8f52200151c801867dde9ce27f829bce00105 Mon Sep 17 00:00:00 2001 From: Sage Moore Date: Mon, 20 Apr 2026 07:22:54 -0700 Subject: [PATCH] [CI][EPLB] Add Async EPLB end-to-end integration test to CI (#40168) Signed-off-by: Sage Moore --- .../qwen30b_a3b_fp8_dp4_async_eplb.sh | 55 +++++++++++++++++++ .buildkite/test_areas/e2e_integration.yaml | 9 +++ 2 files changed, 64 insertions(+) create mode 100755 .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_dp4_async_eplb.sh diff --git a/.buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_dp4_async_eplb.sh b/.buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_dp4_async_eplb.sh new file mode 100755 index 0000000000..06743f16b6 --- /dev/null +++ b/.buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_dp4_async_eplb.sh @@ -0,0 +1,55 @@ +#!/usr/bin/env bash +set -euxo pipefail + +# args: [THRESHOLD] [NUM_QUESTIONS] [START_PORT] +THRESHOLD=${1:-0.8} +NUM_Q=${2:-1319} +PORT=${3:-8050} +OUT_DIR=${OUT_DIR:-/tmp/vllm-scheduled} +mkdir -p "${OUT_DIR}" + +wait_for_server() { + local port=$1 + timeout 600 bash -c ' + until curl -sf "http://127.0.0.1:'"$port"'/health" > /dev/null; do + sleep 1 + done' +} + +MODEL="Qwen/Qwen3-30B-A3B-FP8" +BACK="allgather_reducescatter" + +cleanup() { + if [[ -n "${SERVER_PID:-}" ]] && kill -0 "${SERVER_PID}" 2>/dev/null; then + kill "${SERVER_PID}" 2>/dev/null || true + for _ in {1..20}; do + kill -0 "${SERVER_PID}" 2>/dev/null || break + sleep 0.5 + done + kill -9 "${SERVER_PID}" 2>/dev/null || true + fi +} +trap cleanup EXIT + +VLLM_DEEP_GEMM_WARMUP=skip \ +vllm serve "$MODEL" \ +--enforce-eager \ +--data-parallel-size 4 \ +--enable-expert-parallel \ +--enable-eplb \ +--all2all-backend "$BACK" \ +--eplb-config '{"window_size":20, "step_interval":100, "use_async":true}' \ +--trust-remote-code \ +--max-model-len 2048 \ +--port "$PORT" & +SERVER_PID=$! +wait_for_server "$PORT" + +TAG=$(echo "$MODEL" | tr '/: \\n' '_____') +OUT="${OUT_DIR}/${TAG}_${BACK}.json" +python3 tests/evals/gsm8k/gsm8k_eval.py --host http://127.0.0.1 --port "$PORT" --num-questions "${NUM_Q}" --save-results "${OUT}" +python3 - <= ${THRESHOLD}, f"${MODEL} ${BACK} accuracy {acc}" +PY diff --git a/.buildkite/test_areas/e2e_integration.yaml b/.buildkite/test_areas/e2e_integration.yaml index 5b7f96bc7a..857fefd268 100644 --- a/.buildkite/test_areas/e2e_integration.yaml +++ b/.buildkite/test_areas/e2e_integration.yaml @@ -29,6 +29,15 @@ steps: commands: - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_block_ep_eplb.sh 0.8 200 8020 2 1 +- label: Qwen3-30B-A3B-FP8 DP4 Async EPLB Accuracy + timeout_in_minutes: 60 + device: h100 + optional: true + num_devices: 4 + working_dir: "/vllm-workspace" + commands: + - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_dp4_async_eplb.sh 0.8 200 8050 + - label: DeepSeek V2-Lite Prefetch Offload Accuracy (H100) timeout_in_minutes: 60 device: h100