Skip to content

Commit 0e5e305

Browse files
littledgglittledgg
committed
Merge branch 'develop' of https://github.com/PaddlePaddle/FastDeploy into draft
2 parents d587f21 + 32b3962 commit 0e5e305

File tree

205 files changed

+8975
-1708
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

205 files changed

+8975
-1708
lines changed

‎.github/workflows/_accuracy_test.yml

Lines changed: 174 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,174 @@
1+
name: Accuracy Test
2+
description: "Run Accuracy Tests"
3+
4+
on:
5+
workflow_call:
6+
inputs:
7+
DOCKER_IMAGE:
8+
description: "Build Images"
9+
required: true
10+
type: string
11+
default: "ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddleqa:cuda126-py310"
12+
FASTDEPLOY_ARCHIVE_URL:
13+
description: "URL of the compressed FastDeploy code archive."
14+
required: true
15+
type: string
16+
FASTDEPLOY_WHEEL_URL:
17+
description: "URL of the FastDeploy Wheel."
18+
required: true
19+
type: string
20+
CACHE_DIR:
21+
description: "Cache Dir Use"
22+
required: false
23+
type: string
24+
default: ""
25+
MODEL_CACHE_DIR:
26+
description: "Cache Dir Use"
27+
required: false
28+
type: string
29+
default: ""
30+
31+
jobs:
32+
accuracy_tests:
33+
runs-on: [self-hosted, GPU-h20-1Cards]
34+
steps:
35+
- name: Code Prepare
36+
shell: bash
37+
env:
38+
docker_image: ${{ inputs.DOCKER_IMAGE }}
39+
fd_archive_url: ${{ inputs.FASTDEPLOY_ARCHIVE_URL }}
40+
run: |
41+
set -x
42+
REPO="https://github.com/${{ github.repository }}.git"
43+
FULL_REPO="${{ github.repository }}"
44+
REPO_NAME="${FULL_REPO##*/}"
45+
BASE_BRANCH="${{ github.base_ref }}"
46+
47+
# Clean the repository directory before starting
48+
docker run --rm --net=host -v $(pwd):/workspace -w /workspace \
49+
-e "REPO_NAME=${REPO_NAME}" \
50+
${docker_image} /bin/bash -c '
51+
if [ -d ${REPO_NAME} ]; then
52+
echo "Directory ${REPO_NAME} exists, removing it..."
53+
rm -rf ${REPO_NAME}*
54+
fi
55+
'
56+
57+
wget -q ${fd_archive_url}
58+
tar -xf FastDeploy.tar.gz
59+
rm -rf FastDeploy.tar.gz
60+
cd FastDeploy
61+
git config --global user.name "FastDeployCI"
62+
git config --global user.email "fastdeploy_ci@example.com"
63+
git log -n 3 --oneline
64+
65+
- name: Run FastDeploy Base Tests
66+
shell: bash
67+
env:
68+
docker_image: ${{ inputs.DOCKER_IMAGE }}
69+
fastdeploy_wheel_url: ${{ inputs.FASTDEPLOY_WHEEL_URL }}
70+
CACHE_DIR: ${{ inputs.CACHE_DIR }}
71+
MODEL_CACHE_DIR: ${{ inputs.MODEL_CACHE_DIR }}
72+
run: |
73+
runner_name="${{ runner.name }}"
74+
CARD_ID=$(echo "${runner_name}" | awk -F'-' '{print $NF}')
75+
DEVICES=$(echo "$CARD_ID" | fold -w1 | paste -sd,)
76+
DEVICE_PORT=$(echo "$DEVICES" | cut -d',' -f1)
77+
78+
FLASK_PORT=$((42068 + DEVICE_PORT * 100))
79+
FD_API_PORT=$((42088 + DEVICE_PORT * 100))
80+
FD_ENGINE_QUEUE_PORT=$((42058 + DEVICE_PORT * 100))
81+
FD_METRICS_PORT=$((42078 + DEVICE_PORT * 100))
82+
echo "Test ENV Parameter:"
83+
echo "========================================================="
84+
echo "FLASK_PORT=${FLASK_PORT}"
85+
echo "FD_API_PORT=${FD_API_PORT}"
86+
echo "FD_ENGINE_QUEUE_PORT=${FD_ENGINE_QUEUE_PORT}"
87+
echo "FD_METRICS_PORT=${FD_METRICS_PORT}"
88+
echo "DEVICES=${DEVICES}"
89+
echo "========================================================="
90+
91+
CACHE_DIR="${CACHE_DIR:-$(dirname "$(dirname "${{ github.workspace }}")")}"
92+
echo "CACHE_DIR is set to ${CACHE_DIR}"
93+
if [ ! -f "${CACHE_DIR}/gitconfig" ]; then
94+
touch "${CACHE_DIR}/gitconfig"
95+
fi
96+
if [ ! -d "${MODEL_CACHE_DIR}" ]; then
97+
echo "Error: MODEL_CACHE_DIR '${MODEL_CACHE_DIR}' does not exist."
98+
exit 1
99+
fi
100+
101+
PORTS=($FLASK_PORT $FD_API_PORT $FD_ENGINE_QUEUE_PORT $FD_METRICS_PORT)
102+
LOG_FILE="./port_cleanup_$(date +%Y%m%d_%H%M%S).log"
103+
echo "==== LOG_FILE is ${LOG_FILE} ===="
104+
105+
echo "==== PORT CLEAN BEFORE TASK RUN ====" | tee -a $LOG_FILE
106+
107+
for port in "${PORTS[@]}"; do
108+
PIDS=$(lsof -t -i :$port || true)
109+
if [ -n "$PIDS" ]; then
110+
echo "Port $port is occupied by PID(s): $PIDS" | tee -a $LOG_FILE
111+
echo "$PIDS" | xargs -r kill -9
112+
echo "Port $port cleared" | tee -a $LOG_FILE
113+
else
114+
echo "Port $port is free" | tee -a $LOG_FILE
115+
fi
116+
done
117+
118+
echo "==== PORT CLEAN COMPLETE ====" | tee -a $LOG_FILE
119+
120+
docker run --rm --ipc=host --pid=host --net=host \
121+
-v $(pwd):/workspace \
122+
-w /workspace \
123+
-e fastdeploy_wheel_url=${fastdeploy_wheel_url} \
124+
-e "FD_API_PORT=${FD_API_PORT}" \
125+
-e "FD_ENGINE_QUEUE_PORT=${FD_ENGINE_QUEUE_PORT}" \
126+
-e "FD_METRICS_PORT=${FD_METRICS_PORT}" \
127+
-e "FLASK_PORT=${FLASK_PORT}" \
128+
-v "${MODEL_CACHE_DIR}:/MODELDATA" \
129+
-v "${CACHE_DIR}/gitconfig:/etc/gitconfig:ro" \
130+
-v "${CACHE_DIR}/.cache:/root/.cache" \
131+
-v "${CACHE_DIR}/ConfigDir:/root/.config" \
132+
-e TZ="Asia/Shanghai" \
133+
--gpus '"device='"${DEVICES}"'"' ${docker_image} /bin/bash -xc '
134+
python -m pip install --pre paddlepaddle-gpu -i https://www.paddlepaddle.org.cn/packages/nightly/cu126/
135+
136+
pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
137+
138+
python -m pip install ${fastdeploy_wheel_url}
139+
python -m pip install pytest
140+
141+
wget https://paddle-qa.bj.bcebos.com/zhengtianyu/tools/llm-deploy-linux-amd64
142+
chmod +x ./llm-deploy-linux-amd64
143+
./llm-deploy-linux-amd64 -python python3.10 \
144+
-model_name ERNIE-4.5-0.3B-Paddle \
145+
-model_path /MODELDATA \
146+
--skip install
147+
148+
git config --global --add safe.directory /workspace/FastDeploy
149+
cd FastDeploy
150+
pushd test/ce/deploy
151+
python3.10 deploy.py > dd.log 2>&1 &
152+
sleep 3
153+
curl -X POST http://0.0.0.0:${FLASK_PORT}/start \
154+
-H "Content-Type: application/json" \
155+
-d "{\"--model\": \"/MODELDATA/ERNIE-4.5-0.3B-Paddle\"}"
156+
157+
curl -X POST http://localhost:${FLASK_PORT}/wait_for_infer?timeout=90
158+
popd
159+
160+
pushd test/ce/accuracy_cases
161+
export URL=http://localhost:${FD_API_PORT}/v1/chat/completions
162+
export TEMPLATE=TOKEN_LOGPROB
163+
export MODEL_SIZE=0.3B
164+
TEST_EXIT_CODE=0
165+
python gsm8k.py || TEST_EXIT_CODE=1
166+
popd
167+
echo "TEST_EXIT_CODE=${TEST_EXIT_CODE}" >> /workspace/FastDeploy/exit_code.env
168+
'
169+
if [ -f ./FastDeploy/exit_code.env ]; then
170+
source ./FastDeploy/exit_code.env
171+
cat ./FastDeploy/exit_code.env >> $GITHUB_ENV
172+
fi
173+
echo "TEST_EXIT_CODE=${TEST_EXIT_CODE}"
174+
exit ${TEST_EXIT_CODE}

‎.github/workflows/_base_test.yml

Lines changed: 106 additions & 95 deletions
Original file line numberDiff line numberDiff line change
@@ -70,98 +70,109 @@ jobs:
7070
CACHE_DIR: ${{ inputs.CACHE_DIR }}
7171
MODEL_CACHE_DIR: ${{ inputs.MODEL_CACHE_DIR }}
7272
run: |
73-
runner_name="${{ runner.name }}"
74-
last_char="${runner_name: -1}"
75-
76-
if [[ "$last_char" =~ [0-7] ]]; then
77-
DEVICES="$last_char"
78-
else
79-
DEVICES="0"
80-
fi
81-
82-
FLASK_PORT=$((42068 + DEVICES * 100))
83-
FD_API_PORT=$((42088 + DEVICES * 100))
84-
FD_ENGINE_QUEUE_PORT=$((42058 + DEVICES * 100))
85-
FD_METRICS_PORT=$((42078 + DEVICES * 100))
86-
echo "Test ENV Parameter:"
87-
echo "========================================================="
88-
echo "FLASK_PORT=${FLASK_PORT}"
89-
echo "FD_API_PORT=${FD_API_PORT}"
90-
echo "FD_ENGINE_QUEUE_PORT=${FD_ENGINE_QUEUE_PORT}"
91-
echo "FD_METRICS_PORT=${FD_METRICS_PORT}"
92-
echo "DEVICES=${DEVICES}"
93-
echo "========================================================="
94-
95-
CACHE_DIR="${CACHE_DIR:-$(dirname "$(dirname "${{ github.workspace }}")")}"
96-
echo "CACHE_DIR is set to ${CACHE_DIR}"
97-
if [ ! -f "${CACHE_DIR}/gitconfig" ]; then
98-
touch "${CACHE_DIR}/gitconfig"
99-
fi
100-
if [ ! -d "${MODEL_CACHE_DIR}" ]; then
101-
echo "Error: MODEL_CACHE_DIR '${MODEL_CACHE_DIR}' does not exist."
102-
exit 1
103-
fi
104-
105-
PARENT_DIR=$(dirname "$WORKSPACE")
106-
107-
docker run --rm --ipc=host --pid=host --net=host \
108-
-v $(pwd):/workspace \
109-
-w /workspace \
110-
-e fastdeploy_wheel_url=${fastdeploy_wheel_url} \
111-
-e "FD_API_PORT=${FD_API_PORT}" \
112-
-e "FD_ENGINE_QUEUE_PORT=${FD_ENGINE_QUEUE_PORT}" \
113-
-e "FD_METRICS_PORT=${FD_METRICS_PORT}" \
114-
-e "FLASK_PORT=${FLASK_PORT}" \
115-
-v "${MODEL_CACHE_DIR}:/MODELDATA" \
116-
-v "${CACHE_DIR}/gitconfig:/etc/gitconfig:ro" \
117-
-v "${CACHE_DIR}/.cache:/root/.cache" \
118-
-v "${CACHE_DIR}/ConfigDir:/root/.config" \
119-
-e TZ="Asia/Shanghai" \
120-
--gpus '"device='"${DEVICES}"'"' ${docker_image} /bin/bash -xc '
121-
# python -m pip install --pre paddlepaddle-gpu -i https://www.paddlepaddle.org.cn/packages/nightly/cu126/
122-
python -m pip install paddlepaddle-gpu==3.0.0.dev20250729 -i https://www.paddlepaddle.org.cn/packages/nightly/cu126/
123-
124-
pip config set global.index-url http://pip.baidu.com/root/baidu/+simple/
125-
pip config set install.trusted-host pip.baidu.com
126-
pip config set global.extra-index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
127-
python -m pip install ${fastdeploy_wheel_url}
128-
python -m pip install pytest
129-
130-
wget https://paddle-qa.bj.bcebos.com/zhengtianyu/tools/llm-deploy-linux-amd64
131-
chmod +x ./llm-deploy-linux-amd64
132-
./llm-deploy-linux-amd64 -python python3.10 \
133-
-model_name ERNIE-4.5-0.3B-Paddle \
134-
-model_path /MODELDATA \
135-
--skip install
136-
137-
git config --global --add safe.directory /workspace/FastDeploy
138-
cd FastDeploy
139-
pushd test/ce/deploy
140-
python3.10 deploy.py > dd.log 2>&1 &
141-
sleep 3
142-
curl -X POST http://0.0.0.0:${FLASK_PORT}/start \
143-
-H "Content-Type: application/json" \
144-
-d "{\"--model\": \"/MODELDATA/ERNIE-4.5-0.3B-Paddle\"}"
145-
146-
curl -X POST http://localhost:${FLASK_PORT}/wait_for_infer?timeout=90
147-
popd
148-
149-
pushd test/ce/server
150-
export URL=http://localhost:${FD_API_PORT}/v1/chat/completions
151-
export TEMPLATE=TOKEN_LOGPROB
152-
TEST_EXIT_CODE=0
153-
python -m pytest -sv test_base_chat.py test_compare_top_logprobs.py test_logprobs.py test_params_boundary.py test_seed_usage.py test_stream.py || TEST_EXIT_CODE=1
154-
curl -X POST http://0.0.0.0:${FLASK_PORT}/switch \
155-
-H "Content-Type: application/json" \
156-
-d "{\"--model\": \"/MODELDATA/ERNIE-4.5-0.3B-Paddle\", \"--early-stop-config\": \"{\\\"enable_early_stop\\\":true, \\\"window_size\\\":6, \\\"threshold\\\":0.93}\"}"
157-
curl -X POST http://localhost:${FLASK_PORT}/wait_for_infer?timeout=90
158-
python -m pytest -sv test_repetition_early_stop.py || TEST_EXIT_CODE=1
159-
popd
160-
echo "TEST_EXIT_CODE=${TEST_EXIT_CODE}" >> /workspace/FastDeploy/exit_code.env
161-
'
162-
if [ -f ./FastDeploy/exit_code.env ]; then
163-
source ./FastDeploy/exit_code.env
164-
cat ./FastDeploy/exit_code.env >> $GITHUB_ENV
165-
fi
166-
echo "TEST_EXIT_CODE=${TEST_EXIT_CODE}"
167-
exit ${TEST_EXIT_CODE}
73+
runner_name="${{ runner.name }}"
74+
CARD_ID=$(echo "${runner_name}" | awk -F'-' '{print $NF}')
75+
DEVICES=$(echo "$CARD_ID" | fold -w1 | paste -sd,)
76+
DEVICE_PORT=$(echo "$DEVICES" | cut -d',' -f1)
77+
78+
FLASK_PORT=$((42068 + DEVICE_PORT * 100))
79+
FD_API_PORT=$((42088 + DEVICE_PORT * 100))
80+
FD_ENGINE_QUEUE_PORT=$((42058 + DEVICE_PORT * 100))
81+
FD_METRICS_PORT=$((42078 + DEVICE_PORT * 100))
82+
echo "Test ENV Parameter:"
83+
echo "========================================================="
84+
echo "FLASK_PORT=${FLASK_PORT}"
85+
echo "FD_API_PORT=${FD_API_PORT}"
86+
echo "FD_ENGINE_QUEUE_PORT=${FD_ENGINE_QUEUE_PORT}"
87+
echo "FD_METRICS_PORT=${FD_METRICS_PORT}"
88+
echo "DEVICES=${DEVICES}"
89+
echo "========================================================="
90+
91+
CACHE_DIR="${CACHE_DIR:-$(dirname "$(dirname "${{ github.workspace }}")")}"
92+
echo "CACHE_DIR is set to ${CACHE_DIR}"
93+
if [ ! -f "${CACHE_DIR}/gitconfig" ]; then
94+
touch "${CACHE_DIR}/gitconfig"
95+
fi
96+
if [ ! -d "${MODEL_CACHE_DIR}" ]; then
97+
echo "Error: MODEL_CACHE_DIR '${MODEL_CACHE_DIR}' does not exist."
98+
exit 1
99+
fi
100+
101+
PORTS=($FLASK_PORT $FD_API_PORT $FD_ENGINE_QUEUE_PORT $FD_METRICS_PORT)
102+
LOG_FILE="./port_cleanup_$(date +%Y%m%d_%H%M%S).log"
103+
echo "==== LOG_FILE is ${LOG_FILE} ===="
104+
105+
echo "==== PORT CLEAN BEFORE TASK RUN ====" | tee -a $LOG_FILE
106+
107+
for port in "${PORTS[@]}"; do
108+
PIDS=$(lsof -t -i :$port || true)
109+
if [ -n "$PIDS" ]; then
110+
echo "Port $port is occupied by PID(s): $PIDS" | tee -a $LOG_FILE
111+
echo "$PIDS" | xargs -r kill -9
112+
echo "Port $port cleared" | tee -a $LOG_FILE
113+
else
114+
echo "Port $port is free" | tee -a $LOG_FILE
115+
fi
116+
done
117+
118+
echo "==== PORT CLEAN COMPLETE ====" | tee -a $LOG_FILE
119+
120+
docker run --rm --ipc=host --pid=host --net=host \
121+
-v $(pwd):/workspace \
122+
-w /workspace \
123+
-e fastdeploy_wheel_url=${fastdeploy_wheel_url} \
124+
-e "FD_API_PORT=${FD_API_PORT}" \
125+
-e "FD_ENGINE_QUEUE_PORT=${FD_ENGINE_QUEUE_PORT}" \
126+
-e "FD_METRICS_PORT=${FD_METRICS_PORT}" \
127+
-e "FLASK_PORT=${FLASK_PORT}" \
128+
-v "${MODEL_CACHE_DIR}:/MODELDATA" \
129+
-v "${CACHE_DIR}/gitconfig:/etc/gitconfig:ro" \
130+
-v "${CACHE_DIR}/.cache:/root/.cache" \
131+
-v "${CACHE_DIR}/ConfigDir:/root/.config" \
132+
-e TZ="Asia/Shanghai" \
133+
--gpus '"device='"${DEVICES}"'"' ${docker_image} /bin/bash -xc '
134+
python -m pip install --pre paddlepaddle-gpu -i https://www.paddlepaddle.org.cn/packages/nightly/cu126/
135+
136+
pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
137+
138+
python -m pip install ${fastdeploy_wheel_url}
139+
python -m pip install pytest
140+
141+
wget https://paddle-qa.bj.bcebos.com/zhengtianyu/tools/llm-deploy-linux-amd64
142+
chmod +x ./llm-deploy-linux-amd64
143+
./llm-deploy-linux-amd64 -python python3.10 \
144+
-model_name ERNIE-4.5-0.3B-Paddle \
145+
-model_path /MODELDATA \
146+
--skip install
147+
148+
git config --global --add safe.directory /workspace/FastDeploy
149+
cd FastDeploy
150+
pushd test/ce/deploy
151+
python3.10 deploy.py > dd.log 2>&1 &
152+
sleep 3
153+
curl -X POST http://0.0.0.0:${FLASK_PORT}/start \
154+
-H "Content-Type: application/json" \
155+
-d "{\"--model\": \"/MODELDATA/ERNIE-4.5-0.3B-Paddle\"}"
156+
157+
curl -X POST http://localhost:${FLASK_PORT}/wait_for_infer?timeout=90
158+
popd
159+
160+
pushd test/ce/server
161+
export URL=http://localhost:${FD_API_PORT}/v1/chat/completions
162+
export TEMPLATE=TOKEN_LOGPROB
163+
TEST_EXIT_CODE=0
164+
python -m pytest -sv test_base_chat.py test_compare_top_logprobs.py test_logprobs.py test_params_boundary.py test_seed_usage.py test_stream.py test_evil_cases.py || TEST_EXIT_CODE=1
165+
curl -X POST http://0.0.0.0:${FLASK_PORT}/switch \
166+
-H "Content-Type: application/json" \
167+
-d "{\"--model\": \"/MODELDATA/ERNIE-4.5-0.3B-Paddle\", \"--early-stop-config\": \"{\\\"enable_early_stop\\\":true, \\\"window_size\\\":6, \\\"threshold\\\":0.93}\"}"
168+
curl -X POST http://localhost:${FLASK_PORT}/wait_for_infer?timeout=90
169+
python -m pytest -sv test_repetition_early_stop.py || TEST_EXIT_CODE=1
170+
popd
171+
echo "TEST_EXIT_CODE=${TEST_EXIT_CODE}" >> /workspace/FastDeploy/exit_code.env
172+
'
173+
if [ -f ./FastDeploy/exit_code.env ]; then
174+
source ./FastDeploy/exit_code.env
175+
cat ./FastDeploy/exit_code.env >> $GITHUB_ENV
176+
fi
177+
echo "TEST_EXIT_CODE=${TEST_EXIT_CODE}"
178+
exit ${TEST_EXIT_CODE}

0 commit comments

Comments
 (0)