Skip to content

GitLab

Commit 38d80967 authored by zhuwenwen's avatar zhuwenwen
Browse files

Merge tag 'v0.10.2rc2' into v0.10.2rc2-ori

parents 33650733 880c741b
Expand all Show whitespace changes
Inline Side-by-side
Showing with 1447 additions and 194 deletions
.buildkite/check-wheel-size.py
View file @ 38d80967
......@@ -5,11 +5,11 @@ import os
import sys
import zipfile
# Read the VLLM_MAX_SIZE_MB environment variable, defaulting to 400 MiB
# Note that we have 400 MiB quota, please use it wisely.
# See https://github.com/pypi/support/issues/3792 .
# Read the VLLM_MAX_SIZE_MB environment variable, defaulting to 450 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", 400))
VLLM_MAX_SIZE_MB = int(os.environ.get("VLLM_MAX_SIZE_MB", 450))
def print_top_10_largest_files(zip_file):
......
.buildkite/nightly-benchmarks/scripts/compare-json-results.py
View file @ 38d80967
......@@ -218,7 +218,7 @@ if __name__ == "__main__":
"--xaxis",
type=str,
default="# of max concurrency.",
help="column name to use as X Axis in comparision graph",
help="column name to use as X Axis in comparison graph",
)
args = parser.parse_args()
......
.buildkite/nightly-benchmarks/tests/serving-tests-cpu-snc2.json
View file @ 38d80967
[
{
"test_name": "serving_llama8B_tp1_sharegpt",
"test_name": "serving_llama8B_bf16_tp1_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
......@@ -32,7 +32,7 @@
}
},
{
"test_name": "serving_llama8B_tp2_sharegpt",
"test_name": "serving_llama8B_bf16_tp2_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
......@@ -64,7 +64,7 @@
}
},
{
"test_name": "serving_llama8B_tp4_sharegpt",
"test_name": "serving_llama8B_bf16_tp4_sharegpt",
"qps_list": ["inf"],
"max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
"server_environment_variables": {
......@@ -96,7 +96,7 @@
}
},
{
"test_name": "serving_llama8B_tp1_random_128_128",
"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": {
......@@ -131,7 +131,7 @@
}
},
{
"test_name": "serving_llama8B_tp2_random_128_128",
"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": {
......@@ -166,7 +166,7 @@
}
},
{
"test_name": "serving_llama8B_tp4_random_128_128",
"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": {
......@@ -198,5 +198,413 @@
"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
}
}
]
.buildkite/release-pipeline.yaml
View file @ 38d80967
steps:
# aarch64 + CUDA builds
- label: "Build arm64 wheel - CUDA 12.8"
id: build-wheel-arm64-cuda-12-8
# aarch64 + CUDA builds. PyTorch 2.8 aarch64 + CUDA wheel is only available on CUDA 12.9
- label: "Build arm64 wheel - CUDA 12.9"
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=12.8.1 --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 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"
# x86 + CUDA builds
- 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
......@@ -44,18 +47,14 @@ steps:
env:
DOCKER_BUILDKIT: "1"
# Note(simon): We can always build CUDA 11.8 wheel to ensure the build is working.
# However, this block can be uncommented to save some compute hours.
# - block: "Build CUDA 11.8 wheel"
# key: block-build-cu118-wheel
- label: "Build wheel - CUDA 11.8"
# depends_on: block-build-cu118-wheel
id: build-wheel-cuda-11-8
# x86 + CUDA builds
- label: "Build wheel - CUDA 12.9"
depends_on: ~
id: build-wheel-cuda-12-9
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=11.8.0 --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 --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 ."
- "mkdir artifacts"
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
- "bash .buildkite/scripts/upload-wheels.sh"
......@@ -75,6 +74,7 @@ steps:
- "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
# PyTorch 2.8 aarch64 + CUDA wheel is only available on CUDA 12.9
- label: "Build release image (arm64)"
depends_on: ~
id: build-release-image-arm64
......@@ -82,7 +82,7 @@ steps:
queue: arm64_cpu_queue_postmerge
commands:
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.8.1 --build-arg torch_cuda_arch_list='8.7 9.0 10.0+PTX 12.0' --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list='8.7 9.0 10.0+PTX 12.0' --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
# Add job to create multi-arch manifest
......@@ -103,7 +103,7 @@ steps:
- create-multi-arch-manifest
- build-wheel-cuda-12-8
- build-wheel-cuda-12-6
- build-wheel-cuda-11-8
- build-wheel-cuda-12-9
id: annotate-release-workflow
agents:
queue: cpu_queue_postmerge
......@@ -150,18 +150,24 @@ steps:
env:
DOCKER_BUILDKIT: "1"
- block: "Build Neuron release image"
key: block-neuron-release-image-build
depends_on: ~
- label: "Build and publish Neuron release image"
depends_on: block-neuron-release-image-build
- label: "Build and publish nightly multi-arch image to DockerHub"
depends_on:
- create-multi-arch-manifest
if: build.env("NIGHTLY") == "1"
agents:
queue: neuron-postmerge
queue: cpu_queue_postmerge
commands:
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --tag public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:latest --progress plain -f docker/Dockerfile.neuron ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:latest"
- "docker push public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:$(buildkite-agent meta-data get release-version)"
- "docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:$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"
# Clean up old nightly builds (keep only last 14)
- "bash .buildkite/scripts/cleanup-nightly-builds.sh"
plugins:
- docker-login#v3.0.0:
username: vllmbot
password-env: DOCKERHUB_TOKEN
env:
DOCKER_BUILDKIT: "1"
.buildkite/scripts/cleanup-nightly-builds.sh 0 → 100755
View file @ 38d80967
#!/bin/bash
set -ex
# Clean up old nightly builds from DockerHub, keeping only the last 14 builds
# This script uses DockerHub API to list and delete old tags with "nightly-" prefix
# 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
if [ -z "$DOCKERHUB_TOKEN" ]; then
echo "Error: DOCKERHUB_TOKEN environment variable is not set"
exit 1
fi
# Function to get all tags from DockerHub
get_all_tags() {
local page=1
local all_tags=""
while true; do
local response=$(curl -s -H "Authorization: Bearer $DOCKERHUB_TOKEN" \
"$REPO_API_URL?page=$page&page_size=100")
# Get both last_updated timestamp and tag name, separated by |
local tags=$(echo "$response" | jq -r '.results[] | select(.name | startswith("nightly-")) | "\(.last_updated)|\(.name)"')
if [ -z "$tags" ]; then
break
fi
all_tags="$all_tags$tags"$'\n'
page=$((page + 1))
done
# Sort by timestamp (newest first) and extract just the tag names
echo "$all_tags" | sort -r | cut -d'|' -f2
}
delete_tag() {
local tag_name="$1"
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")
if echo "$response" | jq -e '.detail' > /dev/null 2>&1; then
echo "Warning: Failed to delete tag $tag_name: $(echo "$response" | jq -r '.detail')"
else
echo "Successfully deleted tag: $tag_name"
fi
}
# Get all nightly- prefixed tags, sorted by last_updated timestamp (newest first)
echo "Fetching all tags from DockerHub..."
all_tags=$(get_all_tags)
if [ -z "$all_tags" ]; then
echo "No tags found to clean up"
exit 0
fi
# Count total tags
total_tags=$(echo "$all_tags" | wc -l)
echo "Found $total_tags tags"
# Keep only the last 14 builds (including the current one)
tags_to_keep=14
tags_to_delete=$((total_tags - tags_to_keep))
if [ $tags_to_delete -le 0 ]; then
echo "No tags need to be deleted (only $total_tags tags found, keeping $tags_to_keep)"
exit 0
fi
echo "Will delete $tags_to_delete old tags, keeping the newest $tags_to_keep"
# Get tags to delete (skip the first $tags_to_keep tags)
tags_to_delete_list=$(echo "$all_tags" | tail -n +$((tags_to_keep + 1)))
if [ -z "$tags_to_delete_list" ]; then
echo "No tags to delete"
exit 0
fi
# Delete old tags
echo "Deleting old tags..."
while IFS= read -r tag; do
if [ -n "$tag" ]; then
delete_tag "$tag"
# Add a small delay to avoid rate limiting
sleep 1
fi
done <<< "$tags_to_delete_list"
echo "Cleanup completed successfully"
.buildkite/scripts/hardware_ci/run-neuron-test.sh deleted 100644 → 0
View file @ 33650733
#!/bin/bash
# This script build the Neuron docker image and run the API server inside the container.
# It serves a sanity check for compilation and basic model usage.
set -e
set -v
image_name="neuron/vllm-ci"
container_name="neuron_$(tr -dc A-Za-z0-9 < /dev/urandom | head -c 10; echo)"
HF_CACHE="$(realpath ~)/huggingface"
mkdir -p "${HF_CACHE}"
HF_MOUNT="/root/.cache/huggingface"
HF_TOKEN=$(aws secretsmanager get-secret-value --secret-id "ci/vllm-neuron/hf-token" --region us-west-2 --query 'SecretString' --output text | jq -r .VLLM_NEURON_CI_HF_TOKEN)
NEURON_COMPILE_CACHE_URL="$(realpath ~)/neuron_compile_cache"
mkdir -p "${NEURON_COMPILE_CACHE_URL}"
NEURON_COMPILE_CACHE_MOUNT="/root/.cache/neuron_compile_cache"
# Try building the docker image
aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws
# prune old image and containers to save disk space, and only once a day
# by using a timestamp file in tmp.
if [ -f /tmp/neuron-docker-build-timestamp ]; then
last_build=$(cat /tmp/neuron-docker-build-timestamp)
current_time=$(date +%s)
if [ $((current_time - last_build)) -gt 86400 ]; then
# Remove dangling images (those that are not tagged and not used by any container)
docker image prune -f
# Remove unused volumes / force the system prune for old images as well.
docker volume prune -f && docker system prune -f
echo "$current_time" > /tmp/neuron-docker-build-timestamp
fi
else
date "+%s" > /tmp/neuron-docker-build-timestamp
fi
docker build -t "${image_name}" -f docker/Dockerfile.neuron .
# Setup cleanup
remove_docker_container() {
docker image rm -f "${image_name}" || true;
}
trap remove_docker_container EXIT
# Run the image
docker run --rm -it --device=/dev/neuron0 --network bridge \
-v "${HF_CACHE}:${HF_MOUNT}" \
-e "HF_HOME=${HF_MOUNT}" \
-e "HF_TOKEN=${HF_TOKEN}" \
-v "${NEURON_COMPILE_CACHE_URL}:${NEURON_COMPILE_CACHE_MOUNT}" \
-e "NEURON_COMPILE_CACHE_URL=${NEURON_COMPILE_CACHE_MOUNT}" \
--name "${container_name}" \
${image_name} \
/bin/bash -c "
set -e; # Exit on first error
python3 /workspace/vllm/examples/offline_inference/neuron.py;
python3 -m pytest /workspace/vllm/tests/neuron/1_core/ -v --capture=tee-sys;
for f in /workspace/vllm/tests/neuron/2_core/*.py; do
echo \"Running test file: \$f\";
python3 -m pytest \$f -v --capture=tee-sys;
done
"
\ No newline at end of file
.buildkite/scripts/hardware_ci/run-xpu-test.sh
View file @ 38d80967
......@@ -30,10 +30,12 @@ docker run \
bash -c '
set -e
echo $ZE_AFFINITY_MASK
VLLM_USE_V1=1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
VLLM_USE_V1=1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 -O3 -O.cudagraph_mode=NONE
VLLM_USE_V1=1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend ray
VLLM_USE_V1=1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend mp
pip install tblib==3.1.0
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 -O3 -O.cudagraph_mode=NONE
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend ray
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend mp
VLLM_ATTENTION_BACKEND=TRITON_ATTN_VLLM_V1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
cd tests
pytest -v -s v1/core
pytest -v -s v1/engine
......
.buildkite/scripts/upload-wheels.sh
View file @ 38d80967
......@@ -58,14 +58,15 @@ python3 .buildkite/generate_index.py --wheel "$normal_wheel"
aws s3 cp "$wheel" "s3://vllm-wheels/$BUILDKITE_COMMIT/"
aws s3 cp "$normal_wheel" "s3://vllm-wheels/$BUILDKITE_COMMIT/"
if [[ $normal_wheel == *"cu118"* ]]; then
# if $normal_wheel matches cu118, do not upload the index.html
echo "Skipping index files for cu118 wheels"
elif [[ $normal_wheel == *"cu126"* ]]; then
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 $normal_wheel matches cu128, do not upload the index.html
echo "Skipping index files for cu128 wheels"
else
# only upload index.html for cu128 wheels (default wheels)
# only upload index.html for cu129 wheels (default wheels) as it
# is available on both x86 and arm64
aws s3 cp index.html "s3://vllm-wheels/$BUILDKITE_COMMIT/vllm/index.html"
aws s3 cp "s3://vllm-wheels/nightly/index.html" "s3://vllm-wheels/$BUILDKITE_COMMIT/index.html"
fi
......@@ -74,14 +75,15 @@ fi
aws s3 cp "$wheel" "s3://vllm-wheels/nightly/"
aws s3 cp "$normal_wheel" "s3://vllm-wheels/nightly/"
if [[ $normal_wheel == *"cu118"* ]]; then
# if $normal_wheel matches cu118, do not upload the index.html
echo "Skipping index files for cu118 wheels"
elif [[ $normal_wheel == *"cu126"* ]]; then
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 cu128 wheels (default wheels)
# 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"
fi
......
.buildkite/test-pipeline.yaml
View file @ 38d80967
This diff is collapsed.
.github/.bc-linter.yml 0 → 100644
View file @ 38d80967
# doc: https://github.com/pytorch/test-infra/blob/main/tools/stronghold/docs/bc_linter_config.md
version: 1
paths:
# We temporarily disable globally, and will only enable with `annotations.include`
# include:
# - "vllm/v1/attetion/*.py"
# - "vllm/v1/core/*.py"
exclude:
- "**/*.py"
scan:
functions: true # check free functions and methods
classes: true # check classes/dataclasses
public_only: true # ignore names starting with "_" at any level
annotations:
include: # decorators that force‑include a symbol
- name: "bc_linter_include" # matched by simple name or dotted suffix
propagate_to_members: false # for classes, include methods/inner classes
exclude: # decorators that force‑exclude a symbol
- name: "bc_linter_skip" # matched by simple name or dotted suffix
propagate_to_members: true # for classes, exclude methods/inner classes
excluded_violations: [] # e.g. ["ParameterRenamed", "FieldTypeChanged"]
.github/CODEOWNERS
View file @ 38d80967
......@@ -5,18 +5,21 @@
/vllm/attention/backends/abstract.py @WoosukKwon @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
/vllm/core @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
/vllm/engine/llm_engine.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
/vllm/executor/executor_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
/vllm/worker/worker_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
/vllm/executor/executor_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill @22quinn
/vllm/worker/worker_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill @22quinn
/vllm/worker/worker.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
/vllm/model_executor/layers/sampler.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
/vllm/model_executor/layers/sampler.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill @NickLucche
/vllm/model_executor/layers/quantization @mgoin @robertgshaw2-redhat @tlrmchlsmth @yewentao256
/vllm/model_executor/layers/mamba @tdoublep
/vllm/multimodal @DarkLight1337 @ywang96
/vllm/model_executor/model_loader @22quinn
/vllm/multimodal @DarkLight1337 @ywang96 @NickLucche
/vllm/v1/sample @22quinn @houseroad
/vllm/vllm_flash_attn @LucasWilkinson
/vllm/lora @jeejeelee
/vllm/reasoning @aarnphm
/vllm/entrypoints @aarnphm
/vllm/reasoning @aarnphm @chaunceyjiang
/vllm/entrypoints @aarnphm @chaunceyjiang
/vllm/compilation @zou3519 @youkaichao @ProExpertProg
/vllm/distributed/kv_transfer @NickLucche
CMakeLists.txt @tlrmchlsmth @LucasWilkinson
# Any change to the VllmConfig changes can have a large user-facing impact,
......@@ -25,8 +28,11 @@ CMakeLists.txt @tlrmchlsmth @LucasWilkinson
# vLLM V1
/vllm/v1 @WoosukKwon @robertgshaw2-redhat @njhill @ywang96 @comaniac @alexm-redhat
/vllm/v1/structured_output @mgoin @russellb @aarnphm
/vllm/v1/structured_output @mgoin @russellb @aarnphm @benchislett
/vllm/v1/spec_decode @benchislett @luccafong
/vllm/v1/attention/backends/triton_attn.py @tdoublep
/vllm/v1/core @heheda12345
/vllm/v1/kv_cache_interface.py @heheda12345
# Test ownership
/.buildkite/lm-eval-harness @mgoin @simon-mo
......@@ -34,18 +40,20 @@ CMakeLists.txt @tlrmchlsmth @LucasWilkinson
/tests/distributed/test_multi_node_assignment.py @youkaichao
/tests/distributed/test_pipeline_parallel.py @youkaichao
/tests/distributed/test_same_node.py @youkaichao
/tests/entrypoints @DarkLight1337 @robertgshaw2-redhat @simon-mo @aarnphm
/tests/entrypoints @DarkLight1337 @robertgshaw2-redhat @simon-mo @aarnphm @NickLucche
/tests/kernels @tlrmchlsmth @WoosukKwon @yewentao256
/tests/models @DarkLight1337 @ywang96
/tests/multimodal @DarkLight1337 @ywang96
/tests/multimodal @DarkLight1337 @ywang96 @NickLucche
/tests/prefix_caching @comaniac @KuntaiDu
/tests/quantization @mgoin @robertgshaw2-redhat @yewentao256
/tests/test_inputs.py @DarkLight1337 @ywang96
/tests/v1/entrypoints/llm/test_struct_output_generate.py @mgoin @russellb @aarnphm
/tests/v1/structured_output @mgoin @russellb @aarnphm
/tests/v1/core @heheda12345
/tests/weight_loading @mgoin @youkaichao @yewentao256
/tests/lora @jeejeelee
/tests/models/language/generation/test_hybrid.py @tdoublep
/tests/v1/kv_connector/nixl_integration @NickLucche
# Docs
/docs @hmellor
......@@ -67,6 +75,9 @@ mkdocs.yaml @hmellor
/vllm/attention/backends/dual_chunk_flash_attn.py @sighingnow
/vllm/model_executor/models/qwen* @sighingnow
# MTP-specific files
/vllm/model_executor/models/deepseek_mtp.py @luccafong
# Mistral-specific files
/vllm/model_executor/models/mistral*.py @patrickvonplaten
/vllm/model_executor/models/mixtral*.py @patrickvonplaten
......@@ -86,3 +97,8 @@ mkdocs.yaml @hmellor
/vllm/attention/ops/rocm*.py @gshtras
/vllm/model_executor/layers/fused_moe/rocm*.py @gshtras
# TPU
/vllm/v1/worker/tpu* @NickLucche
/vllm/platforms/tpu.py @NickLucche
/vllm/v1/sample/tpu @NickLucche
/vllm/tests/v1/tpu @NickLucche
\ No newline at end of file
.github/mergify.yml
View file @ 38d80967
......@@ -273,6 +273,20 @@ pull_request_rules:
users:
- "sangstar"
- name: assign reviewer for modelopt changes
conditions:
- or:
- files~=^vllm/model_executor/layers/quantization/modelopt\.py$
- files~=^vllm/model_executor/layers/quantization/__init__\.py$
- files~=^tests/models/quantization/test_modelopt\.py$
- files~=^tests/quantization/test_modelopt\.py$
- files~=^tests/models/quantization/test_nvfp4\.py$
- files~=^docs/features/quantization/modelopt\.md$
actions:
assign:
users:
- "Edwardf0t1"
- name: remove 'needs-rebase' label when conflict is resolved
conditions:
- -conflict
......
.github/workflows/add_label_automerge.yml
View file @ 38d80967
......@@ -10,7 +10,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Add label
uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
with:
script: |
github.rest.issues.addLabels({
......
.github/workflows/bc-lint.yml 0 → 100644
View file @ 38d80967
name: BC Lint
on:
pull_request:
types:
- opened
- synchronize
- reopened
jobs:
bc_lint:
if: github.repository_owner == 'vllm-project'
runs-on: ubuntu-latest
steps:
- name: Run BC Lint Action
uses: pytorch/test-infra/.github/actions/bc-lint@main
with:
repo: ${{ github.event.pull_request.head.repo.full_name }}
base_sha: ${{ github.event.pull_request.base.sha }}
head_sha: ${{ github.event.pull_request.head.sha }}
suppression: ${{ contains(github.event.pull_request.labels.*.name, 'suppress-bc-linter') }}
docs_link: 'https://github.com/pytorch/test-infra/wiki/BC-Linter'
config_dir: .github
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: true
.github/workflows/cleanup_pr_body.yml
View file @ 38d80967
......@@ -16,7 +16,7 @@ jobs:
uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
- name: Set up Python
uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0
uses: actions/setup-python@e797f83bcb11b83ae66e0230d6156d7c80228e7c # v6.0.0
with:
python-version: '3.12'
......
.github/workflows/issue_autolabel.yml
View file @ 38d80967
......@@ -13,7 +13,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Label issues based on keywords
uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
with:
script: |
// Configuration: Add new labels and keywords here
......
.github/workflows/pre-commit.yml
View file @ 38d80967
......@@ -17,7 +17,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
- uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0
- uses: actions/setup-python@e797f83bcb11b83ae66e0230d6156d7c80228e7c # v6.0.0
with:
python-version: "3.12"
- run: echo "::add-matcher::.github/workflows/matchers/actionlint.json"
......
.github/workflows/reminder_comment.yml
View file @ 38d80967
......@@ -9,7 +9,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Remind to run full CI on PR
uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
with:
script: |
try {
......
.github/workflows/stale.yml
View file @ 38d80967
......@@ -13,7 +13,7 @@ jobs:
actions: write
runs-on: ubuntu-latest
steps:
- uses: actions/stale@5bef64f19d7facfb25b37b414482c7164d639639 # v9.1.0
- uses: actions/stale@3a9db7e6a41a89f618792c92c0e97cc736e1b13f # v10.0.0
with:
# Increasing this value ensures that changes to this workflow
# propagate to all issues and PRs in days rather than months
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment