From 7634ffb7094e6c34f487e2d3fa6f5cf78e733950 Mon Sep 17 00:00:00 2001
From: EnflameGCU <118410644+EnflameGCU@users.noreply.github.com>
Date: Fri, 25 Jul 2025 10:59:29 +0800
Subject: [PATCH] [GCU] Add CI (#3006)

---
 .github/workflows/ci_gcu.yml | 89 ++++++++++++++++++++++++++++++++++++
 scripts/run_ci_gcu.sh        | 86 ++++++++++++++++++++++++++++++++++
 test/ci_use/GCU/run_ernie.py | 32 +++++++++++++
 3 files changed, 207 insertions(+)
 create mode 100644 .github/workflows/ci_gcu.yml
 create mode 100644 scripts/run_ci_gcu.sh
 create mode 100644 test/ci_use/GCU/run_ernie.py

diff --git a/.github/workflows/ci_gcu.yml b/.github/workflows/ci_gcu.yml
new file mode 100644
index 000000000..1e918cbdf
--- /dev/null
+++ b/.github/workflows/ci_gcu.yml
@@ -0,0 +1,89 @@
+name: CI_GCU
+
+on:
+  pull_request:
+    branches:
+      - develop
+      - 'release/*'
+  workflow_dispatch:
+
+concurrency:
+  group: ${{ github.event.pull_request.number }}-gcu-ci
+  cancel-in-progress: true
+
+jobs:
+  CI_GCU:
+    runs-on: [self-hosted, GCU-S60-8Card]
+    steps:
+      - name: Print current runner name
+        run: |
+          echo "Current runner name: ${{ runner.name }}"
+
+      - name: Code Checkout
+        env:
+          docker_image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-gcu:topsrider3.5.102-ubuntu20-x86_64-gcc84
+        run: |
+          REPO="https://github.com/${{ github.repository }}.git"
+          FULL_REPO="${{ github.repository }}"
+          REPO_NAME="${FULL_REPO##*/}"
+          BASE_BRANCH="${{ github.base_ref }}"
+          # Clean the repository directory before starting
+          docker run --rm --net=host -v $(pwd):/workspace -w /workspace \
+          -e "REPO_NAME=${REPO_NAME}" \
+          -e "BASE_BRANCH=${BASE_BRANCH}" \
+          ${docker_image} /bin/bash -c '
+            if [ -d ${REPO_NAME} ]; then
+              echo "Directory ${REPO_NAME} exists, removing it..."
+              rm -rf ${REPO_NAME}
+            fi
+          '
+          git config --global user.name "FastDeployCI"
+          git config --global user.email "fastdeploy_ci@example.com"
+          git clone ${REPO} ${REPO_NAME} -b ${BASE_BRANCH}
+          cd FastDeploy
+          if [ "${{ github.event_name }}" = "pull_request" ]; then
+            git fetch origin pull/${{ github.event.pull_request.number }}/head:pr/${{ github.event.pull_request.number }}
+            git merge pr/${{ github.event.pull_request.number }}
+            git log -n 3 --oneline
+          else
+            git checkout ${{ github.sha }}
+            git log -n 3 --oneline
+          fi
+
+      - name: Run CI unittest
+        env:
+          docker_image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-gcu:topsrider3.5.102-ubuntu20-x86_64-gcc84
+        run: |
+          runner_name="${{ runner.name }}"
+          last_char="${runner_name: -1}"
+
+          if [[ "$last_char" =~ [0-3] ]]; then
+            gcu_id="$last_char"
+          else
+            gcu_id="0"
+          fi
+          FD_API_PORT=$((9180 + gcu_id * 100))
+          FD_ENGINE_QUEUE_PORT=$((9150 + gcu_id * 100))
+          FD_METRICS_PORT=$((9170 + gcu_id * 100))
+
+          PARENT_DIR=$(dirname "$WORKSPACE")
+          echo "PARENT_DIR:$PARENT_DIR"
+          echo "Install drivers..."
+          cd /work/deps
+          bash TopsRider_i3x_*_deb_amd64.run --driver --no-auto-load -y
+          cd -
+          docker run --rm --network=host --ipc=host -it --privileged  \
+          -v $(pwd):/workspace -w /workspace \
+          -v "/home:/home" \
+          -v "/work:/work" \
+          -e "MODEL_PATH=/work/models" \
+          -e "http_proxy=$(git config --global --get http.proxy)" \
+          -e "https_proxy=$(git config --global --get https.proxy)" \
+          -e "FD_API_PORT=${FD_API_PORT}" \
+          -e "FD_ENGINE_QUEUE_PORT=${FD_ENGINE_QUEUE_PORT}" \
+          -e "FD_METRICS_PORT=${FD_METRICS_PORT}" \
+           ${docker_image} /bin/bash -c "
+          git config --global --add safe.directory /workspace/FastDeploy
+          cd FastDeploy
+          bash scripts/run_ci_gcu.sh
+          "
diff --git a/scripts/run_ci_gcu.sh b/scripts/run_ci_gcu.sh
new file mode 100644
index 000000000..76d4d1767
--- /dev/null
+++ b/scripts/run_ci_gcu.sh
@@ -0,0 +1,86 @@
+#!/bin/bash
+DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+echo "$DIR"
+
+#先kill一遍
+ps -efww | grep -E 'api_server' | grep -v grep | awk '{print $2}' | xargs kill -9 || true
+ps -efww | grep -E '8188' | grep -v grep | awk '{print $2}' | xargs kill -9 || true
+lsof -t -i :8188 | xargs kill -9 || true
+
+export model_path=${MODEL_PATH}/paddle/ERNIE-4.5-21B-A3B-Paddle
+
+echo "pip install requirements"
+python -m pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
+echo "uninstall org"
+python -m pip uninstall paddlepaddle -y
+python -m pip uninstall paddle-custom-gcu -y
+python -m pip install paddlepaddle==3.1.0a0 -i https://www.paddlepaddle.org.cn/packages/stable/cpu/
+echo "build whl"
+bash build.sh 1 || exit 1
+
+unset http_proxy
+unset https_proxy
+unset no_proxy
+
+# 起服务
+rm -rf log/*
+rm -f core*
+# pkill -9 python #流水线不执行这个
+#清空消息队列
+ipcrm --all=msg
+python -m fastdeploy.entrypoints.openai.api_server \
+    --model ${model_path} \
+    --port 8188 \
+    --metrics-port 8200 \
+    --tensor-parallel-size 4 \
+    --num-gpu-blocks-override 4096 \
+    --max-model-len 32768 \
+    --max-num-seqs 8 \
+    --quantization wint4 > server.log 2>&1 &
+
+sleep 60
+# 探活
+TIMEOUT=$((5 * 60))
+INTERVAL=10            # 检查间隔（秒）
+ENDPOINT="http://0.0.0.0:8188/health"
+START_TIME=$(date +%s) # 记录开始时间戳
+echo "开始服务健康检查，最长等待时间：${TIMEOUT}秒"
+while true; do
+    # 计算已耗时
+    CURRENT_TIME=$(date +%s)
+    ELAPSED=$((CURRENT_TIME - START_TIME))
+
+    # 超时判断
+    if [ $ELAPSED -ge $TIMEOUT ]; then
+        echo -e "\n服务启动超时：经过 $((TIMEOUT/60)) 分钟服务仍未启动！"
+        cat server.log
+        cat log/workerlog.0
+        exit 1
+    fi
+
+    HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" -m 2 "$ENDPOINT" || true)
+
+    if [ "$HTTP_CODE" = "200" ]; then
+        echo -e "\n服务启动成功！耗时 ${ELAPSED} 秒"
+        break
+    else
+        sleep $INTERVAL
+    fi
+done
+
+cat server.log
+
+# 执行服务化推理
+python test/ci_use/GCU/run_ernie.py
+exit_code=$?
+echo exit_code is ${exit_code}
+
+ps -efww | grep -E 'api_server' | grep -v grep | awk '{print $2}' | xargs kill -9 || true
+ps -efww | grep -E '8188' | grep -v grep | awk '{print $2}' | xargs kill -9 || true
+lsof -t -i :8188 | xargs kill -9 || true
+
+if [ ${exit_code} -ne 0 ]; then
+    echo "log/workerlog.0"
+    cat log/workerlog.0
+    exit 1
+fi
diff --git a/test/ci_use/GCU/run_ernie.py b/test/ci_use/GCU/run_ernie.py
new file mode 100644
index 000000000..f4e8a9ef9
--- /dev/null
+++ b/test/ci_use/GCU/run_ernie.py
@@ -0,0 +1,32 @@
+# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import openai
+
+ip = "0.0.0.0"
+service_http_port = "8188"  # 服务配置的
+client = openai.Client(base_url=f"http://{ip}:{service_http_port}/v1", api_key="EMPTY_API_KEY")
+
+# 非流式对话
+response = client.chat.completions.create(
+    model="default",
+    messages=[
+        {"role": "user", "content": "The largest ocean is"},
+    ],
+    temperature=1,
+    top_p=0,
+    max_tokens=64,
+    stream=False,
+)
+print(response)