xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-25 22:20:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into fix_is_usage_stats_enabled_when_disable_it

Browse Source
This commit is contained in:
Ning Xie 2025-12-17 10:00:13 +08:00 committed by GitHub
commit 5d1ee758d8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2883 changed files with 380578 additions and 226476 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
.buildkite/check-wheel-size.py
View File

@ -5,11 +5,11 @@ import os
import sys
import zipfile
# Read the VLLM_MAX_SIZE_MB environment variable, defaulting to 450 MiB
# Read the VLLM_MAX_SIZE_MB environment variable, defaulting to 500 MiB
# Note that we have 800 MiB quota, please use it wisely.
# See https://github.com/pypi/support/issues/6326 .
# Please also sync the value with the one in Dockerfile.
VLLM_MAX_SIZE_MB = int(os.environ.get("VLLM_MAX_SIZE_MB", 450))
VLLM_MAX_SIZE_MB = int(os.environ.get("VLLM_MAX_SIZE_MB", 500))
def print_top_10_largest_files(zip_file):

24
.buildkite/ci_config.yaml Normal file
View File

@ -0,0 +1,24 @@
name: vllm_ci
job_dirs:
- ".buildkite/test_areas"
- ".buildkite/image_build"
run_all_patterns:
- "docker/Dockerfile"
- "CMakeLists.txt"
- "requirements/common.txt"
- "requirements/cuda.txt"
- "requirements/build.txt"
- "requirements/test.txt"
- "setup.py"
- "csrc/"
- "cmake/"
run_all_exclude_patterns:
- "docker/Dockerfile."
- "csrc/cpu/"
- "csrc/rocm/"
- "cmake/hipify.py"
- "cmake/cpu_extension.cmake"
registries: public.ecr.aws/q9t5s3a7
repositories:
main: "vllm-ci-postmerge-repo"
premerge: "vllm-ci-test-repo"

46
.buildkite/generate_index.py
View File

@ -1,46 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
import os
template = """<!DOCTYPE html>
<html>
<body>
<h1>Links for vLLM</h1/>
<a href="../{x86_wheel_html_escaped}">{x86_wheel}</a><br/>
<a href="../{arm_wheel_html_escaped}">{arm_wheel}</a><br/>
</body>
</html>
"""
parser = argparse.ArgumentParser()
parser.add_argument("--wheel", help="The wheel path.", required=True)
args = parser.parse_args()
filename = os.path.basename(args.wheel)
with open("index.html", "w") as f:
print(f"Generated index.html for {args.wheel}")
# sync the abi tag with .buildkite/scripts/upload-wheels.sh
if "x86_64" in filename:
x86_wheel = filename
arm_wheel = filename.replace("x86_64", "aarch64").replace(
"manylinux1", "manylinux2014"
)
elif "aarch64" in filename:
x86_wheel = filename.replace("aarch64", "x86_64").replace(
"manylinux2014", "manylinux1"
)
arm_wheel = filename
else:
raise ValueError(f"Unsupported wheel: {filename}")
# cloudfront requires escaping the '+' character
f.write(
template.format(
x86_wheel=x86_wheel,
x86_wheel_html_escaped=x86_wheel.replace("+", "%2B"),
arm_wheel=arm_wheel,
arm_wheel_html_escaped=arm_wheel.replace("+", "%2B"),
)
)

56
.buildkite/image_build/image_build.sh Executable file
View File

@ -0,0 +1,56 @@
#!/bin/bash
set -e
if [[ $# -lt 8 ]]; then
echo "Usage: $0 <registry> <repo> <commit> <branch> <vllm_use_precompiled> <vllm_merge_base_commit> <cache_from> <cache_to>"
exit 1
fi
REGISTRY=$1
REPO=$2
BUILDKITE_COMMIT=$3
BRANCH=$4
VLLM_USE_PRECOMPILED=$5
VLLM_MERGE_BASE_COMMIT=$6
CACHE_FROM=$7
CACHE_TO=$8
# authenticate with AWS ECR
aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin $REGISTRY
aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 936637512419.dkr.ecr.us-east-1.amazonaws.com
# docker buildx
docker buildx create --name vllm-builder --driver docker-container --use
docker buildx inspect --bootstrap
docker buildx ls
# skip build if image already exists
if [[ -z $(docker manifest inspect $REGISTRY/$REPO:$BUILDKITE_COMMIT) ]]; then
echo "Image not found, proceeding with build..."
else
echo "Image found"
exit 0
fi
if [[ "${VLLM_USE_PRECOMPILED:-0}" == "1" ]]; then
merge_base_commit_build_args="--build-arg VLLM_MERGE_BASE_COMMIT=${VLLM_MERGE_BASE_COMMIT}"
else
merge_base_commit_build_args=""
fi
# build
docker buildx build --file docker/Dockerfile \
--build-arg max_jobs=16 \
--build-arg buildkite_commit=$BUILDKITE_COMMIT \
--build-arg USE_SCCACHE=1 \
--build-arg TORCH_CUDA_ARCH_LIST="8.0 8.9 9.0 10.0" \
--build-arg FI_TORCH_CUDA_ARCH_LIST="8.0 8.9 9.0a 10.0a" \
--build-arg VLLM_USE_PRECOMPILED="${VLLM_USE_PRECOMPILED:-0}" \
${merge_base_commit_build_args} \
--cache-from type=registry,ref=${CACHE_FROM},mode=max \
--cache-to type=registry,ref=${CACHE_TO},mode=max \
--tag ${REGISTRY}/${REPO}:${BUILDKITE_COMMIT} \
$( [[ "${BRANCH}" == "main" ]] && echo "--tag ${REGISTRY}/${REPO}:latest" ) \
--push \
--target test \
--progress plain .

57
.buildkite/image_build/image_build.yaml Normal file
View File

@ -0,0 +1,57 @@
group: Abuild
steps:
- label: ":docker: Build image"
key: image-build
depends_on: []
commands:
- .buildkite/image_build/image_build.sh $REGISTRY $REPO $BUILDKITE_COMMIT $BRANCH $VLLM_USE_PRECOMPILED $VLLM_MERGE_BASE_COMMIT $CACHE_FROM $CACHE_TO
retry:
automatic:
- exit_status: -1 # Agent was lost
limit: 2
- exit_status: -10 # Agent was lost
limit: 2
- label: ":docker: Build CPU image"
key: image-build-cpu
depends_on: []
commands:
- .buildkite/image_build/image_build_cpu.sh $REGISTRY $REPO $BUILDKITE_COMMIT
env:
DOCKER_BUILDKIT: "1"
retry:
automatic:
- exit_status: -1 # Agent was lost
limit: 2
- exit_status: -10 # Agent was lost
limit: 2
- label: ":docker: Build HPU image"
soft_fail: true
depends_on: []
key: image-build-hpu
commands:
- .buildkite/image_build/image_build_hpu.sh $REGISTRY $REPO $BUILDKITE_COMMIT
env:
DOCKER_BUILDKIT: "1"
retry:
automatic:
- exit_status: -1 # Agent was lost
limit: 2
- exit_status: -10 # Agent was lost
limit: 2
- label: ":docker: Build CPU arm64 image"
key: cpu-arm64-image-build
depends_on: []
optional: true
commands:
- .buildkite/image_build/image_build_cpu_arm64.sh $REGISTRY $REPO $BUILDKITE_COMMIT
env:
DOCKER_BUILDKIT: "1"
retry:
automatic:
- exit_status: -1 # Agent was lost
limit: 2
- exit_status: -10 # Agent was lost
limit: 2

36
.buildkite/image_build/image_build_cpu.sh Executable file
View File

@ -0,0 +1,36 @@
#!/bin/bash
set -e
if [[ $# -lt 3 ]]; then
echo "Usage: $0 <registry> <repo> <commit>"
exit 1
fi
REGISTRY=$1
REPO=$2
BUILDKITE_COMMIT=$3
# authenticate with AWS ECR
aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin $REGISTRY
# skip build if image already exists
if [[ -z $(docker manifest inspect $REGISTRY/$REPO:$BUILDKITE_COMMIT-cpu) ]]; then
echo "Image not found, proceeding with build..."
else
echo "Image found"
exit 0
fi
# build
docker build --file docker/Dockerfile.cpu \
--build-arg max_jobs=16 \
--build-arg buildkite_commit=$BUILDKITE_COMMIT \
--build-arg VLLM_CPU_AVX512BF16=true \
--build-arg VLLM_CPU_AVX512VNNI=true \
--build-arg VLLM_CPU_AMXBF16=true \
--tag $REGISTRY/$REPO:$BUILDKITE_COMMIT-cpu \
--target vllm-test \
--progress plain .
# push
docker push $REGISTRY/$REPO:$BUILDKITE_COMMIT-cpu

33
.buildkite/image_build/image_build_cpu_arm64.sh Executable file
View File

@ -0,0 +1,33 @@
#!/bin/bash
set -e
if [[ $# -lt 3 ]]; then
echo "Usage: $0 <registry> <repo> <commit>"
exit 1
fi
REGISTRY=$1
REPO=$2
BUILDKITE_COMMIT=$3
# authenticate with AWS ECR
aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin $REGISTRY
# skip build if image already exists
if [[ -z $(docker manifest inspect $REGISTRY/$REPO:$BUILDKITE_COMMIT-cpu) ]]; then
echo "Image not found, proceeding with build..."
else
echo "Image found"
exit 0
fi
# build
docker build --file docker/Dockerfile.cpu \
--build-arg max_jobs=16 \
--build-arg buildkite_commit=$BUILDKITE_COMMIT \
--tag $REGISTRY/$REPO:$BUILDKITE_COMMIT-cpu \
--target vllm-test \
--progress plain .
# push
docker push $REGISTRY/$REPO:$BUILDKITE_COMMIT-cpu

34
.buildkite/image_build/image_build_hpu.sh Executable file
View File

@ -0,0 +1,34 @@
#!/bin/bash
set -e
if [[ $# -lt 3 ]]; then
echo "Usage: $0 <registry> <repo> <commit>"
exit 1
fi
REGISTRY=$1
REPO=$2
BUILDKITE_COMMIT=$3
# authenticate with AWS ECR
aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin $REGISTRY
# skip build if image already exists
if [[ -z $(docker manifest inspect $REGISTRY/$REPO:$BUILDKITE_COMMIT-hpu) ]]; then
echo "Image not found, proceeding with build..."
else
echo "Image found"
exit 0
fi
# build
docker build \
--file tests/pytorch_ci_hud_benchmark/Dockerfile.hpu \
--build-arg max_jobs=16 \
--build-arg buildkite_commit=$BUILDKITE_COMMIT \
--tag $REGISTRY/$REPO:$BUILDKITE_COMMIT-hpu \
--progress plain \
https://github.com/vllm-project/vllm-gaudi.git
# push
docker push $REGISTRY/$REPO:$BUILDKITE_COMMIT-hpu

8
.buildkite/lm-eval-harness/configs/Qwen2-1.5B-Instruct-W8A16-compressed-tensors.yaml → .buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-QQQ.yaml
View File

@ -1,12 +1,12 @@
# For vllm script, with -t option (tensor parallel size).
# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m nm-testing/Qwen2-1.5B-Instruct-W8A16-Channelwise -b "auto" -l 1000 -f 5 -t 1
model_name: "nm-testing/Qwen2-1.5B-Instruct-W8A16-Channelwise"
# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m HandH1998/QQQ-Llama-3-8b-g128 -b 32 -l 1000 -f 5 -t 1
model_name: "HandH1998/QQQ-Llama-3-8b-g128"
tasks:
- name: "gsm8k"
metrics:
- name: "exact_match,strict-match"
value: 0.595
value: 0.419
- name: "exact_match,flexible-extract"
value: 0.582
value: 0.416
limit: 1000
num_fewshot: 5

12
.buildkite/lm-eval-harness/configs/Meta-Llama-4-Maverick-17B-128E-Instruct-FP8-MM.yaml Normal file
View File

@ -0,0 +1,12 @@
# For hf script, without -t option (tensor parallel size).
# bash .buildkite/lm-eval-harness/run-lm-eval-chartqa-vllm-vlm-baseline.sh -m meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8 -l 100 -t 8
model_name: "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8"
backend: "vllm-vlm"
tasks:
- name: "chartqa"
metrics:
- name: "relaxed_accuracy,none"
# TODO(zhewenl): model card is 0.90, but the actual score is 0.80.
value: 0.80
limit: 100
num_fewshot: 0

11
.buildkite/lm-eval-harness/configs/Meta-Llama-4-Maverick-17B-128E-Instruct-FP8.yaml Normal file
View File

@ -0,0 +1,11 @@
# For hf script, without -t option (tensor parallel size).
# bash .buildkite/lm-eval-harness/run-lm-eval-mmlupro-vllm-baseline.sh -m meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8 -l 250 -t 8 -f 5
model_name: "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8"
tasks:
- name: "mmlu_pro"
metrics:
- name: "exact_match,custom-extract"
value: 0.80
limit: 250 # will run on 250 * 14 subjects = 3500 samples
num_fewshot: 5
rtol: 0.05

3
.buildkite/lm-eval-harness/configs/Qwen2.5-VL-3B-Instruct-FP8-dynamic.yaml
View File

@ -1,4 +1,5 @@
# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic -b auto -l 1319 -f 5 -t 1
# For vllm script, with -t option (tensor parallel size)
# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic -l 1319 -t 1
model_name: "RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic"
tasks:
- name: "gsm8k"

12
.buildkite/lm-eval-harness/configs/Qwen2.5-VL-7B-Instruct.yaml Normal file
View File

@ -0,0 +1,12 @@
# For vllm script, with -t option (tensor parallel size).
# bash .buildkite/lm-eval-harness/run-lm-eval-chartqa-vllm-vlm-baseline.sh -m Qwen/Qwen2.5-VL-7B-Instruct -l 2500 -t 1
model_name: "Qwen/Qwen2.5-VL-7B-Instruct"
backend: "vllm-vlm"
tasks:
- name: "chartqa"
metrics:
- name: "relaxed_accuracy,none"
value: 0.855
limit: 2500
num_fewshot: 0

14
.buildkite/lm-eval-harness/configs/Qwen3-235B-A22B-Instruct-2507-FP8.yaml Normal file
View File

@ -0,0 +1,14 @@
model_name: "Qwen/Qwen3-235B-A22B-Instruct-2507-FP8"
tasks:
- name: "mmlu_pro"
metrics:
- name: "exact_match,custom-extract"
value: 0.82
limit: 250 # will run on 250 * 14 subjects = 3500 samples
num_fewshot: 5
enforce_eager: false # we use false to speed up the eval process
kv_cache_dtype: fp8 # we use fp8 to speed up the eval process
max_model_len: 40960
apply_chat_template: true
fewshot_as_multiturn: true
gen_kwargs: "temperature=0,top_p=1,top_k=0,max_gen_toks=5632,until=<|ENDANSWER|>"

1
.buildkite/lm-eval-harness/configs/models-large-hopper.txt Normal file
View File

@ -0,0 +1 @@
Qwen3-235B-A22B-Instruct-2507-FP8.yaml

1
.buildkite/lm-eval-harness/configs/models-large-rocm.txt Normal file
View File

@ -0,0 +1 @@
Meta-Llama-4-Maverick-17B-128E-Instruct-FP8.yaml

1
.buildkite/lm-eval-harness/configs/models-mm-large-h100.txt Normal file
View File

@ -0,0 +1 @@
Meta-Llama-4-Maverick-17B-128E-Instruct-FP8-MM.yaml

1
.buildkite/lm-eval-harness/configs/models-mm-small.txt Normal file
View File

@ -0,0 +1 @@
Qwen2.5-VL-7B-Instruct.yaml

44
.buildkite/lm-eval-harness/run-lm-eval-chartqa-vllm-vlm-baseline.sh Executable file
View File

@ -0,0 +1,44 @@
#!/bin/bash
# We can use this script to compute baseline accuracy on chartqa for vllm.
#
# Make sure you have lm-eval-harness installed:
# pip install lm-eval==0.4.9
usage() {
echo``
echo "Runs lm eval harness on ChartQA using multimodal vllm."
echo "This pathway is intended to be used to create baselines for "
echo "our correctness tests in vllm's CI."
echo
echo "usage: ${0} <options>"
echo
echo " -m - huggingface stub or local directory of the model"
echo " -l - limit number of samples to run"
echo " -t - tensor parallel size to run at"
echo
}
while getopts "m:l:t:" OPT; do
case ${OPT} in
m )
MODEL="$OPTARG"
;;
l )
LIMIT="$OPTARG"
;;
t )
TP_SIZE="$OPTARG"
;;
\? )
usage
exit 1
;;
esac
done
lm_eval --model vllm-vlm \
--model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE" \
--tasks chartqa \
--batch_size auto \
--apply_chat_template \
--limit $LIMIT

0
.buildkite/lm-eval-harness/run-lm-eval-gsm-hf-baseline.sh Normal file → Executable file
View File

50
.buildkite/lm-eval-harness/run-lm-eval-mmlupro-vllm-baseline.sh Normal file
View File

@ -0,0 +1,50 @@
#!/bin/bash
# We can use this script to compute baseline accuracy on MMLUPRO for vllm.
# We use this for fp8, which HF does not support.
#
# Make sure you have lm-eval-harness installed:
# pip install git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d#egg=lm-eval[api]
usage() {
echo``
echo "Runs lm eval harness on MMLU Pro using huggingface transformers."
echo "This pathway is intended to be used to create baselines for "
echo "our automated nm-test-accuracy workflow"
echo
echo "usage: ${0} <options>"
echo
echo " -m - huggingface stub or local directory of the model"
echo " -l - limit number of samples to run"
echo " -f - number of fewshot samples to use"
echo " -t - tensor parallel size to run at"
echo
}
while getopts "m:b:l:f:t:" OPT; do
case ${OPT} in
m )
MODEL="$OPTARG"
;;
b )
BATCH_SIZE="$OPTARG"
;;
l )
LIMIT="$OPTARG"
;;
f )
FEWSHOT="$OPTARG"
;;
t )
TP_SIZE="$OPTARG"
;;
\? )
usage
exit 1
;;
esac
done
lm_eval --model vllm \
--model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE,add_bos_token=true,trust_remote_code=true,max_model_len=4096" \
--tasks mmlu_pro --num_fewshot "$FEWSHOT" --limit "$LIMIT" \
--batch_size auto

75
.buildkite/lm-eval-harness/test_lm_eval_correctness.py
View File

@ -9,32 +9,78 @@ pytest -s -v test_lm_eval_correctness.py \
--tp-size=1
"""
import os
from contextlib import contextmanager
import lm_eval
import numpy as np
import yaml
RTOL = 0.08
DEFAULT_RTOL = 0.08
@contextmanager
def scoped_env_vars(new_env: dict[str, str]):
if not new_env:
# Fast path: nothing to do
yield
return
old_values = {}
new_keys = []
try:
for key, value in new_env.items():
if key in os.environ:
old_values[key] = os.environ[key]
else:
new_keys.append(key)
os.environ[key] = str(value)
yield
finally:
# Restore / clean up
for key, value in old_values.items():
os.environ[key] = value
for key in new_keys:
os.environ.pop(key, None)
def launch_lm_eval(eval_config, tp_size):
trust_remote_code = eval_config.get("trust_remote_code", False)
max_model_len = eval_config.get("max_model_len", 4096)
batch_size = eval_config.get("batch_size", "auto")
backend = eval_config.get("backend", "vllm")
enforce_eager = eval_config.get("enforce_eager", "true")
kv_cache_dtype = eval_config.get("kv_cache_dtype", "auto")
model_args = (
f"pretrained={eval_config['model_name']},"
f"tensor_parallel_size={tp_size},"
f"enforce_eager=true,"
f"enforce_eager={enforce_eager},"
f"kv_cache_dtype={kv_cache_dtype},"
f"add_bos_token=true,"
f"trust_remote_code={trust_remote_code},"
f"max_model_len={max_model_len}"
)
results = lm_eval.simple_evaluate(
model="vllm",
model_args=model_args,
tasks=[task["name"] for task in eval_config["tasks"]],
num_fewshot=eval_config["num_fewshot"],
limit=eval_config["limit"],
batch_size="auto",
f"max_model_len={max_model_len},"
)
env_vars = eval_config.get("env_vars", None)
with scoped_env_vars(env_vars):
results = lm_eval.simple_evaluate(
model=backend,
model_args=model_args,
tasks=[task["name"] for task in eval_config["tasks"]],
num_fewshot=eval_config["num_fewshot"],
limit=eval_config["limit"],
# TODO(yeq): using chat template w/ fewshot_as_multiturn is supposed help
# text models. however, this is regressing measured strict-match for
# existing text models in CI, so only apply it for mm, or explicitly set
apply_chat_template=eval_config.get(
"apply_chat_template", backend == "vllm-vlm"
),
fewshot_as_multiturn=eval_config.get("fewshot_as_multiturn", False),
# Forward decoding and early-stop controls (e.g., max_gen_toks, until=...)
gen_kwargs=eval_config.get("gen_kwargs"),
batch_size=batch_size,
)
return results
@ -43,6 +89,8 @@ def test_lm_eval_correctness_param(config_filename, tp_size):
results = launch_lm_eval(eval_config, tp_size)
rtol = eval_config.get("rtol", DEFAULT_RTOL)
success = True
for task in eval_config["tasks"]:
for metric in task["metrics"]:
@ -50,8 +98,9 @@ def test_lm_eval_correctness_param(config_filename, tp_size):
measured_value = results["results"][task["name"]][metric["name"]]
print(
f"{task['name']} | {metric['name']}: "
f"ground_truth={ground_truth} | measured={measured_value}"
f"ground_truth={ground_truth:.3f} | "
f"measured={measured_value:.3f} | rtol={rtol}"
)
success = success and np.isclose(ground_truth, measured_value, rtol=RTOL)
success = success and np.isclose(ground_truth, measured_value, rtol=rtol)
assert success

184
.buildkite/nightly-benchmarks/benchmark-pipeline.yaml
View File

@ -1,184 +0,0 @@
steps:
- label: "Wait for container to be ready"
key: wait-for-container-image
agents:
queue: A100
plugins:
- kubernetes:
podSpec:
containers:
- image: badouralix/curl-jq
command:
- sh .buildkite/nightly-benchmarks/scripts/wait-for-image.sh
- label: "Cleanup H100"
agents:
queue: H100
depends_on: ~
command: docker system prune -a --volumes --force
- label: "A100"
# skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
agents:
queue: A100
depends_on: wait-for-container-image
if: build.branch == "main"
plugins:
- kubernetes:
podSpec:
priorityClassName: perf-benchmark
containers:
- image: public.ecr.aws/q9t5s3a7/vllm-ci-postmerge-repo:$BUILDKITE_COMMIT
command:
- bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
resources:
limits:
nvidia.com/gpu: 8
volumeMounts:
- name: devshm
mountPath: /dev/shm
env:
- name: VLLM_USAGE_SOURCE
value: ci-test
- name: HF_TOKEN
valueFrom:
secretKeyRef:
name: hf-token-secret
key: token
nodeSelector:
nvidia.com/gpu.product: NVIDIA-A100-SXM4-80GB
volumes:
- name: devshm
emptyDir:
medium: Memory
- label: "H200"
# skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
agents:
queue: H200
depends_on: wait-for-container-image
if: build.branch == "main"
plugins:
- docker#v5.12.0:
image: public.ecr.aws/q9t5s3a7/vllm-ci-postmerge-repo:$BUILDKITE_COMMIT
command:
- bash
- .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
mount-buildkite-agent: true
propagate-environment: true
ipc: host
gpus: 4,5,6,7
volumes:
- /data/benchmark-hf-cache:/root/.cache/huggingface
environment:
- VLLM_USAGE_SOURCE
- HF_TOKEN
#- block: "Run H100 Benchmark"
#key: block-h100
#depends_on: ~
- label: "H100"
# skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
agents:
queue: H100
depends_on: wait-for-container-image
if: build.branch == "main"
plugins:
- docker#v5.12.0:
image: public.ecr.aws/q9t5s3a7/vllm-ci-postmerge-repo:$BUILDKITE_COMMIT
command:
- bash
- .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
mount-buildkite-agent: true
propagate-environment: true
ipc: host
gpus: all # see CUDA_VISIBLE_DEVICES for actual GPUs used
volumes:
- /data/benchmark-hf-cache:/root/.cache/huggingface
environment:
- VLLM_USAGE_SOURCE
- HF_TOKEN
# Premerge benchmark
- label: "A100"
# skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
agents:
queue: A100
depends_on: wait-for-container-image
if: build.branch != "main"
plugins:
- kubernetes:
podSpec:
priorityClassName: perf-benchmark
containers:
- image: public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT
command:
- bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
resources:
limits:
nvidia.com/gpu: 8
volumeMounts:
- name: devshm
mountPath: /dev/shm
env:
- name: VLLM_USAGE_SOURCE
value: ci-test
- name: HF_TOKEN
valueFrom:
secretKeyRef:
name: hf-token-secret
key: token
nodeSelector:
nvidia.com/gpu.product: NVIDIA-A100-SXM4-80GB
volumes:
- name: devshm
emptyDir:
medium: Memory
- label: "H200"
# skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
agents:
queue: H200
depends_on: wait-for-container-image
if: build.branch != "main"
plugins:
- docker#v5.12.0:
image: public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT
command:
- bash
- .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
mount-buildkite-agent: true
propagate-environment: true
ipc: host
gpus: 4,5,6,7
volumes:
- /data/benchmark-hf-cache:/root/.cache/huggingface
environment:
- VLLM_USAGE_SOURCE
- HF_TOKEN
#- block: "Run H100 Benchmark"
#key: block-h100
#depends_on: ~
- label: "H100"
# skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
agents:
queue: H100
depends_on: wait-for-container-image
if: build.branch != "main"
plugins:
- docker#v5.12.0:
image: public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT
command:
- bash
- .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
mount-buildkite-agent: true
propagate-environment: true
ipc: host
gpus: all # see CUDA_VISIBLE_DEVICES for actual GPUs used
volumes:
- /data/benchmark-hf-cache:/root/.cache/huggingface
environment:
- VLLM_USAGE_SOURCE
- HF_TOKEN

28
.buildkite/nightly-benchmarks/nightly-annotation.md
View File

@ -1,28 +0,0 @@
# Nightly benchmark annotation
## Description
This file contains the downloading link for benchmarking results.
- [benchmarking pipeline](artifact://nightly-pipeline.yaml)
- [benchmarking results](artifact://results.zip)
- [benchmarking code](artifact://nightly-benchmarks.zip)
Please download the visualization scripts in the post
## Results reproduction
- Find the docker we use in `benchmarking pipeline`
- Deploy the docker, and inside the docker:
- Download `nightly-benchmarks.zip`.
- In the same folder, run the following code:
```bash
export HF_TOKEN=<your HF token>
apt update
apt install -y git
unzip nightly-benchmarks.zip
VLLM_SOURCE_CODE_LOC=./ bash .buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh
```
And the results will be inside `./benchmarks/results`.

39
.buildkite/nightly-benchmarks/nightly-descriptions.md
View File

@ -1,39 +0,0 @@
# Nightly benchmark
This benchmark aims to:
- Provide performance clarity: Provide clarity on which one (vllm, tensorrt-llm, lmdeploy and SGLang) leads in performance in what workload.
- Be reproducible: one can run the exact same set of benchmarking commands inside the exact same docker by following reproducing instructions.
Latest results: [results link](https://blog.vllm.ai/2024/09/05/perf-update.html), scroll to the end.
Latest reproduction guilde: [github issue link](https://github.com/vllm-project/vllm/issues/8176)
## Setup
- Docker images:
- vLLM: `vllm/vllm-openai:v0.6.2`
- SGLang: `lmsysorg/sglang:v0.3.2-cu121`
- LMDeploy: `openmmlab/lmdeploy:v0.6.1-cu12`
- TensorRT-LLM: `nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3`
- *NOTE: we use r24.07 as the current implementation only works for this version. We are going to bump this up.*
- Check [nightly-pipeline.yaml](nightly-pipeline.yaml) for the concrete docker images, specs and commands we use for the benchmark.
- Hardware
- 8x Nvidia A100 GPUs
- Workload:
- Dataset
- ShareGPT dataset
- Prefill-heavy dataset (in average 462 input tokens, 16 tokens as output)
- Decode-heavy dataset (in average 462 input tokens, 256 output tokens)
- Check [nightly-tests.json](tests/nightly-tests.json) for the concrete configuration of datasets we use.
- Models: llama-3 8B, llama-3 70B.
- We do not use llama 3.1 as it is incompatible with trt-llm r24.07. ([issue](https://github.com/NVIDIA/TensorRT-LLM/issues/2105)).
- Average QPS (query per second): 2, 4, 8, 16, 32 and inf.
- Queries are randomly sampled, and arrival patterns are determined via Poisson process, but all with fixed random seed.
- Evaluation metrics: Throughput (higher the better), TTFT (time to the first token, lower the better), ITL (inter-token latency, lower the better).
## Known issues
- TRT-LLM crashes with Llama 3.1 8B [issue](https://github.com/NVIDIA/TensorRT-LLM/issues/2105).
- TGI does not support `ignore-eos` flag.

196
.buildkite/nightly-benchmarks/nightly-pipeline.yaml
View File

@ -1,196 +0,0 @@
common_pod_spec: &common_pod_spec
priorityClassName: perf-benchmark
nodeSelector:
nvidia.com/gpu.product: NVIDIA-A100-SXM4-80GB
volumes:
- name: devshm
emptyDir:
medium: Memory
- name: hf-cache
hostPath:
path: /root/.cache/huggingface
type: Directory
common_container_settings: &common_container_settings
command:
- bash .buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh
resources:
limits:
nvidia.com/gpu: 8
volumeMounts:
- name: devshm
mountPath: /dev/shm
- name: hf-cache
mountPath: /root/.cache/huggingface
env:
- name: VLLM_USAGE_SOURCE
value: ci-test
- name: HF_HOME
value: /root/.cache/huggingface
- name: VLLM_SOURCE_CODE_LOC
value: /workspace/build/buildkite/vllm/performance-benchmark
- name: HF_TOKEN
valueFrom:
secretKeyRef:
name: hf-token-secret
key: token
steps:
- block: ":rocket: Ready for comparing vllm against alternatives? This will take 4 hours."
- label: "A100 vllm step 10"
priority: 100
agents:
queue: A100
plugins:
- kubernetes:
podSpec:
<<: *common_pod_spec
containers:
- image: vllm/vllm-openai:v0.6.2
<<: *common_container_settings
- label: "A100 sglang benchmark"
priority: 100
agents:
queue: A100
plugins:
- kubernetes:
podSpec:
<<: *common_pod_spec
containers:
- image: lmsysorg/sglang:v0.3.2-cu121
<<: *common_container_settings
- label: "A100 lmdeploy benchmark"
priority: 100
agents:
queue: A100
plugins:
- kubernetes:
podSpec:
<<: *common_pod_spec
containers:
- image: openmmlab/lmdeploy:v0.6.1-cu12
<<: *common_container_settings
- label: "A100 trt llama-8B"
priority: 100
agents:
queue: A100
plugins:
- kubernetes:
podSpec:
<<: *common_pod_spec
containers:
- image: nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3
<<: *common_container_settings
env:
- name: VLLM_USAGE_SOURCE
value: ci-test
- name: HF_HOME
value: /root/.cache/huggingface
- name: VLLM_SOURCE_CODE_LOC
value: /workspace/build/buildkite/vllm/performance-benchmark
- name: HF_TOKEN
valueFrom:
secretKeyRef:
name: hf-token-secret
key: token
- name: TEST_SELECTOR
value: "llama8B"
- label: "A100 trt llama-70B"
priority: 100
agents:
queue: A100
plugins:
- kubernetes:
podSpec:
<<: *common_pod_spec
containers:
- image: nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3
<<: *common_container_settings
env:
- name: VLLM_USAGE_SOURCE
value: ci-test
- name: HF_HOME
value: /root/.cache/huggingface
- name: VLLM_SOURCE_CODE_LOC
value: /workspace/build/buildkite/vllm/performance-benchmark
- name: HF_TOKEN
valueFrom:
secretKeyRef:
name: hf-token-secret
key: token
- name: TEST_SELECTOR
value: "llama70B"
# FIXME(Kuntai): uncomment this after NVIDIA gives us their test docker image
# - label: "A100 trt benchmark"
# priority: 100
# agents:
# queue: A100
# plugins:
# - kubernetes:
# podSpec:
# <<: *common_pod_spec
# containers:
# - image: nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3
# <<: *common_container_settings
# FIXME(Kuntai): uncomment this after TGI supports `--ignore-eos`.
# - label: "A100 tgi benchmark"
# priority: 100
# agents:
# queue: A100
# plugins:
# - kubernetes:
# podSpec:
# <<: *common_pod_spec
# containers:
# - image: ghcr.io/huggingface/text-generation-inference:2.2.0
# <<: *common_container_settings
- wait
- label: "Collect the results"
priority: 100
agents:
queue: A100
plugins:
- kubernetes:
podSpec:
<<: *common_pod_spec
containers:
- image: vllm/vllm-openai:v0.5.0.post1
command:
- bash .buildkite/nightly-benchmarks/scripts/nightly-annotate.sh
resources:
limits:
nvidia.com/gpu: 8
volumeMounts:
- name: devshm
mountPath: /dev/shm
env:
- name: VLLM_USAGE_SOURCE
value: ci-test
- name: VLLM_SOURCE_CODE_LOC
value: /workspace/build/buildkite/vllm/performance-benchmark
- name: HF_TOKEN
valueFrom:
secretKeyRef:
name: hf-token-secret
key: token
- block: ":rocket: check the results!"

26
.buildkite/nightly-benchmarks/scripts/download-tokenizer.py
View File

@ -1,26 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
from transformers import AutoTokenizer
def main(model, cachedir):
# Load the tokenizer and save it to the specified directory
tokenizer = AutoTokenizer.from_pretrained(model)
tokenizer.save_pretrained(cachedir)
print(f"Tokenizer saved to {cachedir}")
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description="Download and save Hugging Face tokenizer"
)
parser.add_argument("--model", type=str, required=True, help="Name of the model")
parser.add_argument(
"--cachedir", type=str, required=True, help="Directory to save the tokenizer"
)
args = parser.parse_args()
main(args.model, args.cachedir)

97
.buildkite/nightly-benchmarks/scripts/generate-nightly-markdown.py
View File

@ -1,97 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
import json
from pathlib import Path
import numpy as np
import pandas as pd
from tabulate import tabulate
def parse_arguments():
parser = argparse.ArgumentParser(
description="Parse command line arguments for summary-nightly-results script."
)
parser.add_argument(
"--results-folder",
type=str,
required=True,
help="The folder where the results are stored.",
)
parser.add_argument(
"--description", type=str, required=True, help="Description of the results."
)
args = parser.parse_args()
return args
def get_perf(df, method, model, metric):
means = []
for qps in [2, 4, 8, 16, "inf"]:
target = df["Test name"].str.contains(model)
target = target & df["Engine"].str.contains(method)
target = target & df["Test name"].str.contains("qps_" + str(qps))
filtered_df = df[target]
if filtered_df.empty:
means.append(0.0)
else:
means.append(filtered_df[metric].values[0])
return np.array(means)
def get_perf_w_std(df, method, model, metric):
if metric in ["TTFT", "ITL"]:
mean = get_perf(df, method, model, "Mean " + metric + " (ms)")
mean = mean.tolist()
std = get_perf(df, method, model, "Std " + metric + " (ms)")
if std.mean() == 0:
std = None
success = get_perf(df, method, model, "Successful req.")
if std is not None:
std = std / np.sqrt(success)
std = std.tolist()
else:
assert metric == "Tput"
mean = get_perf(df, method, model, "Input Tput (tok/s)") + get_perf(
df, method, model, "Output Tput (tok/s)"
)
mean = mean.tolist()
std = None
return mean, std
def main(args):
results_folder = Path(args.results_folder)
results = []
# collect results
for test_file in results_folder.glob("*_nightly_results.json"):
with open(test_file) as f:
results = results + json.loads(f.read())
# generate markdown table
df = pd.DataFrame.from_dict(results)
md_table = tabulate(df, headers="keys", tablefmt="pipe", showindex=False)
with open(args.description) as f:
description = f.read()
description = description.format(nightly_results_benchmarking_table=md_table)
with open("nightly_results.md", "w") as f:
f.write(description)
if __name__ == "__main__":
args = parse_arguments()
main(args)

9
.buildkite/nightly-benchmarks/scripts/get-lmdeploy-modelname.py
View File

@ -1,9 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from lmdeploy.serve.openai.api_client import APIClient
api_client = APIClient("http://localhost:8000")
model_name = api_client.available_models[0]
print(model_name)

78
.buildkite/nightly-benchmarks/scripts/nightly-annotate.sh
View File

@ -1,78 +0,0 @@
#!/bin/bash
set -ex
set -o pipefail
main() {
(which wget && which curl) || (apt-get update && apt-get install -y wget curl)
(which jq) || (apt-get update && apt-get -y install jq)
(which zip) || (apt-get install -y zip)
if [ ! -f /workspace/buildkite-agent ]; then
echo "buildkite-agent binary not found. Skip plotting the results."
exit 0
fi
# initial annotation
#description="$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/nightly-descriptions.md"
# download results
cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
mkdir -p results/
/workspace/buildkite-agent artifact download 'results/*nightly_results.json' results/
ls
ls results/
# upload benchmark results
zip -r results.zip results/
/workspace/buildkite-agent artifact upload "results.zip"
# upload benchmarking scripts
cd "$VLLM_SOURCE_CODE_LOC/"
zip -r nightly-benchmarks.zip .buildkite/ benchmarks/
/workspace/buildkite-agent artifact upload "nightly-benchmarks.zip"
cd "$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/"
# upload benchmarking pipeline
/workspace/buildkite-agent artifact upload "nightly-pipeline.yaml"
cd "$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/"
/workspace/buildkite-agent annotate --style "success" --context "nightly-benchmarks-results" --append < nightly-annotation.md
# The figures should be generated by a separate process outside the CI/CD pipeline
# # generate figures
# python3 -m pip install tabulate pandas matplotlib
# python3 $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/generate-nightly-markdown.py \
# --description $description \
# --results-folder results/
# python3 $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/plot-nightly-results.py \
# --description $description \
# --results-folder results/ \
# --dataset sharegpt
# python3 $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/plot-nightly-results.py \
# --description $description \
# --results-folder results/ \
# --dataset sonnet_2048_128
# python3 $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/plot-nightly-results.py \
# --description $description \
# --results-folder results/ \
# --dataset sonnet_128_2048
# # upload results and figures
# /workspace/buildkite-agent artifact upload "nightly_results*.png"
# /workspace/buildkite-agent artifact upload $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/nightly-pipeline.yaml
# /workspace/buildkite-agent artifact upload $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/tests/nightly-tests.json
# /workspace/buildkite-agent annotate --style "success" --context "nightly-benchmarks-results" --append < nightly_results.md
}
main "$@"

464
.buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh
View File

@ -1,464 +0,0 @@
#!/bin/bash
set -o pipefail
set -x
check_gpus() {
# check the number of GPUs and GPU type.
declare -g gpu_count=$(nvidia-smi --list-gpus | wc -l)
if [[ $gpu_count -gt 0 ]]; then
echo "GPU found."
else
echo "Need at least 1 GPU to run benchmarking."
exit 1
fi
declare -g gpu_type="$(nvidia-smi --query-gpu=name --format=csv,noheader | awk '{print $2}')"
echo "GPU type is $gpu_type"
}
check_hf_token() {
# check if HF_TOKEN is available and valid
if [[ -z "$HF_TOKEN" ]]; then
echo "Error: HF_TOKEN is not set."
exit 1
elif [[ ! "$HF_TOKEN" =~ ^hf_ ]]; then
echo "Error: HF_TOKEN does not start with 'hf_'."
exit 1
else
echo "HF_TOKEN is set and valid."
fi
}
upload_to_buildkite() {
# upload the benchmarking results to buildkite
# if the agent binary is not found, skip uploading the results, exit 0
if [ ! -f /workspace/buildkite-agent ]; then
echo "buildkite-agent binary not found. Skip uploading the results."
return 0
fi
# /workspace/buildkite-agent annotate --style "success" --context "benchmark-results" --append < $RESULTS_FOLDER/${CURRENT_LLM_SERVING_ENGINE}_nightly_results.md
/workspace/buildkite-agent artifact upload "$RESULTS_FOLDER/*"
}
get_current_llm_serving_engine() {
if which lmdeploy >/dev/null; then
echo "Container: lmdeploy"
export CURRENT_LLM_SERVING_ENGINE=lmdeploy
return
fi
if [ -e /tgi-entrypoint.sh ]; then
echo "Container: tgi"
export CURRENT_LLM_SERVING_ENGINE=tgi
return
fi
if which trtllm-build >/dev/null; then
echo "Container: tensorrt-llm"
export CURRENT_LLM_SERVING_ENGINE=trt
return
fi
if [ -e /sgl-workspace ]; then
echo "Container: sglang"
export CURRENT_LLM_SERVING_ENGINE=sglang
return
fi
if [ -e /vllm-workspace ]; then
echo "Container: vllm"
# move to a completely irrelevant directory, to avoid import vllm from current folder
export CURRENT_LLM_SERVING_ENGINE=vllm
return
fi
}
json2args() {
# transforms the JSON string to command line args, and '_' is replaced to '-'
# example:
# input: { "model": "meta-llama/Llama-2-7b-chat-hf", "tensor_parallel_size": 1 }
# output: --model meta-llama/Llama-2-7b-chat-hf --tensor-parallel-size 1
local json_string=$1
local args=$(
echo "$json_string" | jq -r '
to_entries |
map("--" + (.key | gsub("_"; "-")) + " " + (.value | tostring)) |
join(" ")
'
)
echo "$args"
}
kill_gpu_processes() {
pkill -f '[p]ython'
pkill -f '[p]ython3'
pkill -f '[t]ritonserver'
pkill -f '[p]t_main_thread'
pkill -f '[t]ext-generation'
pkill -f '[l]mdeploy'
# vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
pkill -f '[V]LLM'
while [ "$(nvidia-smi --query-gpu=memory.used --format=csv,noheader,nounits | head -n 1)" -ge 1000 ]; do
sleep 1
done
}
wait_for_server() {
# wait for vllm server to start
# return 1 if vllm server crashes
timeout 1200 bash -c '
until curl -s localhost:8000/v1/completions > /dev/null; do
sleep 1
done' && return 0 || return 1
}
ensure_installed() {
# Ensure that the given command is installed by apt-get
local cmd=$1
if ! which "$cmd" >/dev/null; then
apt-get update && apt-get install -y "$cmd"
fi
}
run_serving_tests() {
# run serving tests using `vllm bench serve` command
# $1: a json file specifying serving test cases
local serving_test_file
serving_test_file=$1
# Iterate over serving tests
jq -c '.[]' "$serving_test_file" | while read -r params; do
# get the test name, and append the GPU type back to it.
test_name=$(echo "$params" | jq -r '.test_name')
# if TEST_SELECTOR is set, only run the test cases that match the selector
if [[ -n "$TEST_SELECTOR" ]] && [[ ! "$test_name" =~ $TEST_SELECTOR ]]; then
echo "Skip test case $test_name."
continue
fi
# prepend the current serving engine to the test name
test_name=${CURRENT_LLM_SERVING_ENGINE}_${test_name}
# get common parameters
common_params=$(echo "$params" | jq -r '.common_parameters')
model=$(echo "$common_params" | jq -r '.model')
tp=$(echo "$common_params" | jq -r '.tp')
dataset_name=$(echo "$common_params" | jq -r '.dataset_name')
dataset_path=$(echo "$common_params" | jq -r '.dataset_path')
port=$(echo "$common_params" | jq -r '.port')
num_prompts=$(echo "$common_params" | jq -r '.num_prompts')
reuse_server=$(echo "$common_params" | jq -r '.reuse_server')
# get client and server arguments
server_params=$(echo "$params" | jq -r ".${CURRENT_LLM_SERVING_ENGINE}_server_parameters")
client_params=$(echo "$params" | jq -r ".${CURRENT_LLM_SERVING_ENGINE}_client_parameters")
client_args=$(json2args "$client_params")
qps_list=$(echo "$params" | jq -r '.qps_list')
qps_list=$(echo "$qps_list" | jq -r '.[] | @sh')
echo "Running over qps list $qps_list"
# check if there is enough GPU to run the test
if [[ $gpu_count -lt $tp ]]; then
echo "Required num-shard $tp but only $gpu_count GPU found. Skip testcase $test_name."
continue
fi
if [[ $reuse_server == "true" ]]; then
echo "Reuse previous server for test case $test_name"
else
kill_gpu_processes
bash "$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/launch-server.sh" \
"$server_params" "$common_params"
fi
if wait_for_server; then
echo ""
echo "$CURRENT_LLM_SERVING_ENGINE server is up and running."
else
echo ""
echo "$CURRENT_LLM_SERVING_ENGINE failed to start within the timeout period."
break
fi
# prepare tokenizer
# this is required for lmdeploy.
cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
rm -rf /tokenizer_cache
mkdir /tokenizer_cache
python3 ../.buildkite/nightly-benchmarks/scripts/download-tokenizer.py \
--model "$model" \
--cachedir /tokenizer_cache
cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
# change model name for lmdeploy (it will not follow standard hf name)
if [[ "$CURRENT_LLM_SERVING_ENGINE" == "lmdeploy" ]]; then
model=$(python ../.buildkite/nightly-benchmarks/scripts/get-lmdeploy-modelname.py)
fi
# iterate over different QPS
for qps in $qps_list; do
# remove the surrounding single quote from qps
if [[ "$qps" == *"inf"* ]]; then
echo "qps was $qps"
qps="inf"
echo "now qps is $qps"
fi
new_test_name=$test_name"_qps_"$qps
backend=$CURRENT_LLM_SERVING_ENGINE
if [[ $backend = "trt" ]]; then
backend="tensorrt-llm"
fi
if [[ "$backend" == *"vllm"* ]]; then
backend="vllm"
fi
if [[ "$dataset_name" = "sharegpt" ]]; then
client_command="vllm bench serve \
--backend $backend \
--tokenizer /tokenizer_cache \
--model $model \
--dataset-name $dataset_name \
--dataset-path $dataset_path \
--num-prompts $num_prompts \
--port $port \
--save-result \
--result-dir $RESULTS_FOLDER \
--result-filename ${new_test_name}.json \
--request-rate $qps \
--ignore-eos \
$client_args"
elif [[ "$dataset_name" = "sonnet" ]]; then
sonnet_input_len=$(echo "$common_params" | jq -r '.sonnet_input_len')
sonnet_output_len=$(echo "$common_params" | jq -r '.sonnet_output_len')
sonnet_prefix_len=$(echo "$common_params" | jq -r '.sonnet_prefix_len')
client_command="vllm bench serve \
--backend $backend \
--tokenizer /tokenizer_cache \
--model $model \
--dataset-name $dataset_name \
--dataset-path $dataset_path \
--num-prompts $num_prompts \
--sonnet-input-len $sonnet_input_len \
--sonnet-output-len $sonnet_output_len \
--sonnet-prefix-len $sonnet_prefix_len \
--port $port \
--save-result \
--result-dir $RESULTS_FOLDER \
--result-filename ${new_test_name}.json \
--request-rate $qps \
--ignore-eos \
$client_args"
else
echo "The dataset name must be either 'sharegpt' or 'sonnet'. Got $dataset_name."
exit 1
fi
echo "Running test case $test_name with qps $qps"
echo "Client command: $client_command"
eval "$client_command"
server_command="None"
# record the benchmarking commands
jq_output=$(jq -n \
--arg server "$server_command" \
--arg client "$client_command" \
--arg gpu "$gpu_type" \
--arg engine "$CURRENT_LLM_SERVING_ENGINE" \
'{
server_command: $server,
client_command: $client,
gpu_type: $gpu,
engine: $engine
}')
echo "$jq_output" >"$RESULTS_FOLDER/${new_test_name}.commands"
done
done
kill_gpu_processes
}
run_genai_perf_tests() {
# run genai-perf tests
# $1: a json file specifying genai-perf test cases
local genai_perf_test_file
genai_perf_test_file=$1
# Iterate over genai-perf tests
jq -c '.[]' "$genai_perf_test_file" | while read -r params; do
# get the test name, and append the GPU type back to it.
test_name=$(echo "$params" | jq -r '.test_name')
# if TEST_SELECTOR is set, only run the test cases that match the selector
if [[ -n "$TEST_SELECTOR" ]] && [[ ! "$test_name" =~ $TEST_SELECTOR ]]; then
echo "Skip test case $test_name."
continue
fi
# prepend the current serving engine to the test name
test_name=${CURRENT_LLM_SERVING_ENGINE}_${test_name}
# get common parameters
common_params=$(echo "$params" | jq -r '.common_parameters')
model=$(echo "$common_params" | jq -r '.model')
tp=$(echo "$common_params" | jq -r '.tp')
dataset_name=$(echo "$common_params" | jq -r '.dataset_name')
dataset_path=$(echo "$common_params" | jq -r '.dataset_path')
port=$(echo "$common_params" | jq -r '.port')
num_prompts=$(echo "$common_params" | jq -r '.num_prompts')
reuse_server=$(echo "$common_params" | jq -r '.reuse_server')
# get client and server arguments
server_params=$(echo "$params" | jq -r ".${CURRENT_LLM_SERVING_ENGINE}_server_parameters")
qps_list=$(echo "$params" | jq -r '.qps_list')
qps_list=$(echo "$qps_list" | jq -r '.[] | @sh')
echo "Running over qps list $qps_list"
# check if there is enough GPU to run the test
if [[ $gpu_count -lt $tp ]]; then
echo "Required num-shard $tp but only $gpu_count GPU found. Skip testcase $test_name."
continue
fi
if [[ $reuse_server == "true" ]]; then
echo "Reuse previous server for test case $test_name"
else
kill_gpu_processes
bash "$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/launch-server.sh" \
"$server_params" "$common_params"
fi
if wait_for_server; then
echo ""
echo "$CURRENT_LLM_SERVING_ENGINE server is up and running."
else
echo ""
echo "$CURRENT_LLM_SERVING_ENGINE failed to start within the timeout period."
break
fi
# iterate over different QPS
for qps in $qps_list; do
# remove the surrounding single quote from qps
if [[ "$qps" == *"inf"* ]]; then
echo "qps was $qps"
qps=$num_prompts
echo "now qps is $qps"
fi
new_test_name=$test_name"_qps_"$qps
backend=$CURRENT_LLM_SERVING_ENGINE
if [[ "$backend" == *"vllm"* ]]; then
backend="vllm"
fi
#TODO: add output dir.
client_command="genai-perf profile \
-m $model \
--service-kind openai \
--backend "$backend" \
--endpoint-type chat \
--streaming \
--url localhost:$port \
--request-rate $qps \
--num-prompts $num_prompts \
"
echo "Client command: $client_command"
eval "$client_command"
#TODO: process/record outputs
done
done
kill_gpu_processes
}
prepare_dataset() {
# download sharegpt dataset
cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
# duplicate sonnet by 4x, to allow benchmarking with input length 2048
cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
echo "" > sonnet_4x.txt
for _ in {1..4}
do
cat sonnet.txt >> sonnet_4x.txt
done
}
main() {
# check if the environment variable is successfully injected from yaml
check_gpus
check_hf_token
get_current_llm_serving_engine
pip install -U transformers
pip install -r requirements/dev.txt
which genai-perf
# check storage
df -h
ensure_installed wget
ensure_installed curl
ensure_installed jq
# genai-perf dependency
ensure_installed libb64-0d
prepare_dataset
cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
declare -g RESULTS_FOLDER=results/
mkdir -p $RESULTS_FOLDER
BENCHMARK_ROOT="$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/"
# run the test
run_serving_tests "$BENCHMARK_ROOT/tests/nightly-tests.json"
# run genai-perf tests
run_genai_perf_tests "$BENCHMARK_ROOT/tests/genai-perf-tests.json"
mv artifacts/ $RESULTS_FOLDER/
# upload benchmark results to buildkite
python3 -m pip install tabulate pandas
python3 "$BENCHMARK_ROOT/scripts/summary-nightly-results.py"
upload_to_buildkite
}
main "$@"

82
.buildkite/nightly-benchmarks/scripts/summary-nightly-results.py
View File

@ -1,82 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import datetime
import json
import os
from pathlib import Path
import pandas as pd
from tabulate import tabulate
results_folder = Path("results/")
# serving results and the keys that will be printed into markdown
serving_results = []
serving_column_mapping = {
"test_name": "Test name",
"gpu_type": "GPU",
"completed": "Successful req.",
"request_throughput": "Tput (req/s)",
"mean_ttft_ms": "Mean TTFT (ms)",
"std_ttft_ms": "Std TTFT (ms)",
"median_ttft_ms": "Median TTFT (ms)",
"mean_itl_ms": "Mean ITL (ms)",
"std_itl_ms": "Std ITL (ms)",
"median_itl_ms": "Median ITL (ms)",
"mean_tpot_ms": "Mean TPOT (ms)",
"std_tpot_ms": "Std TPOT (ms)",
"median_tpot_ms": "Median TPOT (ms)",
"total_token_throughput": "Total Token Tput (tok/s)",
"output_throughput": "Output Tput (tok/s)",
"total_input_tokens": "Total input tokens",
"total_output_tokens": "Total output tokens",
"engine": "Engine",
}
if __name__ == "__main__":
# collect results
for test_file in results_folder.glob("*.json"):
with open(test_file) as f:
raw_result = json.loads(f.read())
# attach the benchmarking command to raw_result
with open(test_file.with_suffix(".commands")) as f:
command = json.loads(f.read())
raw_result.update(command)
# update the test name of this result
raw_result.update({"test_name": test_file.stem})
# add the result to raw_result
serving_results.append(raw_result)
continue
serving_results = pd.DataFrame.from_dict(serving_results)
if not serving_results.empty:
serving_results = serving_results[list(serving_column_mapping.keys())].rename(
columns=serving_column_mapping
)
serving_md_table_with_headers = tabulate(
serving_results, headers="keys", tablefmt="pipe", showindex=False
)
# remove the first line of header
serving_md_table_lines = serving_md_table_with_headers.split("\n")
serving_md_table_without_header = "\n".join(serving_md_table_lines[2:])
prefix = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
prefix = prefix + "_" + os.environ.get("CURRENT_LLM_SERVING_ENGINE")
# document benchmarking results in markdown
with open(results_folder / f"{prefix}_nightly_results.md", "w") as f:
# document results with header.
# for those who wants to reproduce our benchmark.
f.write(serving_md_table_with_headers)
f.write("\n")
# document benchmarking results in json
with open(results_folder / f"{prefix}_nightly_results.json", "w") as f:
results = serving_results.to_dict(orient="records")
f.write(json.dumps(results))

23
.buildkite/nightly-benchmarks/scripts/wait-for-image.sh
View File

@ -1,23 +0,0 @@
#!/bin/sh
TOKEN=$(curl -s -L "https://public.ecr.aws/token?service=public.ecr.aws&scope=repository:q9t5s3a7/vllm-ci-postmerge-repo:pull" | jq -r .token)
if [[ "$BUILDKITE_BRANCH" == "main" ]]; then
URL="https://public.ecr.aws/v2/q9t5s3a7/vllm-ci-postmerge-repo/manifests/$BUILDKITE_COMMIT"
else
URL="https://public.ecr.aws/v2/q9t5s3a7/vllm-ci-test-repo/manifests/$BUILDKITE_COMMIT"
fi
TIMEOUT_SECONDS=10
retries=0
while [ $retries -lt 1000 ]; do
if [ "$(curl -s --max-time "$TIMEOUT_SECONDS" -L -H "Authorization: Bearer $TOKEN" -o /dev/null -w "%{http_code}" "$URL")" -eq 200 ]; then
exit 0
fi
echo "Waiting for image to be available..."
retries=$((retries + 1))
sleep 5
done
exit 1

30
.buildkite/nightly-benchmarks/tests/latency-tests-cpu.json
View File

@ -1,30 +0,0 @@
[
{
"test_name": "latency_llama8B_tp1",
"environment_variables": {
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 1,
"load_format": "dummy",
"num_iters_warmup": 5,
"num_iters": 15
}
},
{
"test_name": "latency_llama8B_tp4",
"environment_variables": {
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 4,
"load_format": "dummy",
"num_iters_warmup": 5,
"num_iters": 15
}
}
]

610
.buildkite/nightly-benchmarks/tests/serving-tests-cpu-snc2.json
View File

@ -1,610 +0,0 @@
[
{
"test_name": "serving_llama8B_bf16_tp1_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_bf16_tp2_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_bf16_tp4_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 4,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_bf16_tp1_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_bf16_tp2_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_bf16_tp4_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 4,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int8_tp1_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int8_tp2_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int8_tp4_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 4,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int8_tp1_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int8_tp2_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int8_tp4_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 4,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int4_tp1_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int4_tp2_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int4_tp4_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 4,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int4_tp1_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int4_tp2_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int4_tp4_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 4,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
}
]

820
.buildkite/nightly-benchmarks/tests/serving-tests-cpu-snc3.json
View File

@ -1,820 +0,0 @@
[
{
"test_name": "serving_llama8B_bf16_pp1_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"pipeline_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_bf16_tp2_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_bf16_pp3_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_bf16_tp2pp3_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 2,
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_bf16_pp1_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"pipeline_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_bf16_tp2_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_bf16_pp3_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_bf16_tp2pp3_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 2,
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int8_pp1_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"pipeline_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int8_tp2_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int8_pp3_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int8_tp2pp3_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 2,
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int8_pp1_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"pipeline_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int8_tp2_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int8_pp3_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int8_tp2pp3_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"tensor_parallel_size": 2,
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int4_pp1_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"pipeline_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int4_tp2_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int4_pp3_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int4_tp2pp3_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 2,
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_int4_pp1_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"pipeline_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int4_tp2_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int4_pp3_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
},
{
"test_name": "serving_llama8B_int4_tp2pp3_random_128_128",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"quantization": "awq",
"tensor_parallel_size": 2,
"pipeline_parallel_size": 3,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 1000
}
}
]

168
.buildkite/nightly-benchmarks/tests/serving-tests-cpu.json
View File

@ -1,168 +0,0 @@
[
{
"test_name": "serving_llama8B_tp1_sharegpt",
"qps_list": [1, 4, 16, "inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_tp2_sharegpt",
"qps_list": [1, 4, 16, "inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_tp4_sharegpt",
"qps_list": [1, 4, 16, "inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 4,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama8B_tp4_random_1024_128",
"qps_list": [1, 4, 16, "inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 4,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 1024,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 100
}
},
{
"test_name": "serving_llama8B_pp6_random_1024_128",
"qps_list": [1, 4, 16, "inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"pipeline_parallel_size": 6,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"enable_chunked_prefill": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 1024,
"random-output-len": 128,
"ignore-eos": "",
"num_prompts": 100
}
}
]

32
.buildkite/nightly-benchmarks/tests/throughput-tests-cpu.json
View File

@ -1,32 +0,0 @@
[
{
"test_name": "throughput_llama8B_tp1",
"environment_variables": {
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 1,
"load_format": "dummy",
"dataset": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200,
"backend": "vllm"
}
},
{
"test_name": "throughput_llama8B_tp4",
"environment_variables": {
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 4,
"load_format": "dummy",
"dataset": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200,
"backend": "vllm"
}
}
]

114
.buildkite/nightly-benchmarks/README.md → .buildkite/performance-benchmarks/README.md
View File

@ -2,40 +2,23 @@
## Introduction
This directory contains two sets of benchmark for vllm.
- Performance benchmark: benchmark vllm's performance under various workload, for **developers** to gain clarity on whether their PR improves/degrades vllm's performance
- Nightly benchmark: compare vllm's performance against alternatives (tgi, trt-llm and lmdeploy), for **the public** to know when to choose vllm.
See [vLLM performance dashboard](https://hud.pytorch.org/benchmark/llms?repoName=vllm-project%2Fvllm) for the latest performance benchmark results and [vLLM GitHub README](https://github.com/vllm-project/vllm/blob/main/README.md) for latest nightly benchmark results.
This directory contains a benchmarking suite for **developers** to run locally and gain clarity on whether their PR improves/degrades vllm's performance.
vLLM also maintains a continuous performance benchmark under [perf.vllm.ai](https://perf.vllm.ai/), hosted under PyTorch CI HUD.
## Performance benchmark quick overview
**Benchmarking Coverage**: latency, throughput and fix-qps serving on A100 (the support for FP8 benchmark on H100 is coming!) and Intel® Xeon® Processors, with different models.
**Benchmarking Coverage**: latency, throughput and fix-qps serving on B200, A100, H100, Intel® Xeon® Processors and Intel® Gaudi® 3 Accelerators with different models.
**Benchmarking Duration**: about 1hr.
**For benchmarking developers**: please try your best to constraint the duration of benchmarking to about 1 hr so that it won't take forever to run.
## Nightly benchmark quick overview
**Benchmarking Coverage**: Fix-qps serving on A100 (the support for FP8 benchmark on H100 is coming!) on Llama-3 8B, 70B and Mixtral 8x7B.
**Benchmarking engines**: vllm, TGI, trt-llm and lmdeploy.
**Benchmarking Duration**: about 3.5hrs.
## Trigger the benchmark
Performance benchmark will be triggered when:
- A PR being merged into vllm.
- Every commit for those PRs with `perf-benchmarks` label AND `ready` label.
Manually Trigger the benchmark
The benchmark needs to be triggered manually:
```bash
bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
bash .buildkite/performance-benchmarks/scripts/run-performance-benchmarks.sh
```
Runtime environment variables:
@ -47,14 +30,11 @@ Runtime environment variables:
- `REMOTE_HOST`: IP for the remote vLLM service to benchmark. Default value is empty string.
- `REMOTE_PORT`: Port for the remote vLLM service to benchmark. Default value is empty string.
Nightly benchmark will be triggered when:
- Every commit for those PRs with `perf-benchmarks` label and `nightly-benchmarks` label.
## Performance benchmark details
See [performance-benchmarks-descriptions.md](performance-benchmarks-descriptions.md) for detailed descriptions, and use `tests/latency-tests.json`, `tests/throughput-tests.json`, `tests/serving-tests.json` to configure the test cases.
> NOTE: For Intel® Xeon® Processors, use `tests/latency-tests-cpu.json`, `tests/throughput-tests-cpu.json`, `tests/serving-tests-cpu.json` instead.
For Intel® Gaudi® 3 Accelerators, use `tests/latency-tests-hpu.json`, `tests/throughput-tests-hpu.json`, `tests/serving-tests-hpu.json` instead.
>
### Latency test
@ -128,6 +108,65 @@ The number of this test is less stable compared to the delay and latency benchma
WARNING: The benchmarking script will save json results by itself, so please do not configure `--save-results` or other results-saving-related parameters in `serving-tests.json`.
#### Default Parameters Field
We can specify default parameters in a JSON field with key `defaults`. Parameters defined in the field are applied globally to all serving tests, and can be overridden in test case fields. Here is an example:
<details>
<summary> An Example of default parameters field </summary>
```json
{
"defaults": {
"qps_list": [
"inf"
],
"server_environment_variables": {
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1
},
"server_parameters": {
"tensor_parallel_size": 1,
"dtype": "bfloat16",
"block_size": 128,
"disable_log_stats": "",
"load_format": "dummy"
},
"client_parameters": {
"backend": "vllm",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128,
"num_prompts": 200,
"ignore-eos": ""
}
},
"tests": [
{
"test_name": "serving_llama3B_tp2_random_128_128",
"server_parameters": {
"model": "meta-llama/Llama-3.2-3B-Instruct",
"tensor_parallel_size": 2,
},
"client_parameters": {
"model": "meta-llama/Llama-3.2-3B-Instruct",
}
},
{
"test_name": "serving_qwen3_tp4_random_128_128",
"server_parameters": {
"model": "Qwen/Qwen3-14B",
"tensor_parallel_size": 4,
},
"client_parameters": {
"model": "Qwen/Qwen3-14B",
}
},
]
}
```
</details>
### Visualizing the results
The `convert-results-json-to-markdown.py` helps you put the benchmarking results inside a markdown table, by formatting [descriptions.md](performance-benchmarks-descriptions.md) with real benchmarking results.
@ -152,26 +191,3 @@ Here is an example using the script to compare result_a and result_b with Model,
A comparison diagram will be generated below the table.
Here is an example to compare between 96c/results_gnr_96c_091_tp2pp3 and 128c/results_gnr_128c_091_tp2pp3
<img width="1886" height="828" alt="image" src="https://github.com/user-attachments/assets/c02a43ef-25d0-4fd6-90e5-2169a28682dd" />
## Nightly test details
See [nightly-descriptions.md](nightly-descriptions.md) for the detailed description on test workload, models and docker containers of benchmarking other llm engines.
### Workflow
- The [nightly-pipeline.yaml](nightly-pipeline.yaml) specifies the docker containers for different LLM serving engines.
- Inside each container, we run [scripts/run-nightly-benchmarks.sh](scripts/run-nightly-benchmarks.sh), which will probe the serving engine of the current container.
- The `scripts/run-nightly-benchmarks.sh` will parse the workload described in [nightly-tests.json](tests/nightly-tests.json) and launch the right benchmark for the specified serving engine via `scripts/launch-server.sh`.
- At last, we run [scripts/summary-nightly-results.py](scripts/summary-nightly-results.py) to collect and plot the final benchmarking results, and update the results to buildkite.
### Nightly tests
In [nightly-tests.json](tests/nightly-tests.json), we include the command line arguments for benchmarking commands, together with the benchmarking test cases. The format is highly similar to performance benchmark.
### Docker containers
The docker containers for benchmarking are specified in `nightly-pipeline.yaml`.
WARNING: the docker versions are HARD-CODED and SHOULD BE ALIGNED WITH `nightly-descriptions.md`. The docker versions need to be hard-coded as there are several version-specific bug fixes inside `scripts/run-nightly-benchmarks.sh` and `scripts/launch-server.sh`.
WARNING: populating `trt-llm` to latest version is not easy, as it requires updating several protobuf files in [tensorrt-demo](https://github.com/neuralmagic/tensorrt-demo.git).

6
.buildkite/nightly-benchmarks/performance-benchmarks-descriptions.md → .buildkite/performance-benchmarks/performance-benchmarks-descriptions.md
View File

@ -5,7 +5,7 @@
- Input length: 32 tokens.
- Output length: 128 tokens.
- Batch size: fixed (8).
- GPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- GPU/HPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- CPU Models: llama-3.1 8B.
- Evaluation metrics: end-to-end latency (mean, median, p99).
@ -16,7 +16,7 @@
- Input length: randomly sample 200 prompts from ShareGPT dataset (with fixed random seed).
- Output length: the corresponding output length of these 200 prompts.
- Batch size: dynamically determined by vllm to achieve maximum throughput.
- GPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- GPU/HPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- CPU Models: llama-3.1 8B.
- Evaluation metrics: throughput.
@ -28,7 +28,7 @@
- Output length: the corresponding output length of these 200 prompts.
- Batch size: dynamically determined by vllm and the arrival pattern of the requests.
- **Average QPS (query per second)**: 1, 4, 16 and inf. QPS = inf means all requests come at once. For other QPS values, the arrival time of each query is determined using a random Poisson process (with fixed random seed).
- GPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- GPU/HPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- We also added a speculative decoding test for llama-3 70B on GPU, under QPS 2
- CPU Models: llama-3.1 8B.
- Evaluation metrics: throughput, TTFT (time to the first token, with mean, median and p99), ITL (inter-token latency, with mean, median and p99).

211
.buildkite/nightly-benchmarks/scripts/compare-json-results.py → .buildkite/performance-benchmarks/scripts/compare-json-results.py
View File

@ -7,6 +7,7 @@ from importlib import util
import pandas as pd
pd.options.display.float_format = "{:.2f}".format
plotly_found = util.find_spec("plotly.express") is not None
@ -109,7 +110,10 @@ def compare_data_columns(
if len(compare_frames) >= 2:
base = compare_frames[0]
current = compare_frames[-1]
ratio = current / base
if "P99" in data_column or "Median" in data_column:
ratio = base / current # for latency
else:
ratio = current / base
ratio = ratio.mask(base == 0) # avoid inf when baseline is 0
ratio.name = f"Ratio 1 vs {len(compare_frames)}"
frames.append(ratio)
@ -199,6 +203,71 @@ def split_json_by_tp_pp(
return saved_paths
def _add_limit_line(fig, y_value, label):
# Visible dashed line + annotation
fig.add_hline(
y=y_value,
line_dash="dash",
line_color="red" if "ttft" in label.lower() else "blue",
annotation_text=f"{label}: {y_value} ms",
annotation_position="top left",
)
# Optional: add a legend item (as a transparent helper trace)
if plot and plotly_found:
import plotly.graph_objects as go
fig.add_trace(
go.Scatter(
x=[None],
y=[None],
mode="lines",
line=dict(
dash="dash", color="red" if "ttft" in label.lower() else "blue"
),
name=f"{label}",
)
)
def _find_concurrency_col(df: pd.DataFrame) -> str:
for c in [
"# of max concurrency.",
"# of max concurrency",
"Max Concurrency",
"max_concurrency",
"Concurrency",
]:
if c in df.columns:
return c
# Fallback: guess an integer-like column (harmless if unused)
for c in df.columns:
if df[c].dtype.kind in "iu" and df[c].nunique() > 1 and df[c].min() >= 1:
return c
return "# of max concurrency."
def _highlight_threshold(
df: pd.DataFrame, threshold: float
) -> "pd.io.formats.style.Styler":
"""Highlight numeric per-configuration columns with value <= threshold."""
conc_col = _find_concurrency_col(df)
key_cols = [
c
for c in ["Model", "Dataset Name", "Input Len", "Output Len", conc_col]
if c in df.columns
]
conf_cols = [
c for c in df.columns if c not in key_cols and not str(c).startswith("Ratio")
]
conf_cols = [c for c in conf_cols if pd.api.types.is_numeric_dtype(df[c])]
return df.style.map(
lambda v: "background-color:#e6ffe6;font-weight:bold;"
if pd.notna(v) and v <= threshold
else "",
subset=conf_cols,
)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
@ -220,6 +289,26 @@ if __name__ == "__main__":
default="# of max concurrency.",
help="column name to use as X Axis in comparison graph",
)
parser.add_argument(
"-l",
"--latency",
type=str,
default="p99",
help="take median|p99 for latency like TTFT/TPOT",
)
parser.add_argument(
"--ttft-max-ms",
type=float,
default=3000.0,
help="Reference limit for TTFT plots (ms)",
)
parser.add_argument(
"--tpot-max-ms",
type=float,
default=100.0,
help="Reference limit for TPOT plots (ms)",
)
args = parser.parse_args()
drop_column = "P99"
@ -234,12 +323,22 @@ if __name__ == "__main__":
"# of max concurrency.",
"qps",
]
data_cols_to_compare = ["Output Tput (tok/s)", "Median TTFT (ms)", "Median"]
html_msgs_for_data_cols = [
"Compare Output Tokens /n",
"Median TTFT /n",
"Median TPOT /n",
]
if "median" in args.latency:
data_cols_to_compare = ["Output Tput (tok/s)", "Median TTFT (ms)", "Median"]
html_msgs_for_data_cols = [
"Compare Output Tokens /n",
"Median TTFT /n",
"Median TPOT /n",
]
drop_column = "P99"
elif "p99" in args.latency:
data_cols_to_compare = ["Output Tput (tok/s)", "P99 TTFT (ms)", "P99"]
html_msgs_for_data_cols = [
"Compare Output Tokens /n",
"P99 TTFT /n",
"P99 TPOT /n",
]
if len(args.file) == 1:
files = split_json_by_tp_pp(args.file[0], output_root="splits")
@ -275,33 +374,83 @@ if __name__ == "__main__":
f"Expected subset: {filtered_info_cols}, "
f"but DataFrame has: {list(output_df.columns)}"
)
output_df_sorted = output_df.sort_values(by=existing_group_cols)
# output_df_sorted = output_df.sort_values(by=existing_group_cols)
output_df_sorted = output_df.sort_values(by=args.xaxis)
output_groups = output_df_sorted.groupby(existing_group_cols, dropna=False)
for name, group in output_groups:
html = group.to_html()
group_name = (
",".join(map(str, name)).replace(",", "_").replace("/", "-")
)
group_html_name = "perf_comparison_" + group_name + ".html"
metric_name = str(data_cols_to_compare[i]).lower()
if "tok/s" in metric_name:
html = group.to_html()
elif "ttft" in metric_name:
styler = _highlight_threshold(group, args.ttft_max_ms).format(
{c: "{:.2f}" for c in group.select_dtypes("number").columns},
na_rep="—",
)
html = styler.to_html(
table_attributes='border="1" class="dataframe"'
)
elif (
"tpot" in metric_name
or "median" in metric_name
or "p99" in metric_name
):
styler = _highlight_threshold(group, args.tpot_max_ms).format(
{c: "{:.2f}" for c in group.select_dtypes("number").columns},
na_rep="—",
)
html = styler.to_html(
table_attributes='border="1" class="dataframe"'
)
text_file.write(html_msgs_for_data_cols[i])
text_file.write(html)
with open(group_html_name, "a+") as sub_text_file:
sub_text_file.write(html_msgs_for_data_cols[i])
sub_text_file.write(html)
if plot and plotly_found:
import plotly.express as px
if plot and plotly_found:
import plotly.express as px
df = group[raw_data_cols]
df_sorted = df.sort_values(by=info_cols[y_axis_index])
# Melt DataFrame for plotting
df_melted = df_sorted.melt(
id_vars=info_cols[y_axis_index],
var_name="Configuration",
value_name=data_cols_to_compare[i],
)
title = data_cols_to_compare[i] + " vs " + info_cols[y_axis_index]
# Create Plotly line chart
fig = px.line(
df_melted,
x=info_cols[y_axis_index],
y=data_cols_to_compare[i],
color="Configuration",
title=title,
markers=True,
)
# Export to HTML
text_file.write(fig.to_html(full_html=True, include_plotlyjs="cdn"))
df = group[raw_data_cols]
df_sorted = df.sort_values(by=info_cols[y_axis_index])
# Melt DataFrame for plotting
df_melted = df_sorted.melt(
id_vars=info_cols[y_axis_index],
var_name="Configuration",
value_name=data_cols_to_compare[i],
)
title = (
data_cols_to_compare[i] + " vs " + info_cols[y_axis_index]
)
# Create Plotly line chart
fig = px.line(
df_melted,
x=info_cols[y_axis_index],
y=data_cols_to_compare[i],
color="Configuration",
title=title,
markers=True,
)
# ---- Add threshold lines based on metric name ----
if "ttft" in metric_name:
_add_limit_line(fig, args.ttft_max_ms, "TTFT limit")
elif (
"tpot" in metric_name
or "median" in metric_name
or "p99" in metric_name
):
_add_limit_line(fig, args.tpot_max_ms, "TPOT limit")
# Export to HTML
text_file.write(
fig.to_html(full_html=True, include_plotlyjs="cdn")
)
sub_text_file.write(
fig.to_html(full_html=True, include_plotlyjs="cdn")
)

6
.buildkite/nightly-benchmarks/scripts/convert-results-json-to-markdown.py → .buildkite/performance-benchmarks/scripts/convert-results-json-to-markdown.py
View File

@ -63,9 +63,11 @@ serving_column_mapping = {
"mean_ttft_ms": "Mean TTFT (ms)",
"median_ttft_ms": "Median TTFT (ms)",
"p99_ttft_ms": "P99 TTFT (ms)",
"std_ttft_ms": "STD TTFT (ms)",
"mean_tpot_ms": "Mean TPOT (ms)",
"median_tpot_ms": "Median",
"p99_tpot_ms": "P99",
"std_tpot_ms": "STD TPOT (ms)",
"mean_itl_ms": "Mean ITL (ms)",
"median_itl_ms": "Median ITL (ms)",
"p99_itl_ms": "P99 ITL (ms)",
@ -368,7 +370,7 @@ if __name__ == "__main__":
# The GPUs sometimes come in format of "GPUTYPE\nGPUTYPE\n...",
# we want to turn it into "8xGPUTYPE"
df["GPU"] = df["GPU"].apply(
lambda x: f"{len(x.split('\n'))}x{x.split('\n')[0]}"
lambda x: "{}x{}".format(len(x.split("\n")), x.split("\n")[0])
)
# get markdown tables
@ -390,7 +392,7 @@ if __name__ == "__main__":
json_file = "benchmark_results.json"
with open(results_folder / md_file, "w") as f:
results = read_markdown(
"../.buildkite/nightly-benchmarks/"
"../.buildkite/performance-benchmarks/"
+ "performance-benchmarks-descriptions.md"
)
results = results.format(

8
.buildkite/nightly-benchmarks/scripts/launch-server.sh → .buildkite/performance-benchmarks/scripts/launch-server.sh
View File

@ -181,18 +181,14 @@ launch_vllm_server() {
if echo "$common_params" | jq -e 'has("fp8")' >/dev/null; then
echo "Key 'fp8' exists in common params. Use neuralmagic fp8 model for convenience."
model=$(echo "$common_params" | jq -r '.neuralmagic_quantized_model')
server_command="python3 \
-m vllm.entrypoints.openai.api_server \
server_command="vllm serve $model \
-tp $tp \
--model $model \
--port $port \
$server_args"
else
echo "Key 'fp8' does not exist in common params."
server_command="python3 \
-m vllm.entrypoints.openai.api_server \
server_command="vllm serve $model \
-tp $tp \
--model $model \
--port $port \
$server_args"
fi

76
.buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh → .buildkite/performance-benchmarks/scripts/run-performance-benchmarks.sh
View File

@ -15,6 +15,8 @@ check_gpus() {
declare -g gpu_count=$(nvidia-smi --list-gpus | wc -l)
elif command -v amd-smi; then
declare -g gpu_count=$(amd-smi list | grep 'GPU' | wc -l)
elif command -v hl-smi; then
declare -g gpu_count=$(hl-smi --list | grep -i "Module ID" | wc -l)
fi
if [[ $gpu_count -gt 0 ]]; then
@ -23,10 +25,16 @@ check_gpus() {
echo "Need at least 1 GPU to run benchmarking."
exit 1
fi
declare -g arch_suffix=''
if command -v nvidia-smi; then
declare -g gpu_type=$(nvidia-smi --query-gpu=name --format=csv,noheader | awk '{print $2}')
elif command -v amd-smi; then
declare -g gpu_type=$(amd-smi static -g 0 -a | grep 'MARKET_NAME' | awk '{print $2}')
elif command -v hl-smi; then
declare -g gpu_type=$(hl-smi -q | grep "Product Name" | head -n 1 | awk -F ':' '{print $2}' | sed 's/^ *//')
arch_suffix='-hpu'
fi
echo "GPU type is $gpu_type"
}
@ -102,7 +110,8 @@ json2envs() {
wait_for_server() {
# wait for vllm server to start
# return 1 if vllm server crashes
timeout 1200 bash -c '
local timeout_val="1200"
timeout "$timeout_val" bash -c '
until curl -X POST localhost:8000/v1/completions; do
sleep 1
done' && return 0 || return 1
@ -138,6 +147,10 @@ kill_gpu_processes() {
while [ "$(amd-smi metric -g 0 | grep 'USED_VRAM' | awk '{print $2}')" -ge 1000 ]; do
sleep 1
done
elif command -v hl-smi; then
while [ "$(hl-smi -q | grep "Used" | head -n 1 | awk '{print $3}')" -ge 1000 ]; do
sleep 1
done
fi
# remove vllm config file
@ -304,12 +317,44 @@ run_throughput_tests() {
run_serving_tests() {
# run serving tests using `vllm bench serve` command
# $1: a json file specifying serving test cases
#
# Supported JSON formats:
# 1) Plain format: top-level array
# [ { "test_name": "...", "server_parameters": {...}, ... }, ... ]
#
# 2) Default parameters field + plain format tests
# {
# "defaults": { ... },
# "tests": [ { "test_name": "...", "server_parameters": {...}, ... }, ... ]
# }
local serving_test_file
serving_test_file=$1
# Iterate over serving tests
jq -c '.[]' "$serving_test_file" | while read -r params; do
jq -c '
if type == "array" then
# Plain format: test cases array
.[]
elif (type == "object" and has("tests")) then
# merge the default parameters into each test cases
. as $root
| ($root.defaults // {}) as $d
| ($root.tests // [])[]
# default qps / max_concurrency from defaults if missing
| .qps_list = (.qps_list // $d.qps_list)
| .max_concurrency_list = (.max_concurrency_list // $d.max_concurrency_list)
# merge envs / params: test overrides defaults
| .server_environment_variables =
(($d.server_environment_variables // {}) + (.server_environment_variables // {}))
| .server_parameters =
(($d.server_parameters // {}) + (.server_parameters // {}))
| .client_parameters =
(($d.client_parameters // {}) + (.client_parameters // {}))
else
error("Unsupported serving test file format: must be array or object with .tests")
end
' "$serving_test_file" | while read -r params; do
# get the test name, and append the GPU type back to it.
test_name=$(echo "$params" | jq -r '.test_name')
if [[ ! "$test_name" =~ ^serving_ ]]; then
@ -323,20 +368,25 @@ run_serving_tests() {
continue
fi
# get client and server arguments
# get client and server arguments (after merged the default parameters)
server_params=$(echo "$params" | jq -r '.server_parameters')
server_envs=$(echo "$params" | jq -r '.server_environment_variables')
client_params=$(echo "$params" | jq -r '.client_parameters')
server_args=$(json2args "$server_params")
server_envs=$(json2envs "$server_envs")
client_args=$(json2args "$client_params")
# qps_list
qps_list=$(echo "$params" | jq -r '.qps_list')
qps_list=$(echo "$qps_list" | jq -r '.[] | @sh')
echo "Running over qps list $qps_list"
# max_concurrency_list (fallback to num_prompts if missing)
max_concurrency_list=$(echo "$params" | jq -r '.max_concurrency_list')
if [[ -z "$max_concurrency_list" || "$max_concurrency_list" == "null" ]]; then
num_prompts=$(echo "$client_params" | jq -r '.num_prompts')
max_concurrency_list="[$num_prompts]"
num_prompts=$(echo "$client_params" | jq -r '.num_prompts')
max_concurrency_list="[$num_prompts]"
fi
max_concurrency_list=$(echo "$max_concurrency_list" | jq -r '.[] | @sh')
echo "Running over max concurrency list $max_concurrency_list"
@ -365,8 +415,7 @@ run_serving_tests() {
continue
fi
server_command="$server_envs python3 \
-m vllm.entrypoints.openai.api_server \
server_command="$server_envs vllm serve \
$server_args"
# run the server
@ -452,14 +501,10 @@ main() {
ARCH='-cpu'
else
check_gpus
ARCH="$arch_suffix"
fi
check_hf_token
# Set to v1 to run v1 benchmark
if [[ "${ENGINE_VERSION:-v0}" == "v1" ]]; then
export VLLM_USE_V1=1
fi
# dependencies
(which wget && which curl) || (apt-get update && apt-get install -y wget curl)
(which jq) || (apt-get update && apt-get -y install jq)
@ -475,7 +520,12 @@ main() {
ensure_sharegpt_downloaded
declare -g RESULTS_FOLDER=results/
mkdir -p $RESULTS_FOLDER
QUICK_BENCHMARK_ROOT=../.buildkite/nightly-benchmarks/
QUICK_BENCHMARK_ROOT=../.buildkite/performance-benchmarks/
# dump vllm info via vllm collect-env
env_output=$(vllm collect-env)
echo "$env_output" >"$RESULTS_FOLDER/vllm_env.txt"
# benchmarking
run_serving_tests $QUICK_BENCHMARK_ROOT/tests/"${SERVING_JSON:-serving-tests$ARCH.json}"

0
.buildkite/nightly-benchmarks/tests/genai-perf-tests.json → .buildkite/performance-benchmarks/tests/genai-perf-tests.json
View File

26
.buildkite/performance-benchmarks/tests/latency-tests-cpu.json Normal file
View File

@ -0,0 +1,26 @@
[
{
"test_name": "latency_llama8B_tp2",
"environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"num_iters_warmup": 5,
"num_iters": 15
}
}
]

55
.buildkite/performance-benchmarks/tests/latency-tests-hpu.json Normal file
View File

@ -0,0 +1,55 @@
[
{
"test_name": "latency_llama8B_tp1",
"environment_variables": {
"PT_HPU_LAZY_MODE": 1,
"VLLM_CONTIGUOUS_PA": 1,
"VLLM_DEFRAG": 1
},
"parameters": {
"model": "meta-llama/Meta-Llama-3.1-8B-Instruct",
"tensor_parallel_size": 1,
"load_format": "dummy",
"num-iters-warmup": 5,
"num-iters": 15,
"max-model-len": 256,
"async-scheduling": ""
}
},
{
"test_name": "latency_llama70B_tp4",
"environment_variables": {
"PT_HPU_LAZY_MODE": 1,
"PT_HPU_ENABLE_LAZY_COLLECTIVES": 1,
"VLLM_CONTIGUOUS_PA": 1,
"VLLM_DEFRAG": 1
},
"parameters": {
"model": "meta-llama/Meta-Llama-3.1-70B-Instruct",
"tensor_parallel_size": 4,
"load_format": "dummy",
"num-iters-warmup": 5,
"num-iters": 15,
"max-model-len": 256,
"async-scheduling": ""
}
},
{
"test_name": "latency_mixtral8x7B_tp2",
"environment_variables": {
"PT_HPU_LAZY_MODE": 1,
"PT_HPU_ENABLE_LAZY_COLLECTIVES": 1,
"VLLM_CONTIGUOUS_PA": 1,
"VLLM_DEFRAG": 1
},
"parameters": {
"model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
"tensor_parallel_size": 2,
"load_format": "dummy",
"num-iters-warmup": 5,
"num-iters": 15,
"max-model-len": 256,
"async-scheduling": ""
}
}
]

0
.buildkite/nightly-benchmarks/tests/latency-tests.json → .buildkite/performance-benchmarks/tests/latency-tests.json
View File

0
.buildkite/nightly-benchmarks/tests/nightly-tests.json → .buildkite/performance-benchmarks/tests/nightly-tests.json
View File

246
.buildkite/performance-benchmarks/tests/serving-tests-cpu.json Normal file
View File

@ -0,0 +1,246 @@
{
"defaults": {
"qps_list": [
"inf"
],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"server_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 1,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"load_format": "dummy"
},
"client_parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"backend": "vllm",
"ignore-eos": "",
"num_prompts": 200
}
},
"tests": [
{
"test_name": "serving_llama8B_tp1_sharegpt",
"server_parameters": {
"tensor_parallel_size": 1
},
"client_parameters": {
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json"
}
},
{
"test_name": "serving_llama8B_tp2_sharegpt",
"server_parameters": {
"tensor_parallel_size": 2
},
"client_parameters": {
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json"
}
},
{
"test_name": "serving_llama8B_tp1_random_128_128",
"server_parameters": {
"tensor_parallel_size": 1
},
"client_parameters": {
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
},
{
"test_name": "serving_llama8B_tp2_random_128_128",
"server_parameters": {
"tensor_parallel_size": 2
},
"client_parameters": {
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
},
{
"test_name": "serving_llama8B_tp4_random_128_128",
"server_parameters": {
"tensor_parallel_size": 4
},
"client_parameters": {
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
},
{
"test_name": "serving_llama8B_tp1_random_128_2048",
"server_parameters": {
"tensor_parallel_size": 1
},
"client_parameters": {
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 2048
}
},
{
"test_name": "serving_llama8B_tp2_random_128_2048",
"server_parameters": {
"tensor_parallel_size": 2
},
"client_parameters": {
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 2048
}
},
{
"test_name": "serving_llama8B_tp4_random_128_2048",
"server_parameters": {
"tensor_parallel_size": 4
},
"client_parameters": {
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 2048
}
},
{
"test_name": "serving_llama8B_tp1_random_2048_128",
"server_parameters": {
"tensor_parallel_size": 1
},
"client_parameters": {
"dataset_name": "random",
"random-input-len": 2048,
"random-output-len": 128
}
},
{
"test_name": "serving_llama8B_tp2_random_2048_128",
"server_parameters": {
"tensor_parallel_size": 2
},
"client_parameters": {
"dataset_name": "random",
"random-input-len": 2048,
"random-output-len": 128
}
},
{
"test_name": "serving_llama8B_tp4_random_2048_128",
"server_parameters": {
"tensor_parallel_size": 4
},
"client_parameters": {
"dataset_name": "random",
"random-input-len": 2048,
"random-output-len": 128
}
},
{
"test_name": "serving_llama3B_tp1_random_128_128",
"server_parameters": {
"model": "meta-llama/Llama-3.2-3B-Instruct",
"tensor_parallel_size": 1
},
"client_parameters": {
"model": "meta-llama/Llama-3.2-3B-Instruct",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
},
{
"test_name": "serving_granite2B_tp1_random_128_128",
"server_parameters": {
"model": "ibm-granite/granite-3.2-2b-instruct",
"tensor_parallel_size": 1
},
"client_parameters": {
"model": "ibm-granite/granite-3.2-2b-instruct",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
},
{
"test_name": "serving_qwen1.7B_tp1_random_128_128",
"server_parameters": {
"model": "Qwen/Qwen3-1.7B",
"tensor_parallel_size": 1
},
"client_parameters": {
"model": "Qwen/Qwen3-1.7B",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
},
{
"test_name": "serving_qwen4B_tp1_random_128_128",
"server_parameters": {
"model": "Qwen/Qwen3-4B",
"tensor_parallel_size": 1
},
"client_parameters": {
"model": "Qwen/Qwen3-4B",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
},
{
"test_name": "serving_qwen8B_tp1_random_128_128",
"server_parameters": {
"model": "Qwen/Qwen3-8B",
"tensor_parallel_size": 1
},
"client_parameters": {
"model": "Qwen/Qwen3-8B",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
},
{
"test_name": "serving_glm9B_tp1_random_128_128",
"server_parameters": {
"model": "zai-org/glm-4-9b-hf",
"tensor_parallel_size": 1
},
"client_parameters": {
"model": "zai-org/glm-4-9b-hf",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
},
{
"test_name": "serving_gemma7B_tp1_random_128_128",
"server_parameters": {
"model": "google/gemma-7b",
"tensor_parallel_size": 1
},
"client_parameters": {
"model": "google/gemma-7b",
"dataset_name": "random",
"random-input-len": 128,
"random-output-len": 128
}
}
]
}

82
.buildkite/performance-benchmarks/tests/serving-tests-hpu.json Normal file
View File

@ -0,0 +1,82 @@
[
{
"test_name": "serving_llama8B_tp1_sharegpt",
"qps_list": [1, 4, 16, "inf"],
"server_environment_variables": {
"PT_HPU_LAZY_MODE": 1,
"VLLM_CONTIGUOUS_PA": 1,
"VLLM_DEFRAG": 1
},
"server_parameters": {
"model": "meta-llama/Meta-Llama-3.1-8B-Instruct",
"tensor_parallel_size": 1,
"swap_space": 16,
"disable_log_stats": "",
"load_format": "dummy",
"max-model-len": 2048,
"max-num-seqs": 256,
"async-scheduling": ""
},
"client_parameters": {
"model": "meta-llama/Meta-Llama-3.1-8B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_llama70B_tp4_sharegpt",
"qps_list": [1, 4, 16, "inf"],
"server_environment_variables": {
"PT_HPU_LAZY_MODE": 1,
"PT_HPU_ENABLE_LAZY_COLLECTIVES": 1,
"VLLM_CONTIGUOUS_PA": 1,
"VLLM_DEFRAG": 1
},
"server_parameters": {
"model": "meta-llama/Meta-Llama-3.1-70B-Instruct",
"tensor_parallel_size": 4,
"swap_space": 16,
"disable_log_stats": "",
"load_format": "dummy",
"max-model-len": 2048,
"max-num-seqs": 256,
"async-scheduling": ""
},
"client_parameters": {
"model": "meta-llama/Meta-Llama-3.1-70B-Instruct",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
},
{
"test_name": "serving_mixtral8x7B_tp2_sharegpt",
"qps_list": [1, 4, 16, "inf"],
"server_environment_variables": {
"PT_HPU_LAZY_MODE": 1,
"PT_HPU_ENABLE_LAZY_COLLECTIVES": 1,
"VLLM_CONTIGUOUS_PA": 1,
"VLLM_DEFRAG": 1
},
"server_parameters": {
"model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
"tensor_parallel_size": 2,
"swap_space": 16,
"disable_log_stats": "",
"load_format": "dummy",
"max-model-len": 2048,
"max-num-seqs": 256,
"async-scheduling": ""
},
"client_parameters": {
"model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
"backend": "vllm",
"dataset_name": "sharegpt",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200
}
}
]

0
.buildkite/nightly-benchmarks/tests/serving-tests.json → .buildkite/performance-benchmarks/tests/serving-tests.json
View File

27
.buildkite/performance-benchmarks/tests/throughput-tests-cpu.json Normal file
View File

@ -0,0 +1,27 @@
[
{
"test_name": "throughput_llama8B_tp2",
"environment_variables": {
"VLLM_RPC_TIMEOUT": 100000,
"VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
"VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
"VLLM_CPU_SGL_KERNEL": 1,
"VLLM_CPU_KVCACHE_SPACE": 40
},
"parameters": {
"model": "meta-llama/Llama-3.1-8B-Instruct",
"tensor_parallel_size": 2,
"dtype": "bfloat16",
"distributed_executor_backend": "mp",
"block_size": 128,
"trust_remote_code": "",
"disable_log_stats": "",
"enforce_eager": "",
"max_num_batched_tokens": 2048,
"max_num_seqs": 256,
"dataset": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 200,
"backend": "vllm"
}
}
]

61
.buildkite/performance-benchmarks/tests/throughput-tests-hpu.json Normal file
View File

@ -0,0 +1,61 @@
[
{
"test_name": "throughput_llama8B_tp1",
"environment_variables": {
"PT_HPU_LAZY_MODE": 1,
"VLLM_CONTIGUOUS_PA": 1,
"VLLM_DEFRAG": 1
},
"parameters": {
"model": "meta-llama/Meta-Llama-3.1-8B-Instruct",
"tensor_parallel_size": 1,
"load_format": "dummy",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 1000,
"backend": "vllm",
"max-model-len": 2048,
"max-num-seqs": 512,
"async-scheduling": ""
}
},
{
"test_name": "throughput_llama70B_tp4",
"environment_variables": {
"PT_HPU_LAZY_MODE": 1,
"PT_HPU_ENABLE_LAZY_COLLECTIVES": 1,
"VLLM_CONTIGUOUS_PA": 1,
"VLLM_DEFRAG": 1
},
"parameters": {
"model": "meta-llama/Meta-Llama-3.1-70B-Instruct",
"tensor_parallel_size": 4,
"load_format": "dummy",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 1000,
"backend": "vllm",
"max-model-len": 2048,
"max-num-seqs": 512,
"async-scheduling": ""
}
},
{
"test_name": "throughput_mixtral8x7B_tp2",
"environment_variables": {
"PT_HPU_LAZY_MODE": 1,
"PT_HPU_ENABLE_LAZY_COLLECTIVES": 1,
"VLLM_CONTIGUOUS_PA": 1,
"VLLM_DEFRAG": 1
},
"parameters": {
"model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
"tensor_parallel_size": 2,
"load_format": "dummy",
"dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
"num_prompts": 1000,
"backend": "vllm",
"max-model-len": 2048,
"max-num-seqs": 512,
"async-scheduling": ""
}
}
]

0
.buildkite/nightly-benchmarks/tests/throughput-tests.json → .buildkite/performance-benchmarks/tests/throughput-tests.json
View File

46
.buildkite/pyproject.toml
View File

@ -1,46 +0,0 @@
# This local pyproject file is part of the migration from yapf to ruff format.
# It uses the same core rules as the main pyproject.toml file, but with the
# following differences:
# - ruff line length is overridden to 88
# - deprecated typing ignores (UP006, UP035) have been removed
[tool.ruff]
line-length = 88
[tool.ruff.lint.per-file-ignores]
"vllm/third_party/**" = ["ALL"]
"vllm/version.py" = ["F401"]
"vllm/_version.py" = ["ALL"]
[tool.ruff.lint]
select = [
# pycodestyle
"E",
# Pyflakes
"F",
# pyupgrade
"UP",
# flake8-bugbear
"B",
# flake8-simplify
"SIM",
# isort
"I",
# flake8-logging-format
"G",
]
ignore = [
# star imports
"F405", "F403",
# lambda expression assignment
"E731",
# Loop control variable not used within loop body
"B007",
# f-string format
"UP032",
# Can remove once 3.10+ is the minimum Python version
"UP007",
]
[tool.ruff.format]
docstring-code-format = true

143
.buildkite/release-pipeline.yaml
View File

@ -1,49 +1,46 @@
steps:
# aarch64 + CUDA builds. PyTorch 2.8 aarch64 + CUDA wheel is only available on CUDA 12.9
# aarch64 + CUDA builds
- label: "Build arm64 wheel - CUDA 12.9"
depends_on: ~
id: build-wheel-arm64-cuda-12-9
agents:
queue: arm64_cpu_queue_postmerge
commands:
# #NOTE: torch_cuda_arch_list is derived from upstream PyTorch build files here:
# https://github.com/pytorch/pytorch/blob/main/.ci/aarch64_linux/aarch64_ci_build.sh#L7
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list='8.7 9.0 10.0+PTX 12.0' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
- "mkdir artifacts"
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
- "bash .buildkite/scripts/upload-wheels.sh"
env:
DOCKER_BUILDKIT: "1"
- block: "Build CUDA 12.8 wheel"
key: block-build-cu128-wheel
- label: "Build wheel - CUDA 12.8"
depends_on: block-build-cu128-wheel
id: build-wheel-cuda-12-8
agents:
queue: cpu_queue_postmerge
commands:
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.8.1 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
- "mkdir artifacts"
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
- "bash .buildkite/scripts/upload-wheels.sh"
env:
DOCKER_BUILDKIT: "1"
- block: "Build CUDA 12.6 wheel"
key: block-build-cu126-wheel
- label: "Build arm64 wheel - CUDA 13.0"
depends_on: ~
- label: "Build wheel - CUDA 12.6"
depends_on: block-build-cu126-wheel
id: build-wheel-cuda-12-6
id: build-wheel-arm64-cuda-13-0
agents:
queue: cpu_queue_postmerge
queue: arm64_cpu_queue_postmerge
commands:
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.6.3 --build-arg torch_cuda_arch_list='7.0 7.5 8.0 8.9 9.0+PTX' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
# #NOTE: torch_cuda_arch_list is derived from upstream PyTorch build files here:
# https://github.com/pytorch/pytorch/blob/main/.ci/aarch64_linux/aarch64_ci_build.sh#L7
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
- "mkdir artifacts"
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
- "bash .buildkite/scripts/upload-wheels.sh"
- "bash .buildkite/scripts/upload-wheels.sh manylinux_2_35"
env:
DOCKER_BUILDKIT: "1"
# aarch64 build
- label: "Build arm64 CPU wheel"
depends_on: ~
id: build-wheel-arm64-cpu
agents:
queue: arm64_cpu_queue_postmerge
commands:
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --build-arg VLLM_BUILD_ACL=ON --tag vllm-ci:build-image --target vllm-build --progress plain -f docker/Dockerfile.cpu ."
- "mkdir artifacts"
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
- "bash .buildkite/scripts/upload-wheels.sh manylinux_2_35"
env:
DOCKER_BUILDKIT: "1"
@ -54,13 +51,41 @@ steps:
agents:
queue: cpu_queue_postmerge
commands:
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list='7.0 7.5 8.0 8.9 9.0+PTX' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
- "mkdir artifacts"
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
- "bash .buildkite/scripts/upload-wheels.sh"
- "bash .buildkite/scripts/upload-wheels.sh manylinux_2_31"
env:
DOCKER_BUILDKIT: "1"
- label: "Build wheel - CUDA 13.0"
depends_on: ~
id: build-wheel-cuda-13-0
agents:
queue: cpu_queue_postmerge
commands:
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
- "mkdir artifacts"
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
- "bash .buildkite/scripts/upload-wheels.sh manylinux_2_35"
env:
DOCKER_BUILDKIT: "1"
# x86 CPU wheel build
- label: "Build x86 CPU wheel"
depends_on: ~
id: build-wheel-x86-cpu
agents:
queue: cpu_queue_postmerge
commands:
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --build-arg VLLM_CPU_AVX512BF16=true --build-arg VLLM_CPU_AVX512VNNI=true --build-arg VLLM_CPU_AMXBF16=true --tag vllm-ci:build-image --target vllm-build --progress plain -f docker/Dockerfile.cpu ."
- "mkdir artifacts"
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
- "bash .buildkite/scripts/upload-wheels.sh manylinux_2_35"
env:
DOCKER_BUILDKIT: "1"
# Build release images (12.9)
- label: "Build release image (x86)"
depends_on: ~
id: build-release-image-x86
@ -68,13 +93,12 @@ steps:
queue: cpu_queue_postmerge
commands:
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.8.1 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
# re-tag to default image tag and push, just in case arm64 build fails
- "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
# PyTorch 2.8 aarch64 + CUDA wheel is only available on CUDA 12.9
- label: "Build release image (arm64)"
depends_on: ~
id: build-release-image-arm64
@ -82,7 +106,7 @@ steps:
queue: arm64_cpu_queue_postmerge
commands:
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list='8.7 9.0 10.0+PTX 12.0' --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
# Add job to create multi-arch manifest
@ -101,33 +125,12 @@ steps:
- label: "Annotate release workflow"
depends_on:
- create-multi-arch-manifest
- build-wheel-cuda-12-8
- build-wheel-cuda-12-6
- build-wheel-cuda-12-9
id: annotate-release-workflow
agents:
queue: cpu_queue_postmerge
commands:
- "bash .buildkite/scripts/annotate-release.sh"
- label: "Build and publish TPU release image"
depends_on: ~
if: build.env("NIGHTLY") == "1"
agents:
queue: tpu_queue_postmerge
commands:
- "yes | docker system prune -a"
- "git fetch --all"
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --tag vllm/vllm-tpu:nightly --tag vllm/vllm-tpu:$BUILDKITE_COMMIT --progress plain -f docker/Dockerfile.tpu ."
- "docker push vllm/vllm-tpu:nightly"
- "docker push vllm/vllm-tpu:$BUILDKITE_COMMIT"
plugins:
- docker-login#v3.0.0:
username: vllmbot
password-env: DOCKERHUB_TOKEN
env:
DOCKER_BUILDKIT: "1"
- input: "Provide Release version here"
id: input-release-version
fields:
@ -144,12 +147,28 @@ steps:
queue: cpu_queue_postmerge
commands:
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --build-arg VLLM_CPU_AVX512BF16=true --build-arg VLLM_CPU_AVX512VNNI=true --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest --progress plain --target vllm-openai -f docker/Dockerfile.cpu ."
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --build-arg VLLM_CPU_AVX512BF16=true --build-arg VLLM_CPU_AVX512VNNI=true --build-arg VLLM_CPU_AMXBF16=true --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest --progress plain --target vllm-openai -f docker/Dockerfile.cpu ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest"
- "docker push public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(buildkite-agent meta-data get release-version)"
env:
DOCKER_BUILDKIT: "1"
- block: "Build arm64 CPU release image"
key: block-arm64-cpu-release-image-build
depends_on: ~
- label: "Build and publish arm64 CPU release image"
depends_on: block-arm64-cpu-release-image-build
agents:
queue: arm64_cpu_queue_postmerge
commands:
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --tag public.ecr.aws/q9t5s3a7/vllm-arm64-cpu-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-arm64-cpu-release-repo:latest --progress plain --target vllm-openai -f docker/Dockerfile.cpu ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-arm64-cpu-release-repo:latest"
- "docker push public.ecr.aws/q9t5s3a7/vllm-arm64-cpu-release-repo:$(buildkite-agent meta-data get release-version)"
env:
DOCKER_BUILDKIT: "1"
- label: "Build and publish nightly multi-arch image to DockerHub"
depends_on:
- create-multi-arch-manifest
@ -158,11 +177,16 @@ steps:
queue: cpu_queue_postmerge
commands:
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
- "docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
- "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT vllm/vllm-openai:nightly"
- "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT vllm/vllm-openai:nightly-$BUILDKITE_COMMIT"
- "docker push vllm/vllm-openai:nightly"
- "docker push vllm/vllm-openai:nightly-$BUILDKITE_COMMIT"
- "docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64"
- "docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64"
- "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64 vllm/vllm-openai:nightly-x86_64"
- "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64 vllm/vllm-openai:nightly-aarch64"
- "docker push vllm/vllm-openai:nightly-x86_64"
- "docker push vllm/vllm-openai:nightly-aarch64"
- "docker manifest create vllm/vllm-openai:nightly vllm/vllm-openai:nightly-x86_64 vllm/vllm-openai:nightly-aarch64 --amend"
- "docker manifest create vllm/vllm-openai:nightly-$BUILDKITE_COMMIT vllm/vllm-openai:nightly-x86_64 vllm/vllm-openai:nightly-aarch64 --amend"
- "docker manifest push vllm/vllm-openai:nightly"
- "docker manifest push vllm/vllm-openai:nightly-$BUILDKITE_COMMIT"
# Clean up old nightly builds (keep only last 14)
- "bash .buildkite/scripts/cleanup-nightly-builds.sh"
plugins:
@ -171,3 +195,4 @@ steps:
password-env: DOCKERHUB_TOKEN
env:
DOCKER_BUILDKIT: "1"
DOCKERHUB_USERNAME: "vllmbot"

55
.buildkite/scripts/annotate-release.sh
View File

@ -2,30 +2,53 @@
set -ex
# Get release version and strip leading 'v' if present
RELEASE_VERSION=$(buildkite-agent meta-data get release-version | sed 's/^v//')
if [ -z "$RELEASE_VERSION" ]; then
echo "Error: RELEASE_VERSION is empty. 'release-version' metadata might not be set or is invalid."
exit 1
# Get release version, default to 1.0.0.dev for nightly/per-commit builds
RELEASE_VERSION=$(buildkite-agent meta-data get release-version 2>/dev/null | sed 's/^v//')
if [ -z "${RELEASE_VERSION}" ]; then
RELEASE_VERSION="1.0.0.dev"
fi
buildkite-agent annotate --style 'info' --context 'release-workflow' << EOF
To download the wheel:
To download the wheel (by commit):
\`\`\`
aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux1_x86_64.whl .
aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux2014_aarch64.whl .
aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}+cu129-cp38-abi3-manylinux1_x86_64.whl .
aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}+cu129-cp38-abi3-manylinux1_x86_64.whl .
\`\`\`
To download the wheel (by version):
\`\`\`
aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux1_x86_64.whl .
aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}+cu126/vllm-${RELEASE_VERSION}+cu126-cp38-abi3-manylinux1_x86_64.whl .
aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}+cu118/vllm-${RELEASE_VERSION}+cu118-cp38-abi3-manylinux1_x86_64.whl .
aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux2014_aarch64.whl .
aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}+cu129/vllm-${RELEASE_VERSION}+cu129-cp38-abi3-manylinux1_x86_64.whl .
aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}+cu130/vllm-${RELEASE_VERSION}+cu130-cp38-abi3-manylinux1_x86_64.whl .
\`\`\`
To download and upload the image:
\`\`\`
docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}
docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT} vllm/vllm-openai
docker tag vllm/vllm-openai vllm/vllm-openai:latest
docker tag vllm/vllm-openai vllm/vllm-openai:v${RELEASE_VERSION}
docker push vllm/vllm-openai:latest
docker push vllm/vllm-openai:v${RELEASE_VERSION}
docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-x86_64
docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64
docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-x86_64 vllm/vllm-openai:x86_64
docker tag vllm/vllm-openai:x86_64 vllm/vllm-openai:latest-x86_64
docker tag vllm/vllm-openai:x86_64 vllm/vllm-openai:v${RELEASE_VERSION}-x86_64
docker push vllm/vllm-openai:latest-x86_64
docker push vllm/vllm-openai:v${RELEASE_VERSION}-x86_64
docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64 vllm/vllm-openai:aarch64
docker tag vllm/vllm-openai:aarch64 vllm/vllm-openai:latest-aarch64
docker tag vllm/vllm-openai:aarch64 vllm/vllm-openai:v${RELEASE_VERSION}-aarch64
docker push vllm/vllm-openai:latest-aarch64
docker push vllm/vllm-openai:v${RELEASE_VERSION}-aarch64
docker manifest rm vllm/vllm-openai:latest
docker manifest create vllm/vllm-openai:latest vllm/vllm-openai:latest-x86_64 vllm/vllm-openai:latest-aarch64
docker manifest create vllm/vllm-openai:v${RELEASE_VERSION} vllm/vllm-openai:v${RELEASE_VERSION}-x86_64 vllm/vllm-openai:v${RELEASE_VERSION}-aarch64
docker manifest push vllm/vllm-openai:latest
docker manifest push vllm/vllm-openai:v${RELEASE_VERSION}
\`\`\`
EOF
EOF

29
.buildkite/scripts/cleanup-nightly-builds.sh
View File

@ -8,20 +8,41 @@ set -ex
# DockerHub API endpoint for vllm/vllm-openai repository
REPO_API_URL="https://hub.docker.com/v2/repositories/vllm/vllm-openai/tags"
# Get DockerHub token from environment
# Get DockerHub credentials from environment
if [ -z "$DOCKERHUB_TOKEN" ]; then
echo "Error: DOCKERHUB_TOKEN environment variable is not set"
exit 1
fi
if [ -z "$DOCKERHUB_USERNAME" ]; then
echo "Error: DOCKERHUB_USERNAME environment variable is not set"
exit 1
fi
# Get DockerHub bearer token
echo "Getting DockerHub bearer token..."
set +x
BEARER_TOKEN=$(curl -s -X POST \
-H "Content-Type: application/json" \
-d "{\"username\": \"$DOCKERHUB_USERNAME\", \"password\": \"$DOCKERHUB_TOKEN\"}" \
"https://hub.docker.com/v2/users/login" | jq -r '.token')
set -x
if [ -z "$BEARER_TOKEN" ] || [ "$BEARER_TOKEN" = "null" ]; then
echo "Error: Failed to get DockerHub bearer token"
exit 1
fi
# Function to get all tags from DockerHub
get_all_tags() {
local page=1
local all_tags=""
while true; do
local response=$(curl -s -H "Authorization: Bearer $DOCKERHUB_TOKEN" \
set +x
local response=$(curl -s -H "Authorization: Bearer $BEARER_TOKEN" \
"$REPO_API_URL?page=$page&page_size=100")
set -x
# Get both last_updated timestamp and tag name, separated by |
local tags=$(echo "$response" | jq -r '.results[] | select(.name | startswith("nightly-")) | "\(.last_updated)|\(.name)"')
@ -43,7 +64,9 @@ delete_tag() {
echo "Deleting tag: $tag_name"
local delete_url="https://hub.docker.com/v2/repositories/vllm/vllm-openai/tags/$tag_name"
local response=$(curl -s -X DELETE -H "Authorization: Bearer $DOCKERHUB_TOKEN" "$delete_url")
set +x
local response=$(curl -s -X DELETE -H "Authorization: Bearer $BEARER_TOKEN" "$delete_url")
set -x
if echo "$response" | jq -e '.detail' > /dev/null 2>&1; then
echo "Warning: Failed to delete tag $tag_name: $(echo "$response" | jq -r '.detail')"

400
.buildkite/scripts/generate-nightly-index.py Normal file
View File

@ -0,0 +1,400 @@
#!/usr/bin/env python3
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
# do not complain about line length (for docstring)
# ruff: noqa: E501
import argparse
import json
import sys
from dataclasses import asdict, dataclass
from datetime import datetime
from pathlib import Path
from typing import Any
from urllib.parse import quote
import regex as re
if not sys.version_info >= (3, 12):
raise RuntimeError("This script requires Python 3.12 or higher.")
INDEX_HTML_TEMPLATE = """<!DOCTYPE html>
<html>
<!-- {comment} -->
<meta name="pypi:repository-version" content="1.0">
<body>
{items}
</body>
</html>
"""
@dataclass
class WheelFileInfo:
package_name: str
version: str
build_tag: str | None
python_tag: str
abi_tag: str
platform_tag: str
variant: str | None
filename: str
def parse_from_filename(file: str) -> WheelFileInfo:
"""
Parse wheel file name to extract metadata.
The format of wheel names:
{package_name}-{version}(-{build_tag})?-{python_tag}-{abi_tag}-{platform_tag}.whl
All versions could contain a variant like '+cu129' or '.cpu' or `.rocm` (or not).
Example:
vllm-0.11.0-cp38-abi3-manylinux1_x86_64.whl
vllm-0.10.2rc2+cu129-cp38-abi3-manylinux2014_aarch64.whl
vllm-0.11.1rc8.dev14+gaa384b3c0-cp38-abi3-manylinux2014_aarch64.whl
vllm-0.11.1rc8.dev14+gaa384b3c0.cu130-cp38-abi3-manylinux1_x86_64.whl
"""
wheel_file_re = re.compile(
r"^(?P<package_name>.+)-(?P<version>[^-]+?)(-(?P<build_tag>[^-]+))?-(?P<python_tag>[^-]+)-(?P<abi_tag>[^-]+)-(?P<platform_tag>[^-]+)\.whl$"
)
match = wheel_file_re.match(file)
if not match:
raise ValueError(f"Invalid wheel file name: {file}")
package_name = match.group("package_name")
version = match.group("version")
build_tag = match.group("build_tag")
python_tag = match.group("python_tag")
abi_tag = match.group("abi_tag")
platform_tag = match.group("platform_tag")
# extract variant from version
variant = None
if "dev" in version:
ver_after_dev = version.split("dev")[-1]
if "." in ver_after_dev:
variant = ver_after_dev.split(".")[-1]
version = version.removesuffix("." + variant)
else:
if "+" in version:
version, variant = version.split("+")
return WheelFileInfo(
package_name=package_name,
version=version,
build_tag=build_tag,
python_tag=python_tag,
abi_tag=abi_tag,
platform_tag=platform_tag,
variant=variant,
filename=file,
)
def generate_project_list(subdir_names: list[str], comment: str = "") -> str:
"""
Generate project list HTML content linking to each project & variant sub-directory.
"""
href_tags = []
for name in sorted(subdir_names):
name = name.strip("/").strip(".")
href_tags.append(f' <a href="{name}/">{name}/</a><br/>')
return INDEX_HTML_TEMPLATE.format(items="\n".join(href_tags), comment=comment)
def generate_package_index_and_metadata(
wheel_files: list[WheelFileInfo],
wheel_base_dir: Path,
index_base_dir: Path,
comment: str = "",
) -> tuple[str, str]:
"""
Generate package index HTML content for a specific package, linking to actual wheel files.
"""
href_tags = []
metadata = []
for file in sorted(wheel_files, key=lambda x: x.filename):
relative_path = (
wheel_base_dir.relative_to(index_base_dir, walk_up=True) / file.filename
)
# handle with '+' in URL, and avoid double-encoding '/' and already-encoded '%2B'
# NOTE: this is AWS S3 specific behavior!
file_path_quoted = quote(relative_path.as_posix(), safe=":%/")
href_tags.append(f' <a href="{file_path_quoted}">{file.filename}</a><br/>')
file_meta = asdict(file)
file_meta["path"] = file_path_quoted
metadata.append(file_meta)
index_str = INDEX_HTML_TEMPLATE.format(items="\n".join(href_tags), comment=comment)
metadata_str = json.dumps(metadata, indent=2)
return index_str, metadata_str
def generate_index_and_metadata(
whl_files: list[str],
wheel_base_dir: Path,
index_base_dir: Path,
default_variant: str | None = None,
alias_to_default: str | None = None,
comment: str = "",
):
"""
Generate index for all wheel files.
Args:
whl_files (list[str]): List of wheel files (must be directly under `wheel_base_dir`).
wheel_base_dir (Path): Base directory for wheel files.
index_base_dir (Path): Base directory to store index files.
default_variant (str | None): The default variant name, if any.
alias_to_default (str | None): Alias variant name for the default variant, if any.
comment (str | None): Optional comment to include in the generated HTML files.
First, parse all wheel files to extract metadata.
We need to collect all wheel files for each variant, and generate an index for it (in a sub-directory).
The index for the default variant (if any) is generated in the root index directory.
If `default_variant` is provided, all wheels must have variant suffixes, and the default variant index
is purely a copy of the corresponding variant index, with only the links adjusted.
Otherwise, all wheels without variant suffixes are treated as the default variant.
If `alias_to_default` is provided, an additional alias sub-directory is created, it has the same content
as the default variant index, but the links are adjusted accordingly.
Index directory structure:
index_base_dir/ (hosted at wheels.vllm.ai/{nightly,$commit,$version}/)
index.html # project list, linking to "vllm/" and other packages, and all variant sub-directories
vllm/
index.html # package index, pointing to actual files in wheel_base_dir (relative path)
metadata.json # machine-readable metadata for all wheels in this package
cpu/ # cpu variant sub-directory
index.html
vllm/
index.html
metadata.json
cu129/ # cu129 is actually the alias to default variant
index.html
vllm/
index.html
metadata.json
cu130/ # cu130 variant sub-directory
index.html
vllm/
index.html
metadata.json
...
metadata.json stores a dump of all wheel files' metadata in a machine-readable format:
[
{
"package_name": "vllm",
"version": "0.10.2rc2",
"build_tag": null,
"python_tag": "cp38",
"abi_tag": "abi3",
"platform_tag": "manylinux2014_aarch64",
"variant": "cu129",
"filename": "vllm-0.10.2rc2+cu129-cp38-abi3-manylinux2014_aarch64.whl",
"path": "../vllm-0.10.2rc2%2Bcu129-cp38-abi3-manylinux2014_aarch64.whl" # to be concatenated with the directory URL and URL-encoded
},
...
]
"""
parsed_files = [parse_from_filename(f) for f in whl_files]
if not parsed_files:
print("No wheel files found, skipping index generation.")
return
# Group by variant
variant_to_files: dict[str, list[WheelFileInfo]] = {}
for file in parsed_files:
variant = file.variant or "default"
if variant not in variant_to_files:
variant_to_files[variant] = []
variant_to_files[variant].append(file)
print(f"Found variants: {list(variant_to_files.keys())}")
# sanity check for default variant
if default_variant:
if "default" in variant_to_files:
raise ValueError(
"All wheel files must have variant suffixes when `default_variant` is specified."
)
if default_variant not in variant_to_files:
raise ValueError(
f"Default variant '{default_variant}' not found among wheel files."
)
if alias_to_default:
if "default" not in variant_to_files:
# e.g. only some wheels are uploaded to S3 currently
print(
"[WARN] Alias to default variant specified, but no default variant found."
)
elif alias_to_default in variant_to_files:
raise ValueError(
f"Alias variant name '{alias_to_default}' already exists among wheel files."
)
else:
variant_to_files[alias_to_default] = variant_to_files["default"].copy()
print(f"Alias variant '{alias_to_default}' created for default variant.")
# Generate comment in HTML header
comment_str = f" ({comment})" if comment else ""
comment_tmpl = f"Generated on {datetime.now().isoformat()}{comment_str}"
# Generate index for each variant
subdir_names = set()
for variant, files in variant_to_files.items():
if variant == "default":
variant_dir = index_base_dir
else:
variant_dir = index_base_dir / variant
subdir_names.add(variant)
variant_dir.mkdir(parents=True, exist_ok=True)
# gather all package names in this variant
packages = set(f.package_name for f in files)
if variant == "default":
# these packages should also appear in the "project list"
# generate after all variants are processed
subdir_names = subdir_names.union(packages)
else:
# generate project list for this variant directly
project_list_str = generate_project_list(sorted(packages), comment_tmpl)
with open(variant_dir / "index.html", "w") as f:
f.write(project_list_str)
for package in packages:
# filter files belonging to this package only
package_files = [f for f in files if f.package_name == package]
package_dir = variant_dir / package
package_dir.mkdir(parents=True, exist_ok=True)
index_str, metadata_str = generate_package_index_and_metadata(
package_files, wheel_base_dir, package_dir, comment
)
with open(package_dir / "index.html", "w") as f:
f.write(index_str)
with open(package_dir / "metadata.json", "w") as f:
f.write(metadata_str)
# Generate top-level project list index
project_list_str = generate_project_list(sorted(subdir_names), comment_tmpl)
with open(index_base_dir / "index.html", "w") as f:
f.write(project_list_str)
if __name__ == "__main__":
"""
Arguments:
--version <version> : version string for the current build (e.g., commit hash)
--current-objects <path_to_json> : path to JSON file containing current S3 objects listing in this version directory
--output-dir <output_directory> : directory to store generated index files
--alias-to-default <alias_variant_name> : (optional) alias variant name for the default variant
--comment <comment_string> : (optional) comment string to include in generated HTML files
"""
parser = argparse.ArgumentParser(
description="Process nightly build wheel files to generate indices."
)
parser.add_argument(
"--version",
type=str,
required=True,
help="Version string for the current build (e.g., commit hash)",
)
parser.add_argument(
"--current-objects",
type=str,
required=True,
help="Path to JSON file containing current S3 objects listing in this version directory",
)
parser.add_argument(
"--output-dir",
type=str,
required=True,
help="Directory to store generated index files",
)
parser.add_argument(
"--alias-to-default",
type=str,
default=None,
help="Alias variant name for the default variant",
)
parser.add_argument(
"--comment",
type=str,
default="",
help="Optional comment string to include in generated HTML files",
)
args = parser.parse_args()
version = args.version
if "/" in version or "\\" in version:
raise ValueError("Version string must not contain slashes.")
current_objects_path = Path(args.current_objects)
output_dir = Path(args.output_dir)
if not output_dir.exists():
output_dir.mkdir(parents=True, exist_ok=True)
# Read current objects JSON
with open(current_objects_path) as f:
current_objects: dict[str, list[dict[str, Any]]] = json.load(f)
# current_objects looks like from list_objects_v2 S3 API:
"""
"Contents": [
{
"Key": "e2f56c309d2a28899c68975a7e104502d56deb8f/vllm-0.11.2.dev363+ge2f56c309-cp38-abi3-manylinux1_x86_64.whl",
"LastModified": "2025-11-28T14:00:32+00:00",
"ETag": "\"37a38339c7cdb61ca737021b968075df-52\"",
"ChecksumAlgorithm": [
"CRC64NVME"
],
"ChecksumType": "FULL_OBJECT",
"Size": 435649349,
"StorageClass": "STANDARD"
},
...
]
"""
# Extract wheel file keys
wheel_files = []
for item in current_objects.get("Contents", []):
key: str = item["Key"]
if key.endswith(".whl"):
wheel_files.append(key.split("/")[-1]) # only the filename is used
print(f"Found {len(wheel_files)} wheel files for version {version}: {wheel_files}")
# keep only "official" files for a non-nightly version (specifed by cli args)
PY_VERSION_RE = re.compile(r"^\d+\.\d+\.\d+([a-zA-Z0-9.+-]*)?$")
if PY_VERSION_RE.match(version):
# upload-wheels.sh ensures no "dev" is in args.version
wheel_files = list(
filter(lambda x: version in x and "dev" not in x, wheel_files)
)
print(f"Non-nightly version detected, wheel files used: {wheel_files}")
else:
print("Nightly version detected, keeping all wheel files.")
# Generate index and metadata, assuming wheels and indices are stored as:
# s3://vllm-wheels/{version}/<wheel files>
# s3://vllm-wheels/<anything>/<index files>
wheel_base_dir = Path(output_dir).parent / version
index_base_dir = Path(output_dir)
generate_index_and_metadata(
whl_files=wheel_files,
wheel_base_dir=wheel_base_dir,
index_base_dir=index_base_dir,
default_variant=None,
alias_to_default=args.alias_to_default,
comment=args.comment.strip(),
)
print(f"Successfully generated index and metadata in {output_dir}")

39
.buildkite/scripts/hardware_ci/run-amd-test.sh
View File

@ -59,7 +59,7 @@ while true; do
fi
done
echo "--- Pulling container"
echo "--- Pulling container"
image_name="rocm/vllm-ci:${BUILDKITE_COMMIT}"
container_name="rocm_${BUILDKITE_COMMIT}_$(tr -dc A-Za-z0-9 < /dev/urandom | head -c 10; echo)"
docker pull "${image_name}"
@ -78,21 +78,13 @@ HF_MOUNT="/root/.cache/huggingface"
commands=$@
echo "Commands:$commands"
if [[ $commands == *"pytest -v -s basic_correctness/test_basic_correctness.py"* ]]; then
commands=${commands//"pytest -v -s basic_correctness/test_basic_correctness.py"/"VLLM_USE_TRITON_FLASH_ATTN=0 pytest -v -s basic_correctness/test_basic_correctness.py"}
fi
commands=${commands//"pytest -v -s basic_correctness/test_basic_correctness.py"/"pytest -v -s basic_correctness/test_basic_correctness.py"}
if [[ $commands == *"pytest -v -s models/test_registry.py"* ]]; then
commands=${commands//"pytest -v -s models/test_registry.py"/"pytest -v -s models/test_registry.py -k 'not BambaForCausalLM and not GritLM and not Mamba2ForCausalLM and not Zamba2ForCausalLM'"}
fi
if [[ $commands == *"VLLM_USE_V1=0 pytest -v -s models/test_initialization.py -k 'not llama4 and not plamo2'"* ]]; then
commands=${commands//"VLLM_USE_V1=0 pytest -v -s models/test_initialization.py -k 'not llama4 and not plamo2'"/"VLLM_USE_V1=0 pytest -v -s models/test_initialization.py -k 'not llama4 and not plamo2 and not BambaForCausalLM and not Gemma2ForCausalLM and not Grok1ModelForCausalLM and not Zamba2ForCausalLM and not Gemma2Model and not GritLM'"}
fi
if [[ $commands == *"pytest -v -s compile/test_basic_correctness.py"* ]]; then
commands=${commands//"pytest -v -s compile/test_basic_correctness.py"/"VLLM_USE_TRITON_FLASH_ATTN=0 pytest -v -s compile/test_basic_correctness.py"}
fi
commands=${commands//"pytest -v -s compile/test_basic_correctness.py"/"pytest -v -s compile/test_basic_correctness.py"}
if [[ $commands == *"pytest -v -s lora"* ]]; then
commands=${commands//"pytest -v -s lora"/"VLLM_ROCM_CUSTOM_PAGED_ATTN=0 pytest -v -s lora"}
@ -167,12 +159,6 @@ if [[ $commands == *" entrypoints/llm "* ]]; then
--ignore=entrypoints/llm/test_prompt_validation.py "}
fi
#Obsolete currently
##ignore certain Entrypoints/llm tests
#if [[ $commands == *" && pytest -v -s entrypoints/llm/test_guided_generate.py"* ]]; then
# commands=${commands//" && pytest -v -s entrypoints/llm/test_guided_generate.py"/" "}
#fi
# --ignore=entrypoints/openai/test_encoder_decoder.py \
# --ignore=entrypoints/openai/test_embedding.py \
# --ignore=entrypoints/openai/test_oot_registration.py
@ -183,19 +169,28 @@ fi
PARALLEL_JOB_COUNT=8
MYPYTHONPATH=".."
# check if the command contains shard flag, we will run all shards in parallel because the host have 8 GPUs.
# Test that we're launching on the machine that has
# proper access to GPUs
render_gid=$(getent group render | cut -d: -f3)
if [[ -z "$render_gid" ]]; then
echo "Error: 'render' group not found. This is required for GPU access." >&2
exit 1
fi
# check if the command contains shard flag, we will run all shards in parallel because the host have 8 GPUs.
if [[ $commands == *"--shard-id="* ]]; then
# assign job count as the number of shards used
commands=${commands//"--num-shards= "/"--num-shards=${PARALLEL_JOB_COUNT} "}
# assign job count as the number of shards used
commands=$(echo "$commands" | sed -E "s/--num-shards[[:blank:]]*=[[:blank:]]*[0-9]*/--num-shards=${PARALLEL_JOB_COUNT} /g" | sed 's/ \\ / /g')
for GPU in $(seq 0 $(($PARALLEL_JOB_COUNT-1))); do
# assign shard-id for each shard
commands_gpu=${commands//"--shard-id= "/"--shard-id=${GPU} "}
commands_gpu=$(echo "$commands" | sed -E "s/--shard-id[[:blank:]]*=[[:blank:]]*[0-9]*/--shard-id=${GPU} /g" | sed 's/ \\ / /g')
echo "Shard ${GPU} commands:$commands_gpu"
echo "Render devices: $BUILDKITE_AGENT_META_DATA_RENDER_DEVICES"
docker run \
--device /dev/kfd $BUILDKITE_AGENT_META_DATA_RENDER_DEVICES \
--network=host \
--shm-size=16gb \
--group-add "$render_gid" \
--rm \
-e HIP_VISIBLE_DEVICES="${GPU}" \
-e HF_TOKEN \
@ -227,8 +222,8 @@ else
--device /dev/kfd $BUILDKITE_AGENT_META_DATA_RENDER_DEVICES \
--network=host \
--shm-size=16gb \
--group-add "$render_gid" \
--rm \
-e HIP_VISIBLE_DEVICES=0 \
-e HF_TOKEN \
-e AWS_ACCESS_KEY_ID \
-e AWS_SECRET_ACCESS_KEY \

68
.buildkite/scripts/hardware_ci/run-cpu-test-arm.sh Executable file
View File

@ -0,0 +1,68 @@
#!/bin/bash
# This script build the CPU docker image and run the offline inference inside the container.
# It serves a sanity check for compilation and basic model usage.
set -ex
# allow to bind to different cores
CORE_RANGE=${CORE_RANGE:-0-16}
OMP_CORE_RANGE=${OMP_CORE_RANGE:-0-16}
export CMAKE_BUILD_PARALLEL_LEVEL=16
# Setup cleanup
remove_docker_container() {
set -e;
docker rm -f cpu-test || true;
}
trap remove_docker_container EXIT
remove_docker_container
# Try building the docker image
docker build --tag cpu-test --target vllm-test -f docker/Dockerfile.cpu .
# Run the image
docker run -itd --cpuset-cpus="$CORE_RANGE" --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=16 --env VLLM_CPU_CI_ENV=1 -e E2E_OMP_THREADS="$OMP_CORE_RANGE" --shm-size=4g --name cpu-test cpu-test
function cpu_tests() {
set -e
docker exec cpu-test bash -c "
set -e
pip list"
# offline inference
docker exec cpu-test bash -c "
set -e
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m"
# Run model tests
docker exec cpu-test bash -c "
set -e
pytest -x -v -s tests/models/multimodal/generation/test_whisper.py -m cpu_model"
# Run kernel tests
docker exec cpu-test bash -c "
set -e
pytest -x -v -s tests/kernels/test_onednn.py
pytest -x -v -s tests/kernels/attention/test_cpu_attn.py
pytest -x -v -s tests/kernels/moe/test_moe.py -k test_cpu_fused_moe_basic"
# basic online serving
docker exec cpu-test bash -c '
set -e
VLLM_CPU_OMP_THREADS_BIND=$E2E_OMP_THREADS vllm serve Qwen/Qwen3-0.6B --max-model-len 2048 &
server_pid=$!
timeout 600 bash -c "until curl localhost:8000/v1/models; do sleep 1; done" || exit 1
vllm bench serve \
--backend vllm \
--dataset-name random \
--model Qwen/Qwen3-0.6B \
--num-prompts 20 \
--endpoint /v1/completions
kill -s SIGTERM $server_pid &'
}
# All of CPU tests are expected to be finished less than 40 mins.
export -f cpu_tests
timeout 2h bash -c cpu_tests

25
.buildkite/scripts/hardware_ci/run-cpu-test-ppc64le.sh
View File

@ -25,25 +25,30 @@ function cpu_tests() {
# offline inference
podman exec -it "$container_id" bash -c "
set -e
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m"
export TORCH_COMPILE_DISABLE=1
set -xve
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m" >> $HOME/test_basic.log
# Run basic model test
podman exec -it "$container_id" bash -c "
set -e
export TORCH_COMPILE_DISABLE=1
set -evx
pip install pytest pytest-asyncio einops peft Pillow soundfile transformers_stream_generator matplotlib
pip install sentence-transformers datamodel_code_generator
pytest -v -s tests/models/language/generation/test_bart.py -m cpu_model
pytest -v -s tests/models/language/generation/test_common.py::test_models[False-5-32-openai-community/gpt2]
pytest -v -s tests/models/language/generation/test_common.py::test_models[False-5-32-facebook/opt-125m]
pytest -v -s tests/models/language/generation/test_common.py::test_models[False-5-32-google/gemma-1.1-2b-it]
pip install sentence-transformers datamodel_code_generator tblib
# Note: disable Bart until supports V1
# pytest -v -s tests/models/language/generation/test_bart.py -m cpu_model
pytest -v -s tests/models/language/generation/test_common.py::test_models[False-False-5-32-openai-community/gpt2]
pytest -v -s tests/models/language/generation/test_common.py::test_models[False-False-5-32-facebook/opt-125m]
pytest -v -s tests/models/language/generation/test_common.py::test_models[False-False-5-32-google/gemma-1.1-2b-it]
pytest -v -s tests/models/language/pooling/test_classification.py::test_models[float-jason9693/Qwen2.5-1.5B-apeach]
pytest -v -s tests/models/language/pooling/test_embedding.py -m cpu_model"
# TODO: Below test case tests/models/language/pooling/test_embedding.py::test_models[True-ssmits/Qwen2-7B-Instruct-embed-base] fails on ppc64le. Disabling it for time being.
# pytest -v -s tests/models/language/pooling/test_embedding.py -m cpu_model" >> $HOME/test_rest.log
}
# All of CPU tests are expected to be finished less than 40 mins.
export container_id
export -f cpu_tests
timeout 40m bash -c cpu_tests
timeout 120m bash -c cpu_tests

28
.buildkite/scripts/hardware_ci/run-cpu-test.sh
View File

@ -21,8 +21,8 @@ trap remove_docker_container EXIT
remove_docker_container
# Try building the docker image
numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --tag cpu-test-"$NUMA_NODE" --target vllm-test -f docker/Dockerfile.cpu .
numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --build-arg VLLM_CPU_DISABLE_AVX512="true" --tag cpu-test-"$NUMA_NODE"-avx2 --target vllm-test -f docker/Dockerfile.cpu .
numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --progress plain --tag cpu-test-"$NUMA_NODE" --target vllm-test -f docker/Dockerfile.cpu .
numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --progress plain --build-arg VLLM_CPU_DISABLE_AVX512="true" --tag cpu-test-"$NUMA_NODE"-avx2 --target vllm-test -f docker/Dockerfile.cpu .
# Run the image, setting --shm-size=4g for tensor parallel.
docker run -itd --cpuset-cpus="$CORE_RANGE" --cpuset-mems="$NUMA_NODE" --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --privileged=true -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=16 --env VLLM_CPU_CI_ENV=1 -e E2E_OMP_THREADS="$OMP_CORE_RANGE" --shm-size=4g --name cpu-test-"$NUMA_NODE" cpu-test-"$NUMA_NODE"
@ -49,6 +49,7 @@ function cpu_tests() {
# Run kernel tests
docker exec cpu-test-"$NUMA_NODE" bash -c "
set -e
pytest -x -v -s tests/kernels/attention/test_cpu_attn.py
pytest -x -v -s tests/kernels/test_onednn.py"
# Run basic model test
@ -58,15 +59,11 @@ function cpu_tests() {
# pytest -x -v -s tests/kernels/attention/test_cache.py -m cpu_model
# pytest -x -v -s tests/kernels/attention/test_mla_decode_cpu.py -m cpu_model
# Note: disable Bart until supports V1
pytest -x -v -s tests/models/language/generation -m cpu_model \
--ignore=tests/models/language/generation/test_bart.py
VLLM_CPU_SGL_KERNEL=1 pytest -x -v -s tests/models/language/generation -m cpu_model \
--ignore=tests/models/language/generation/test_bart.py
pytest -x -v -s tests/models/language/generation -m cpu_model
VLLM_CPU_SGL_KERNEL=1 pytest -x -v -s tests/models/language/generation -m cpu_model
pytest -x -v -s tests/models/language/pooling -m cpu_model
pytest -x -v -s tests/models/multimodal/generation \
--ignore=tests/models/multimodal/generation/test_mllama.py \
--ignore=tests/models/multimodal/generation/test_pixtral.py \
-m cpu_model"
@ -74,14 +71,13 @@ function cpu_tests() {
docker exec cpu-test-"$NUMA_NODE" bash -c "
set -e
pytest -x -s -v \
tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_logprobs[False-10-32-neuralmagic/Llama-3.2-1B-quantized.w8a8]"
tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_logprobs"
# Note: disable it until supports V1
# Run AWQ test
# docker exec cpu-test-"$NUMA_NODE" bash -c "
# set -e
# VLLM_USE_V1=0 pytest -x -s -v \
# tests/quantization/test_ipex_quant.py"
# Run AWQ/GPTQ test
docker exec cpu-test-"$NUMA_NODE" bash -c "
set -e
pytest -x -s -v \
tests/quantization/test_cpu_wna16.py"
# Run multi-lora tests
docker exec cpu-test-"$NUMA_NODE" bash -c "
@ -120,4 +116,4 @@ function cpu_tests() {
# All of CPU tests are expected to be finished less than 40 mins.
export -f cpu_tests
timeout 2h bash -c "cpu_tests $CORE_RANGE $NUMA_NODE"
timeout 2.5h bash -c "cpu_tests $CORE_RANGE $NUMA_NODE"

192
.buildkite/scripts/hardware_ci/run-npu-test.sh Normal file
View File

@ -0,0 +1,192 @@
#!/bin/bash
# This script build the Ascend NPU docker image and run the offline inference inside the container.
# It serves a sanity check for compilation and basic model usage.
set -ex
# Base ubuntu image with basic ascend development libraries and python installed
VLLM_ASCEND_REPO="https://github.com/vllm-project/vllm-ascend.git"
CONFIG_FILE_REMOTE_PATH="tests/e2e/vllm_interface/vllm_test.cfg"
TEST_RUN_CONFIG_FILE="vllm_test.cfg"
VLLM_ASCEND_TMP_DIR=
# Get the test run configuration file from the vllm-ascend repository
fetch_vllm_test_cfg() {
VLLM_ASCEND_TMP_DIR=$(mktemp -d)
# Ensure that the temporary directory is cleaned up when an exception occurs during configuration file retrieval
cleanup() {
rm -rf "${VLLM_ASCEND_TMP_DIR}"
}
trap cleanup EXIT
GIT_TRACE=1 git clone -v --depth 1 "${VLLM_ASCEND_REPO}" "${VLLM_ASCEND_TMP_DIR}"
if [ ! -f "${VLLM_ASCEND_TMP_DIR}/${CONFIG_FILE_REMOTE_PATH}" ]; then
echo "Error: file '${CONFIG_FILE_REMOTE_PATH}' does not exist in the warehouse" >&2
exit 1
fi
# If the file already exists locally, just overwrite it
cp "${VLLM_ASCEND_TMP_DIR}/${CONFIG_FILE_REMOTE_PATH}" "${TEST_RUN_CONFIG_FILE}"
echo "Copied ${CONFIG_FILE_REMOTE_PATH} to ${TEST_RUN_CONFIG_FILE}"
# Since the trap will be overwritten later, and when it is executed here, the task of cleaning up resources
# when the trap is abnormal has been completed, so the temporary resources are manually deleted here.
rm -rf "${VLLM_ASCEND_TMP_DIR}"
trap - EXIT
}
# Downloads test run configuration file from a remote URL.
# Loads the configuration into the current script environment.
get_config() {
if [ ! -f "${TEST_RUN_CONFIG_FILE}" ]; then
echo "Error: file '${TEST_RUN_CONFIG_FILE}' does not exist in the warehouse" >&2
exit 1
fi
source "${TEST_RUN_CONFIG_FILE}"
echo "Base docker image name that get from configuration: ${BASE_IMAGE_NAME}"
return 0
}
# get test running configuration.
fetch_vllm_test_cfg
get_config
# Check if the function call was successful. If not, exit the script.
if [ $? -ne 0 ]; then
exit 1
fi
image_name="npu/vllm-ci:${BUILDKITE_COMMIT}_${EPOCHSECONDS}"
container_name="npu_${BUILDKITE_COMMIT}_$(tr -dc A-Za-z0-9 < /dev/urandom | head -c 10; echo)"
# BUILDKITE_AGENT_NAME format is {hostname}-{agent_idx}-{npu_card_num}cards
agent_idx=$(echo "${BUILDKITE_AGENT_NAME}" | awk -F'-' '{print $(NF-1)}')
echo "agent_idx: ${agent_idx}"
builder_name="cachebuilder${agent_idx}"
builder_cache_dir="/mnt/docker-cache${agent_idx}"
mkdir -p ${builder_cache_dir}
# Try building the docker image
cat <<EOF | DOCKER_BUILDKIT=1 docker build \
--add-host cache-service-vllm.nginx-pypi-cache.svc.cluster.local:${PYPI_CACHE_HOST} \
--builder ${builder_name} --cache-from type=local,src=${builder_cache_dir} \
--cache-to type=local,dest=${builder_cache_dir},mode=max \
--progress=plain --load -t ${image_name} -f - .
FROM ${BASE_IMAGE_NAME}
# Define environments
ENV DEBIAN_FRONTEND=noninteractive
ENV SOC_VERSION="ascend910b1"
RUN pip config set global.index-url http://cache-service-vllm.nginx-pypi-cache.svc.cluster.local:${PYPI_CACHE_PORT}/pypi/simple && \
pip config set global.trusted-host cache-service-vllm.nginx-pypi-cache.svc.cluster.local && \
apt-get update -y && \
apt-get install -y python3-pip git vim wget net-tools gcc g++ cmake libnuma-dev && \
rm -rf /var/cache/apt/* && \
rm -rf /var/lib/apt/lists/*
# Install for pytest to make the docker build cache layer always valid
RUN --mount=type=cache,target=/root/.cache/pip \
pip install pytest>=6.0 modelscope
WORKDIR /workspace/vllm
# Install vLLM dependencies in advance. Effect: As long as common.txt remains unchanged, the docker cache layer will be valid.
COPY requirements/common.txt /workspace/vllm/requirements/common.txt
RUN --mount=type=cache,target=/root/.cache/pip \
pip install -r requirements/common.txt
COPY . .
# Install vLLM
RUN --mount=type=cache,target=/root/.cache/pip \
VLLM_TARGET_DEVICE="empty" python3 -m pip install -v -e /workspace/vllm/ --extra-index https://download.pytorch.org/whl/cpu/ && \
python3 -m pip uninstall -y triton
# Install vllm-ascend
WORKDIR /workspace
ARG VLLM_ASCEND_REPO=https://github.com/vllm-project/vllm-ascend.git
ARG VLLM_ASCEND_TAG=main
RUN git config --global url."https://gh-proxy.test.osinfra.cn/https://github.com/".insteadOf "https://github.com/" && \
git clone --depth 1 \$VLLM_ASCEND_REPO --branch \$VLLM_ASCEND_TAG /workspace/vllm-ascend
# Install vllm dependencies in advance. Effect: As long as common.txt remains unchanged, the docker cache layer will be valid.
RUN --mount=type=cache,target=/root/.cache/pip \
pip install -r /workspace/vllm-ascend/requirements.txt
RUN --mount=type=cache,target=/root/.cache/pip \
export PIP_EXTRA_INDEX_URL=https://mirrors.huaweicloud.com/ascend/repos/pypi && \
source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
source /usr/local/Ascend/nnal/atb/set_env.sh && \
export LD_LIBRARY_PATH=\$LD_LIBRARY_PATH:/usr/local/Ascend/ascend-toolkit/latest/`uname -i`-linux/devlib && \
python3 -m pip install -v -e /workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/
ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
ENV VLLM_USE_MODELSCOPE=True
WORKDIR /workspace/vllm-ascend
CMD ["/bin/bash"]
EOF
# Setup cleanup
remove_docker_container() {
docker rm -f "${container_name}" || true;
docker image rm -f "${image_name}" || true;
docker system prune -f || true;
}
trap remove_docker_container EXIT
# Generate corresponding --device args based on BUILDKITE_AGENT_NAME
# Ascend NPU BUILDKITE_AGENT_NAME format is {hostname}-{agent_idx}-{npu_card_num}cards, and agent_idx starts from 1.
# e.g. atlas-a2-001-1-2cards means this is the 1-th agent on atlas-a2-001 host, and it has 2 NPU cards.
# returns --device /dev/davinci0 --device /dev/davinci1
parse_and_gen_devices() {
local input="$1"
local index cards_num
if [[ "$input" =~ ([0-9]+)-([0-9]+)cards$ ]]; then
index="${BASH_REMATCH[1]}"
cards_num="${BASH_REMATCH[2]}"
else
echo "parse error" >&2
return 1
fi
local devices=""
local i=0
while (( i < cards_num )); do
local dev_idx=$(((index - 1)*cards_num + i ))
devices="$devices --device /dev/davinci${dev_idx}"
((i++))
done
# trim leading space
devices="${devices#"${devices%%[![:space:]]*}"}"
# Output devices: assigned to the caller variable
printf '%s' "$devices"
}
devices=$(parse_and_gen_devices "${BUILDKITE_AGENT_NAME}") || exit 1
# Run the image and execute the Out-Of-Tree (OOT) platform interface test case on Ascend NPU hardware.
# This test checks whether the OOT platform interface is functioning properly in conjunction with
# the hardware plugin vllm-ascend.
model_cache_dir=/mnt/modelscope${agent_idx}
mkdir -p ${model_cache_dir}
docker run \
${devices} \
--device /dev/davinci_manager \
--device /dev/devmm_svm \
--device /dev/hisi_hdc \
-v /usr/local/dcmi:/usr/local/dcmi \
-v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
-v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ \
-v /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info \
-v /etc/ascend_install.info:/etc/ascend_install.info \
-v ${model_cache_dir}:/root/.cache/modelscope \
--entrypoint="" \
--name "${container_name}" \
"${image_name}" \
bash -c '
set -e
pytest -v -s tests/e2e/vllm_interface/
'

5
.buildkite/scripts/hardware_ci/run-tpu-v1-test-part2.sh
View File

@ -62,12 +62,11 @@ echo "--- Installing Python dependencies ---"
python3 -m pip install --progress-bar off git+https://github.com/thuml/depyf.git \
&& python3 -m pip install --progress-bar off pytest pytest-asyncio tpu-info \
&& python3 -m pip install --progress-bar off "lm-eval @ git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d" \
&& python3 -m pip install --progress-bar off hf-transfer
&& python3 -m pip install --progress-bar off hf-transfer tblib==3.1.0
echo "--- Python dependencies installed ---"
export VLLM_USE_V1=1
export VLLM_XLA_CHECK_RECOMPILATION=1
export VLLM_XLA_CACHE_PATH=
echo "Using VLLM V1"
echo "--- Hardware Information ---"
# tpu-info

5
.buildkite/scripts/hardware_ci/run-tpu-v1-test.sh
View File

@ -62,12 +62,11 @@ echo "--- Installing Python dependencies ---"
python3 -m pip install --progress-bar off git+https://github.com/thuml/depyf.git \
&& python3 -m pip install --progress-bar off pytest pytest-asyncio tpu-info \
&& python3 -m pip install --progress-bar off "lm-eval @ git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d" \
&& python3 -m pip install --progress-bar off hf-transfer
&& python3 -m pip install --progress-bar off hf-transfer tblib==3.1.0
echo "--- Python dependencies installed ---"
export VLLM_USE_V1=1
export VLLM_XLA_CHECK_RECOMPILATION=1
export VLLM_XLA_CACHE_PATH=
echo "Using VLLM V1"
echo "--- Hardware Information ---"
# tpu-info

16
.buildkite/scripts/hardware_ci/run-xpu-test.sh
View File

@ -20,7 +20,10 @@ trap remove_docker_container EXIT
# Run the image and test offline inference/tensor parallel
docker run \
--device /dev/dri \
--device /dev/dri:/dev/dri \
--net=host \
--ipc=host \
--privileged \
-v /dev/dri/by-path:/dev/dri/by-path \
--entrypoint="" \
-e "HF_TOKEN=${HF_TOKEN}" \
@ -32,19 +35,18 @@ docker run \
echo $ZE_AFFINITY_MASK
pip install tblib==3.1.0
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 -O3 -O.cudagraph_mode=NONE
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 -O3 -cc.cudagraph_mode=NONE
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend ray
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend mp
VLLM_ATTENTION_BACKEND=TRITON_ATTN_VLLM_V1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
python3 examples/offline_inference/basic/generate.py --model Intel/Qwen2.5-0.5B-W4A16-G128-AutoRound-LLMC-TEST-ONLY --enforce-eager
VLLM_ATTENTION_BACKEND=TRITON_ATTN python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
cd tests
pytest -v -s v1/core
pytest -v -s v1/engine
pytest -v -s v1/sample --ignore=v1/sample/test_logprobs.py --ignore=v1/sample/test_logprobs_e2e.py
pytest -v -s v1/worker --ignore=v1/worker/test_gpu_model_runner.py
pytest -v -s v1/structured_output
pytest -v -s v1/spec_decode --ignore=v1/spec_decode/test_max_len.py --ignore=v1/spec_decode/test_eagle.py --ignore=v1/spec_decode/test_tree_attention.py
pytest -v -s v1/kv_connector/unit --ignore=v1/kv_connector/unit/test_multi_connector.py --ignore=v1/kv_connector/unit/test_nixl_connector.py --ignore=v1/kv_connector/unit/test_shared_storage_connector.py
pytest -v -s v1/spec_decode --ignore=v1/spec_decode/test_max_len.py --ignore=v1/spec_decode/test_tree_attention.py --ignore=v1/spec_decode/test_speculators_eagle3.py
pytest -v -s v1/kv_connector/unit --ignore=v1/kv_connector/unit/test_multi_connector.py --ignore=v1/kv_connector/unit/test_nixl_connector.py --ignore=v1/kv_connector/unit/test_example_connector.py --ignore=v1/kv_connector/unit/test_lmcache_integration.py
pytest -v -s v1/test_serial_utils.py
pytest -v -s v1/test_utils.py
pytest -v -s v1/test_metrics_reader.py
'

2
.buildkite/scripts/run-benchmarks.sh
View File

@ -18,7 +18,7 @@ vllm bench throughput --input-len 256 --output-len 256 --output-json throughput_
bench_throughput_exit_code=$?
# run server-based benchmarks and upload the result to buildkite
python3 -m vllm.entrypoints.openai.api_server --model meta-llama/Llama-2-7b-chat-hf &
vllm serve meta-llama/Llama-2-7b-chat-hf &
server_pid=$!
wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json

64
.buildkite/scripts/run-prime-rl-test.sh Executable file
View File

@ -0,0 +1,64 @@
#!/bin/bash
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
# Setup script for Prime-RL integration tests
# This script prepares the environment for running Prime-RL tests with nightly vLLM
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
REPO_ROOT="$(cd "${SCRIPT_DIR}/../.." && pwd)"
PRIME_RL_REPO="https://github.com/PrimeIntellect-ai/prime-rl.git"
PRIME_RL_DIR="${REPO_ROOT}/prime-rl"
if command -v rocm-smi &> /dev/null || command -v rocminfo &> /dev/null; then
echo "AMD GPU detected. Prime-RL currently only supports NVIDIA. Skipping..."
exit 0
fi
echo "Setting up Prime-RL integration test environment..."
# Clean up any existing Prime-RL directory
if [ -d "${PRIME_RL_DIR}" ]; then
echo "Removing existing Prime-RL directory..."
rm -rf "${PRIME_RL_DIR}"
fi
# Install UV if not available
if ! command -v uv &> /dev/null; then
echo "Installing UV package manager..."
curl -LsSf https://astral.sh/uv/install.sh | sh
source $HOME/.local/bin/env
fi
# Clone Prime-RL repository at specific branch for reproducible tests
PRIME_RL_BRANCH="integ-vllm-main"
echo "Cloning Prime-RL repository at branch: ${PRIME_RL_BRANCH}..."
git clone --branch "${PRIME_RL_BRANCH}" --single-branch "${PRIME_RL_REPO}" "${PRIME_RL_DIR}"
cd "${PRIME_RL_DIR}"
echo "Setting up UV project environment..."
export UV_PROJECT_ENVIRONMENT=/usr/local
ln -s /usr/bin/python3 /usr/local/bin/python
# Remove vllm pin from pyproject.toml
echo "Removing vllm pin from pyproject.toml..."
sed -i '/vllm==/d' pyproject.toml
# Sync Prime-RL dependencies
echo "Installing Prime-RL dependencies..."
uv sync --inexact && uv sync --inexact --all-extras
# Verify installation
echo "Verifying installations..."
uv run python -c "import vllm; print(f'vLLM version: {vllm.__version__}')"
uv run python -c "import prime_rl; print('Prime-RL imported successfully')"
echo "Prime-RL integration test environment setup complete!"
echo "Running Prime-RL integration tests..."
export WANDB_MODE=offline # this makes this test not require a WANDB_API_KEY
uv run pytest -vs tests/integration/test_rl.py -m gpu
echo "Prime-RL integration tests completed!"

72
.buildkite/scripts/scheduled_integration_test/deepseek_v2_lite_ep_eplb.sh Normal file
View File

@ -0,0 +1,72 @@
#!/usr/bin/env bash
set -euxo pipefail
# args: [THRESHOLD] [NUM_QUESTIONS] [START_PORT]
THRESHOLD=${1:-0.25}
NUM_Q=${2:-1319}
PORT=${3:-8010}
OUT_DIR=${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="deepseek-ai/DeepSeek-V2-lite"
# Set BACKENDS based on platform
if command -v rocm-smi &> /dev/null || [[ -d /opt/rocm ]] || [[ -n "${ROCM_PATH:-}" ]]; then
# ROCm platform
BACKENDS=("allgather_reducescatter")
# Disable MOE padding for ROCm since it is causing eplb to fail
export VLLM_ROCM_MOE_PADDING=0
else
# Non-ROCm platform (CUDA/other)
BACKENDS=("deepep_high_throughput" "deepep_low_latency")
fi
cleanup() {
if [[ -n "${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
for BACK in "${BACKENDS[@]}"; do
VLLM_DEEP_GEMM_WARMUP=skip \
VLLM_ALL2ALL_BACKEND=$BACK \
vllm serve "$MODEL" \
--enforce-eager \
--tensor-parallel-size 2 \
--data-parallel-size 2 \
--enable-expert-parallel \
--enable-eplb \
--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 ${OUT}
python3 - <<PY
import json; acc=json.load(open('${OUT}'))['accuracy']
print(f"${MODEL} ${BACK}: accuracy {acc:.3f}")
assert acc >= ${THRESHOLD}, f"${MODEL} ${BACK} accuracy {acc}"
PY
cleanup
SERVER_PID=
sleep 1
PORT=$((PORT+1))
done

74
.buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_block_ep_eplb.sh Normal file
View File

@ -0,0 +1,74 @@
#!/usr/bin/env bash
set -euxo pipefail
# args: [THRESHOLD] [NUM_QUESTIONS] [START_PORT] [DATA_PARALLEL_SIZE] [TENSOR_PARALLEL_SIZE]
THRESHOLD=${1:-0.8}
NUM_Q=${2:-1319}
PORT=${3:-8020}
DATA_PARALLEL_SIZE=${4:-2}
TENSOR_PARALLEL_SIZE=${5:-2}
OUT_DIR=${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"
# Set BACKENDS based on platform
if command -v rocm-smi &> /dev/null || [[ -d /opt/rocm ]] || [[ -n "${ROCM_PATH:-}" ]]; then
# ROCm platform
BACKENDS=("allgather_reducescatter")
# Disable MOE padding for ROCm since it is causing eplb to fail
export VLLM_ROCM_MOE_PADDING=0
else
# Non-ROCm platform (CUDA/other)
BACKENDS=("deepep_high_throughput" "deepep_low_latency")
fi
cleanup() {
if [[ -n "${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
for BACK in "${BACKENDS[@]}"; do
VLLM_DEEP_GEMM_WARMUP=skip \
VLLM_ALL2ALL_BACKEND=$BACK \
vllm serve "$MODEL" \
--enforce-eager \
--enable-eplb \
--eplb-config '{"window_size":10, "step_interval":100, "num_redundant_experts":0, "log_balancedness":true}' \
--tensor-parallel-size ${TENSOR_PARALLEL_SIZE} \
--data-parallel-size ${DATA_PARALLEL_SIZE} \
--enable-expert-parallel \
--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 ${OUT}
python3 - <<PY
import json; acc=json.load(open('${OUT}'))['accuracy']
print(f"${MODEL} ${BACK}: accuracy {acc:.3f}")
assert acc >= ${THRESHOLD}, f"${MODEL} ${BACK} accuracy {acc}"
PY
cleanup
SERVER_PID=
sleep 1
PORT=$((PORT+1))
done

74
.buildkite/scripts/scheduled_integration_test/qwen3_next_mtp_async_eplb.sh Normal file
View File

@ -0,0 +1,74 @@
#!/usr/bin/env bash
set -euxo pipefail
# args: [THRESHOLD] [NUM_QUESTIONS] [START_PORT]
THRESHOLD=${1:-0.25}
NUM_Q=${2:-1319}
PORT=${3:-8040}
OUT_DIR=${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-Next-80B-A3B-Instruct"
# Set BACKENDS based on platform
if command -v rocm-smi &> /dev/null || [[ -d /opt/rocm ]] || [[ -n "${ROCM_PATH:-}" ]]; then
# ROCm platform
BACKENDS=("allgather_reducescatter")
# Disable MOE padding for ROCm since it is causing eplb to fail
export VLLM_ROCM_MOE_PADDING=0
else
# Non-ROCm platform (CUDA/other)
BACKENDS=("deepep_high_throughput" "deepep_low_latency")
fi
cleanup() {
if [[ -n "${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
for BACK in "${BACKENDS[@]}"; do
VLLM_DEEP_GEMM_WARMUP=skip \
VLLM_ALL2ALL_BACKEND=$BACK \
vllm serve "$MODEL" \
--enforce-eager \
--tensor-parallel-size 4 \
--enable-expert-parallel \
--enable-eplb \
--eplb-config '{"window_size":200,"step_interval":600,"use_async":true}' \
--speculative-config '{"method":"qwen3_next_mtp","num_speculative_tokens":1}' \
--trust-remote-code \
--max-model-len 2048 \
--gpu-memory-utilization 0.9 \
--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 ${OUT}
python3 - <<PY
import json; acc=json.load(open('${OUT}'))['accuracy']
print(f"${MODEL} ${BACK}: accuracy {acc:.3f}")
assert acc >= ${THRESHOLD}, f"${MODEL} ${BACK} accuracy {acc}"
PY
cleanup
SERVER_PID=
sleep 1
PORT=$((PORT+1))
done

2
.buildkite/scripts/tpu/quantized_v6e_1.env
View File

@ -9,6 +9,6 @@ MAX_NUM_BATCHED_TOKENS=1024
TENSOR_PARALLEL_SIZE=1
MAX_MODEL_LEN=2048
DOWNLOAD_DIR=/mnt/disks/persist
EXPECTED_THROUGHPUT=10.0
EXPECTED_THROUGHPUT=8.7
INPUT_LEN=1800
OUTPUT_LEN=128

2
.buildkite/scripts/tpu/run_bm.sh
View File

@ -42,7 +42,7 @@ echo "lanching vllm..."
echo "logging to $VLLM_LOG"
echo
VLLM_USE_V1=1 vllm serve $MODEL \
vllm serve $MODEL \
--seed 42 \
--max-num-seqs $MAX_NUM_SEQS \
--max-num-batched-tokens $MAX_NUM_BATCHED_TOKENS \

134
.buildkite/scripts/upload-wheels.sh
View File

@ -2,6 +2,28 @@
set -ex
# ======== part 0: setup ========
BUCKET="vllm-wheels"
INDICES_OUTPUT_DIR="indices"
DEFAULT_VARIANT_ALIAS="cu129" # align with vLLM_MAIN_CUDA_VERSION in vllm/envs.py
PYTHON=${PYTHON_PROG:=python3} # try to read from env var, otherwise use python3
SUBPATH=$BUILDKITE_COMMIT
S3_COMMIT_PREFIX="s3://$BUCKET/$SUBPATH/"
# detect if python3.10+ is available
has_new_python=$($PYTHON -c "print(1 if __import__('sys').version_info >= (3,12) else 0)")
if [[ "$has_new_python" -eq 0 ]]; then
# use new python from docker
docker pull python:3-slim
PYTHON="docker run --rm -v $(pwd):/app -w /app python:3-slim python3"
fi
echo "Using python interpreter: $PYTHON"
echo "Python version: $($PYTHON --version)"
# ========= part 1: collect, rename & upload the wheel ==========
# Assume wheels are in artifacts/dist/*.whl
wheel_files=(artifacts/dist/*.whl)
@ -10,82 +32,76 @@ if [[ ${#wheel_files[@]} -ne 1 ]]; then
echo "Error: Expected exactly one wheel file in artifacts/dist/, but found ${#wheel_files[@]}"
exit 1
fi
# Get the single wheel file
wheel="${wheel_files[0]}"
# Detect architecture and rename 'linux' to appropriate manylinux version
arch=$(uname -m)
if [[ $arch == "x86_64" ]]; then
manylinux_version="manylinux1"
elif [[ $arch == "aarch64" ]]; then
manylinux_version="manylinux2014"
else
echo "Warning: Unknown architecture $arch, using manylinux1 as default"
manylinux_version="manylinux1"
fi
# default build image uses ubuntu 20.04, which corresponds to manylinux_2_31
# we also accept params as manylinux tag
# refer to https://github.com/mayeut/pep600_compliance?tab=readme-ov-file#acceptable-distros-to-build-wheels
manylinux_version="${1:-manylinux_2_31}"
# Rename 'linux' to the appropriate manylinux version in the wheel filename
if [[ "$wheel" != *"linux"* ]]; then
echo "Error: Wheel filename does not contain 'linux': $wheel"
exit 1
fi
new_wheel="${wheel/linux/$manylinux_version}"
mv -- "$wheel" "$new_wheel"
wheel="$new_wheel"
echo "Renamed wheel to: $wheel"
# Extract the version from the wheel
version=$(unzip -p "$wheel" '**/METADATA' | grep '^Version: ' | cut -d' ' -f2)
echo "Version: $version"
echo "Version in wheel: $version"
pure_version="${version%%+*}"
echo "Pure version (without variant): $pure_version"
normal_wheel="$wheel" # Save the original wheel filename
# copy wheel to its own bucket
aws s3 cp "$wheel" "$S3_COMMIT_PREFIX"
# If the version contains "dev", rename it to v1.0.0.dev for consistency
if [[ $version == *dev* ]]; then
suffix="${version##*.}"
if [[ $suffix == cu* ]]; then
new_version="1.0.0.dev+${suffix}"
else
new_version="1.0.0.dev"
fi
new_wheel="${wheel/$version/$new_version}"
# use cp to keep both files in the artifacts directory
cp -- "$wheel" "$new_wheel"
wheel="$new_wheel"
version="$new_version"
fi
# ========= part 2: generate and upload indices ==========
# generate indices for all existing wheels in the commit directory
# this script might be run multiple times if there are multiple variants being built
# so we need to guarantee there is little chance for "TOCTOU" issues
# i.e., one process is generating indices while another is uploading a new wheel
# so we need to ensure no time-consuming operations happen below
# Upload the wheel to S3
python3 .buildkite/generate_index.py --wheel "$normal_wheel"
# list all wheels in the commit directory
echo "Existing wheels on S3:"
aws s3 ls "$S3_COMMIT_PREFIX"
obj_json="objects.json"
aws s3api list-objects-v2 --bucket "$BUCKET" --prefix "$SUBPATH/" --delimiter / --output json > "$obj_json"
mkdir -p "$INDICES_OUTPUT_DIR"
# generate index for this commit
aws s3 cp "$wheel" "s3://vllm-wheels/$BUILDKITE_COMMIT/"
aws s3 cp "$normal_wheel" "s3://vllm-wheels/$BUILDKITE_COMMIT/"
if [[ $normal_wheel == *"cu126"* ]]; then
# if $normal_wheel matches cu126, do not upload the index.html
echo "Skipping index files for cu126 wheels"
elif [[ $normal_wheel == *"cu128"* ]]; then
# if $normal_wheel matches cu128, do not upload the index.html
echo "Skipping index files for cu128 wheels"
# call script to generate indicies for all existing wheels
# this indices have relative paths that could work as long as it is next to the wheel directory in s3
# i.e., the wheels are always in s3://vllm-wheels/<commit>/
# and indices can be placed in /<commit>/, or /nightly/, or /<version>/
if [[ ! -z "$DEFAULT_VARIANT_ALIAS" ]]; then
alias_arg="--alias-to-default $DEFAULT_VARIANT_ALIAS"
else
# only upload index.html for cu129 wheels (default wheels) as it
# is available on both x86 and arm64
aws s3 cp index.html "s3://vllm-wheels/$BUILDKITE_COMMIT/vllm/index.html"
aws s3 cp "s3://vllm-wheels/nightly/index.html" "s3://vllm-wheels/$BUILDKITE_COMMIT/index.html"
alias_arg=""
fi
# generate index for nightly
aws s3 cp "$wheel" "s3://vllm-wheels/nightly/"
aws s3 cp "$normal_wheel" "s3://vllm-wheels/nightly/"
# HACK: we do not need regex module here, but it is required by pre-commit hook
# To avoid any external dependency, we simply replace it back to the stdlib re module
sed -i 's/import regex as re/import re/g' .buildkite/scripts/generate-nightly-index.py
$PYTHON .buildkite/scripts/generate-nightly-index.py --version "$SUBPATH" --current-objects "$obj_json" --output-dir "$INDICES_OUTPUT_DIR" --comment "commit $BUILDKITE_COMMIT" $alias_arg
if [[ $normal_wheel == *"cu126"* ]]; then
# if $normal_wheel matches cu126, do not upload the index.html
echo "Skipping index files for cu126 wheels"
elif [[ $normal_wheel == *"cu128"* ]]; then
# if $normal_wheel matches cu128, do not upload the index.html
echo "Skipping index files for cu128 wheels"
else
# only upload index.html for cu129 wheels (default wheels) as it
# is available on both x86 and arm64
aws s3 cp index.html "s3://vllm-wheels/nightly/vllm/index.html"
# copy indices to /<commit>/ unconditionally
echo "Uploading indices to $S3_COMMIT_PREFIX"
aws s3 cp --recursive "$INDICES_OUTPUT_DIR/" "$S3_COMMIT_PREFIX"
# copy to /nightly/ only if it is on the main branch and not a PR
if [[ "$BUILDKITE_BRANCH" == "main" && "$BUILDKITE_PULL_REQUEST" == "false" ]]; then
echo "Uploading indices to overwrite /nightly/"
aws s3 cp --recursive "$INDICES_OUTPUT_DIR/" "s3://$BUCKET/nightly/"
fi
aws s3 cp "$wheel" "s3://vllm-wheels/$version/"
aws s3 cp index.html "s3://vllm-wheels/$version/vllm/index.html"
# re-generate and copy to /<pure_version>/ only if it does not have "dev" in the version
if [[ "$version" != *"dev"* ]]; then
echo "Re-generating indices for /$pure_version/"
rm -rf "$INDICES_OUTPUT_DIR/*"
mkdir -p "$INDICES_OUTPUT_DIR"
$PYTHON .buildkite/scripts/generate-nightly-index.py --version "$pure_version" --current-objects "$obj_json" --output-dir "$INDICES_OUTPUT_DIR" --comment "version $pure_version" $alias_arg
aws s3 cp --recursive "$INDICES_OUTPUT_DIR/" "s3://$BUCKET/$pure_version/"
fi

1626
.buildkite/test-amd.yaml Normal file
View File

File diff suppressed because it is too large Load Diff

682
.buildkite/test-pipeline.yaml
View File

File diff suppressed because it is too large Load Diff

21
.buildkite/test_areas/attention.yaml Normal file
View File

@ -0,0 +1,21 @@
group: Attention
depends_on:
- image-build
steps:
- label: V1 attention (H100)
timeout_in_minutes: 30
gpu: h100
source_file_dependencies:
- vllm/v1/attention
- tests/v1/attention
commands:
- pytest -v -s v1/attention
- label: V1 attention (B200)
timeout_in_minutes: 30
gpu: b200
source_file_dependencies:
- vllm/v1/attention
- tests/v1/attention
commands:
- VLLM_DISABLE_FLASHINFER_PREFILL=1 pytest -v -s v1/attention # TODO: FI prefill is bugged and causes incorrectness, fix this

16
.buildkite/test_areas/basic_correctness.yaml Normal file
View File

@ -0,0 +1,16 @@
group: Basic Correctness
depends_on:
- image-build
steps:
- label: Basic Correctness
timeout_in_minutes: 30
source_file_dependencies:
- vllm/
- tests/basic_correctness/test_basic_correctness
- tests/basic_correctness/test_cpu_offload
- tests/basic_correctness/test_cumem.py
commands:
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
- pytest -v -s basic_correctness/test_cumem.py
- pytest -v -s basic_correctness/test_basic_correctness.py
- pytest -v -s basic_correctness/test_cpu_offload.py

19
.buildkite/test_areas/benchmarks.yaml Normal file
View File

@ -0,0 +1,19 @@
group: Benchmarks
depends_on:
- image-build
steps:
- label: Benchmarks
timeout_in_minutes: 20
working_dir: "/vllm-workspace/.buildkite"
source_file_dependencies:
- benchmarks/
commands:
- bash scripts/run-benchmarks.sh
- label: Benchmarks CLI Test
timeout_in_minutes: 20
source_file_dependencies:
- vllm/
- tests/benchmarks/
commands:
- pytest -v -s benchmarks/

57
.buildkite/test_areas/compile.yaml Normal file
View File

@ -0,0 +1,57 @@
group: Compile
depends_on:
- image-build
steps:
- label: Fusion and Compile Tests (B200)
timeout_in_minutes: 40
working_dir: "/vllm-workspace/"
gpu: b200
source_file_dependencies:
- csrc/quantization/fp4/
- vllm/model_executor/layers/quantization/utils/flashinfer_utils.py
- vllm/v1/attention/backends/flashinfer.py
- vllm/v1/worker/
- vllm/v1/cudagraph_dispatcher.py
- vllm/compilation/
# can affect pattern matching
- vllm/model_executor/layers/layernorm.py
- vllm/model_executor/layers/activation.py
- vllm/model_executor/layers/quantization/input_quant_fp8.py
- tests/compile/test_fusion_attn.py
- tests/compile/test_silu_mul_quant_fusion.py
- tests/compile/distributed/test_fusion_all_reduce.py
- tests/compile/distributed/test_fusions_e2e.py
- tests/compile/fullgraph/test_full_graph.py
commands:
- nvidia-smi
- pytest -v -s tests/compile/test_fusion_attn.py
- pytest -v -s tests/compile/test_silu_mul_quant_fusion.py
# this runner has 2 GPUs available even though num_gpus=2 is not set
- pytest -v -s tests/compile/distributed/test_fusion_all_reduce.py
# Limit to Inductor partition, no custom ops, and allreduce & attn fusion to reduce running time
# Wrap with quotes to escape yaml
- "pytest -v -s tests/compile/distributed/test_fusions_e2e.py::test_tp2_attn_quant_allreduce_rmsnorm -k 'True and not +quant_fp8 and not +rms_norm'"
# test_fp8_kv_scale_compile requires FlashAttention (not supported on default L4/L40)
- pytest -v -s tests/compile/fullgraph/test_full_graph.py::test_fp8_kv_scale_compile
- label: Fusion E2E (2 GPUs)(B200)
timeout_in_minutes: 40
working_dir: "/vllm-workspace/"
gpu: b200
optional: true
num_gpus: 2
source_file_dependencies:
- csrc/quantization/fp4/
- vllm/model_executor/layers/quantization/utils/flashinfer_utils.py
- vllm/v1/attention/backends/flashinfer.py
- vllm/compilation/
# can affect pattern matching
- vllm/model_executor/layers/layernorm.py
- vllm/model_executor/layers/activation.py
- vllm/model_executor/layers/quantization/input_quant_fp8.py
- tests/compile/distributed/test_fusions_e2e.py
commands:
- nvidia-smi
# Run all e2e fusion tests
- pytest -v -s tests/compile/distributed/test_fusions_e2e.py

22
.buildkite/test_areas/cuda.yaml Normal file
View File

@ -0,0 +1,22 @@
group: CUDA
depends_on:
- image-build
steps:
- label: Platform Tests (CUDA)
timeout_in_minutes: 15
source_file_dependencies:
- vllm/
- tests/cuda
commands:
- pytest -v -s cuda/test_cuda_context.py
- label: Cudagraph
timeout_in_minutes: 20
source_file_dependencies:
- tests/v1/cudagraph
- vllm/v1/cudagraph_dispatcher.py
- vllm/config/compilation.py
- vllm/compilation
commands:
- pytest -v -s v1/cudagraph/test_cudagraph_dispatch.py
- pytest -v -s v1/cudagraph/test_cudagraph_mode.py

199
.buildkite/test_areas/distributed.yaml Normal file
View File

@ -0,0 +1,199 @@
group: Distributed
depends_on:
- image-build
steps:
- label: Distributed Comm Ops
timeout_in_minutes: 20
working_dir: "/vllm-workspace/tests"
num_gpus: 2
source_file_dependencies:
- vllm/distributed
- tests/distributed
commands:
- pytest -v -s distributed/test_comm_ops.py
- pytest -v -s distributed/test_shm_broadcast.py
- pytest -v -s distributed/test_shm_buffer.py
- pytest -v -s distributed/test_shm_storage.py
- label: Distributed (2 GPUs)
timeout_in_minutes: 90
working_dir: "/vllm-workspace/tests"
num_gpus: 2
source_file_dependencies:
- vllm/compilation/
- vllm/distributed/
- vllm/engine/
- vllm/executor/
- vllm/worker/worker_base.py
- vllm/v1/engine/
- vllm/v1/worker/
- tests/compile/fullgraph/test_basic_correctness.py
- tests/compile/test_wrapper.py
- tests/distributed/
- tests/entrypoints/llm/test_collective_rpc.py
- tests/v1/distributed
- tests/v1/entrypoints/openai/test_multi_api_servers.py
- tests/v1/shutdown
- tests/v1/worker/test_worker_memory_snapshot.py
commands:
# https://github.com/NVIDIA/nccl/issues/1838
- export NCCL_CUMEM_HOST_ENABLE=0
- TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_async_llm_dp.py
- TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_eagle_dp.py
- TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_external_lb_dp.py
- DP_SIZE=2 pytest -v -s v1/entrypoints/openai/test_multi_api_servers.py
- pytest -v -s entrypoints/llm/test_collective_rpc.py
- pytest -v -s ./compile/fullgraph/test_basic_correctness.py
- pytest -v -s ./compile/test_wrapper.py
- VLLM_TEST_SAME_HOST=1 torchrun --nproc-per-node=4 distributed/test_same_node.py | grep 'Same node test passed'
- VLLM_TEST_SAME_HOST=1 VLLM_TEST_WITH_DEFAULT_DEVICE_SET=1 torchrun --nproc-per-node=4 distributed/test_same_node.py | grep 'Same node test passed'
- pytest -v -s distributed/test_sequence_parallel.py
- CUDA_VISIBLE_DEVICES=0,1 pytest -v -s v1/shutdown
- pytest -v -s v1/worker/test_worker_memory_snapshot.py
- label: Distributed Tests (4 GPUs)
timeout_in_minutes: 50
working_dir: "/vllm-workspace/tests"
num_gpus: 4
source_file_dependencies:
- vllm/distributed/
- tests/distributed/test_utils
- tests/distributed/test_pynccl
- tests/distributed/test_events
- tests/compile/fullgraph/test_basic_correctness.py
- examples/offline_inference/rlhf.py
- examples/offline_inference/rlhf_colocate.py
- tests/examples/offline_inference/data_parallel.py
- tests/v1/distributed
- tests/v1/engine/test_engine_core_client.py
- tests/distributed/test_symm_mem_allreduce.py
commands:
# https://github.com/NVIDIA/nccl/issues/1838
- export NCCL_CUMEM_HOST_ENABLE=0
# test with torchrun tp=2 and external_dp=2
- torchrun --nproc-per-node=4 distributed/test_torchrun_example.py
# test with torchrun tp=2 and pp=2
- PP_SIZE=2 torchrun --nproc-per-node=4 distributed/test_torchrun_example.py
# test with torchrun tp=4 and dp=1
- TP_SIZE=4 torchrun --nproc-per-node=4 distributed/test_torchrun_example_moe.py
# test with torchrun tp=2, pp=2 and dp=1
- PP_SIZE=2 TP_SIZE=2 torchrun --nproc-per-node=4 distributed/test_torchrun_example_moe.py
# test with torchrun tp=1 and dp=4 with ep
- DP_SIZE=4 ENABLE_EP=1 torchrun --nproc-per-node=4 distributed/test_torchrun_example_moe.py
# test with torchrun tp=2 and dp=2 with ep
- TP_SIZE=2 DP_SIZE=2 ENABLE_EP=1 torchrun --nproc-per-node=4 distributed/test_torchrun_example_moe.py
# test with internal dp
- python3 ../examples/offline_inference/data_parallel.py --enforce-eager
- TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/distributed/test_async_llm_dp.py
- TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/distributed/test_eagle_dp.py
- TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/distributed/test_external_lb_dp.py
- TP_SIZE=1 DP_SIZE=4 pytest -v -s v1/distributed/test_internal_lb_dp.py
- TP_SIZE=1 DP_SIZE=4 pytest -v -s v1/distributed/test_hybrid_lb_dp.py
- pytest -v -s v1/engine/test_engine_core_client.py::test_kv_cache_events_dp
- pytest -v -s distributed/test_utils.py
- pytest -v -s compile/fullgraph/test_basic_correctness.py
- pytest -v -s distributed/test_pynccl.py
- pytest -v -s distributed/test_events.py
- pytest -v -s distributed/test_symm_mem_allreduce.py
# TODO: create a dedicated test section for multi-GPU example tests
# when we have multiple distributed example tests
- cd ../examples/offline_inference
- VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 rlhf.py
- VLLM_ALLOW_INSECURE_SERIALIZATION=1 RAY_DEDUP_LOGS=0 python3 rlhf_colocate.py
- label: Distributed Tests (8 GPUs)(H100)
timeout_in_minutes: 10
gpu: h100
num_gpus: 8
working_dir: "/vllm-workspace/tests"
source_file_dependencies:
- examples/offline_inference/torchrun_dp_example.py
- vllm/config/parallel.py
- vllm/distributed/
- vllm/v1/engine/llm_engine.py
- vllm/v1/executor/uniproc_executor.py
- vllm/v1/worker/gpu_worker.py
commands:
# https://github.com/NVIDIA/nccl/issues/1838
- export NCCL_CUMEM_HOST_ENABLE=0
# test with torchrun tp=2 and dp=4 with ep
- torchrun --nproc-per-node=8 ../examples/offline_inference/torchrun_dp_example.py --tp-size=2 --pp-size=1 --dp-size=4 --enable-ep
- label: Distributed Tests (4 GPUs)(A100)
gpu: a100
optional: true
num_gpus: 4
source_file_dependencies:
- vllm/
commands:
# NOTE: don't test llama model here, it seems hf implementation is buggy
# see https://github.com/vllm-project/vllm/pull/5689 for details
- pytest -v -s distributed/test_custom_all_reduce.py
- torchrun --nproc_per_node=2 distributed/test_ca_buffer_sharing.py
- TARGET_TEST_SUITE=A100 pytest basic_correctness/ -v -s -m 'distributed(num_gpus=2)'
- pytest -v -s -x lora/test_mixtral.py
- label: Distributed Tests (2 GPUs)(H200)
gpu: h200
optional: true
working_dir: "/vllm-workspace/"
num_gpus: 2
commands:
- VLLM_TEST_CLEAN_GPU_MEMORY=1 pytest -v -s tests/compile/distributed/test_async_tp.py
- pytest -v -s tests/compile/distributed/test_sequence_parallelism.py
- pytest -v -s tests/compile/distributed/test_fusion_all_reduce.py
- VLLM_TEST_CLEAN_GPU_MEMORY=1 pytest -v -s tests/compile/distributed/test_fusions_e2e.py -k 'not Llama-4'
- VLLM_TEST_CLEAN_GPU_MEMORY=1 pytest -v -s tests/distributed/test_sequence_parallel.py
- pytest -v -s tests/distributed/test_context_parallel.py
- CUDA_VISIBLE_DEVICES=1,2 VLLM_ALL2ALL_BACKEND=deepep_high_throughput VLLM_USE_DEEP_GEMM=1 VLLM_LOGGING_LEVEL=DEBUG python3 examples/offline_inference/data_parallel.py --model Qwen/Qwen1.5-MoE-A2.7B --tp-size=1 --dp-size=2 --max-model-len 2048
- pytest -v -s tests/v1/distributed/test_dbo.py
- label: Distributed Tests (2 GPUs)(B200)
gpu: b200
optional: true
working_dir: "/vllm-workspace/"
num_gpus: 2
commands:
- pytest -v -s tests/distributed/test_context_parallel.py
- pytest -v -s tests/distributed/test_nccl_symm_mem_allreduce.py
- pytest -v -s tests/v1/distributed/test_dbo.py
- label: 2 Node Test (4 GPUs)
timeout_in_minutes: 30
working_dir: "/vllm-workspace/tests"
num_gpus: 2
num_nodes: 2
source_file_dependencies:
- vllm/distributed/
- vllm/engine/
- vllm/executor/
- vllm/model_executor/models/
- tests/distributed/
- tests/examples/offline_inference/data_parallel.py
commands:
- ./.buildkite/scripts/run-multi-node-test.sh /vllm-workspace/tests 2 2 public.ecr.aws/q9t5s3a7/vllm-ci-postmerge-repo:0bec63fa317e1fbd62e19b0fc31c43c81bf89077 "VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed' && NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed' && python3 ../examples/offline_inference/data_parallel.py --dp-size=2 --tp-size=1 --node-size=2 --node-rank=0 --master-addr=192.168.10.10 --master-port=12345 --enforce-eager --trust-remote-code && VLLM_MULTI_NODE=1 pytest -v -s distributed/test_multi_node_assignment.py && VLLM_MULTI_NODE=1 pytest -v -s distributed/test_pipeline_parallel.py" "VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed' && NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed' && python3 ../examples/offline_inference/data_parallel.py --dp-size=2 --tp-size=1 --node-size=2 --node-rank=1 --master-addr=192.168.10.10 --master-port=12345 --enforce-eager --trust-remote-code"
- label: Distributed NixlConnector PD accuracy (4 GPUs)
timeout_in_minutes: 30
working_dir: "/vllm-workspace/tests"
num_gpus: 4
source_file_dependencies:
- vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
- tests/v1/kv_connector/nixl_integration/
commands:
- uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
- bash v1/kv_connector/nixl_integration/tp_config_sweep_accuracy_test.sh
- label: Pipeline + Context Parallelism (4 GPUs))
timeout_in_minutes: 60
working_dir: "/vllm-workspace/tests"
num_gpus: 4
source_file_dependencies:
- vllm/distributed/
- vllm/engine/
- vllm/executor/
- vllm/model_executor/models/
- tests/distributed/
commands:
- pytest -v -s distributed/test_pp_cudagraph.py
- pytest -v -s distributed/test_pipeline_parallel.py

59
.buildkite/test_areas/e2e_integration.yaml Normal file
View File

@ -0,0 +1,59 @@
group: E2E Integration
depends_on:
- image-build
steps:
- label: DeepSeek V2-Lite Accuracy
timeout_in_minutes: 60
gpu: h100
optional: true
num_gpus: 4
working_dir: "/vllm-workspace"
commands:
- bash .buildkite/scripts/scheduled_integration_test/deepseek_v2_lite_ep_eplb.sh 0.25 200 8010
- label: Qwen3-30B-A3B-FP8-block Accuracy
timeout_in_minutes: 60
gpu: h100
optional: true
num_gpus: 4
working_dir: "/vllm-workspace"
commands:
- bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_block_ep_eplb.sh 0.8 200 8020
- label: Qwen3-30B-A3B-FP8-block Accuracy (B200)
timeout_in_minutes: 60
gpu: b200
optional: true
num_gpus: 2
working_dir: "/vllm-workspace"
commands:
- bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_block_ep_eplb.sh 0.8 200 8020 2 1
- label: Prime-RL Integration (2 GPUs)
timeout_in_minutes: 30
optional: true
num_gpus: 2
working_dir: "/vllm-workspace"
source_file_dependencies:
- vllm/
- .buildkite/scripts/run-prime-rl-test.sh
commands:
- bash .buildkite/scripts/run-prime-rl-test.sh
- label: DeepSeek V2-Lite Async EPLB Accuracy
timeout_in_minutes: 60
gpu: h100
optional: true
num_gpus: 4
working_dir: "/vllm-workspace"
commands:
- bash .buildkite/scripts/scheduled_integration_test/deepseek_v2_lite_ep_async_eplb.sh 0.25 1319 8030
- label: Qwen3-Next-80B-A3B-Instruct MTP Async EPLB Accuracy
timeout_in_minutes: 60
gpu: h100
optional: true
num_gpus: 4
working_dir: "/vllm-workspace"
commands:
- bash .buildkite/scripts/scheduled_integration_test/qwen3_next_mtp_async_eplb.sh 0.8 1319 8040

26
.buildkite/test_areas/engine.yaml Normal file
View File

@ -0,0 +1,26 @@
group: Engine
depends_on:
- image-build
steps:
- label: Engine
timeout_in_minutes: 15
source_file_dependencies:
- vllm/
- tests/engine
- tests/test_sequence
- tests/test_config
- tests/test_logger
- tests/test_vllm_port
commands:
- pytest -v -s engine test_sequence.py test_config.py test_logger.py test_vllm_port.py
- label: V1 e2e + engine
timeout_in_minutes: 45
source_file_dependencies:
- vllm/
- tests/v1
commands:
# TODO: accuracy does not match, whether setting
# VLLM_USE_FLASHINFER_SAMPLER or not on H100.
- pytest -v -s v1/e2e
- pytest -v -s v1/engine

68
.buildkite/test_areas/entrypoints.yaml Normal file
View File

@ -0,0 +1,68 @@
group: Entrypoints
depends_on:
- image-build
steps:
- label: Entrypoints Unit Tests
timeout_in_minutes: 10
working_dir: "/vllm-workspace/tests"
source_file_dependencies:
- vllm/entrypoints
- tests/entrypoints/
commands:
- pytest -v -s entrypoints/openai/tool_parsers
- pytest -v -s entrypoints/ --ignore=entrypoints/llm --ignore=entrypoints/openai --ignore=entrypoints/offline_mode --ignore=entrypoints/test_chat_utils.py --ignore=entrypoints/pooling
- label: Entrypoints Integration (LLM)
timeout_in_minutes: 40
working_dir: "/vllm-workspace/tests"
source_file_dependencies:
- vllm/
- tests/entrypoints/llm
- tests/entrypoints/offline_mode
commands:
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
- pytest -v -s entrypoints/llm --ignore=entrypoints/llm/test_generate.py --ignore=entrypoints/llm/test_collective_rpc.py
- pytest -v -s entrypoints/llm/test_generate.py # it needs a clean process
- pytest -v -s entrypoints/offline_mode # Needs to avoid interference with other tests
- label: Entrypoints Integration (API Server)
timeout_in_minutes: 130
working_dir: "/vllm-workspace/tests"
source_file_dependencies:
- vllm/
- tests/entrypoints/openai
- tests/entrypoints/test_chat_utils
commands:
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
- PYTHONPATH=/vllm-workspace pytest -v -s entrypoints/openai/test_collective_rpc.py # PYTHONPATH is needed to import custom Worker extension
- pytest -v -s entrypoints/openai --ignore=entrypoints/openai/test_chat_with_tool_reasoning.py --ignore=entrypoints/openai/test_oot_registration.py --ignore=entrypoints/openai/test_tensorizer_entrypoint.py --ignore=entrypoints/openai/correctness/ --ignore=entrypoints/openai/test_collective_rpc.py --ignore=entrypoints/openai/tool_parsers/
- pytest -v -s entrypoints/test_chat_utils.py
- label: Entrypoints Integration (Pooling)
timeout_in_minutes: 50
working_dir: "/vllm-workspace/tests"
source_file_dependencies:
- vllm/
- tests/entrypoints/pooling
commands:
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
- pytest -v -s entrypoints/pooling
- label: Entrypoints V1
timeout_in_minutes: 50
source_file_dependencies:
- vllm/
- tests/v1
commands:
- pytest -v -s v1/entrypoints
- label: OpenAI API Correctness
timeout_in_minutes: 30
source_file_dependencies:
- csrc/
- vllm/entrypoints/openai/
- vllm/model_executor/models/whisper.py
commands: # LMEval+Transcription WER check
- pytest -s entrypoints/openai/correctness/

23
.buildkite/test_areas/expert_parallelism.yaml Normal file
View File

@ -0,0 +1,23 @@
group: Expert Parallelism
depends_on:
- image-build
steps:
- label: EPLB Algorithm
timeout_in_minutes: 15
working_dir: "/vllm-workspace/tests"
source_file_dependencies:
- vllm/distributed/eplb
- tests/distributed/test_eplb_algo.py
commands:
- pytest -v -s distributed/test_eplb_algo.py
- label: EPLB Execution
timeout_in_minutes: 20
working_dir: "/vllm-workspace/tests"
num_gpus: 4
source_file_dependencies:
- vllm/distributed/eplb
- tests/distributed/test_eplb_execute.py
commands:
- pytest -v -s distributed/test_eplb_execute.py
- pytest -v -s distributed/test_eplb_spec_decode.py

117
.buildkite/test_areas/kernels.yaml Normal file
View File

@ -0,0 +1,117 @@
group: Kernels
depends_on:
- image-build
steps:
- label: Kernels Core Operation Test
timeout_in_minutes: 75
source_file_dependencies:
- csrc/
- tests/kernels/core
- tests/kernels/test_top_k_per_row.py
commands:
- pytest -v -s kernels/core kernels/test_top_k_per_row.py
- label: Kernels Attention Test %N
timeout_in_minutes: 35
source_file_dependencies:
- csrc/attention/
- vllm/attention
- vllm/v1/attention
- tests/kernels/attention
commands:
- pytest -v -s kernels/attention --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
parallelism: 2
- label: Kernels Quantization Test %N
timeout_in_minutes: 90
source_file_dependencies:
- csrc/quantization/
- vllm/model_executor/layers/quantization
- tests/kernels/quantization
commands:
- pytest -v -s kernels/quantization --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
parallelism: 2
- label: Kernels MoE Test %N
timeout_in_minutes: 60
source_file_dependencies:
- csrc/quantization/cutlass_w8a8/moe/
- csrc/moe/
- tests/kernels/moe
- vllm/model_executor/layers/fused_moe/
- vllm/distributed/device_communicators/
- vllm/envs.py
- vllm/config
commands:
- pytest -v -s kernels/moe --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
parallelism: 2
- label: Kernels Mamba Test
timeout_in_minutes: 45
source_file_dependencies:
- csrc/mamba/
- tests/kernels/mamba
- vllm/model_executor/layers/mamba/ops
commands:
- pytest -v -s kernels/mamba
- label: Kernels DeepGEMM Test (H100)
timeout_in_minutes: 45
gpu: h100
num_gpus: 1
source_file_dependencies:
- tools/install_deepgemm.sh
- vllm/utils/deep_gemm.py
- vllm/model_executor/layers/fused_moe
- vllm/model_executor/layers/quantization
- tests/kernels/quantization/test_block_fp8.py
- tests/kernels/moe/test_deepgemm.py
- tests/kernels/moe/test_batched_deepgemm.py
- tests/kernels/attention/test_deepgemm_attention.py
commands:
- pytest -v -s kernels/quantization/test_block_fp8.py -k deep_gemm
- pytest -v -s kernels/moe/test_deepgemm.py
- pytest -v -s kernels/moe/test_batched_deepgemm.py
- pytest -v -s kernels/attention/test_deepgemm_attention.py
- label: Kernels (B200)
timeout_in_minutes: 30
working_dir: "/vllm-workspace/"
gpu: b200
# optional: true
source_file_dependencies:
- csrc/quantization/fp4/
- csrc/attention/mla/
- csrc/quantization/cutlass_w8a8/moe/
- vllm/model_executor/layers/fused_moe/cutlass_moe.py
- vllm/model_executor/layers/fused_moe/flashinfer_cutlass_moe.py
- vllm/model_executor/layers/fused_moe/flashinfer_cutlass_prepare_finalize.py
- vllm/model_executor/layers/quantization/utils/flashinfer_utils.py
- vllm/v1/attention/backends/flashinfer.py
- vllm/v1/attention/backends/mla/cutlass_mla.py
- vllm/v1/attention/backends/mla/flashinfer_mla.py
- vllm/platforms/cuda.py
- vllm/attention/selector.py
commands:
- nvidia-smi
- python3 examples/offline_inference/basic/chat.py
# Attention
# num_heads2 broken by https://github.com/flashinfer-ai/flashinfer/issues/1353
- pytest -v -s tests/kernels/attention/test_attention_selector.py
- pytest -v -s tests/kernels/attention/test_flashinfer.py -k 'not num_heads2'
- pytest -v -s tests/kernels/attention/test_flashinfer_trtllm_attention.py
- pytest -v -s tests/kernels/attention/test_cutlass_mla_decode.py
- pytest -v -s tests/kernels/attention/test_flashinfer_mla_decode.py
# Quantization
- pytest -v -s tests/kernels/quantization/test_cutlass_scaled_mm.py -k 'fp8'
- pytest -v -s tests/kernels/quantization/test_nvfp4_quant.py
- pytest -v -s tests/kernels/quantization/test_silu_mul_nvfp4_quant.py
- pytest -v -s tests/kernels/quantization/test_nvfp4_scaled_mm.py
- pytest -v -s tests/kernels/quantization/test_flashinfer_scaled_mm.py
- pytest -v -s tests/kernels/quantization/test_flashinfer_nvfp4_scaled_mm.py
- pytest -v -s tests/kernels/quantization/test_nvfp4_qutlass.py
- pytest -v -s tests/kernels/quantization/test_mxfp4_qutlass.py
- pytest -v -s tests/kernels/moe/test_nvfp4_moe.py
- pytest -v -s tests/kernels/moe/test_ocp_mx_moe.py
- pytest -v -s tests/kernels/moe/test_flashinfer.py
- pytest -v -s tests/kernels/moe/test_cutedsl_moe.py

46
.buildkite/test_areas/lm_eval.yaml Normal file
View File

@ -0,0 +1,46 @@
group: LM Eval
depends_on:
- image-build
steps:
- label: LM Eval Small Models
timeout_in_minutes: 75
source_file_dependencies:
- csrc/
- vllm/model_executor/layers/quantization
autorun_on_main: true
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-small.txt --tp-size=1
- label: LM Eval Large Models (4 GPUs)(A100)
gpu: a100
optional: true
num_gpus: 4
working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
source_file_dependencies:
- csrc/
- vllm/model_executor/layers/quantization
commands:
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
- pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large.txt --tp-size=4
- label: LM Eval Large Models (4 GPUs)(H100)
gpu: h100
optional: true
num_gpus: 4
working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
source_file_dependencies:
- csrc/
- vllm/model_executor/layers/quantization
commands:
- export VLLM_USE_DEEP_GEMM=0 # We found Triton is faster than DeepGEMM for H100
- pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large-hopper.txt --tp-size=4
- label: LM Eval Small Models (B200)
timeout_in_minutes: 120
gpu: b200
optional: true
source_file_dependencies:
- csrc/
- vllm/model_executor/layers/quantization
commands:
- pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-blackwell.txt --tp-size=1

31
.buildkite/test_areas/lora.yaml Normal file
View File

@ -0,0 +1,31 @@
group: LoRA
depends_on:
- image-build
steps:
- label: LoRA %N
timeout_in_minutes: 30
source_file_dependencies:
- vllm/lora
- tests/lora
commands:
- pytest -v -s lora --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --ignore=lora/test_chatglm3_tp.py --ignore=lora/test_llama_tp.py --ignore=lora/test_llm_with_multi_loras.py --ignore=lora/test_olmoe_tp.py --ignore=lora/test_deepseekv2_tp.py --ignore=lora/test_gptoss_tp.py --ignore=lora/test_qwen3moe_tp.py
parallelism: 4
- label: LoRA TP (Distributed)
timeout_in_minutes: 30
num_gpus: 4
source_file_dependencies:
- vllm/lora
- tests/lora
commands:
# FIXIT: find out which code initialize cuda before running the test
# before the fix, we need to use spawn to test it
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
# There is some Tensor Parallelism related processing logic in LoRA that
# requires multi-GPU testing for validation.
- pytest -v -s -x lora/test_chatglm3_tp.py
- pytest -v -s -x lora/test_llama_tp.py
- pytest -v -s -x lora/test_llm_with_multi_loras.py
- pytest -v -s -x lora/test_olmoe_tp.py
- pytest -v -s -x lora/test_gptoss_tp.py

165
.buildkite/test_areas/misc.yaml Normal file
View File

@ -0,0 +1,165 @@
group: Miscellaneous
depends_on:
- image-build
steps:
- label: V1 Others
timeout_in_minutes: 60
source_file_dependencies:
- vllm/
- tests/v1
commands:
- uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
# split the test to avoid interference
- pytest -v -s -m 'not cpu_test' v1/core
- pytest -v -s v1/executor
- pytest -v -s v1/kv_offload
- pytest -v -s v1/sample
- pytest -v -s v1/logits_processors
- pytest -v -s v1/worker
- pytest -v -s v1/spec_decode
- pytest -v -s -m 'not cpu_test' v1/kv_connector/unit
- pytest -v -s -m 'not cpu_test' v1/metrics
- pytest -v -s v1/test_oracle.py
- pytest -v -s v1/test_request.py
- pytest -v -s v1/test_outputs.py
# Integration test for streaming correctness (requires special branch).
- pip install -U git+https://github.com/robertgshaw2-redhat/lm-evaluation-harness.git@streaming-api
- pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
- label: V1 Others (CPU)
depends_on: ~
source_file_dependencies:
- vllm/
- tests/v1
no_gpu: true
commands:
# split the test to avoid interference
- pytest -v -s -m 'cpu_test' v1/core
- pytest -v -s v1/structured_output
- pytest -v -s v1/test_serial_utils.py
- pytest -v -s -m 'cpu_test' v1/kv_connector/unit
- pytest -v -s -m 'cpu_test' v1/metrics
- label: Regression
timeout_in_minutes: 20
source_file_dependencies:
- vllm/
- tests/test_regression
commands:
- pip install modelscope
- pytest -v -s test_regression.py
working_dir: "/vllm-workspace/tests" # optional
- label: Examples
timeout_in_minutes: 45
working_dir: "/vllm-workspace/examples"
source_file_dependencies:
- vllm/entrypoints
- vllm/multimodal
- examples/
commands:
- pip install tensorizer # for tensorizer test
- python3 offline_inference/basic/chat.py # for basic
- python3 offline_inference/basic/generate.py --model facebook/opt-125m
- python3 offline_inference/basic/generate.py --model meta-llama/Llama-2-13b-chat-hf --cpu-offload-gb 10
- python3 offline_inference/basic/classify.py
- python3 offline_inference/basic/embed.py
- python3 offline_inference/basic/score.py
# for multi-modal models
- python3 offline_inference/audio_language.py --seed 0
- python3 offline_inference/vision_language.py --seed 0
- python3 offline_inference/vision_language_multi_image.py --seed 0
- python3 offline_inference/encoder_decoder_multimodal.py --model-type whisper --seed 0
# for pooling models
- python3 pooling/pooling/vision_language_pooling.py --seed 0
# for features demo
- python3 offline_inference/prefix_caching.py
- python3 offline_inference/llm_engine_example.py
- python3 others/tensorize_vllm_model.py --model facebook/opt-125m serialize --serialized-directory /tmp/ --suffix v1 && python3 others/tensorize_vllm_model.py --model facebook/opt-125m deserialize --path-to-tensors /tmp/vllm/facebook/opt-125m/v1/model.tensors
- python3 offline_inference/spec_decode.py --test --method eagle --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 2048
# https://github.com/vllm-project/vllm/pull/26682 uses slightly more memory in PyTorch 2.9+ causing this test to OOM in 1xL4 GPU
- python3 offline_inference/spec_decode.py --test --method eagle3 --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 1536
- label: Metrics, Tracing (2 GPUs)
timeout_in_minutes: 20
num_gpus: 2
source_file_dependencies:
- vllm/
- tests/v1/tracing
commands:
- "pip install \
'opentelemetry-sdk>=1.26.0' \
'opentelemetry-api>=1.26.0' \
'opentelemetry-exporter-otlp>=1.26.0' \
'opentelemetry-semantic-conventions-ai>=0.4.1'"
- pytest -v -s v1/tracing
- label: Python-only Installation
depends_on: ~
timeout_in_minutes: 20
source_file_dependencies:
- tests/standalone_tests/python_only_compile.sh
- setup.py
commands:
- bash standalone_tests/python_only_compile.sh
- label: Async Engine, Inputs, Utils, Worker
timeout_in_minutes: 50
source_file_dependencies:
- vllm/
- tests/multimodal
- tests/utils_
commands:
- pytest -v -s -m 'not cpu_test' multimodal
- pytest -v -s utils_
- label: Async Engine, Inputs, Utils, Worker, Config (CPU)
depends_on: ~
timeout_in_minutes: 30
source_file_dependencies:
- vllm/
- tests/test_inputs.py
- tests/test_outputs.py
- tests/multimodal
- tests/standalone_tests/lazy_imports.py
- tests/tokenizers_
- tests/tool_parsers
- tests/transformers_utils
- tests/config
no_gpu: true
commands:
- python3 standalone_tests/lazy_imports.py
- pytest -v -s test_inputs.py
- pytest -v -s test_outputs.py
- pytest -v -s -m 'cpu_test' multimodal
- pytest -v -s tokenizers_
- pytest -v -s tool_parsers
- pytest -v -s transformers_utils
- pytest -v -s config
- label: GPT-OSS Eval (B200)
timeout_in_minutes: 60
working_dir: "/vllm-workspace/"
gpu: b200
optional: true
source_file_dependencies:
- tests/evals/gpt_oss
- vllm/model_executor/models/gpt_oss.py
- vllm/model_executor/layers/quantization/mxfp4.py
- vllm/v1/attention/backends/flashinfer.py
commands:
- uv pip install --system 'gpt-oss[eval]==0.0.5'
- pytest -s -v tests/evals/gpt_oss/test_gpqa_correctness.py --model openai/gpt-oss-20b --metric 0.58
- label: Batch Invariance (H100)
timeout_in_minutes: 25
gpu: h100
source_file_dependencies:
- vllm/v1/attention
- vllm/model_executor/layers
- tests/v1/determinism/
commands:
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
- pip install pytest-timeout pytest-forked
- pytest -v -s v1/determinism/test_batch_invariance.py
- pytest -v -s v1/determinism/test_rms_norm_batch_invariant.py

17
.buildkite/test_areas/model_executor.yaml Normal file
View File

@ -0,0 +1,17 @@
group: Model Executor
depends_on:
- image-build
steps:
- label: Model Executor
timeout_in_minutes: 35
source_file_dependencies:
- vllm/engine/arg_utils.py
- vllm/config/model.py
- vllm/model_executor
- tests/model_executor
- tests/entrypoints/openai/test_tensorizer_entrypoint.py
commands:
- apt-get update && apt-get install -y curl libsodium23
- export VLLM_WORKER_MULTIPROC_METHOD=spawn
- pytest -v -s model_executor
- pytest -v -s entrypoints/openai/test_tensorizer_entrypoint.py

62
.buildkite/test_areas/models_basic.yaml Normal file
View File

@ -0,0 +1,62 @@
group: Models - Basic
depends_on:
- image-build
steps:
- label: Basic Models Tests (Initialization)
timeout_in_minutes: 45
mirror_hardwares: [amdexperimental]
torch_nightly: true
source_file_dependencies:
- vllm/
- tests/models/test_initialization.py
commands:
# Run a subset of model initialization tests
- pytest -v -s models/test_initialization.py::test_can_initialize_small_subset
- label: Basic Models Tests (Extra Initialization) %N
timeout_in_minutes: 45
mirror_hardwares: [amdexperimental]
torch_nightly: true
source_file_dependencies:
- vllm/model_executor/models/
- tests/models/test_initialization.py
commands:
# Only when vLLM model source is modified - test initialization of a large
# subset of supported models (the complement of the small subset in the above
# test.) Also run if model initialization test file is modified
- pytest -v -s models/test_initialization.py -k 'not test_can_initialize_small_subset' --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
parallelism: 2
- label: Basic Models Tests (Other)
timeout_in_minutes: 45
source_file_dependencies:
- vllm/
- tests/models/test_transformers.py
- tests/models/test_registry.py
commands:
- pytest -v -s models/test_transformers.py models/test_registry.py
- label: Basic Models Test (Other CPU) # 5min
timeout_in_minutes: 10
source_file_dependencies:
- vllm/
- tests/models/test_utils.py
- tests/models/test_vision.py
no_gpu: true
commands:
- pytest -v -s models/test_utils.py models/test_vision.py
- label: Transformers Nightly Models
working_dir: "/vllm-workspace/"
optional: true
soft_fail: true
commands:
- pip install --upgrade git+https://github.com/huggingface/transformers
- pytest -v -s tests/models/test_initialization.py
- pytest -v -s tests/models/test_transformers.py
- pytest -v -s tests/models/multimodal/processing/
- pytest -v -s tests/models/multimodal/test_mapping.py
- python3 examples/offline_inference/basic/chat.py
- python3 examples/offline_inference/vision_language.py --model-type qwen2_5_vl
# Whisper needs spawn method to avoid deadlock
- VLLM_WORKER_MULTIPROC_METHOD=spawn python3 examples/offline_inference/audio_language.py --model-type whisper

22
.buildkite/test_areas/models_distributed.yaml Normal file
View File

@ -0,0 +1,22 @@
group: Models - Distributed
depends_on:
- image-build
steps:
- label: Distributed Model Tests (2 GPUs)
timeout_in_minutes: 50
working_dir: "/vllm-workspace/tests"
num_gpus: 2
source_file_dependencies:
- vllm/model_executor/model_loader/sharded_state_loader.py
- vllm/model_executor/models/
- tests/basic_correctness/
- tests/model_executor/model_loader/test_sharded_state_loader.py
- tests/models/
commands:
- TARGET_TEST_SUITE=L4 pytest basic_correctness/ -v -s -m 'distributed(num_gpus=2)'
- CUDA_VISIBLE_DEVICES=0,1 pytest -v -s model_executor/model_loader/test_sharded_state_loader.py
# Avoid importing model tests that cause CUDA reinitialization error
- pytest models/test_transformers.py -v -s -m 'distributed(num_gpus=2)'
- pytest models/language -v -s -m 'distributed(num_gpus=2)'
- pytest models/multimodal -v -s -m 'distributed(num_gpus=2)' --ignore models/multimodal/generation/test_whisper.py
- VLLM_WORKER_MULTIPROC_METHOD=spawn pytest models/multimodal/generation/test_whisper.py -v -s -m 'distributed(num_gpus=2)'

91
.buildkite/test_areas/models_language.yaml Normal file
View File

@ -0,0 +1,91 @@
group: Models - Language
depends_on:
- image-build
steps:
- label: Language Models Tests (Standard)
timeout_in_minutes: 25
mirror_hardwares: [amdexperimental]
torch_nightly: true
source_file_dependencies:
- vllm/
- tests/models/language
commands:
# Test standard language models, excluding a subset of slow tests
- pip freeze | grep -E 'torch'
- pytest -v -s models/language -m 'core_model and (not slow_test)'
- label: Language Models Tests (Extra Standard) %N
timeout_in_minutes: 45
mirror_hardwares: [amdexperimental]
torch_nightly: true
source_file_dependencies:
- vllm/model_executor/models/
- tests/models/language/pooling/test_embedding.py
- tests/models/language/generation/test_common.py
- tests/models/language/pooling/test_classification.py
commands:
# Shard slow subset of standard language models tests. Only run when model
# source is modified, or when specified test files are modified
- pip freeze | grep -E 'torch'
- pytest -v -s models/language -m 'core_model and slow_test' --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
parallelism: 2
- label: Language Models Tests (Hybrid) %N
timeout_in_minutes: 75
mirror_hardwares: [amdexperimental]
torch_nightly: true
source_file_dependencies:
- vllm/
- tests/models/language/generation
commands:
# Install fast path packages for testing against transformers
# Note: also needed to run plamo2 model in vLLM
- uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.2.5'
- uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.5.2'
# Shard hybrid language model tests
- pytest -v -s models/language/generation -m hybrid_model --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB
parallelism: 2
- label: Language Models Test (Extended Generation) # 80min
timeout_in_minutes: 110
mirror_hardwares: [amdexperimental]
optional: true
source_file_dependencies:
- vllm/
- tests/models/language/generation
commands:
# Install fast path packages for testing against transformers
# Note: also needed to run plamo2 model in vLLM
- uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.2.5'
- uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.5.2'
- pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
- label: Language Models Test (PPL)
timeout_in_minutes: 110
mirror_hardwares: [amdexperimental]
optional: true
source_file_dependencies:
- vllm/
- tests/models/language/generation_ppl_test
commands:
- pytest -v -s models/language/generation_ppl_test
- label: Language Models Test (Extended Pooling) # 36min
timeout_in_minutes: 50
mirror_hardwares: [amdexperimental]
optional: true
source_file_dependencies:
- vllm/
- tests/models/language/pooling
commands:
- pytest -v -s models/language/pooling -m 'not core_model'
- label: Language Models Test (MTEB)
timeout_in_minutes: 110
mirror_hardwares: [amdexperimental]
optional: true
source_file_dependencies:
- vllm/
- tests/models/language/pooling_mteb_test
commands:
- pytest -v -s models/language/pooling_mteb_test

79
.buildkite/test_areas/models_multimodal.yaml Normal file
View File

@ -0,0 +1,79 @@
group: Models - Multimodal
depends_on:
- image-build
steps:
- label: Multi-Modal Models (Standard) # 60min
timeout_in_minutes: 80
source_file_dependencies:
- vllm/
- tests/models/multimodal
commands:
- pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
- pip freeze | grep -E 'torch'
- pytest -v -s models/multimodal -m core_model --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/processing
- cd .. && VLLM_WORKER_MULTIPROC_METHOD=spawn pytest -v -s tests/models/multimodal/generation/test_whisper.py -m core_model # Otherwise, mp_method="spawn" doesn't work
- label: Multi-Modal Processor Test (CPU)
timeout_in_minutes: 60
source_file_dependencies:
- vllm/
- tests/models/multimodal
no_gpu: true
commands:
- pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
- pytest -v -s models/multimodal/processing --ignore models/multimodal/processing/test_tensor_schema.py
- label: Multi-Modal Processor # 44min
timeout_in_minutes: 60
source_file_dependencies:
- vllm/
- tests/models/multimodal
commands:
- pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
- pytest -v -s models/multimodal/processing/test_tensor_schema.py
- label: Multi-Modal Accuracy Eval (Small Models) # 50min
timeout_in_minutes: 70
working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
source_file_dependencies:
- vllm/multimodal/
- vllm/inputs/
- vllm/v1/core/
commands:
- pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-mm-small.txt --tp-size=1
- label: Multi-Modal Models (Extended) 1
optional: true
source_file_dependencies:
- vllm/
- tests/models/multimodal
commands:
- pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
- pytest -v -s models/multimodal -m 'not core_model' --ignore models/multimodal/generation/test_common.py --ignore models/multimodal/processing
- label: Multi-Modal Models (Extended) 2
optional: true
source_file_dependencies:
- vllm/
- tests/models/multimodal
commands:
- pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
- pytest -v -s models/multimodal/generation/test_common.py -m 'split(group=0) and not core_model'
- label: Multi-Modal Models (Extended) 3
optional: true
source_file_dependencies:
- vllm/
- tests/models/multimodal
commands:
- pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
- pytest -v -s models/multimodal/generation/test_common.py -m 'split(group=1) and not core_model'
# This test is used only in PR development phase to test individual models and should never run on main
- label: Custom Models
optional: true
commands:
- echo 'Testing custom models...'
# PR authors can temporarily add commands below to test individual models
# e.g. pytest -v -s models/encoder_decoder/vision_language/test_mllama.py
# *To avoid merge conflicts, remember to REMOVE (not just comment out) them before merging the PR*

34
.buildkite/test_areas/plugins.yaml Normal file
View File

@ -0,0 +1,34 @@
group: Plugins
depends_on:
- image-build
steps:
- label: Plugin Tests (2 GPUs)
timeout_in_minutes: 60
working_dir: "/vllm-workspace/tests"
num_gpus: 2
source_file_dependencies:
- vllm/plugins/
- tests/plugins/
commands:
# begin platform plugin and general plugin tests, all the code in-between runs on dummy platform
- pip install -e ./plugins/vllm_add_dummy_platform
- pytest -v -s plugins_tests/test_platform_plugins.py
- pip uninstall vllm_add_dummy_platform -y
# end platform plugin tests
# begin io_processor plugins test, all the code in between uses the prithvi_io_processor plugin
- pip install -e ./plugins/prithvi_io_processor_plugin
- pytest -v -s plugins_tests/test_io_processor_plugins.py
- pip uninstall prithvi_io_processor_plugin -y
# end io_processor plugins test
# begin stat_logger plugins test
- pip install -e ./plugins/vllm_add_dummy_stat_logger
- pytest -v -s plugins_tests/test_stats_logger_plugins.py
- pip uninstall dummy_stat_logger -y
# end stat_logger plugins test
# other tests continue here:
- pytest -v -s plugins_tests/test_scheduler_plugins.py
- pip install -e ./plugins/vllm_add_dummy_model
- pytest -v -s distributed/test_distributed_oot.py
- pytest -v -s entrypoints/openai/test_oot_registration.py # it needs a clean process
- pytest -v -s models/test_oot_registration.py # it needs a clean process
- pytest -v -s plugins/lora_resolvers # unit tests for in-tree lora resolver plugins

50
.buildkite/test_areas/pytorch.yaml Normal file
View File

@ -0,0 +1,50 @@
group: PyTorch
depends_on:
- image-build
steps:
- label: PyTorch Compilation Unit Tests
timeout_in_minutes: 30
source_file_dependencies:
- vllm/
- tests/compile
commands:
# Run unit tests defined directly under compile/,
# not including subdirectories, which are usually heavier
# tests covered elsewhere.
# Use `find` to launch multiple instances of pytest so that
# they do not suffer from https://github.com/vllm-project/vllm/issues/28965
- "find compile/ -maxdepth 1 -name 'test_*.py' -exec pytest -s -v {} \\;"
- label: PyTorch Fullgraph Smoke Test
timeout_in_minutes: 30
source_file_dependencies:
- vllm/
- tests/compile
commands:
# Run smoke tests under fullgraph directory, except test_full_graph.py
# as it is a heavy test that is covered in other steps.
# Use `find` to launch multiple instances of pytest so that
# they do not suffer from https://github.com/vllm-project/vllm/issues/28965
- "find compile/fullgraph/ -name 'test_*.py' -not -name 'test_full_graph.py' -exec pytest -s -v {} \\;"
- label: PyTorch Fullgraph
timeout_in_minutes: 40
source_file_dependencies:
- vllm/
- tests/compile
commands:
# fp8 kv scales not supported on sm89, tested on Blackwell instead
- pytest -v -s compile/fullgraph/test_full_graph.py -k 'not test_fp8_kv_scale_compile'
# Limit to no custom ops to reduce running time
# Wrap with quotes to escape yaml and avoid starting -k string with a -
- "pytest -v -s compile/distributed/test_fusions_e2e.py -k 'TRITON and not +quant_fp8 and not Llama-4'"
- label: Pytorch Nightly Dependency Override Check # 2min
# if this test fails, it means the nightly torch version is not compatible with some
# of the dependencies. Please check the error message and add the package to whitelist
# in /vllm/tools/pre_commit/generate_nightly_torch_test.py
soft_fail: true
source_file_dependencies:
- requirements/nightly_torch_test.txt
commands:
- bash standalone_tests/pytorch_nightly_dependency.sh

46
.buildkite/test_areas/quantization.yaml Normal file
View File

@ -0,0 +1,46 @@
group: Quantization
depends_on:
- image-build
steps:
- label: Quantization
timeout_in_minutes: 90
source_file_dependencies:
- csrc/
- vllm/model_executor/layers/quantization
- tests/quantization
commands:
# temporary install here since we need nightly, will move to requirements/test.in
# after torchao 0.12 release, and pin a working version of torchao nightly here
# since torchao nightly is only compatible with torch nightly currently
# https://github.com/pytorch/ao/issues/2919, we'll have to skip new torchao tests for now
# we can only upgrade after this is resolved
# TODO(jerryzh168): resolve the above comment
- uv pip install --system torchao==0.13.0 --index-url https://download.pytorch.org/whl/cu129
- uv pip install --system conch-triton-kernels
- VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s quantization/ --ignore quantization/test_blackwell_moe.py
- label: Quantized MoE Test (B200)
timeout_in_minutes: 60
working_dir: "/vllm-workspace/"
gpu: b200
source_file_dependencies:
- tests/quantization/test_blackwell_moe.py
- vllm/model_executor/models/deepseek_v2.py
- vllm/model_executor/models/gpt_oss.py
- vllm/model_executor/models/llama4.py
- vllm/model_executor/layers/fused_moe
- vllm/model_executor/layers/quantization/compressed_tensors
- vllm/model_executor/layers/quantization/modelopt.py
- vllm/model_executor/layers/quantization/mxfp4.py
- vllm/v1/attention/backends/flashinfer.py
commands:
- pytest -s -v tests/quantization/test_blackwell_moe.py
- label: Quantized Models Test
timeout_in_minutes: 60
source_file_dependencies:
- vllm/model_executor/layers/quantization
- tests/models/quantization
commands:
- pytest -v -s models/quantization

Some files were not shown because too many files have changed in this diff Show More