Compare commits

..

1 Commits

Author SHA1 Message Date
79acf80471 Fast decode prepare path for prepare_inputs logic
Signed-off-by: Alexander Matveev <alexm@neuralmagic.com>
2025-05-08 17:26:00 +00:00
2638 changed files with 125763 additions and 312473 deletions

View File

@ -1,20 +1,19 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import os
import sys
import zipfile
# 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 .
# 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 .
# 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', 400))
def print_top_10_largest_files(zip_file):
"""Print the top 10 largest files in the given zip file."""
with zipfile.ZipFile(zip_file, "r") as z:
with zipfile.ZipFile(zip_file, 'r') as z:
file_sizes = [(f, z.getinfo(f).file_size) for f in z.namelist()]
file_sizes.sort(key=lambda x: x[1], reverse=True)
for f, size in file_sizes[:10]:
@ -29,18 +28,14 @@ def check_wheel_size(directory):
wheel_path = os.path.join(root, file_name)
wheel_size_mb = os.path.getsize(wheel_path) / (1024 * 1024)
if wheel_size_mb > VLLM_MAX_SIZE_MB:
print(
f"Not allowed: Wheel {wheel_path} is larger "
f"({wheel_size_mb:.2f} MB) than the limit "
f"({VLLM_MAX_SIZE_MB} MB)."
)
print(f"Not allowed: Wheel {wheel_path} is larger "
f"({wheel_size_mb:.2f} MB) than the limit "
f"({VLLM_MAX_SIZE_MB} MB).")
print_top_10_largest_files(wheel_path)
return 1
else:
print(
f"Wheel {wheel_path} is within the allowed size "
f"({wheel_size_mb:.2f} MB)."
)
print(f"Wheel {wheel_path} is within the allowed size "
f"({wheel_size_mb:.2f} MB).")
return 0
@ -50,4 +45,4 @@ if __name__ == "__main__":
sys.exit(1)
directory = sys.argv[1]
sys.exit(check_wheel_size(directory))
sys.exit(check_wheel_size(directory))

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
import os
@ -8,8 +7,7 @@ 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/>
<a href="../{wheel_html_escaped}">{wheel}</a><br/>
</body>
</html>
"""
@ -22,25 +20,7 @@ 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"),
)
)
template.format(wheel=filename,
wheel_html_escaped=filename.replace("+", "%2B")))

View File

@ -0,0 +1,12 @@
# For vllm script, with -t option (tensor parallel size).
# 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: 0.419
- name: "exact_match,flexible-extract"
value: 0.416
limit: 1000
num_fewshot: 5

View File

@ -1,11 +0,0 @@
# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m RedHatAI/Llama-3.2-1B-Instruct-FP8 -b "auto" -l 1319 -f 5 -t 1
model_name: "RedHatAI/Llama-3.2-1B-Instruct-FP8"
tasks:
- name: "gsm8k"
metrics:
- name: "exact_match,strict-match"
value: 0.335
- name: "exact_match,flexible-extract"
value: 0.323
limit: 1319
num_fewshot: 5

View File

@ -1,11 +0,0 @@
# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m Qwen/Qwen2.5-1.5B-Instruct -b auto -l 1319 -f 5 -t 1
model_name: "Qwen/Qwen2.5-1.5B-Instruct"
tasks:
- name: "gsm8k"
metrics:
- name: "exact_match,strict-match"
value: 0.54
- name: "exact_match,flexible-extract"
value: 0.59
limit: 1319
num_fewshot: 5

View File

@ -1,11 +0,0 @@
# 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
model_name: "RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic"
tasks:
- name: "gsm8k"
metrics:
- name: "exact_match,strict-match"
value: 0.47
- name: "exact_match,flexible-extract"
value: 0.64
limit: 1319
num_fewshot: 5

View File

@ -1,6 +1,10 @@
Qwen2.5-1.5B-Instruct.yaml
Meta-Llama-3-8B-Instruct.yaml
Meta-Llama-3-8B-Instruct-FP8-compressed-tensors.yaml
Meta-Llama-3.2-1B-Instruct-INT8-compressed-tensors.yaml
Meta-Llama-3-8B-Instruct-INT8-compressed-tensors-asym.yaml
Meta-Llama-3-8B-Instruct-nonuniform-compressed-tensors.yaml
Qwen2.5-VL-3B-Instruct-FP8-dynamic.yaml
Meta-Llama-3-8B-Instruct-Channelwise-compressed-tensors.yaml
Qwen1.5-MoE-W4A16-compressed-tensors.yaml
Qwen2-1.5B-Instruct-INT8-compressed-tensors.yaml
Qwen2-1.5B-Instruct-FP8W8.yaml
Meta-Llama-3-8B-QQQ.yaml

View File

@ -1,44 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from pathlib import Path
import pytest
def pytest_addoption(parser):
parser.addoption(
"--config-list-file",
action="store",
help="Path to the file listing model config YAMLs (one per line)",
)
parser.addoption(
"--tp-size",
action="store",
default="1",
help="Tensor parallel size to use for evaluation",
)
@pytest.fixture(scope="session")
def config_list_file(pytestconfig, config_dir):
rel_path = pytestconfig.getoption("--config-list-file")
return config_dir / rel_path
@pytest.fixture(scope="session")
def tp_size(pytestconfig):
return pytestconfig.getoption("--tp-size")
def pytest_generate_tests(metafunc):
if "config_filename" in metafunc.fixturenames:
rel_path = metafunc.config.getoption("--config-list-file")
config_list_file = Path(rel_path).resolve()
config_dir = config_list_file.parent
with open(config_list_file, encoding="utf-8") as f:
configs = [
config_dir / line.strip()
for line in f
if line.strip() and not line.startswith("#")
]
metafunc.parametrize("config_filename", configs)

View File

@ -2,7 +2,7 @@
# We can use this script to compute baseline accuracy on GSM for transformers.
#
# Make sure you have lm-eval-harness installed:
# pip install git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d#egg=lm-eval[api]
# pip install lm-eval==0.4.4
usage() {
echo``

View File

@ -3,7 +3,7 @@
# 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]
# pip install lm-eval==0.4.4
usage() {
echo``
@ -46,6 +46,6 @@ while getopts "m:b:l:f:t:" OPT; do
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" \
--model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE,distributed_executor_backend=ray,trust_remote_code=true,max_model_len=4096" \
--tasks gsm8k --num_fewshot "$FEWSHOT" --limit "$LIMIT" \
--batch_size "$BATCH_SIZE"

View File

@ -0,0 +1,59 @@
#!/bin/bash
usage() {
echo``
echo "Runs lm eval harness on GSM8k using vllm and compares to "
echo "precomputed baseline (measured by HF transformers.)"
echo
echo "usage: ${0} <options>"
echo
echo " -c - path to the test data config (e.g. configs/small-models.txt)"
echo " -t - tensor parallel size"
echo
}
SUCCESS=0
while getopts "c:t:" OPT; do
case ${OPT} in
c )
CONFIG="$OPTARG"
;;
t )
TP_SIZE="$OPTARG"
;;
\? )
usage
exit 1
;;
esac
done
# Parse list of configs.
IFS=$'\n' read -d '' -r -a MODEL_CONFIGS < "$CONFIG"
for MODEL_CONFIG in "${MODEL_CONFIGS[@]}"
do
LOCAL_SUCCESS=0
echo "=== RUNNING MODEL: $MODEL_CONFIG WITH TP SIZE: $TP_SIZE==="
export LM_EVAL_TEST_DATA_FILE=$PWD/configs/${MODEL_CONFIG}
export LM_EVAL_TP_SIZE=$TP_SIZE
pytest -s test_lm_eval_correctness.py || LOCAL_SUCCESS=$?
if [[ $LOCAL_SUCCESS == 0 ]]; then
echo "=== PASSED MODEL: ${MODEL_CONFIG} ==="
else
echo "=== FAILED MODEL: ${MODEL_CONFIG} ==="
fi
SUCCESS=$((SUCCESS + LOCAL_SUCCESS))
done
if [ "${SUCCESS}" -eq "0" ]; then
exit 0
else
exit 1
fi

View File

@ -1,57 +1,69 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
"""
LM eval harness on model to compare vs HF baseline computed offline.
Configs are found in configs/$MODEL.yaml
pytest -s -v test_lm_eval_correctness.py \
--config-list-file=configs/models-small.txt \
--tp-size=1
* export LM_EVAL_TEST_DATA_FILE=configs/Meta-Llama-3-70B-Instruct.yaml
* export LM_EVAL_TP_SIZE=4
* pytest -s test_lm_eval_correctness.py
"""
import os
from pathlib import Path
import lm_eval
import numpy as np
import numpy
import pytest
import yaml
RTOL = 0.08
TEST_DATA_FILE = os.environ.get(
"LM_EVAL_TEST_DATA_FILE",
".buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-Instruct.yaml")
TP_SIZE = os.environ.get("LM_EVAL_TP_SIZE", 1)
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)
model_args = (
f"pretrained={eval_config['model_name']},"
f"tensor_parallel_size={tp_size},"
f"enforce_eager=true,"
f"add_bos_token=true,"
f"trust_remote_code={trust_remote_code},"
f"max_model_len={max_model_len}"
)
def launch_lm_eval(eval_config):
trust_remote_code = eval_config.get('trust_remote_code', False)
model_args = f"pretrained={eval_config['model_name']}," \
f"tensor_parallel_size={TP_SIZE}," \
f"add_bos_token=true," \
f"trust_remote_code={trust_remote_code}"
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",
)
batch_size="auto")
return results
def test_lm_eval_correctness_param(config_filename, tp_size):
eval_config = yaml.safe_load(config_filename.read_text(encoding="utf-8"))
def test_lm_eval_correctness():
eval_config = yaml.safe_load(
Path(TEST_DATA_FILE).read_text(encoding="utf-8"))
results = launch_lm_eval(eval_config, tp_size)
if eval_config[
"model_name"] == "nm-testing/Meta-Llama-3-70B-Instruct-FBGEMM-nonuniform": #noqa: E501
pytest.skip("FBGEMM is currently failing on main.")
# Launch eval requests.
results = launch_lm_eval(eval_config)
# Confirm scores match ground truth.
success = True
for task in eval_config["tasks"]:
for metric in task["metrics"]:
ground_truth = metric["value"]
measured_value = results["results"][task["name"]][metric["name"]]
print(
f"{task['name']} | {metric['name']}: "
f"ground_truth={ground_truth} | measured={measured_value}"
)
success = success and np.isclose(ground_truth, measured_value, rtol=RTOL)
print(f'{task["name"]} | {metric["name"]}: '
f'ground_truth={ground_truth} | measured={measured_value}')
success = success and numpy.isclose(
ground_truth, measured_value, rtol=RTOL)
# Assert at the end, print all scores even on failure for debugging.
assert success

View File

@ -7,11 +7,11 @@ 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.
See [vLLM performance dashboard](https://perf.vllm.ai) 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.
## 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 A100 (the support for FP8 benchmark on H100 is coming!), with different models.
**Benchmarking Duration**: about 1hr.
@ -28,34 +28,16 @@ See [vLLM performance dashboard](https://hud.pytorch.org/benchmark/llms?repoName
## 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
```bash
bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
```
Runtime environment variables:
- `ON_CPU`: set the value to '1' on Intel® Xeon® Processors. Default value is 0.
- `SERVING_JSON`: JSON file to use for the serving tests. Default value is empty string (use default file).
- `LATENCY_JSON`: JSON file to use for the latency tests. Default value is empty string (use default file).
- `THROUGHPUT_JSON`: JSON file to use for the throughout tests. Default value is empty string (use default file).
- `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.
>
### Latency test
Here is an example of one test inside `latency-tests.json`:
@ -78,7 +60,7 @@ Here is an example of one test inside `latency-tests.json`:
In this example:
- The `test_name` attributes is a unique identifier for the test. In `latency-tests.json`, it must start with `latency_`.
- The `parameters` attribute control the command line arguments to be used for `vllm bench latency`. Note that please use underline `_` instead of the dash `-` when specifying the command line arguments, and `run-performance-benchmarks.sh` will convert the underline to dash when feeding the arguments to `vllm bench latency`. For example, the corresponding command line arguments for `vllm bench latency` will be `--model meta-llama/Meta-Llama-3-8B --tensor-parallel-size 1 --load-format dummy --num-iters-warmup 5 --num-iters 15`
- The `parameters` attribute control the command line arguments to be used for `benchmark_latency.py`. Note that please use underline `_` instead of the dash `-` when specifying the command line arguments, and `run-performance-benchmarks.sh` will convert the underline to dash when feeding the arguments to `benchmark_latency.py`. For example, the corresponding command line arguments for `benchmark_latency.py` will be `--model meta-llama/Meta-Llama-3-8B --tensor-parallel-size 1 --load-format dummy --num-iters-warmup 5 --num-iters 15`
Note that the performance numbers are highly sensitive to the value of the parameters. Please make sure the parameters are set correctly.
@ -86,13 +68,13 @@ WARNING: The benchmarking script will save json results by itself, so please do
### Throughput test
The tests are specified in `throughput-tests.json`. The syntax is similar to `latency-tests.json`, except for that the parameters will be fed forward to `vllm bench throughput`.
The tests are specified in `throughput-tests.json`. The syntax is similar to `latency-tests.json`, except for that the parameters will be fed forward to `benchmark_throughput.py`.
The number of this test is also stable -- a slight change on the value of this number might vary the performance numbers by a lot.
### Serving test
We test the throughput by using `vllm bench serve` with request rate = inf to cover the online serving overhead. The corresponding parameters are in `serving-tests.json`, and here is an example:
We test the throughput by using `benchmark_serving.py` with request rate = inf to cover the online serving overhead. The corresponding parameters are in `serving-tests.json`, and here is an example:
```json
[
@ -104,6 +86,7 @@ We test the throughput by using `vllm bench serve` with request rate = inf to co
"tensor_parallel_size": 1,
"swap_space": 16,
"disable_log_stats": "",
"disable_log_requests": "",
"load_format": "dummy"
},
"client_parameters": {
@ -121,8 +104,8 @@ Inside this example:
- The `test_name` attribute is also a unique identifier for the test. It must start with `serving_`.
- The `server-parameters` includes the command line arguments for vLLM server.
- The `client-parameters` includes the command line arguments for `vllm bench serve`.
- The `qps_list` controls the list of qps for test. It will be used to configure the `--request-rate` parameter in `vllm bench serve`
- The `client-parameters` includes the command line arguments for `benchmark_serving.py`.
- The `qps_list` controls the list of qps for test. It will be used to configure the `--request-rate` parameter in `benchmark_serving.py`
The number of this test is less stable compared to the delay and latency benchmarks (due to randomized sharegpt dataset sampling inside `benchmark_serving.py`), but a large change on this number (e.g. 5% change) still vary the output greatly.
@ -130,29 +113,12 @@ WARNING: The benchmarking script will save json results by itself, so please do
### 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.
The `convert-results-json-to-markdown.py` helps you put the benchmarking results inside a markdown table, by formatting [descriptions.md](tests/descriptions.md) with real benchmarking results.
You can find the result presented as a table inside the `buildkite/performance-benchmark` job page.
If you do not see the table, please wait till the benchmark finish running.
The json version of the table (together with the json version of the benchmark) will be also attached to the markdown file.
The raw benchmarking results (in the format of json files) are in the `Artifacts` tab of the benchmarking.
The `compare-json-results.py` helps to compare benchmark results JSON files converted using `convert-results-json-to-markdown.py`.
When run, benchmark script generates results under `benchmark/results` folder, along with the `benchmark_results.md` and `benchmark_results.json`.
`compare-json-results.py` compares two `benchmark_results.json` files and provides performance ratio e.g. for Output Tput, Median TTFT and Median TPOT.
If only one benchmark_results.json is passed, `compare-json-results.py` compares different TP and PP configurations in the benchmark_results.json instead.
Here is an example using the script to compare result_a and result_b with Model, Dataset name, input/output length, max concurrency and qps.
`python3 compare-json-results.py -f results_a/benchmark_results.json -f results_b/benchmark_results.json`
| | Model | Dataset Name | Input Len | Output Len | # of max concurrency | qps | results_a/benchmark_results.json | results_b/benchmark_results.json | perf_ratio |
|----|---------------------------------------|--------|-----|-----|------|-----|-----------|----------|----------|
| 0 | meta-llama/Meta-Llama-3.1-8B-Instruct | random | 128 | 128 | 1000 | 1 | 142.633982 | 156.526018 | 1.097396 |
| 1 | meta-llama/Meta-Llama-3.1-8B-Instruct | random | 128 | 128 | 1000 | inf| 241.620334 | 294.018783 | 1.216863 |
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.
@ -160,9 +126,9 @@ See [nightly-descriptions.md](nightly-descriptions.md) for the detailed descript
### 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.
- Inside each container, we run [run-nightly-suite.sh](run-nightly-suite.sh), which will probe the serving engine of the current container.
- The `run-nightly-suite.sh` will redirect the request to `tests/run-[llm serving engine name]-nightly.sh`, which parses the workload described in [nightly-tests.json](tests/nightly-tests.json) and performs the benchmark.
- At last, we run [scripts/plot-nightly-results.py](scripts/plot-nightly-results.py) to collect and plot the final benchmarking results, and update the results to buildkite.
### Nightly tests
@ -172,6 +138,6 @@ In [nightly-tests.json](tests/nightly-tests.json), we include the command line a
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: 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 `tests/run-[llm serving engine name]-nightly.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).

View File

@ -1,4 +1,3 @@
# Nightly benchmark annotation
## Description
@ -14,15 +13,15 @@ Please download the visualization scripts in the post
- 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:
- 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
```
```console
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`.

View File

@ -13,25 +13,25 @@ Latest reproduction guilde: [github issue link](https://github.com/vllm-project/
## 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.
- 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 uses 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
- 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).
- 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

View File

@ -1,12 +1,10 @@
# Performance benchmarks descriptions
## Latency tests
- Input length: 32 tokens.
- Output length: 128 tokens.
- Batch size: fixed (8).
- GPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- CPU Models: llama-3.1 8B.
- Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- Evaluation metrics: end-to-end latency (mean, median, p99).
{latency_tests_markdown_table}
@ -16,8 +14,7 @@
- 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.
- CPU Models: llama-3.1 8B.
- Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- Evaluation metrics: throughput.
{throughput_tests_markdown_table}
@ -28,18 +25,12 @@
- 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.
- We also added a speculative decoding test for llama-3 70B on GPU, under QPS 2
- CPU Models: llama-3.1 8B.
- Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
- We also added a speculative decoding test for llama-3 70B, under QPS 2
- Evaluation metrics: throughput, TTFT (time to the first token, with mean, median and p99), ITL (inter-token latency, with mean, median and p99).
- For CPU, we added random dataset tests to benchmark fixed input/output length with 100 prompts.
{serving_tests_markdown_table}
## Platform Information
{platform_markdown_table}
## json version of the benchmarking tables
This section contains the data of the markdown tables above in JSON format.

View File

@ -1,307 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
import json
import os
from importlib import util
import pandas as pd
plotly_found = util.find_spec("plotly.express") is not None
def compare_data_columns(
files, name_column, data_column, info_cols, drop_column, debug=False
):
"""
Align concatenation by keys derived from info_cols instead of row order.
- Pick one canonical key list: subset of info_cols present in ALL files.
- For each file: set index to those keys, aggregate duplicates
- (mean for metric, first for names).
- Concat along axis=1 (indexes align), then reset_index so callers can
- group by columns.
- If --debug, add a <file_label>_name column per file.
"""
print("\ncompare_data_column:", data_column)
frames = []
raw_data_cols = []
compare_frames = []
# 1) choose a canonical key list from info_cols that exists in ALL files
cols_per_file = []
for f in files:
try:
df_tmp = pd.read_json(f, orient="records")
except Exception as err:
raise ValueError(f"Failed to read {f}") from err
cols_per_file.append(set(df_tmp.columns))
key_cols = [c for c in info_cols if all(c in cset for cset in cols_per_file)]
if not key_cols:
# soft fallback: use any info_cols present in the first file
key_cols = [c for c in info_cols if c in list(cols_per_file[0])]
if not key_cols:
raise ValueError(
"No common key columns found from info_cols across the input files."
)
# 2) build a single "meta" block (keys as columns) once, aligned by the key index
meta_added = False
for file in files:
df = pd.read_json(file, orient="records")
# Keep rows that actually have the compared metric (same as original behavior)
if drop_column in df.columns:
df = df.dropna(subset=[drop_column], ignore_index=True)
# Stabilize numeric key columns (harmless if missing)
for c in (
"Input Len",
"Output Len",
"TP Size",
"PP Size",
"# of max concurrency.",
"qps",
):
if c in df.columns:
df[c] = pd.to_numeric(df[c], errors="coerce")
# Ensure all key columns exist
for c in key_cols:
if c not in df.columns:
df[c] = pd.NA
# Set index = key_cols and aggregate duplicates → unique MultiIndex
df_idx = df.set_index(key_cols, drop=False)
# meta (key columns), unique per key
meta = df_idx[key_cols]
if not meta.index.is_unique:
meta = meta.groupby(level=key_cols, dropna=False).first()
# metric series for this file, aggregated to one row per key
file_label = "/".join(file.split("/")[:-1]) or os.path.basename(file)
s = df_idx[data_column]
if not s.index.is_unique:
s = s.groupby(level=key_cols, dropna=False).mean()
s.name = file_label # column label like original
# add meta once (from first file) so keys are the leftmost columns
if not meta_added:
frames.append(meta)
meta_added = True
# (NEW) debug: aligned test-name column per file
if debug and name_column in df_idx.columns:
name_s = df_idx[name_column]
if not name_s.index.is_unique:
name_s = name_s.groupby(level=key_cols, dropna=False).first()
name_s.name = f"{file_label}_name"
frames.append(name_s)
frames.append(s)
raw_data_cols.append(file_label)
compare_frames.append(s)
# Generalize ratio: for any file N>=2, add ratio (fileN / file1)
if len(compare_frames) >= 2:
base = compare_frames[0]
current = compare_frames[-1]
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)
# 4) concat on columns with aligned MultiIndex;
# then reset_index to return keys as columns
concat_df = pd.concat(frames, axis=1)
concat_df = concat_df.reset_index(drop=True).reset_index()
if "index" in concat_df.columns:
concat_df = concat_df.drop(columns=["index"])
# Ensure key/info columns appear first (in your info_cols order)
front = [c for c in info_cols if c in concat_df.columns]
rest = [c for c in concat_df.columns if c not in front]
concat_df = concat_df[front + rest]
print(raw_data_cols)
return concat_df, raw_data_cols
def split_json_by_tp_pp(
input_file: str = "benchmark_results.json", output_root: str = "."
) -> list[str]:
"""
Split a benchmark JSON into separate folders by (TP Size, PP Size).
Creates: <output_root>/tp{TP}_pp{PP}/benchmark_results.json
Returns: list of file paths written.
"""
# Load JSON data into DataFrame
with open(input_file, encoding="utf-8") as f:
data = json.load(f)
# If the JSON is a dict with a list under common keys, use that list
if isinstance(data, dict):
for key in ("results", "serving_results", "benchmarks", "data"):
if isinstance(data.get(key), list):
data = data[key]
break
df = pd.DataFrame(data)
# Keep only "serving" tests
name_col = next(
(c for c in ["Test name", "test_name", "Test Name"] if c in df.columns), None
)
if name_col:
df = df[
df[name_col].astype(str).str.contains(r"serving", case=False, na=False)
].copy()
# Handle alias column names
rename_map = {
"tp_size": "TP Size",
"tensor_parallel_size": "TP Size",
"pp_size": "PP Size",
"pipeline_parallel_size": "PP Size",
}
df.rename(
columns={k: v for k, v in rename_map.items() if k in df.columns}, inplace=True
)
# Ensure TP/PP columns exist (default to 1 if missing)
if "TP Size" not in df.columns:
df["TP Size"] = 1
if "PP Size" not in df.columns:
df["PP Size"] = 1
# make sure TP/PP are numeric ints with no NaN
df["TP Size"] = (
pd.to_numeric(df.get("TP Size", 1), errors="coerce").fillna(1).astype(int)
)
df["PP Size"] = (
pd.to_numeric(df.get("PP Size", 1), errors="coerce").fillna(1).astype(int)
)
# Split into separate folders
saved_paths: list[str] = []
for (tp, pp), group_df in df.groupby(["TP Size", "PP Size"], dropna=False):
folder_name = os.path.join(output_root, f"tp{int(tp)}_pp{int(pp)}")
os.makedirs(folder_name, exist_ok=True)
filepath = os.path.join(folder_name, "benchmark_results.json")
group_df.to_json(filepath, orient="records", indent=2, force_ascii=False)
print(f"Saved: {filepath}")
saved_paths.append(filepath)
return saved_paths
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"-f", "--file", action="append", type=str, help="input file name"
)
parser.add_argument(
"--debug", action="store_true", help="show all information for debugging"
)
parser.add_argument(
"--plot",
action=argparse.BooleanOptionalAction,
default=True,
help="plot perf diagrams or not --no-plot --plot",
)
parser.add_argument(
"-x",
"--xaxis",
type=str,
default="# of max concurrency.",
help="column name to use as X Axis in comparison graph",
)
args = parser.parse_args()
drop_column = "P99"
name_column = "Test name"
info_cols = [
"Model",
"Dataset Name",
"Input Len",
"Output Len",
"TP Size",
"PP Size",
"# 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 len(args.file) == 1:
files = split_json_by_tp_pp(args.file[0], output_root="splits")
info_cols = [c for c in info_cols if c not in ("TP Size", "PP Size")]
else:
files = args.file
print("comparing : " + ", ".join(files))
debug = args.debug
plot = args.plot
# For Plot feature, assign y axis from one of info_cols
y_axis_index = info_cols.index(args.xaxis) if args.xaxis in info_cols else 6
with open("perf_comparison.html", "w") as text_file:
for i in range(len(data_cols_to_compare)):
output_df, raw_data_cols = compare_data_columns(
files,
name_column,
data_cols_to_compare[i],
info_cols,
drop_column,
debug=debug,
)
# For Plot feature, insert y axis from one of info_cols
raw_data_cols.insert(0, info_cols[y_axis_index])
filtered_info_cols = info_cols[:-2]
existing_group_cols = [
c for c in filtered_info_cols if c in output_df.columns
]
if not existing_group_cols:
raise ValueError(
f"No valid group-by columns "
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_groups = output_df_sorted.groupby(existing_group_cols, dropna=False)
for name, group in output_groups:
html = group.to_html()
text_file.write(html_msgs_for_data_cols[i])
text_file.write(html)
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"))

View File

@ -1,19 +1,14 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
import json
import os
import shlex
from importlib import util
from pathlib import Path
from typing import Any
import pandas as pd
import psutil
import regex as re
from tabulate import tabulate
results_folder = Path("results/")
# latency results and the keys that will be printed into markdown
latency_results = []
latency_column_mapping = {
@ -33,39 +28,28 @@ throughput_results = []
throughput_results_column_mapping = {
"test_name": "Test name",
"gpu_type": "GPU",
"num_requests": "# of req.",
"total_num_tokens": "Total # of tokens",
"elapsed_time": "Elapsed time (s)",
# "num_requests": "# of req.",
# "total_num_tokens": "Total # of tokens",
# "elapsed_time": "Elapsed time (s)",
"requests_per_second": "Tput (req/s)",
"tokens_per_second": "Tput (tok/s)",
# "tokens_per_second": "Tput (tok/s)",
}
# serving results and the keys that will be printed into markdown
serving_results = []
serving_column_mapping = {
"test_name": "Test name",
"model_id": "Model",
"dataset_name": "Dataset Name",
"input_len": "Input Len",
"output_len": "Output Len",
"tp_size": "TP Size",
"pp_size": "PP Size",
"dtype": "dtype",
"gpu_type": "GPU",
"completed": "# of req.",
"qps": "qps",
"max_concurrency": "# of max concurrency.",
# "completed": "# of req.",
"request_throughput": "Tput (req/s)",
"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",
# "input_throughput": "Input Tput (tok/s)",
# "output_throughput": "Output Tput (tok/s)",
"mean_ttft_ms": "Mean TTFT (ms)",
"median_ttft_ms": "Median TTFT (ms)",
"p99_ttft_ms": "P99 TTFT (ms)",
"mean_tpot_ms": "Mean TPOT (ms)",
"median_tpot_ms": "Median",
"p99_tpot_ms": "P99",
# "mean_tpot_ms": "Mean TPOT (ms)",
# "median_tpot_ms": "Median",
# "p99_tpot_ms": "P99",
"mean_itl_ms": "Mean ITL (ms)",
"median_itl_ms": "Median ITL (ms)",
"p99_itl_ms": "P99 ITL (ms)",
@ -81,134 +65,24 @@ def read_markdown(file):
def results_to_json(latency, throughput, serving):
return json.dumps(
{
"latency": latency.to_dict(),
"throughput": throughput.to_dict(),
"serving": serving.to_dict(),
}
)
def get_size_with_unit(bytes, suffix="B"):
"""
Scale bytes to its proper format
e.g:
1253656 => '1.20MB'
1253656678 => '1.17GB'
"""
factor = 1024
for unit in ["", "K", "M", "G", "T", "P"]:
if bytes < factor:
return f"{bytes:.2f}{unit}{suffix}"
bytes /= factor
def _coerce(val: str) -> Any:
"""Best-effort type coercion from string to Python types."""
low = val.lower()
if low == "null":
return None
if low == "true":
return True
if low == "false":
return False
# integers
if re.fullmatch(r"[+-]?\d+", val):
try:
return int(val)
except ValueError:
pass
# floats (keep 'inf'/'-inf'/'nan' as strings)
if re.fullmatch(r"[+-]?\d*\.\d+", val):
try:
return float(val)
except ValueError:
pass
return val
def parse_client_command(cmd: str) -> dict[str, Any]:
"""Parse the client_command shell string into {executable, script, args}."""
toks = shlex.split(cmd)
if len(toks) < 2:
raise ValueError("client_command must include an executable and a script")
executable, script = toks[0], toks[1]
args: dict[str, Any] = {}
i = 2
while i < len(toks):
t = toks[i]
if t.startswith("--"):
# --key=value or --key (value) or boolean flag
if "=" in t:
key, val = t.split("=", 1)
if key == "--metadata":
md = {}
if val:
if "=" in val:
k, v = val.split("=", 1)
md[k] = _coerce(v)
else:
md[val] = True
args[key] = md
else:
args[key] = _coerce(val)
i += 1
continue
key = t
# Special: consume metadata k=v pairs until next --flag
if key == "--metadata":
i += 1
md = {}
while i < len(toks) and not toks[i].startswith("--"):
pair = toks[i]
if "=" in pair:
k, v = pair.split("=", 1)
md[k] = _coerce(v)
else:
md[pair] = True
i += 1
args[key] = md
continue
# Standard: check if next token is a value (not a flag)
if i + 1 < len(toks) and not toks[i + 1].startswith("--"):
args[key] = _coerce(toks[i + 1])
i += 2
else:
# lone flag -> True
args[key] = True
i += 1
else:
# unexpected positional; skip
i += 1
return {"executable": executable, "script": script, "args": args}
return json.dumps({
'latency': latency.to_dict(),
'throughput': throughput.to_dict(),
'serving': serving.to_dict()
})
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"-r",
"--result",
type=str,
default="results",
help="Folder name for benchmark output results.",
)
args = parser.parse_args()
results_folder = Path(args.result)
if not results_folder.exists():
raise FileNotFoundError(f"results folder does not exist: {results_folder}")
# collect results
for test_file in results_folder.glob("*.json"):
with open(test_file) as f:
raw_result = json.loads(f.read())
if "serving" in str(test_file):
# this result is generated via `vllm bench serve` command
# this result is generated via `benchmark_serving.py`
# attach the benchmarking command to raw_result
try:
with open(test_file.with_suffix(".commands")) as f:
@ -216,50 +90,18 @@ if __name__ == "__main__":
except OSError as e:
print(e)
continue
# Parse Server Command Arg
out: dict[str, Any] = {
"server_command": parse_client_command(command["server_command"])
}
parse_args = [
"--tensor-parallel-size",
"--pipeline-parallel-size",
"--dtype",
]
col_mapping = ["tp_size", "pp_size", "dtype"]
for index, arg in enumerate(parse_args):
if arg in out["server_command"]["args"]:
raw_result.update(
{col_mapping[index]: out["server_command"]["args"][arg]}
)
# Parse Client Command Arg
out: dict[str, Any] = {
"client_command": parse_client_command(command["client_command"])
}
parse_args = [
"--dataset-name",
"--random-input-len",
"--random-output-len",
"--request-rate",
]
col_mapping = ["dataset_name", "input_len", "output_len", "qps"]
for index, arg in enumerate(parse_args):
if arg in out["client_command"]["args"]:
raw_result.update(
{col_mapping[index]: out["client_command"]["args"][arg]}
)
# Add Server, Client command
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
elif "latency" in f.name:
# this result is generated via `vllm bench latency` command
# this result is generated via `benchmark_latency.py`
# attach the benchmarking command to raw_result
try:
@ -278,8 +120,7 @@ if __name__ == "__main__":
for perc in [10, 25, 50, 75, 90, 99]:
# Multiply 1000 to convert the time unit from s to ms
raw_result.update(
{f"P{perc}": 1000 * raw_result["percentiles"][str(perc)]}
)
{f"P{perc}": 1000 * raw_result["percentiles"][str(perc)]})
raw_result["avg_latency"] = raw_result["avg_latency"] * 1000
# add the result to raw_result
@ -287,7 +128,7 @@ if __name__ == "__main__":
continue
elif "throughput" in f.name:
# this result is generated via `vllm bench throughput` command
# this result is generated via `benchmark_throughput.py`
# attach the benchmarking command to raw_result
try:
@ -312,51 +153,26 @@ if __name__ == "__main__":
serving_results = pd.DataFrame.from_dict(serving_results)
throughput_results = pd.DataFrame.from_dict(throughput_results)
svmem = psutil.virtual_memory()
platform_data = {
"Physical cores": [psutil.cpu_count(logical=False)],
"Total cores": [psutil.cpu_count(logical=True)],
"Total Memory": [get_size_with_unit(svmem.total)],
}
if util.find_spec("numa") is not None:
from numa import info
platform_data["Total NUMA nodes"] = [info.get_num_configured_nodes()]
if util.find_spec("cpuinfo") is not None:
from cpuinfo import get_cpu_info
platform_data["CPU Brand"] = [get_cpu_info()["brand_raw"]]
platform_results = pd.DataFrame.from_dict(
platform_data, orient="index", columns=["Platform Info"]
)
raw_results_json = results_to_json(
latency_results, throughput_results, serving_results
)
raw_results_json = results_to_json(latency_results, throughput_results,
serving_results)
# remapping the key, for visualization purpose
if not latency_results.empty:
latency_results = latency_results[list(latency_column_mapping.keys())].rename(
columns=latency_column_mapping
)
latency_results = latency_results[list(
latency_column_mapping.keys())].rename(
columns=latency_column_mapping)
if not serving_results.empty:
valid_columns = [
col for col in serving_column_mapping if col in serving_results.columns
]
serving_results = serving_results[valid_columns].rename(
columns=serving_column_mapping
)
serving_results = serving_results[list(
serving_column_mapping.keys())].rename(
columns=serving_column_mapping)
if not throughput_results.empty:
throughput_results = throughput_results[
list(throughput_results_column_mapping.keys())
].rename(columns=throughput_results_column_mapping)
throughput_results = throughput_results[list(
throughput_results_column_mapping.keys())].rename(
columns=throughput_results_column_mapping)
processed_results_json = results_to_json(
latency_results, throughput_results, serving_results
)
processed_results_json = results_to_json(latency_results,
throughput_results,
serving_results)
for df in [latency_results, serving_results, throughput_results]:
if df.empty:
@ -368,45 +184,38 @@ 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: f"{len(x.split('\n'))}x{x.split('\n')[0]}")
# get markdown tables
latency_md_table = tabulate(
latency_results, headers="keys", tablefmt="pipe", showindex=False
)
serving_md_table = tabulate(
serving_results, headers="keys", tablefmt="pipe", showindex=False
)
throughput_md_table = tabulate(
throughput_results, headers="keys", tablefmt="pipe", showindex=False
)
platform_md_table = tabulate(
platform_results, headers="keys", tablefmt="pipe", showindex=True
)
latency_md_table = tabulate(latency_results,
headers='keys',
tablefmt='pipe',
showindex=False)
serving_md_table = tabulate(serving_results,
headers='keys',
tablefmt='pipe',
showindex=False)
throughput_md_table = tabulate(throughput_results,
headers='keys',
tablefmt='pipe',
showindex=False)
# document the result
md_file = "benchmark_results.md"
json_file = "benchmark_results.json"
with open(results_folder / md_file, "w") as f:
results = read_markdown(
"../.buildkite/nightly-benchmarks/"
+ "performance-benchmarks-descriptions.md"
)
with open(results_folder / "benchmark_results.md", "w") as f:
results = read_markdown("../.buildkite/nightly-benchmarks/" +
"performance-benchmarks-descriptions.md")
results = results.format(
latency_tests_markdown_table=latency_md_table,
throughput_tests_markdown_table=throughput_md_table,
serving_tests_markdown_table=serving_md_table,
platform_markdown_table=platform_md_table,
benchmarking_results_in_json_string=processed_results_json,
)
benchmarking_results_in_json_string=processed_results_json)
f.write(results)
# document benchmarking results in json
with open(results_folder / json_file, "w") as f:
results = (
latency_results.to_dict(orient="records")
+ throughput_results.to_dict(orient="records")
+ serving_results.to_dict(orient="records")
)
with open(results_folder / "benchmark_results.json", "w") as f:
results = latency_results.to_dict(
orient='records') + throughput_results.to_dict(
orient='records') + serving_results.to_dict(orient='records')
f.write(json.dumps(results))

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
@ -15,12 +14,15 @@ def main(model, 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"
)
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)

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
import json
@ -12,33 +11,33 @@ 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."
)
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))
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)
means.append(0.)
else:
means.append(filtered_df[metric].values[0])
@ -46,6 +45,7 @@ def get_perf(df, method, model, metric):
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()
@ -60,8 +60,7 @@ def get_perf_w_std(df, method, model, metric):
else:
assert metric == "Tput"
mean = get_perf(df, method, model, "Input Tput (tok/s)") + get_perf(
df, method, model, "Output Tput (tok/s)"
)
df, method, model, "Output Tput (tok/s)")
mean = mean.tolist()
std = None
@ -81,17 +80,18 @@ def main(args):
# generate markdown table
df = pd.DataFrame.from_dict(results)
md_table = tabulate(df, headers="keys", tablefmt="pipe", showindex=False)
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)
description = description.format(
nightly_results_benchmarking_table=md_table)
with open("nightly_results.md", "w") as f:
f.write(description)
if __name__ == "__main__":
if __name__ == '__main__':
args = parse_arguments()
main(args)

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from lmdeploy.serve.openai.api_client import APIClient

View File

@ -73,7 +73,7 @@ get_current_llm_serving_engine() {
echo "Container: vllm"
# move to a completely irrelevant directory, to avoid import vllm from current folder
export CURRENT_LLM_SERVING_ENGINE=vllm
return
fi
}
@ -95,14 +95,12 @@ json2args() {
}
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'
pkill -f python
pkill -f python3
pkill -f tritonserver
pkill -f pt_main_thread
pkill -f text-generation
pkill -f lmdeploy
while [ "$(nvidia-smi --query-gpu=memory.used --format=csv,noheader,nounits | head -n 1)" -ge 1000 ]; do
sleep 1
@ -127,7 +125,7 @@ ensure_installed() {
}
run_serving_tests() {
# run serving tests using `vllm bench serve` command
# run serving tests using `benchmark_serving.py`
# $1: a json file specifying serving test cases
local serving_test_file
@ -227,7 +225,7 @@ run_serving_tests() {
if [[ "$dataset_name" = "sharegpt" ]]; then
client_command="vllm bench serve \
client_command="python3 benchmark_serving.py \
--backend $backend \
--tokenizer /tokenizer_cache \
--model $model \
@ -248,7 +246,7 @@ run_serving_tests() {
sonnet_output_len=$(echo "$common_params" | jq -r '.sonnet_output_len')
sonnet_prefix_len=$(echo "$common_params" | jq -r '.sonnet_prefix_len')
client_command="vllm bench serve \
client_command="python3 benchmark_serving.py \
--backend $backend \
--tokenizer /tokenizer_cache \
--model $model \
@ -267,13 +265,13 @@ run_serving_tests() {
$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"
@ -304,7 +302,7 @@ run_serving_tests() {
}
run_genai_perf_tests() {
# run genai-perf tests
# run genai-perf tests
# $1: a json file specifying genai-perf test cases
local genai_perf_test_file
@ -313,14 +311,14 @@ run_genai_perf_tests() {
# Iterate over genai-perf tests
jq -c '.[]' "$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')
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}_${test_name}
@ -371,10 +369,10 @@ run_genai_perf_tests() {
qps=$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
@ -382,7 +380,7 @@ run_genai_perf_tests() {
client_command="genai-perf profile \
-m $model \
--service-kind openai \
--backend "$backend" \
--backend vllm \
--endpoint-type chat \
--streaming \
--url localhost:$port \
@ -415,7 +413,7 @@ prepare_dataset() {
do
cat sonnet.txt >> sonnet_4x.txt
done
}
main() {

View File

@ -31,20 +31,6 @@ check_gpus() {
echo "GPU type is $gpu_type"
}
check_cpus() {
# check the number of CPUs and NUMA Node and GPU type.
declare -g numa_count=$(lscpu | grep "NUMA node(s):" | awk '{print $3}')
if [[ $numa_count -gt 0 ]]; then
echo "NUMA found."
echo $numa_count
else
echo "Need at least 1 NUMA to run benchmarking."
exit 1
fi
declare -g gpu_type="cpu"
echo "GPU type is $gpu_type"
}
check_hf_token() {
# check if HF_TOKEN is available and valid
if [[ -z "$HF_TOKEN" ]]; then
@ -83,22 +69,6 @@ json2args() {
echo "$args"
}
json2envs() {
# transforms the JSON string to environment variables.
# example:
# input: { "VLLM_CPU_KVCACHE_SPACE": 5 }
# output: VLLM_CPU_KVCACHE_SPACE=5
local json_string=$1
local args=$(
echo "$json_string" | jq -r '
to_entries |
map((.key ) + "=" + (.value | tostring)) |
join(" ")
'
)
echo "$args"
}
wait_for_server() {
# wait for vllm server to start
# return 1 if vllm server crashes
@ -126,8 +96,7 @@ kill_gpu_processes() {
ps -aux
lsof -t -i:8000 | xargs -r kill -9
pgrep python3 | xargs -r kill -9
# vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
pgrep VLLM | xargs -r kill -9
# wait until GPU memory usage smaller than 1GB
if command -v nvidia-smi; then
@ -165,7 +134,7 @@ upload_to_buildkite() {
}
run_latency_tests() {
# run latency tests using `vllm bench latency` command
# run latency tests using `benchmark_latency.py`
# $1: a json file specifying latency test cases
local latency_test_file
@ -189,26 +158,15 @@ run_latency_tests() {
# get arguments
latency_params=$(echo "$params" | jq -r '.parameters')
latency_args=$(json2args "$latency_params")
latency_environment_variables=$(echo "$params" | jq -r '.environment_variables')
latency_envs=$(json2envs "$latency_environment_variables")
# check if there is enough GPU to run the test
tp=$(echo "$latency_params" | jq -r '.tensor_parallel_size')
if [ "$ON_CPU" == "1" ]; then
pp=$(echo "$latency_params" | jq -r '.pipeline_parallel_size')
world_size=$(($tp*$pp))
if [[ $numa_count -lt $world_size && -z "${REMOTE_HOST}" ]]; then
echo "Required world-size $world_size but only $numa_count NUMA nodes found. Skip testcase $test_name."
continue
fi
else
if [[ $gpu_count -lt $tp ]]; then
echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
continue
fi
if [[ $gpu_count -lt $tp ]]; then
echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
continue
fi
latency_command=" $latency_envs vllm bench latency \
latency_command="python3 benchmark_latency.py \
--output-json $RESULTS_FOLDER/${test_name}.json \
$latency_args"
@ -234,7 +192,7 @@ run_latency_tests() {
}
run_throughput_tests() {
# run throughput tests using `vllm bench throughput`
# run throughput tests using `benchmark_throughput.py`
# $1: a json file specifying throughput test cases
local throughput_test_file
@ -258,26 +216,15 @@ run_throughput_tests() {
# get arguments
throughput_params=$(echo "$params" | jq -r '.parameters')
throughput_args=$(json2args "$throughput_params")
throughput_environment_variables=$(echo "$params" | jq -r '.environment_variables')
throughput_envs=$(json2envs "$throughput_environment_variables")
# check if there is enough GPU to run the test
tp=$(echo "$throughput_params" | jq -r '.tensor_parallel_size')
if [ "$ON_CPU" == "1" ]; then
pp=$(echo "$throughput_params" | jq -r '.pipeline_parallel_size')
world_size=$(($tp*$pp))
if [[ $numa_count -lt $world_size && -z "${REMOTE_HOST}" ]]; then
echo "Required world-size $world_size but only $numa_count NUMA nodes found. Skip testcase $test_name."
continue
fi
else
if [[ $gpu_count -lt $tp ]]; then
echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
continue
fi
if [[ $gpu_count -lt $tp ]]; then
echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
continue
fi
throughput_command=" $throughput_envs vllm bench throughput \
throughput_command="python3 benchmark_throughput.py \
--output-json $RESULTS_FOLDER/${test_name}.json \
$throughput_args"
@ -302,7 +249,7 @@ run_throughput_tests() {
}
run_serving_tests() {
# run serving tests using `vllm bench serve` command
# run serving tests using `benchmark_serving.py`
# $1: a json file specifying serving test cases
local serving_test_file
@ -325,36 +272,18 @@ run_serving_tests() {
# get client and server arguments
server_params=$(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 "$client_params")
qps_list=$(echo "$params" | jq -r '.qps_list')
qps_list=$(echo "$qps_list" | jq -r '.[] | @sh')
echo "Running over qps list $qps_list"
max_concurrency_list=$(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]"
fi
max_concurrency_list=$(echo "$max_concurrency_list" | jq -r '.[] | @sh')
echo "Running over max concurrency list $max_concurrency_list"
# check if there is enough resources to run the test
# check if there is enough GPU to run the test
tp=$(echo "$server_params" | jq -r '.tensor_parallel_size')
if [ "$ON_CPU" == "1" ]; then
pp=$(echo "$server_params" | jq -r '.pipeline_parallel_size')
world_size=$(($tp*$pp))
if [[ $numa_count -lt $world_size && -z "${REMOTE_HOST}" ]]; then
echo "Required world-size $world_size but only $numa_count NUMA nodes found. Skip testcase $test_name."
continue
fi
else
if [[ $gpu_count -lt $tp ]]; then
echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
continue
fi
if [[ $gpu_count -lt $tp ]]; then
echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
continue
fi
# check if server model and client model is aligned
@ -365,33 +294,23 @@ run_serving_tests() {
continue
fi
server_command="$server_envs python3 \
server_command="python3 \
-m vllm.entrypoints.openai.api_server \
$server_args"
# run the server
echo "Running test case $test_name"
echo "Server command: $server_command"
# support remote vllm server
client_remote_args=""
if [[ -z "${REMOTE_HOST}" ]]; then
bash -c "$server_command" &
server_pid=$!
# wait until the server is alive
if wait_for_server; then
echo ""
echo "vLLM server is up and running."
else
echo ""
echo "vLLM failed to start within the timeout period."
fi
bash -c "$server_command" &
server_pid=$!
# wait until the server is alive
if wait_for_server; then
echo ""
echo "vllm server is up and running."
else
server_command="Using Remote Server $REMOTE_HOST $REMOTE_PORT"
if [[ ${REMOTE_PORT} ]]; then
client_remote_args=" --host=$REMOTE_HOST --port=$REMOTE_PORT "
else
client_remote_args=" --host=$REMOTE_HOST "
fi
echo ""
echo "vllm failed to start within the timeout period."
fi
# iterate over different QPS
@ -403,39 +322,35 @@ run_serving_tests() {
echo "now qps is $qps"
fi
# iterate over different max_concurrency
for max_concurrency in $max_concurrency_list; do
new_test_name=$test_name"_qps_"$qps"_concurrency_"$max_concurrency
echo " new test name $new_test_name"
# pass the tensor parallel size to the client so that it can be displayed
# on the benchmark dashboard
client_command="vllm bench serve \
--save-result \
--result-dir $RESULTS_FOLDER \
--result-filename ${new_test_name}.json \
--request-rate $qps \
--max-concurrency $max_concurrency \
--metadata "tensor_parallel_size=$tp" \
$client_args $client_remote_args "
new_test_name=$test_name"_qps_"$qps
echo "Running test case $test_name with qps $qps"
echo "Client command: $client_command"
# pass the tensor parallel size to the client so that it can be displayed
# on the benchmark dashboard
client_command="python3 benchmark_serving.py \
--save-result \
--result-dir $RESULTS_FOLDER \
--result-filename ${new_test_name}.json \
--request-rate $qps \
--metadata "tensor_parallel_size=$tp" \
$client_args"
bash -c "$client_command"
echo "Running test case $test_name with qps $qps"
echo "Client command: $client_command"
# record the benchmarking commands
jq_output=$(jq -n \
--arg server "$server_command" \
--arg client "$client_command" \
--arg gpu "$gpu_type" \
'{
server_command: $server,
client_command: $client,
gpu_type: $gpu
}')
echo "$jq_output" >"$RESULTS_FOLDER/${new_test_name}.commands"
bash -c "$client_command"
# record the benchmarking commands
jq_output=$(jq -n \
--arg server "$server_command" \
--arg client "$client_command" \
--arg gpu "$gpu_type" \
'{
server_command: $server,
client_command: $client,
gpu_type: $gpu
}')
echo "$jq_output" >"$RESULTS_FOLDER/${new_test_name}.commands"
done
done
# clean up
@ -445,14 +360,7 @@ run_serving_tests() {
}
main() {
local ARCH
ARCH=''
if [ "$ON_CPU" == "1" ];then
check_cpus
ARCH='-cpu'
else
check_gpus
fi
check_gpus
check_hf_token
# Set to v1 to run v1 benchmark
@ -465,7 +373,7 @@ main() {
(which jq) || (apt-get update && apt-get -y install jq)
(which lsof) || (apt-get update && apt-get install -y lsof)
# get the current IP address, required by `vllm bench serve` command
# get the current IP address, required by benchmark_serving.py
export VLLM_HOST_IP=$(hostname -I | awk '{print $1}')
# turn of the reporting of the status of each request, to clean up the terminal output
export VLLM_LOGGING_LEVEL="WARNING"
@ -478,9 +386,9 @@ main() {
QUICK_BENCHMARK_ROOT=../.buildkite/nightly-benchmarks/
# benchmarking
run_serving_tests $QUICK_BENCHMARK_ROOT/tests/"${SERVING_JSON:-serving-tests$ARCH.json}"
run_latency_tests $QUICK_BENCHMARK_ROOT/tests/"${LATENCY_JSON:-latency-tests$ARCH.json}"
run_throughput_tests $QUICK_BENCHMARK_ROOT/tests/"${THROUGHPUT_JSON:-throughput-tests$ARCH.json}"
run_serving_tests $QUICK_BENCHMARK_ROOT/tests/serving-tests.json
run_latency_tests $QUICK_BENCHMARK_ROOT/tests/latency-tests.json
run_throughput_tests $QUICK_BENCHMARK_ROOT/tests/throughput-tests.json
# postprocess benchmarking results
pip install tabulate pandas

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import datetime
import json
@ -35,8 +34,10 @@ serving_column_mapping = {
}
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())
@ -55,16 +56,17 @@ if __name__ == "__main__":
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_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
)
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:])
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")
@ -74,9 +76,10 @@ if __name__ == "__main__":
# document results with header.
# for those who wants to reproduce our benchmark.
f.write(serving_md_table_with_headers)
f.write("\n")
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")
results = serving_results.to_dict(orient='records')
f.write(json.dumps(results))

View File

@ -11,7 +11,9 @@
},
"vllm_server_parameters": {
"disable_log_stats": "",
"disable_log_requests": "",
"gpu_memory_utilization": 0.9,
"num_scheduler_steps": 10,
"max_num_seqs": 512,
"dtype": "bfloat16"
},

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
}
}
]

View File

@ -35,7 +35,9 @@
},
"vllm_server_parameters": {
"disable_log_stats": "",
"disable_log_requests": "",
"gpu_memory_utilization": 0.9,
"num_scheduler_steps": 10,
"max_num_seqs": 512,
"dtype": "bfloat16"
},
@ -88,7 +90,9 @@
},
"vllm_server_parameters": {
"disable_log_stats": "",
"disable_log_requests": "",
"gpu_memory_utilization": 0.9,
"num_scheduler_steps": 10,
"max_num_seqs": 512,
"dtype": "bfloat16"
},
@ -141,7 +145,9 @@
},
"vllm_server_parameters": {
"disable_log_stats": "",
"disable_log_requests": "",
"gpu_memory_utilization": 0.9,
"num_scheduler_steps": 10,
"max_num_seqs": 512,
"dtype": "bfloat16"
},
@ -191,7 +197,9 @@
},
"vllm_server_parameters": {
"disable_log_stats": "",
"disable_log_requests": "",
"gpu_memory_utilization": 0.9,
"num_scheduler_steps": 10,
"max_num_seqs": 512,
"dtype": "bfloat16"
},
@ -243,7 +251,9 @@
},
"vllm_server_parameters": {
"disable_log_stats": "",
"disable_log_requests": "",
"gpu_memory_utilization": 0.9,
"num_scheduler_steps": 10,
"max_num_seqs": 512,
"dtype": "bfloat16"
},
@ -295,7 +305,9 @@
},
"vllm_server_parameters": {
"disable_log_stats": "",
"disable_log_requests": "",
"gpu_memory_utilization": 0.9,
"num_scheduler_steps": 10,
"max_num_seqs": 512,
"dtype": "bfloat16"
},

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
}
}
]

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
}
}
]

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
}
}
]

View File

@ -7,6 +7,7 @@
"tensor_parallel_size": 1,
"swap_space": 16,
"disable_log_stats": "",
"disable_log_requests": "",
"load_format": "dummy"
},
"client_parameters": {
@ -25,6 +26,7 @@
"tensor_parallel_size": 4,
"swap_space": 16,
"disable_log_stats": "",
"disable_log_requests": "",
"load_format": "dummy"
},
"client_parameters": {
@ -43,6 +45,7 @@
"tensor_parallel_size": 2,
"swap_space": 16,
"disable_log_stats": "",
"disable_log_requests": "",
"load_format": "dummy"
},
"client_parameters": {
@ -57,7 +60,8 @@
"test_name": "serving_llama70B_tp4_sharegpt_specdecode",
"qps_list": [2],
"server_parameters": {
"model": "meta-llama/Meta-Llama-3.1-70B-Instruct",
"model": "meta-llama/Meta-Llama-3.1-70B-Instruct",
"disable_log_requests": "",
"tensor_parallel_size": 4,
"swap_space": 16,
"speculative_config": {

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"
}
}
]

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

View File

@ -1,138 +1,76 @@
steps:
# 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=12.9.1 --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"
- 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
- label: "Build wheel - CUDA 12.4"
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 ."
- "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=12.4.0 --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 12.6 wheel"
key: block-build-cu126-wheel
depends_on: ~
- label: "Build wheel - CUDA 12.6"
depends_on: block-build-cu126-wheel
id: build-wheel-cuda-12-6
- label: "Build wheel - CUDA 12.1"
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=12.6.3 --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 ."
- "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=12.1.0 --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
- label: "Build wheel - CUDA 12.9"
depends_on: ~
id: build-wheel-cuda-12-9
# 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
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=11.8.0 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
- "mkdir artifacts"
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
- "bash .buildkite/scripts/upload-wheels.sh"
env:
DOCKER_BUILDKIT: "1"
- label: "Build release image (x86)"
- block: "Build release image"
depends_on: ~
id: build-release-image-x86
key: block-release-image-build
- label: "Build release image"
depends_on: block-release-image-build
agents:
queue: cpu_queue_postmerge
commands:
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.8.1 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$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_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.4.0 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT --target vllm-openai --progress plain -f docker/Dockerfile ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$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
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 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
- label: "Create multi-arch manifest"
depends_on:
- build-release-image-x86
- build-release-image-arm64
id: create-multi-arch-manifest
agents:
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 manifest create public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64 --amend"
- "docker manifest push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
- 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
username: vllm
password-env: DOCKERHUB_TOKEN
env:
DOCKER_BUILDKIT: "1"
- input: "Provide Release version here"
id: input-release-version
fields:
- text: "What is the release version?"
key: release-version
key: "release-version"
- block: "Build CPU release image"
key: block-cpu-release-image-build
@ -144,30 +82,22 @@ 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:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest --progress plain --target vllm-openai -f docker/Dockerfile.cpu ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest"
- "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(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:$(buildkite-agent meta-data get release-version)"
env:
DOCKER_BUILDKIT: "1"
- label: "Build and publish nightly multi-arch image to DockerHub"
depends_on:
- create-multi-arch-manifest
if: build.env("NIGHTLY") == "1"
- block: "Build Neuron release image"
key: block-neuron-release-image-build
depends_on: ~
- label: "Build and publish Neuron release image"
depends_on: block-neuron-release-image-build
agents:
queue: cpu_queue_postmerge
queue: neuron-postmerge
commands:
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
- "docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:$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"
# 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
- "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:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:latest --progress plain -f docker/Dockerfile.neuron ."
- "docker push public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:$(buildkite-agent meta-data get release-version)"
env:
DOCKER_BUILDKIT: "1"

View File

@ -1,31 +0,0 @@
#!/bin/bash
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
fi
buildkite-agent annotate --style 'info' --context 'release-workflow' << EOF
To download the wheel:
\`\`\`
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 .
\`\`\`
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}
\`\`\`
EOF

View File

@ -1,17 +0,0 @@
#!/bin/bash
# Usage: ./ci_clean_log.sh ci.log
# This script strips timestamps and color codes from CI log files.
# Check if argument is given
if [ $# -lt 1 ]; then
echo "Usage: $0 ci.log"
exit 1
fi
INPUT_FILE="$1"
# Strip timestamps
sed -i 's/^\[[0-9]\{4\}-[0-9]\{2\}-[0-9]\{2\}T[0-9]\{2\}:[0-9]\{2\}:[0-9]\{2\}Z\] //' "$INPUT_FILE"
# Strip colorization
sed -i -r 's/\x1B\[[0-9;]*[mK]//g' "$INPUT_FILE"

View File

@ -1,97 +0,0 @@
#!/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"

View File

@ -3,9 +3,6 @@
# This script runs test inside the corresponding ROCm docker container.
set -o pipefail
# Export Python path
export PYTHONPATH=".."
# Print ROCm version
echo "--- Confirming Clean Initial State"
while true; do
@ -77,70 +74,38 @@ 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
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
if [[ $commands == *"pytest -v -s lora"* ]]; then
commands=${commands//"pytest -v -s lora"/"VLLM_ROCM_CUSTOM_PAGED_ATTN=0 pytest -v -s lora"}
fi
#ignore certain kernels tests
if [[ $commands == *" kernels/core"* ]]; then
if [[ $commands == *" kernels "* ]]; then
commands="${commands} \
--ignore=kernels/core/test_fused_quant_layernorm.py \
--ignore=kernels/core/test_permute_cols.py"
fi
if [[ $commands == *" kernels/attention"* ]]; then
commands="${commands} \
--ignore=kernels/attention/test_attention_selector.py \
--ignore=kernels/attention/test_encoder_decoder_attn.py \
--ignore=kernels/attention/test_flash_attn.py \
--ignore=kernels/attention/test_flashinfer.py \
--ignore=kernels/attention/test_prefix_prefill.py \
--ignore=kernels/attention/test_cascade_flash_attn.py \
--ignore=kernels/attention/test_mha_attn.py \
--ignore=kernels/attention/test_lightning_attn.py \
--ignore=kernels/attention/test_attention.py"
fi
if [[ $commands == *" kernels/quantization"* ]]; then
commands="${commands} \
--ignore=kernels/quantization/test_int8_quant.py \
--ignore=kernels/quantization/test_machete_mm.py \
--ignore=kernels/quantization/test_block_fp8.py \
--ignore=kernels/quantization/test_block_int8.py \
--ignore=kernels/quantization/test_marlin_gemm.py \
--ignore=kernels/quantization/test_cutlass_scaled_mm.py \
--ignore=kernels/quantization/test_int8_kernel.py"
fi
if [[ $commands == *" kernels/mamba"* ]]; then
commands="${commands} \
--ignore=kernels/mamba/test_mamba_mixer2.py \
--ignore=kernels/mamba/test_causal_conv1d.py \
--ignore=kernels/mamba/test_mamba_ssm_ssd.py"
fi
if [[ $commands == *" kernels/moe"* ]]; then
commands="${commands} \
--ignore=kernels/moe/test_moe.py \
--ignore=kernels/moe/test_cutlass_moe.py \
--ignore=kernels/moe/test_triton_moe_ptpc_fp8.py"
--ignore=kernels/test_attention_selector.py \
--ignore=kernels/test_blocksparse_attention.py \
--ignore=kernels/test_causal_conv1d.py \
--ignore=kernels/test_cutlass.py \
--ignore=kernels/test_encoder_decoder_attn.py \
--ignore=kernels/test_flash_attn.py \
--ignore=kernels/test_flashinfer.py \
--ignore=kernels/test_int8_quant.py \
--ignore=kernels/test_machete_gemm.py \
--ignore=kernels/test_mamba_ssm.py \
--ignore=kernels/test_marlin_gemm.py \
--ignore=kernels/test_moe.py \
--ignore=kernels/test_prefix_prefill.py \
--ignore=kernels/test_rand.py \
--ignore=kernels/test_sampler.py \
--ignore=kernels/test_cascade_flash_attn.py \
--ignore=kernels/test_mamba_mixer2.py \
--ignore=kernels/test_aqlm.py \
--ignore=kernels/test_machete_mm.py \
--ignore=kernels/test_mha_attn.py \
--ignore=kernels/test_block_fp8.py \
--ignore=kernels/test_cutlass_moe.py \
--ignore=kernels/test_mamba_ssm_ssd.py \
--ignore=kernels/test_attention.py \
--ignore=kernels/test_block_int8.py \
--ignore=kernels/test_fused_quant_layernorm.py \
--ignore=kernels/test_int8_kernel.py \
--ignore=kernels/test_triton_moe_ptpc_fp8.py \
--ignore=kernels/test_permute_cols.py"
fi
#ignore certain Entrypoints/openai tests
@ -164,6 +129,7 @@ if [[ $commands == *" entrypoints/llm "* ]]; then
--ignore=entrypoints/llm/test_chat.py \
--ignore=entrypoints/llm/test_accuracy.py \
--ignore=entrypoints/llm/test_init.py \
--ignore=entrypoints/llm/test_generate_multiple_loras.py \
--ignore=entrypoints/llm/test_prompt_validation.py "}
fi
@ -181,8 +147,6 @@ 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.
if [[ $commands == *"--shard-id="* ]]; then
# assign job count as the number of shards used
@ -203,7 +167,6 @@ if [[ $commands == *"--shard-id="* ]]; then
-e AWS_SECRET_ACCESS_KEY \
-v "${HF_CACHE}:${HF_MOUNT}" \
-e "HF_HOME=${HF_MOUNT}" \
-e "PYTHONPATH=${MYPYTHONPATH}" \
--name "${container_name}_${GPU}" \
"${image_name}" \
/bin/bash -c "${commands_gpu}" \
@ -234,7 +197,6 @@ else
-e AWS_SECRET_ACCESS_KEY \
-v "${HF_CACHE}:${HF_MOUNT}" \
-e "HF_HOME=${HF_MOUNT}" \
-e "PYTHONPATH=${MYPYTHONPATH}" \
--name "${container_name}" \
"${image_name}" \
/bin/bash -c "${commands}"

View File

@ -7,7 +7,6 @@ set -ex
# Setup cleanup
remove_docker_container() {
if [[ -n "$container_id" ]]; then
podman stop --all -t0
podman rm -f "$container_id" || true
fi
podman system prune -f
@ -33,12 +32,9 @@ function cpu_tests() {
set -e
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]
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"
pytest -v -s tests/models/embedding/language/test_cls_models.py::test_classification_models[float-jason9693/Qwen2.5-1.5B-apeach]
pytest -v -s tests/models/embedding/language/test_embedding.py::test_models[half-BAAI/bge-base-en-v1.5]
pytest -v -s tests/models/encoder_decoder/language -m cpu_model"
}
# All of CPU tests are expected to be finished less than 40 mins.

View File

@ -6,118 +6,89 @@ set -ex
# allow to bind to different cores
CORE_RANGE=${CORE_RANGE:-48-95}
# used for TP/PP E2E test
OMP_CORE_RANGE=${OMP_CORE_RANGE:-48-95}
NUMA_NODE=${NUMA_NODE:-1}
export CMAKE_BUILD_PARALLEL_LEVEL=32
# Setup cleanup
remove_docker_container() {
set -e;
docker rm -f cpu-test-"$NUMA_NODE" cpu-test-"$NUMA_NODE"-avx2 || true;
remove_docker_container() {
set -e;
docker rm -f cpu-test-"$BUILDKITE_BUILD_NUMBER"-"$NUMA_NODE" cpu-test-"$BUILDKITE_BUILD_NUMBER"-avx2-"$NUMA_NODE" || true;
docker image rm cpu-test-"$BUILDKITE_BUILD_NUMBER" cpu-test-"$BUILDKITE_BUILD_NUMBER"-avx2 || true;
}
trap remove_docker_container EXIT
remove_docker_container
# Try building the docker image
numactl -C "$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 "$NUMA_NODE" docker build --tag cpu-test-"$BUILDKITE_BUILD_NUMBER" --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-"$BUILDKITE_BUILD_NUMBER"-avx2 --target vllm-test -f docker/Dockerfile.cpu .
# Run the image, setting --shm-size=4g for tensor parallel.
docker run -itd --cpuset-cpus="$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"
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"-avx2 cpu-test-"$NUMA_NODE"-avx2
docker run -itd --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --cpuset-cpus="$CORE_RANGE" \
--cpuset-mems="$NUMA_NODE" --privileged=true -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --shm-size=4g --name cpu-test-"$BUILDKITE_BUILD_NUMBER"-"$NUMA_NODE" cpu-test-"$BUILDKITE_BUILD_NUMBER"
docker run -itd --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --cpuset-cpus="$CORE_RANGE" \
--cpuset-mems="$NUMA_NODE" --privileged=true -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --shm-size=4g --name cpu-test-"$BUILDKITE_BUILD_NUMBER"-avx2-"$NUMA_NODE" cpu-test-"$BUILDKITE_BUILD_NUMBER"-avx2
function cpu_tests() {
set -e
export NUMA_NODE=$2
# list packages
docker exec cpu-test-"$NUMA_NODE"-avx2 bash -c "
set -e
pip list"
docker exec cpu-test-"$NUMA_NODE" bash -c "
set -e
pip list"
export BUILDKITE_BUILD_NUMBER=$3
# offline inference
docker exec cpu-test-"$NUMA_NODE"-avx2 bash -c "
docker exec cpu-test-"$BUILDKITE_BUILD_NUMBER"-avx2-"$NUMA_NODE" bash -c "
set -e
python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m"
# Run kernel tests
docker exec cpu-test-"$NUMA_NODE" bash -c "
set -e
pytest -x -v -s tests/kernels/test_onednn.py"
# Run basic model test
docker exec cpu-test-"$NUMA_NODE" bash -c "
docker exec cpu-test-"$BUILDKITE_BUILD_NUMBER"-"$NUMA_NODE" bash -c "
set -e
# Note: disable until supports V1
# 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/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"
pytest -v -s tests/kernels/test_cache.py -m cpu_model
pytest -v -s tests/kernels/test_mla_decode_cpu.py -m cpu_model
pytest -v -s tests/models/decoder_only/language -m cpu_model
pytest -v -s tests/models/embedding/language -m cpu_model
pytest -v -s tests/models/encoder_decoder/language -m cpu_model
pytest -v -s tests/models/decoder_only/audio_language -m cpu_model
pytest -v -s tests/models/decoder_only/vision_language -m cpu_model"
# Run compressed-tensor test
docker exec cpu-test-"$NUMA_NODE" bash -c "
docker exec cpu-test-"$BUILDKITE_BUILD_NUMBER"-"$NUMA_NODE" bash -c "
set -e
pytest -x -s -v \
tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_logprobs[False-10-32-neuralmagic/Llama-3.2-1B-quantized.w8a8]"
pytest -s -v \
tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_static_setup \
tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_dynamic_per_token"
# Note: disable it until supports V1
# Run AWQ test
# docker exec cpu-test-"$NUMA_NODE" bash -c "
# set -e
# VLLM_USE_V1=0 pytest -x -s -v \
# tests/quantization/test_ipex_quant.py"
docker exec cpu-test-"$BUILDKITE_BUILD_NUMBER"-"$NUMA_NODE" bash -c "
set -e
pytest -s -v \
tests/quantization/test_ipex_quant.py"
# Run chunked-prefill and prefix-cache test
docker exec cpu-test-"$BUILDKITE_BUILD_NUMBER"-"$NUMA_NODE" bash -c "
set -e
pytest -s -v -k cpu_model \
tests/basic_correctness/test_chunked_prefill.py"
# online serving
docker exec cpu-test-"$BUILDKITE_BUILD_NUMBER"-"$NUMA_NODE" bash -c "
set -e
export VLLM_CPU_KVCACHE_SPACE=10
export VLLM_CPU_OMP_THREADS_BIND=$1
python3 -m vllm.entrypoints.openai.api_server --model facebook/opt-125m --dtype half &
timeout 600 bash -c 'until curl localhost:8000/v1/models; do sleep 1; done' || exit 1
python3 benchmarks/benchmark_serving.py \
--backend vllm \
--dataset-name random \
--model facebook/opt-125m \
--num-prompts 20 \
--endpoint /v1/completions \
--tokenizer facebook/opt-125m"
# Run multi-lora tests
docker exec cpu-test-"$NUMA_NODE" bash -c "
docker exec cpu-test-"$BUILDKITE_BUILD_NUMBER"-"$NUMA_NODE" bash -c "
set -e
pytest -x -s -v \
pytest -s -v \
tests/lora/test_qwen2vl.py"
# online serving: tp+pp
docker exec cpu-test-"$NUMA_NODE" bash -c '
set -e
VLLM_CPU_OMP_THREADS_BIND=$E2E_OMP_THREADS VLLM_CPU_SGL_KERNEL=1 vllm serve meta-llama/Llama-3.2-3B-Instruct -tp=2 -pp=2 &
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 meta-llama/Llama-3.2-3B-Instruct \
--num-prompts 20 \
--endpoint /v1/completions
kill -s SIGTERM $server_pid &'
# online serving: tp+dp
docker exec cpu-test-"$NUMA_NODE" bash -c '
set -e
VLLM_CPU_OMP_THREADS_BIND=$E2E_OMP_THREADS VLLM_CPU_SGL_KERNEL=1 vllm serve meta-llama/Llama-3.2-3B-Instruct -tp=2 -dp=2 &
server_pid=$!
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 meta-llama/Llama-3.2-3B-Instruct \
--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 $CORE_RANGE $NUMA_NODE"
timeout 40m bash -c "cpu_tests $CORE_RANGE $NUMA_NODE $BUILDKITE_BUILD_NUMBER"

View File

@ -16,7 +16,8 @@ DOCKER_BUILDKIT=1 docker build . \
--build-arg max_jobs=66 \
--build-arg nvcc_threads=2 \
--build-arg RUN_WHEEL_CHECK=false \
--build-arg torch_cuda_arch_list="9.0+PTX"
--build-arg torch_cuda_arch_list="9.0+PTX" \
--build-arg vllm_fa_cmake_gpu_arches="90-real"
# Setup cleanup
remove_docker_container() { docker rm -f gh200-test || true; }

View File

@ -2,55 +2,23 @@
# 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 -exuo pipefail
set -ex
# Try building the docker image
cat <<EOF | docker build -t hpu-plugin-v1-test-env -f - .
FROM gaudi-base-image:latest
COPY ./ /workspace/vllm
WORKDIR /workspace/vllm
ENV no_proxy=localhost,127.0.0.1
ENV PT_HPU_ENABLE_LAZY_COLLECTIVES=true
RUN VLLM_TARGET_DEVICE=empty pip install .
RUN pip install git+https://github.com/vllm-project/vllm-gaudi.git
# install development dependencies (for testing)
RUN python3 -m pip install -e tests/vllm_test_utils
WORKDIR /workspace/
RUN git clone https://github.com/vllm-project/vllm-gaudi.git
RUN ln -s /workspace/vllm/tests && ln -s /workspace/vllm/examples && ln -s /workspace/vllm/benchmarks
EOF
docker build -t hpu-test-env -f docker/Dockerfile.hpu .
# Setup cleanup
# certain versions of HPU software stack have a bug that can
# override the exit code of the script, so we need to use
# separate remove_docker_containers and remove_docker_containers_and_exit
# separate remove_docker_container and remove_docker_container_and_exit
# functions, while other platforms only need one remove_docker_container
# function.
EXITCODE=1
remove_docker_containers() { docker rm -f hpu-plugin-v1-test || true; }
trap 'remove_docker_containers; exit $EXITCODE;' EXIT
remove_docker_containers
echo "Running HPU plugin v1 test"
docker run --rm --runtime=habana --name=hpu-plugin-v1-test --network=host \
-e HABANA_VISIBLE_DEVICES=all \
hpu-plugin-v1-test-env \
/bin/bash "/workspace/vllm-gaudi/tests/upstream_tests/ci_tests.sh"
remove_docker_container() { docker rm -f hpu-test || true; }
remove_docker_container_and_exit() { remove_docker_container; exit $EXITCODE; }
trap remove_docker_container_and_exit EXIT
remove_docker_container
# Run the image and launch offline inference
docker run --runtime=habana --name=hpu-test --network=host -e HABANA_VISIBLE_DEVICES=all -e VLLM_SKIP_WARMUP=true --entrypoint="" hpu-test-env python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m
EXITCODE=$?
if [ $EXITCODE -eq 0 ]; then
echo "Test with basic model passed"
else
echo "Test with basic model FAILED with exit code: $EXITCODE" >&2
fi
# The trap will handle the container removal and final exit.

View File

@ -0,0 +1,54 @@
#!/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 "${HF_CACHE}"
HF_MOUNT="/root/.cache/huggingface"
NEURON_COMPILE_CACHE_URL="$(realpath ~)/neuron_compile_cache"
mkdir -p "${NEURON_COMPILE_CACHE_URL}"
NEURON_COMPILE_CACHE_MOUNT="/root/.cache/neuron_compile_cache"
# Try building the docker image
aws ecr get-login-password --region us-west-2 | docker login --username AWS --password-stdin 763104351884.dkr.ecr.us-west-2.amazonaws.com
# prune old image and containers to save disk space, and only once a day
# by using a timestamp file in tmp.
if [ -f /tmp/neuron-docker-build-timestamp ]; then
last_build=$(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=${HF_MOUNT}" \
-v "${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 "python3 /workspace/vllm/examples/offline_inference/neuron.py && python3 -m pytest /workspace/vllm/tests/neuron/1_core/ -v --capture=tee-sys && python3 -m pytest /workspace/vllm/tests/neuron/2_core/ -v --capture=tee-sys"

View File

@ -1,167 +0,0 @@
#!/bin/bash
set -xu
remove_docker_container() {
docker rm -f tpu-test || true;
}
trap remove_docker_container EXIT
# Remove the container that might not be cleaned up in the previous run.
remove_docker_container
# Build the docker image.
docker build -f docker/Dockerfile.tpu -t vllm-tpu .
# Set up cleanup.
cleanup_docker() {
# Get Docker's root directory
docker_root=$(docker info -f '{{.DockerRootDir}}')
if [ -z "$docker_root" ]; then
echo "Failed to determine Docker root directory."
exit 1
fi
echo "Docker root directory: $docker_root"
# Check disk usage of the filesystem where Docker's root directory is located
disk_usage=$(df "$docker_root" | tail -1 | awk '{print $5}' | sed 's/%//')
# Define the threshold
threshold=70
if [ "$disk_usage" -gt "$threshold" ]; then
echo "Disk usage is above $threshold%. Cleaning up Docker images and volumes..."
# 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 --force --filter "until=72h" --all
echo "Docker images and volumes cleanup completed."
else
echo "Disk usage is below $threshold%. No cleanup needed."
fi
}
cleanup_docker
# For HF_TOKEN.
source /etc/environment
docker run --privileged --net host --shm-size=16G -it \
-e "HF_TOKEN=$HF_TOKEN" --name tpu-test \
vllm-tpu /bin/bash -c '
set -e # Exit immediately if a command exits with a non-zero status.
set -u # Treat unset variables as an error.
echo "--- Starting script inside Docker container ---"
# Create results directory
RESULTS_DIR=$(mktemp -d)
# If mktemp fails, set -e will cause the script to exit.
echo "Results will be stored in: $RESULTS_DIR"
# Install dependencies
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
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
echo "--- Starting Tests ---"
set +e
overall_script_exit_code=0
# --- Test Definitions ---
# If a test fails, this function will print logs and will not cause the main script to exit.
run_test() {
local test_num=$1
local test_name=$2
local test_command=$3
local log_file="$RESULTS_DIR/test_${test_num}.log"
local actual_exit_code
echo "--- TEST_$test_num: Running $test_name ---"
# Execute the test command.
eval "$test_command" > >(tee -a "$log_file") 2> >(tee -a "$log_file" >&2)
actual_exit_code=$?
echo "TEST_${test_num}_COMMAND_EXIT_CODE: $actual_exit_code" # This goes to main log
echo "TEST_${test_num}_COMMAND_EXIT_CODE: $actual_exit_code" >> "$log_file" # Also to per-test log
if [ "$actual_exit_code" -ne 0 ]; then
echo "TEST_$test_num ($test_name) FAILED with exit code $actual_exit_code." >&2
echo "--- Log for failed TEST_$test_num ($test_name) ---" >&2
if [ -f "$log_file" ]; then
cat "$log_file" >&2
else
echo "Log file $log_file not found for TEST_$test_num ($test_name)." >&2
fi
echo "--- End of log for TEST_$test_num ($test_name) ---" >&2
return "$actual_exit_code" # Return the failure code
else
echo "TEST_$test_num ($test_name) PASSED."
return 0 # Return success
fi
}
# Helper function to call run_test and update the overall script exit code
run_and_track_test() {
local test_num_arg="$1"
local test_name_arg="$2"
local test_command_arg="$3"
# Run the test
run_test "$test_num_arg" "$test_name_arg" "$test_command_arg"
local test_specific_exit_code=$?
# If the test failed, set the overall script exit code to 1
if [ "$test_specific_exit_code" -ne 0 ]; then
# No need for extra echo here, run_test already logged the failure.
overall_script_exit_code=1
fi
}
# --- Actual Test Execution ---
run_and_track_test 1 "test_struct_output_generate.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/entrypoints/llm/test_struct_output_generate.py -k \"not test_structured_output_with_reasoning_matrices\""
run_and_track_test 2 "test_moe_pallas.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/tpu/test_moe_pallas.py"
run_and_track_test 3 "test_lora.py" \
"VLLM_XLA_CHECK_RECOMPILATION=0 python3 -m pytest -s -v /workspace/vllm/tests/tpu/lora/test_lora.py"
run_and_track_test 4 "test_tpu_qkv_linear.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_tpu_qkv_linear.py"
run_and_track_test 5 "test_spmd_model_weight_loading.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_spmd_model_weight_loading.py"
run_and_track_test 6 "test_kv_cache_update_kernel.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_kv_cache_update_kernel.py"
run_and_track_test 7 "test_tpu_int8.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_tpu_int8.py"
# After all tests have been attempted, exit with the overall status.
if [ "$overall_script_exit_code" -ne 0 ]; then
echo "--- One or more tests FAILED. Overall script exiting with failure code 1. ---"
else
echo "--- All tests have completed and PASSED. Overall script exiting with success code 0. ---"
fi
exit "$overall_script_exit_code"
' # IMPORTANT: This is the closing single quote for the bash -c "..." command. Ensure it is present and correct.
# Capture the exit code of the docker run command
DOCKER_RUN_EXIT_CODE=$?
# The trap will run for cleanup.
# Exit the main script with the Docker run command's exit code.
if [ "$DOCKER_RUN_EXIT_CODE" -ne 0 ]; then
echo "Docker run command failed with exit code $DOCKER_RUN_EXIT_CODE."
exit "$DOCKER_RUN_EXIT_CODE"
else
echo "Docker run command completed successfully."
exit 0
fi
# TODO: This test fails because it uses RANDOM_SEED sampling
# pytest -v -s /workspace/vllm/tests/tpu/test_custom_dispatcher.py \

View File

@ -1,175 +1,54 @@
#!/bin/bash
set -xu
remove_docker_container() {
docker rm -f tpu-test || true;
}
trap remove_docker_container EXIT
# Remove the container that might not be cleaned up in the previous run.
remove_docker_container
set -xue
# Build the docker image.
docker build -f docker/Dockerfile.tpu -t vllm-tpu .
# Set up cleanup.
cleanup_docker() {
# Get Docker's root directory
docker_root=$(docker info -f '{{.DockerRootDir}}')
if [ -z "$docker_root" ]; then
echo "Failed to determine Docker root directory."
exit 1
fi
echo "Docker root directory: $docker_root"
# Check disk usage of the filesystem where Docker's root directory is located
disk_usage=$(df "$docker_root" | tail -1 | awk '{print $5}' | sed 's/%//')
# Define the threshold
threshold=70
if [ "$disk_usage" -gt "$threshold" ]; then
echo "Disk usage is above $threshold%. Cleaning up Docker images and volumes..."
# 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 --force --filter "until=72h" --all
echo "Docker images and volumes cleanup completed."
else
echo "Disk usage is below $threshold%. No cleanup needed."
fi
}
cleanup_docker
remove_docker_container() { docker rm -f tpu-test || true; }
trap remove_docker_container EXIT
# Remove the container that might not be cleaned up in the previous run.
remove_docker_container
# For HF_TOKEN.
source /etc/environment
# Run a simple end-to-end example.
docker run --privileged --net host --shm-size=16G -it \
-e "HF_TOKEN=$HF_TOKEN" --name tpu-test \
vllm-tpu /bin/bash -c '
set -e # Exit immediately if a command exits with a non-zero status.
set -u # Treat unset variables as an error.
vllm-tpu /bin/bash -c "python3 -m pip install git+https://github.com/thuml/depyf.git \
&& python3 -m pip install pytest pytest-asyncio tpu-info \
&& python3 -m pip install lm_eval[api]==0.4.4 \
&& export VLLM_XLA_CACHE_PATH= \
&& export VLLM_USE_V1=1 \
&& export VLLM_XLA_CHECK_RECOMPILATION=1 \
&& echo HARDWARE \
&& tpu-info \
&& echo TEST_0 \
&& pytest -v -s /workspace/vllm/tests/v1/tpu/test_perf.py \
&& echo TEST_1 \
&& pytest -v -s /workspace/vllm/tests/tpu/test_compilation.py \
&& echo TEST_2 \
&& pytest -v -s /workspace/vllm/tests/v1/tpu/test_basic.py \
&& echo TEST_3 \
&& pytest -v -s /workspace/vllm/tests/entrypoints/llm/test_accuracy.py::test_lm_eval_accuracy_v1_engine \
&& echo TEST_4 \
&& pytest -s -v /workspace/vllm/tests/tpu/test_quantization_accuracy.py \
&& echo TEST_5 \
&& python3 /workspace/vllm/examples/offline_inference/tpu.py \
&& echo TEST_6 \
&& pytest -s -v /workspace/vllm/tests/v1/tpu/worker/test_tpu_model_runner.py \
&& echo TEST_7 \
&& pytest -s -v /workspace/vllm/tests/v1/tpu/test_sampler.py \
&& echo TEST_8 \
&& pytest -s -v /workspace/vllm/tests/v1/tpu/test_topk_topp_sampler.py \
&& echo TEST_9 \
&& pytest -s -v /workspace/vllm/tests/v1/tpu/test_multimodal.py \
&& echo TEST_10 \
&& pytest -s -v /workspace/vllm/tests/v1/tpu/test_pallas.py \
&& echo TEST_11 \
&& pytest -s -v /workspace/vllm/tests/v1/entrypoints/llm/test_struct_output_generate.py" \
echo "--- Starting script inside Docker container ---"
# Create results directory
RESULTS_DIR=$(mktemp -d)
# If mktemp fails, set -e will cause the script to exit.
echo "Results will be stored in: $RESULTS_DIR"
# Install dependencies
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
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
echo "--- Starting Tests ---"
set +e
overall_script_exit_code=0
# --- Test Definitions ---
# If a test fails, this function will print logs and will not cause the main script to exit.
run_test() {
local test_num=$1
local test_name=$2
local test_command=$3
local log_file="$RESULTS_DIR/test_${test_num}.log"
local actual_exit_code
echo "--- TEST_$test_num: Running $test_name ---"
# Execute the test command.
eval "$test_command" > >(tee -a "$log_file") 2> >(tee -a "$log_file" >&2)
actual_exit_code=$?
echo "TEST_${test_num}_COMMAND_EXIT_CODE: $actual_exit_code" # This goes to main log
echo "TEST_${test_num}_COMMAND_EXIT_CODE: $actual_exit_code" >> "$log_file" # Also to per-test log
if [ "$actual_exit_code" -ne 0 ]; then
echo "TEST_$test_num ($test_name) FAILED with exit code $actual_exit_code." >&2
echo "--- Log for failed TEST_$test_num ($test_name) ---" >&2
if [ -f "$log_file" ]; then
cat "$log_file" >&2
else
echo "Log file $log_file not found for TEST_$test_num ($test_name)." >&2
fi
echo "--- End of log for TEST_$test_num ($test_name) ---" >&2
return "$actual_exit_code" # Return the failure code
else
echo "TEST_$test_num ($test_name) PASSED."
return 0 # Return success
fi
}
# Helper function to call run_test and update the overall script exit code
run_and_track_test() {
local test_num_arg="$1"
local test_name_arg="$2"
local test_command_arg="$3"
# Run the test
run_test "$test_num_arg" "$test_name_arg" "$test_command_arg"
local test_specific_exit_code=$?
# If the test failed, set the overall script exit code to 1
if [ "$test_specific_exit_code" -ne 0 ]; then
# No need for extra echo here, run_test already logged the failure.
overall_script_exit_code=1
fi
}
# --- Actual Test Execution ---
run_and_track_test 0 "test_perf.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_perf.py"
run_and_track_test 1 "test_compilation.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/tpu/test_compilation.py"
run_and_track_test 2 "test_basic.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_basic.py"
run_and_track_test 3 "test_accuracy.py::test_lm_eval_accuracy_v1_engine" \
"python3 -m pytest -s -v /workspace/vllm/tests/entrypoints/llm/test_accuracy.py::test_lm_eval_accuracy_v1_engine"
run_and_track_test 4 "test_quantization_accuracy.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/tpu/test_quantization_accuracy.py"
run_and_track_test 5 "examples/offline_inference/tpu.py" \
"python3 /workspace/vllm/examples/offline_inference/tpu.py"
run_and_track_test 6 "test_tpu_model_runner.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/worker/test_tpu_model_runner.py"
run_and_track_test 7 "test_sampler.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_sampler.py"
run_and_track_test 8 "test_topk_topp_sampler.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_topk_topp_sampler.py"
run_and_track_test 9 "test_multimodal.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_multimodal.py"
run_and_track_test 10 "test_pallas.py" \
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_pallas.py"
# After all tests have been attempted, exit with the overall status.
if [ "$overall_script_exit_code" -ne 0 ]; then
echo "--- One or more tests FAILED. Overall script exiting with failure code 1. ---"
else
echo "--- All tests have completed and PASSED. Overall script exiting with success code 0. ---"
fi
exit "$overall_script_exit_code"
' # IMPORTANT: This is the closing single quote for the bash -c "..." command. Ensure it is present and correct.
# Capture the exit code of the docker run command
DOCKER_RUN_EXIT_CODE=$?
# The trap will run for cleanup.
# Exit the main script with the Docker run command's exit code.
if [ "$DOCKER_RUN_EXIT_CODE" -ne 0 ]; then
echo "Docker run command failed with exit code $DOCKER_RUN_EXIT_CODE."
exit "$DOCKER_RUN_EXIT_CODE"
else
echo "Docker run command completed successfully."
exit 0
fi
# TODO: This test fails because it uses RANDOM_SEED sampling
# pytest -v -s /workspace/vllm/tests/tpu/test_custom_dispatcher.py \
# && VLLM_USE_V1=1 pytest -v -s /workspace/vllm/tests/tpu/test_custom_dispatcher.py \

View File

@ -11,8 +11,8 @@ container_name="xpu_${BUILDKITE_COMMIT}_$(tr -dc A-Za-z0-9 < /dev/urandom | head
docker build -t ${image_name} -f docker/Dockerfile.xpu .
# Setup cleanup
remove_docker_container() {
docker rm -f "${container_name}" || true;
remove_docker_container() {
docker rm -f "${container_name}" || true;
docker image rm -f "${image_name}" || true;
docker system prune -f || true;
}
@ -23,27 +23,9 @@ docker run \
--device /dev/dri \
-v /dev/dri/by-path:/dev/dri/by-path \
--entrypoint="" \
-e "HF_TOKEN=${HF_TOKEN}" \
-e "ZE_AFFINITY_MASK=${ZE_AFFINITY_MASK}" \
--name "${container_name}" \
"${image_name}" \
bash -c '
set -e
echo $ZE_AFFINITY_MASK
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
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/test_serial_utils.py
pytest -v -s v1/test_utils.py
pytest -v -s v1/test_metrics_reader.py
sh -c '
VLLM_USE_V1=0 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m
VLLM_USE_V1=0 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m -tp 2
'

View File

@ -1,18 +0,0 @@
#!/bin/bash
# Usage: ./rerun_test.sh path/to/test.py::test_name
# Check if argument is given
if [ $# -lt 1 ]; then
echo "Usage: $0 path/to/test.py::test_name"
echo "Example: $0 tests/v1/engine/test_engine_core_client.py::test_kv_cache_events[True-tcp]"
exit 1
fi
TEST=$1
COUNT=1
while pytest -sv "$TEST"; do
COUNT=$((COUNT + 1))
echo "RUN NUMBER ${COUNT}"
done

View File

@ -11,10 +11,10 @@ cd "$(dirname "${BASH_SOURCE[0]}")/../.."
(which wget && which curl) || (apt-get update && apt-get install -y wget curl)
# run python-based benchmarks and upload the result to buildkite
vllm bench latency --output-json latency_results.json 2>&1 | tee benchmark_latency.txt
python3 benchmarks/benchmark_latency.py --output-json latency_results.json 2>&1 | tee benchmark_latency.txt
bench_latency_exit_code=$?
vllm bench throughput --input-len 256 --output-len 256 --output-json throughput_results.json 2>&1 | tee benchmark_throughput.txt
python3 benchmarks/benchmark_throughput.py --input-len 256 --output-len 256 --output-json throughput_results.json 2>&1 | tee benchmark_throughput.txt
bench_throughput_exit_code=$?
# run server-based benchmarks and upload the result to buildkite
@ -24,7 +24,7 @@ wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/r
# wait for server to start, timeout after 600 seconds
timeout 600 bash -c 'until curl localhost:8000/v1/models; do sleep 1; done' || exit 1
vllm bench serve \
python3 benchmarks/benchmark_serving.py \
--backend vllm \
--dataset-name sharegpt \
--dataset-path ./ShareGPT_V3_unfiltered_cleaned_split.json \

View File

@ -1,24 +0,0 @@
#!/bin/bash
set -euo pipefail
docker_root=$(docker info -f '{{.DockerRootDir}}')
if [ -z "$docker_root" ]; then
echo "Failed to determine Docker root directory."
exit 1
fi
echo "Docker root directory: $docker_root"
# Check disk usage of the filesystem where Docker's root directory is located
disk_usage=$(df "$docker_root" | tail -1 | awk '{print $5}' | sed 's/%//')
# Define the threshold
threshold=70
if [ "$disk_usage" -gt "$threshold" ]; then
echo "Disk usage is above $threshold%. Cleaning up Docker images and volumes..."
# 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 --force --filter "until=24h" --all
echo "Docker images and volumes cleanup completed."
else
echo "Disk usage is below $threshold%. No cleanup needed."
fi

View File

@ -1,14 +0,0 @@
# Environment config
TEST_NAME=llama8b
CONTAINER_NAME=tpu-test
# vllm config
MODEL=meta-llama/Llama-3.1-8B-Instruct
MAX_NUM_SEQS=256
MAX_NUM_BATCHED_TOKENS=1024
TENSOR_PARALLEL_SIZE=1
MAX_MODEL_LEN=2048
DOWNLOAD_DIR=/mnt/disks/persist
EXPECTED_THROUGHPUT=8.0
INPUT_LEN=1800
OUTPUT_LEN=128

View File

@ -1,90 +0,0 @@
#!/bin/bash
if [ ! -f "$1" ]; then
echo "Error: The env file '$1' does not exist."
exit 1 # Exit the script with a non-zero status to indicate an error
fi
ENV_FILE=$1
# For testing on local vm, use `set -a` to export all variables
source /etc/environment
source $ENV_FILE
remove_docker_container() {
docker rm -f $CONTAINER_NAME || true;
}
trap remove_docker_container EXIT
# Remove the container that might not be cleaned up in the previous run.
remove_docker_container
LOG_ROOT=$(mktemp -d)
# If mktemp fails, set -e will cause the script to exit.
echo "Results will be stored in: $LOG_ROOT"
if [ -z "$HF_TOKEN" ]; then
echo "Error: HF_TOKEN is not set or is empty."
exit 1
fi
# Make sure mounted disk or dir exists
if [ ! -d "$DOWNLOAD_DIR" ]; then
echo "Error: Folder $DOWNLOAD_DIR does not exist. This is useually a mounted drive. If no mounted drive, just create a folder."
exit 1
fi
echo "Run model $MODEL"
echo
echo "starting docker...$CONTAINER_NAME"
echo
docker run \
-v $DOWNLOAD_DIR:$DOWNLOAD_DIR \
--env-file $ENV_FILE \
-e HF_TOKEN="$HF_TOKEN" \
-e TARGET_COMMIT=$BUILDKITE_COMMIT \
-e MODEL=$MODEL \
-e WORKSPACE=/workspace \
--name $CONTAINER_NAME \
-d \
--privileged \
--network host \
-v /dev/shm:/dev/shm \
vllm/vllm-tpu-bm tail -f /dev/null
echo "run script..."
echo
docker exec "$CONTAINER_NAME" /bin/bash -c ".buildkite/scripts/tpu/run_bm.sh"
echo "copy result back..."
VLLM_LOG="$LOG_ROOT/$TEST_NAME"_vllm_log.txt
BM_LOG="$LOG_ROOT/$TEST_NAME"_bm_log.txt
docker cp "$CONTAINER_NAME:/workspace/vllm_log.txt" "$VLLM_LOG"
docker cp "$CONTAINER_NAME:/workspace/bm_log.txt" "$BM_LOG"
throughput=$(grep "Request throughput (req/s):" "$BM_LOG" | sed 's/[^0-9.]//g')
echo "throughput for $TEST_NAME at $BUILDKITE_COMMIT: $throughput"
if [ "$BUILDKITE" = "true" ]; then
echo "Running inside Buildkite"
buildkite-agent artifact upload "$VLLM_LOG"
buildkite-agent artifact upload "$BM_LOG"
else
echo "Not running inside Buildkite"
fi
#
# compare the throughput with EXPECTED_THROUGHPUT
# and assert meeting the expectation
#
if [[ -z "$throughput" || ! "$throughput" =~ ^[0-9]+([.][0-9]+)?$ ]]; then
echo "Failed to get the throughput"
exit 1
fi
if (( $(echo "$throughput < $EXPECTED_THROUGHPUT" | bc -l) )); then
echo "Error: throughput($throughput) is less than expected($EXPECTED_THROUGHPUT)"
exit 1
fi

View File

@ -1,14 +0,0 @@
# Environment config
TEST_NAME=llama8bw8a8
CONTAINER_NAME=tpu-test
# vllm config
MODEL=RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8
MAX_NUM_SEQS=128
MAX_NUM_BATCHED_TOKENS=1024
TENSOR_PARALLEL_SIZE=1
MAX_MODEL_LEN=2048
DOWNLOAD_DIR=/mnt/disks/persist
EXPECTED_THROUGHPUT=10.0
INPUT_LEN=1800
OUTPUT_LEN=128

View File

@ -1,93 +0,0 @@
#!/bin/bash
set -euo pipefail
VLLM_LOG="$WORKSPACE/vllm_log.txt"
BM_LOG="$WORKSPACE/bm_log.txt"
if [ -n "$TARGET_COMMIT" ]; then
head_hash=$(git rev-parse HEAD)
if [ "$TARGET_COMMIT" != "$head_hash" ]; then
echo "Error: target commit $TARGET_COMMIT does not match HEAD: $head_hash"
exit 1
fi
fi
echo "model: $MODEL"
echo
#
# create a log folder
#
mkdir "$WORKSPACE/log"
# TODO: Move to image building.
pip install pandas
pip install datasets
#
# create sonnet_4x
#
echo "Create sonnet_4x.txt"
echo "" > benchmarks/sonnet_4x.txt
for _ in {1..4}
do
cat benchmarks/sonnet.txt >> benchmarks/sonnet_4x.txt
done
#
# start vllm service in backend
#
echo "lanching vllm..."
echo "logging to $VLLM_LOG"
echo
VLLM_USE_V1=1 vllm serve $MODEL \
--seed 42 \
--max-num-seqs $MAX_NUM_SEQS \
--max-num-batched-tokens $MAX_NUM_BATCHED_TOKENS \
--tensor-parallel-size $TENSOR_PARALLEL_SIZE \
--no-enable-prefix-caching \
--download_dir $DOWNLOAD_DIR \
--max-model-len $MAX_MODEL_LEN > "$VLLM_LOG" 2>&1 &
echo "wait for 20 minutes.."
echo
# sleep 1200
# wait for 10 minutes...
for i in {1..120}; do
# TODO: detect other type of errors.
if grep -Fq "raise RuntimeError" "$VLLM_LOG"; then
echo "Detected RuntimeError, exiting."
exit 1
elif grep -Fq "Application startup complete" "$VLLM_LOG"; then
echo "Application started"
break
else
echo "wait for 10 seconds..."
sleep 10
fi
done
#
# run test
#
echo "run benchmark test..."
echo "logging to $BM_LOG"
echo
vllm bench serve \
--backend vllm \
--model $MODEL \
--dataset-name sonnet \
--dataset-path benchmarks/sonnet_4x.txt \
--sonnet-input-len $INPUT_LEN \
--sonnet-output-len $OUTPUT_LEN \
--ignore-eos > "$BM_LOG"
echo "completed..."
echo
throughput=$(grep "Request throughput (req/s):" "$BM_LOG" | sed 's/[^0-9.]//g')
echo "throughput: $throughput"
echo

View File

@ -14,19 +14,8 @@ 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 $arch, using manylinux1 as default"
manylinux_version="manylinux1"
fi
# Rename 'linux' to the appropriate manylinux version in the wheel filename
new_wheel="${wheel/linux/$manylinux_version}"
# Rename 'linux' to 'manylinux1' in the wheel filename
new_wheel="${wheel/linux/manylinux1}"
mv -- "$wheel" "$new_wheel"
wheel="$new_wheel"
@ -58,15 +47,14 @@ python3 .buildkite/generate_index.py --wheel "$normal_wheel"
aws s3 cp "$wheel" "s3://vllm-wheels/$BUILDKITE_COMMIT/"
aws s3 cp "$normal_wheel" "s3://vllm-wheels/$BUILDKITE_COMMIT/"
if [[ $normal_wheel == *"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"
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 == *"cu121"* ]]; then
# if $normal_wheel matches cu121, do not upload the index.html
echo "Skipping index files for cu121 wheels"
else
# only upload index.html for cu129 wheels (default wheels) as it
# is available on both x86 and arm64
# only upload index.html for cu124 wheels (default wheels)
aws s3 cp index.html "s3://vllm-wheels/$BUILDKITE_COMMIT/vllm/index.html"
aws s3 cp "s3://vllm-wheels/nightly/index.html" "s3://vllm-wheels/$BUILDKITE_COMMIT/index.html"
fi
@ -75,17 +63,15 @@ fi
aws s3 cp "$wheel" "s3://vllm-wheels/nightly/"
aws s3 cp "$normal_wheel" "s3://vllm-wheels/nightly/"
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"
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 == *"cu121"* ]]; then
# if $normal_wheel matches cu121, do not upload the index.html
echo "Skipping index files for cu121 wheels"
else
# only upload index.html for cu129 wheels (default wheels) as it
# is available on both x86 and arm64
# only upload index.html for cu124 wheels (default wheels)
aws s3 cp index.html "s3://vllm-wheels/nightly/vllm/index.html"
fi
aws s3 cp "$wheel" "s3://vllm-wheels/$version/"
aws s3 cp index.html "s3://vllm-wheels/$version/vllm/index.html"
aws s3 cp "$wheel" "s3://vllm-wheels/$version/"

File diff suppressed because it is too large Load Diff

View File

@ -1,6 +0,0 @@
# https://developers.google.com/gemini-code-assist/docs/customize-gemini-behavior-github
have_fun: false # Just review the code
code_review:
comment_severity_threshold: HIGH # Reduce quantity of comments
pull_request_opened:
summary: false # Don't summarize the PR in a separate comment

85
.github/CODEOWNERS vendored
View File

@ -5,89 +5,38 @@
/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 @22quinn
/vllm/worker/worker_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill @22quinn
/vllm/executor/executor_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
/vllm/worker/worker_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
/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/quantization @mgoin @robertgshaw2-redhat @tlrmchlsmth @yewentao256
/vllm/model_executor/layers/mamba @tdoublep
/vllm/model_executor/model_loader @22quinn
/vllm/model_executor/layers/quantization @mgoin @robertgshaw2-redhat @tlrmchlsmth
/vllm/model_executor/guided_decoding @mgoin @russellb
/vllm/multimodal @DarkLight1337 @ywang96
/vllm/v1/sample @22quinn @houseroad
/vllm/vllm_flash_attn @LucasWilkinson
/vllm/lora @jeejeelee
/vllm/reasoning @aarnphm
/vllm/entrypoints @aarnphm
/vllm/compilation @zou3519 @youkaichao @ProExpertProg
CMakeLists.txt @tlrmchlsmth @LucasWilkinson
# Any change to the VllmConfig changes can have a large user-facing impact,
# so spam a lot of people
/vllm/config @simon-mo @WoosukKwon @youkaichao @robertgshaw2-redhat @mgoin @tlrmchlsmth @houseroad @hmellor @yewentao256 @ProExpertProg
CMakeLists.txt @tlrmchlsmth
# vLLM V1
/vllm/v1 @WoosukKwon @robertgshaw2-redhat @njhill @ywang96 @comaniac @alexm-redhat
/vllm/v1/structured_output @mgoin @russellb @aarnphm @benchislett
/vllm/v1/spec_decode @benchislett @luccafong
/vllm/v1/attention/backends/triton_attn.py @tdoublep
/vllm/v1/structured_output @mgoin @russellb
# Test ownership
/.buildkite/lm-eval-harness @mgoin @simon-mo
/tests/async_engine @njhill @robertgshaw2-redhat @simon-mo
/tests/basic_correctness/test_chunked_prefill @rkooo567 @comaniac
/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/kernels @tlrmchlsmth @WoosukKwon @yewentao256
/tests/entrypoints @DarkLight1337 @robertgshaw2-redhat @simon-mo
/tests/entrypoints/llm/test_guided_generate.py @mgoin @russellb
/tests/kernels @tlrmchlsmth @WoosukKwon
/tests/model_executor/test_guided_processors.py @mgoin @russellb
/tests/models @DarkLight1337 @ywang96
/tests/multi_step @alexm-redhat @comaniac
/tests/multimodal @DarkLight1337 @ywang96
/tests/prefix_caching @comaniac @KuntaiDu
/tests/quantization @mgoin @robertgshaw2-redhat @yewentao256
/tests/quantization @mgoin @robertgshaw2-redhat
/tests/spec_decode @njhill @LiuXiaoxuanPKU
/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/weight_loading @mgoin @youkaichao @yewentao256
/tests/lora @jeejeelee
/tests/models/language/generation/test_hybrid.py @tdoublep
# Docs
/docs @hmellor
mkdocs.yaml @hmellor
# CPU
/vllm/v1/worker/^cpu @bigPYJ1151
/csrc/cpu @bigPYJ1151
/vllm/platforms/cpu.py @bigPYJ1151
/cmake/cpu_extension.cmake @bigPYJ1151
/docker/Dockerfile.cpu @bigPYJ1151
# Intel GPU
/vllm/v1/worker/^xpu @jikunshang
/vllm/platforms/xpu.py @jikunshang
/docker/Dockerfile.xpu @jikunshang
# Qwen-specific files
/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
/vllm/model_executor/models/voxtral*.py @patrickvonplaten
/vllm/model_executor/models/pixtral*.py @patrickvonplaten
/vllm/transformers_utils/configs/mistral.py @patrickvonplaten
/vllm/transformers_utils/tokenizers/mistral.py @patrickvonplaten
# Kernels
/vllm/attention/ops/chunked_prefill_paged_decode.py @tdoublep
/vllm/attention/ops/triton_unified_attention.py @tdoublep
# ROCm related: specify owner with write access to notify AMD folks for careful code review
/docker/Dockerfile.rocm* @gshtras
/vllm/v1/attention/backends/rocm*.py @gshtras
/vllm/v1/attention/backends/mla/rocm*.py @gshtras
/vllm/attention/ops/rocm*.py @gshtras
/vllm/model_executor/layers/fused_moe/rocm*.py @gshtras
/tests/v1/entrypoints/llm/test_struct_output_generate.py @mgoin @russellb
/tests/v1/structured_output @mgoin @russellb
/tests/weight_loading @mgoin @youkaichao

View File

@ -8,16 +8,6 @@ body:
attributes:
value: >
#### Before submitting an issue, please make sure the issue hasn't been already addressed by searching through [the existing and past issues](https://github.com/vllm-project/vllm/issues?q=is%3Aissue+sort%3Acreated-desc+).
- type: markdown
attributes:
value: |
⚠️ **SECURITY WARNING:** Please review any text you paste to ensure it does not contain sensitive information such as:
- API tokens or keys (e.g., Hugging Face tokens, OpenAI API keys)
- Passwords or authentication credentials
- Private URLs or endpoints
- Personal or confidential data
Consider redacting or replacing sensitive values with placeholders like `<YOUR_TOKEN_HERE>` when sharing configuration or code examples.
- type: textarea
attributes:
label: Your current environment
@ -31,12 +21,12 @@ body:
It is suggested to download and execute the latest script, as vllm might frequently update the diagnosis information needed for accurately and quickly responding to issues.
value: |
<details>
<summary>The output of <code>python collect_env.py</code></summary>
<summary>The output of `python collect_env.py`</summary>
```text
Your output of `python collect_env.py` here
```
</details>
validations:
required: true
@ -85,20 +75,20 @@ body:
```
```
The error message you got, with the full traceback and the error logs with [dump_input.py:##] if present.
The error message you got, with the full traceback.
```
validations:
required: true
- type: markdown
attributes:
value: |
⚠️ Please separate bugs of `transformers` implementation or usage from bugs of `vllm`. If you think anything is wrong with the model's output:
value: >
⚠️ Please separate bugs of `transformers` implementation or usage from bugs of `vllm`. If you think anything is wrong with the models' output:
- Try the counterpart of `transformers` first. If the error appears, please go to [their issues](https://github.com/huggingface/transformers/issues?q=is%3Aissue+is%3Aopen+sort%3Aupdated-desc).
- If the error only appears in vllm, please provide the detailed script of how you run `transformers` and `vllm`, also highlight the difference and what you expect.
Thanks for reporting 🙏!
Thanks for contributing 🎉!
- type: checkboxes
id: askllm
attributes:

View File

@ -1,69 +0,0 @@
name: 🧪 CI failure report
description: Report a failing test.
title: "[CI Failure]: "
labels: ["ci-failure"]
body:
- type: markdown
attributes:
value: >
#### Include the name of the failing Buildkite step and test file in the title.
- type: input
attributes:
label: Name of failing test
description: |
Paste in the fully-qualified name of the failing test from the logs.
placeholder: |
`path/to/test_file.py::test_name[params]`
validations:
required: true
- type: checkboxes
attributes:
label: Basic information
description: Select all items that apply to the failing test.
options:
- label: Flaky test
- label: Can reproduce locally
- label: Caused by external libraries (e.g. bug in `transformers`)
- type: textarea
attributes:
label: 🧪 Describe the failing test
description: |
Please provide a clear and concise description of the failing test.
placeholder: |
A clear and concise description of the failing test.
```
The error message you got, with the full traceback and the error logs with [dump_input.py:##] if present.
```
validations:
required: true
- type: textarea
attributes:
label: 📝 History of failing test
description: |
Since when did the test start to fail?
You can look up its history via [Buildkite Test Suites](https://buildkite.com/organizations/vllm/analytics/suites/ci-1/tests?branch=main).
If you have time, identify the PR that caused the test to fail on main. You can do so via the following methods:
- Use Buildkite Test Suites to find the PR where the test failure first occurred, and reproduce the failure locally.
- Run [`git bisect`](https://git-scm.com/docs/git-bisect) locally.
- Manually unblock Buildkite steps for suspected PRs on main and check the results. (authorized users only)
placeholder: |
Approximate timeline and/or problematic PRs
A link to the Buildkite analytics of the failing test (if available)
validations:
required: true
- type: textarea
attributes:
label: CC List.
description: >
The list of people you want to CC. Usually, this includes those who worked on the PR that failed the test.
- type: markdown
attributes:
value: >
Thanks for reporting 🙏!

View File

@ -46,7 +46,7 @@ body:
- type: markdown
attributes:
value: >
Thanks for contributing 🎉! The vLLM core team hosts a biweekly RFC review session at 9:30AM Pacific Time, while most RFCs can be discussed online, you can optionally sign up for a slot to discuss your RFC online [here](https://docs.google.com/document/d/1CiLVBZeIVfR7_PNAKVSusxpceywkoOOB78qoWqHvSZc/edit).
Thanks for contributing 🎉!
- type: checkboxes
id: askllm
attributes:

View File

@ -1,21 +1,6 @@
<!-- markdownlint-disable -->
PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.
FILL IN THE PR DESCRIPTION HERE
## Purpose
FIX #xxxx (*link existing issues this PR will resolve*)
## Test Plan
## Test Result
---
<details>
<summary> Essential Elements of an Effective PR Description Checklist </summary>
- [ ] The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
- [ ] The test plan, such as providing test command.
- [ ] The test results, such as pasting the results comparison before and after, or e2e results
- [ ] (Optional) The necessary documentation update, such as updating `supported_models.md` and `examples` for a new model.
- [ ] (Optional) Release notes update. If your change is user facing, please update the release notes draft in the [Google Doc](https://docs.google.com/document/d/1YyVqrgX4gHTtrstbq8oWUImOyPCKSGnJ7xtTpmXzlRs/edit?tab=t.0).
</details>
**BEFORE SUBMITTING, PLEASE READ <https://docs.vllm.ai/en/latest/contributing>** (anything written below this line will be removed by GitHub Actions)
<!--- pyml disable-next-line no-emphasis-as-heading -->
**BEFORE SUBMITTING, PLEASE READ <https://docs.vllm.ai/en/latest/contributing/overview.html>** (anything written below this line will be removed by GitHub Actions)

148
.github/mergify.yml vendored
View File

@ -27,22 +27,6 @@ pull_request_rules:
add:
- ci/build
- name: label-deepseek
description: Automatically apply deepseek label
conditions:
- or:
- files~=^examples/.*deepseek.*\.py
- files~=^tests/.*deepseek.*\.py
- files~=^vllm/entrypoints/openai/tool_parsers/.*deepseek.*\.py
- files~=^vllm/model_executor/models/.*deepseek.*\.py
- files~=^vllm/reasoning/.*deepseek.*\.py
- files~=^vllm/transformers_utils/.*deepseek.*\.py
- title~=(?i)DeepSeek
actions:
label:
add:
- deepseek
- name: label-frontend
description: Automatically apply frontend label
conditions:
@ -52,21 +36,6 @@ pull_request_rules:
add:
- frontend
- name: label-llama
description: Automatically apply llama label
conditions:
- or:
- files~=^examples/.*llama.*\.py
- files~=^tests/.*llama.*\.py
- files~=^vllm/entrypoints/openai/tool_parsers/llama.*\.py
- files~=^vllm/model_executor/models/.*llama.*\.py
- files~=^vllm/transformers_utils/configs/.*llama.*\.py
- title~=(?i)llama
actions:
label:
add:
- llama
- name: label-multi-modality
description: Automatically apply multi-modality label
conditions:
@ -74,84 +43,14 @@ pull_request_rules:
- files~=^vllm/multimodal/
- files~=^tests/multimodal/
- files~=^tests/models/multimodal/
- files~=^tests/models/*/audio_language/
- files~=^tests/models/*/vision_language/
- files=tests/models/test_vision.py
actions:
label:
add:
- multi-modality
- name: label-new-model
description: Automatically apply new-model label
conditions:
- and:
- files~=^vllm/model_executor/models/
- files=vllm/model_executor/models/registry.py
actions:
label:
add:
- new-model
- name: label-performance
description: Automatically apply performance label
conditions:
- or:
- files~=^benchmarks/
- files~=^vllm/benchmarks/
- files~=^tests/benchmarks/
- files~=^\.buildkite/nightly-benchmarks/
actions:
label:
add:
- performance
- name: label-qwen
description: Automatically apply qwen label
conditions:
- or:
- files~=^examples/.*qwen.*\.py
- files~=^tests/.*qwen.*\.py
- files~=^vllm/model_executor/models/.*qwen.*\.py
- files~=^vllm/reasoning/.*qwen.*\.py
- title~=(?i)Qwen
actions:
label:
add:
- qwen
- name: label-gpt-oss
description: Automatically apply gpt-oss label
conditions:
- or:
- files~=^examples/.*gpt[-_]?oss.*\.py
- files~=^tests/.*gpt[-_]?oss.*\.py
- files~=^vllm/model_executor/models/.*gpt[-_]?oss.*\.py
- files~=^vllm/model_executor/layers/.*gpt[-_]?oss.*\.py
- title~=(?i)gpt[-_]?oss
actions:
label:
add:
- gpt-oss
- name: label-rocm
description: Automatically apply rocm label
conditions:
- or:
- files~=^csrc/rocm/
- files~=^docker/Dockerfile.rocm
- files~=^requirements/rocm.*\.txt
- files~=^vllm/attention/backends/rocm.*\.py
- files~=^vllm/attention/ops/rocm.*\.py
- files~=^vllm/model_executor/layers/fused_moe/rocm.*\.py
- files~=^vllm/v1/attention/backends/mla/rocm.*\.py
- files~=^tests/kernels/.*_rocm.*\.py
- files=vllm/platforms/rocm.py
- title~=(?i)AMD
- title~=(?i)ROCm
actions:
label:
add:
- rocm
- name: label-structured-output
description: Automatically apply structured-output label
conditions:
@ -159,10 +58,13 @@ pull_request_rules:
- files~=^benchmarks/structured_schemas/
- files=benchmarks/benchmark_serving_structured_output.py
- files=benchmarks/run_structured_output_benchmark.sh
- files=docs/features/structured_outputs.md
- files=docs/source/features/structured_outputs.md
- files=examples/offline_inference/structured_outputs.py
- files=examples/online_serving/openai_chat_completion_structured_outputs.py
- files=examples/online_serving/openai_chat_completion_structured_outputs_with_reasoning.py
- files~=^vllm/model_executor/guided_decoding/
- files=tests/model_executor/test_guided_processors.py
- files=tests/entrypoints/llm/test_guided_generate.py
- files~=^tests/v1/structured_output/
- files=tests/v1/entrypoints/llm/test_guided_generate.py
- files~=^vllm/v1/structured_output/
@ -175,12 +77,9 @@ pull_request_rules:
description: Automatically apply speculative-decoding label
conditions:
- or:
- files~=^vllm/v1/spec_decode/
- files~=^tests/v1/spec_decode/
- files~=^examples/.*(spec_decode|mlpspeculator|eagle|speculation).*\.py
- files~=^vllm/model_executor/models/.*eagle.*\.py
- files=vllm/model_executor/models/mlp_speculator.py
- files~=^vllm/transformers_utils/configs/(eagle|medusa|mlp_speculator)\.py
- files~=^vllm/spec_decode/
- files=vllm/model_executor/layers/spec_decode_base_sampler.py
- files~=^tests/spec_decode/
actions:
label:
add:
@ -236,7 +135,9 @@ pull_request_rules:
- files~=^tests/entrypoints/openai/tool_parsers/
- files=tests/entrypoints/openai/test_chat_with_tool_reasoning.py
- files~=^vllm/entrypoints/openai/tool_parsers/
- files=docs/features/tool_calling.md
- files=docs/source/features/tool_calling.md
- files=docs/source/getting_started/examples/openai_chat_completion_client_with_tools.md
- files=docs/source/getting_started/examples/chat_with_tools.md
- files~=^examples/tool_chat_*
- files=examples/offline_inference/chat_with_tools.py
- files=examples/online_serving/openai_chat_completion_client_with_tools_required.py
@ -262,31 +163,6 @@ pull_request_rules:
https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork
- name: assign reviewer for tensorizer changes
conditions:
- files~=^vllm/model_executor/model_loader/tensorizer.py
- files~=^vllm/model_executor/model_loader/tensorizer_loader.py
- files~=^tests/entrypoints/openai/test_tensorizer_entrypoint.py
- files~=^tests/tensorizer_loader/
actions:
assign:
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

View File

@ -1,21 +0,0 @@
# scale-config.yml:
# Powers what instance types are available for GHA auto-scaled
# runners. Runners listed here will be available as self hosted
# runners, configuration is directly pulled from the main branch.
# runner_types:
# runner_label:
# instance_type: m4.large
# os: linux
# # min_available defaults to the global cfg in the ALI Terraform
# min_available: undefined
# # when max_available value is not defined, no max runners is enforced
# max_available: undefined
# disk_size: 50
# is_ephemeral: true
runner_types:
linux.2xlarge:
disk_size: 150
instance_type: c5.2xlarge
is_ephemeral: true
os: linux

View File

@ -15,18 +15,18 @@ NEW=/tmp/new_pr_body.txt
gh pr view --json body --template "{{.body}}" "${PR_NUMBER}" > "${OLD}"
cp "${OLD}" "${NEW}"
# Remove markdown comments (like the <!-- markdownlint-disable --> at the start)
sed -i '/<!--.*-->$/d' "${NEW}"
# Remove "FIX #xxxx (*link existing issues this PR will resolve*)"
sed -i '/FIX #xxxx.*$/d' "${NEW}"
# Remove "PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED."
sed -i '/PLEASE FILL IN THE PR DESCRIPTION HERE.*$/d' "${NEW}"
# Remove "FILL IN THE PR DESCRIPTION HERE"
sed -i '/FILL IN THE PR DESCRIPTION HERE/d' "${NEW}"
# Remove all lines after and including "**BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE DESCRIPTION ABOVE**"
sed -i '/\*\*BEFORE SUBMITTING, PLEASE READ.*\*\*/,$d' "${NEW}"
# Remove HTML <details> section that includes <summary> text of "PR Checklist (Click to Expand)"
python3 - <<EOF
import regex as re
import re
with open("${NEW}", "r") as file:
content = file.read()

View File

@ -1,6 +1,4 @@
name: Add label on auto-merge enabled
permissions:
pull-requests: write
on:
pull_request_target:
types:

View File

@ -16,16 +16,11 @@ jobs:
uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
- name: Set up Python
uses: actions/setup-python@e797f83bcb11b83ae66e0230d6156d7c80228e7c # v6.0.0
uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0
with:
python-version: '3.12'
- name: Install Python dependencies
run: |
python3 -m pip install --upgrade pip
python3 -m pip install regex
- name: Update PR description
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: bash .github/scripts/cleanup_pr_body.sh "${{ github.event.number }}"
run: .github/scripts/cleanup_pr_body.sh "${{ github.event.number }}"

View File

@ -1,309 +0,0 @@
name: Label issues based on keywords
on:
issues:
types: [opened, edited, reopened]
permissions:
issues: write # needed so the workflow can add labels
contents: read
concurrency:
group: issue-labeler-${{ github.event.issue.number }}
cancel-in-progress: true
jobs:
add-labels:
runs-on: ubuntu-latest
steps:
- name: Label issues based on keywords
uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
with:
script: |
// Configuration: Add new labels and keywords here
const labelConfig = {
rocm: {
// Keyword search - matches whole words only (with word boundaries)
keywords: [
{
term: "composable kernel",
searchIn: "both"
},
{
term: "rccl",
searchIn: "body" // only search in body
},
{
term: "migraphx",
searchIn: "title" // only search in title
},
{
term: "hipgraph",
searchIn: "both"
},
{
term: "ROCm System Management Interface",
searchIn: "body"
},
],
// Substring search - matches anywhere in text (partial matches)
substrings: [
{
term: "VLLM_ROCM_",
searchIn: "both"
},
{
term: "aiter",
searchIn: "title"
},
{
term: "rocm",
searchIn: "title"
},
{
term: "amd",
searchIn: "title"
},
{
term: "hip-",
searchIn: "both"
},
{
term: "gfx",
searchIn: "both"
},
{
term: "cdna",
searchIn: "both"
},
{
term: "rdna",
searchIn: "both"
},
{
term: "torch_hip",
searchIn: "body" // only in body
},
{
term: "_hip",
searchIn: "both"
},
{
term: "hip_",
searchIn: "both"
},
// ROCm tools and libraries
{
term: "hipify",
searchIn: "both"
},
],
// Regex patterns - for complex pattern matching
regexPatterns: [
{
pattern: "\\bmi\\d{3}[a-z]*\\b",
description: "AMD GPU names (mi + 3 digits + optional letters)",
flags: "gi",
searchIn: "both" // "title", "body", or "both"
}
],
},
};
// Helper function to create regex based on search type
function createSearchRegex(term, type) {
// Escape special regex characters in the term
const escapedTerm = term.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
switch (type) {
case 'keyword':
// Word boundary search - matches whole words only
return new RegExp(`\\b${escapedTerm}\\b`, "gi");
case 'substring':
// Substring search - matches anywhere in the text
return new RegExp(escapedTerm, "gi");
default:
throw new Error(`Unknown search type: ${type}`);
}
}
// Helper function to find matching terms in text with line information
function findMatchingTermsWithLines(text, searchTerms = [], searchType = 'keyword', searchLocation = '') {
const matches = [];
const lines = text.split('\n');
for (const termConfig of searchTerms) {
let regex;
let term, searchIn, pattern, description, flags;
// Handle different input formats (string or object)
if (typeof termConfig === 'string') {
term = termConfig;
searchIn = 'both'; // default
} else {
term = termConfig.term;
searchIn = termConfig.searchIn || 'both';
pattern = termConfig.pattern;
description = termConfig.description;
flags = termConfig.flags;
}
// Skip if this term shouldn't be searched in the current location
if (searchIn !== 'both' && searchIn !== searchLocation) {
continue;
}
// Create appropriate regex
if (searchType === 'regex') {
regex = new RegExp(pattern, flags || "gi");
} else {
regex = createSearchRegex(term, searchType);
}
const termMatches = [];
// Check each line for matches
lines.forEach((line, lineIndex) => {
const lineMatches = line.match(regex);
if (lineMatches) {
lineMatches.forEach(match => {
termMatches.push({
match: match,
lineNumber: lineIndex + 1,
lineContent: line.trim(),
searchType: searchType,
searchLocation: searchLocation,
originalTerm: term || pattern,
description: description,
// Show context around the match in the line
context: line.length > 100 ?
line.substring(Math.max(0, line.toLowerCase().indexOf(match.toLowerCase()) - 30),
line.toLowerCase().indexOf(match.toLowerCase()) + match.length + 30) + '...'
: line.trim()
});
});
}
});
if (termMatches.length > 0) {
matches.push({
term: term || (description || pattern),
searchType: searchType,
searchLocation: searchLocation,
searchIn: searchIn,
pattern: pattern,
matches: termMatches,
count: termMatches.length
});
}
}
return matches;
}
// Helper function to check if label should be added
async function processLabel(labelName, config) {
const body = context.payload.issue.body || "";
const title = context.payload.issue.title || "";
core.notice(`Processing label: ${labelName}`);
core.notice(`Issue Title: "${title}"`);
core.notice(`Issue Body length: ${body.length} characters`);
let shouldAddLabel = false;
let allMatches = [];
let reason = '';
const keywords = config.keywords || [];
const substrings = config.substrings || [];
const regexPatterns = config.regexPatterns || [];
core.notice(`Searching with ${keywords.length} keywords, ${substrings.length} substrings, and ${regexPatterns.length} regex patterns`);
// Search in title
if (title.trim()) {
core.notice(`Searching in title: "${title}"`);
const titleKeywordMatches = findMatchingTermsWithLines(title, keywords, 'keyword', 'title');
const titleSubstringMatches = findMatchingTermsWithLines(title, substrings, 'substring', 'title');
const titleRegexMatches = findMatchingTermsWithLines(title, regexPatterns, 'regex', 'title');
allMatches.push(...titleKeywordMatches, ...titleSubstringMatches, ...titleRegexMatches);
}
// Search in body
if (body.trim()) {
core.notice(`Searching in body (${body.length} characters)`);
const bodyKeywordMatches = findMatchingTermsWithLines(body, keywords, 'keyword', 'body');
const bodySubstringMatches = findMatchingTermsWithLines(body, substrings, 'substring', 'body');
const bodyRegexMatches = findMatchingTermsWithLines(body, regexPatterns, 'regex', 'body');
allMatches.push(...bodyKeywordMatches, ...bodySubstringMatches, ...bodyRegexMatches);
}
if (allMatches.length > 0) {
core.notice(`Found ${allMatches.length} matching term(s):`);
for (const termMatch of allMatches) {
const locationText = termMatch.searchLocation === 'title' ? 'title' : 'body';
const searchInText = termMatch.searchIn === 'both' ? 'both' : termMatch.searchIn;
if (termMatch.searchType === 'regex') {
core.notice(` 📍 Regex: "${termMatch.term}" (pattern: ${termMatch.pattern}) found ${termMatch.count} time(s) in ${locationText} (configured to search in: ${searchInText}):`);
} else {
core.notice(` 📍 Term: "${termMatch.term}" (${termMatch.searchType} search) found ${termMatch.count} time(s) in ${locationText} (configured to search in: ${searchInText}):`);
}
// Show details for each match
termMatch.matches.forEach((match, index) => {
core.notice(` ${index + 1}. Line ${match.lineNumber} in ${match.searchLocation}: "${match.match}" [${match.searchType}]`);
if (match.description) {
core.notice(` Description: ${match.description}`);
}
core.notice(` Context: ${match.context}`);
if (match.lineContent !== match.context) {
core.notice(` Full line: ${match.lineContent}`);
}
});
}
shouldAddLabel = true;
const totalMatches = allMatches.reduce((sum, t) => sum + t.count, 0);
const titleMatches = allMatches.filter(t => t.searchLocation === 'title').reduce((sum, t) => sum + t.count, 0);
const bodyMatches = allMatches.filter(t => t.searchLocation === 'body').reduce((sum, t) => sum + t.count, 0);
const keywordMatches = allMatches.filter(t => t.searchType === 'keyword').reduce((sum, t) => sum + t.count, 0);
const substringMatches = allMatches.filter(t => t.searchType === 'substring').reduce((sum, t) => sum + t.count, 0);
const regexMatches = allMatches.filter(t => t.searchType === 'regex').reduce((sum, t) => sum + t.count, 0);
reason = `Found ${totalMatches} total matches (${titleMatches} in title, ${bodyMatches} in body) - ${keywordMatches} keyword matches, ${substringMatches} substring matches, ${regexMatches} regex matches`;
}
core.notice(`Final decision: ${shouldAddLabel ? 'ADD LABEL' : 'DO NOT ADD LABEL'}`);
core.notice(`Reason: ${reason || 'No matching terms found'}`);
if (shouldAddLabel) {
const existingLabels = context.payload.issue.labels.map(l => l.name);
if (!existingLabels.includes(labelName)) {
await github.rest.issues.addLabels({
owner: context.repo.owner,
repo: context.repo.repo,
issue_number: context.issue.number,
labels: [labelName],
});
core.notice(`Label "${labelName}" added. ${reason}`);
return true;
}
core.notice(`Label "${labelName}" already present.`);
return false;
}
core.notice(`No matching terms found for label "${labelName}".`);
return false;
}
// Process all configured labels
const processLabels = Object.entries(labelConfig)
.map(([labelName, config]) => processLabel(labelName, config));
const labelsAdded = await Promise.all(processLabels);
const numLabelsAdded = labelsAdded.reduce((x, y) => x + y, 0);
core.notice(`Processing complete. ${numLabelsAdded} label(s) added.`);

82
.github/workflows/lint-and-deploy.yaml vendored Normal file
View File

@ -0,0 +1,82 @@
name: Lint and Deploy Charts
on: pull_request
jobs:
lint-and-deploy:
runs-on: ubuntu-latest
steps:
- name: Checkout
uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
with:
fetch-depth: 0
- name: Set up Helm
uses: azure/setup-helm@b9e51907a09c216f16ebe8536097933489208112 # v4.3.0
with:
version: v3.14.4
#Python is required because ct lint runs Yamale and yamllint which require Python.
- uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0
with:
python-version: '3.13'
- name: Set up chart-testing
uses: helm/chart-testing-action@0d28d3144d3a25ea2cc349d6e59901c4ff469b3b # v2.7.0
with:
version: v3.10.1
- name: Run chart-testing (lint)
run: ct lint --target-branch ${{ github.event.repository.default_branch }} --chart-dirs examples/online_serving/chart-helm --charts examples/online_serving/chart-helm
- name: Setup minio
run: |
docker network create vllm-net
docker run -d -p 9000:9000 --name minio --net vllm-net \
-e "MINIO_ACCESS_KEY=minioadmin" \
-e "MINIO_SECRET_KEY=minioadmin" \
-v /tmp/data:/data \
-v /tmp/config:/root/.minio \
minio/minio server /data
export AWS_ACCESS_KEY_ID=minioadmin
export AWS_SECRET_ACCESS_KEY=minioadmin
export AWS_EC2_METADATA_DISABLED=true
mkdir opt-125m
cd opt-125m && curl -O -Ls "https://huggingface.co/facebook/opt-125m/resolve/main/{pytorch_model.bin,config.json,generation_config.json,merges.txt,special_tokens_map.json,tokenizer_config.json,vocab.json}" && cd ..
aws --endpoint-url http://127.0.0.1:9000/ s3 mb s3://testbucket
aws --endpoint-url http://127.0.0.1:9000/ s3 cp opt-125m/ s3://testbucket/opt-125m --recursive
- name: Create kind cluster
uses: helm/kind-action@a1b0e391336a6ee6713a0583f8c6240d70863de3 # v1.12.0
- name: Build the Docker image vllm cpu
run: docker buildx build -f docker/Dockerfile.cpu -t vllm-cpu-env .
- name: Configuration of docker images, network and namespace for the kind cluster
run: |
docker pull amazon/aws-cli:2.6.4
kind load docker-image amazon/aws-cli:2.6.4 --name chart-testing
kind load docker-image vllm-cpu-env:latest --name chart-testing
docker network connect vllm-net "$(docker ps -aqf "name=chart-testing-control-plane")"
kubectl create ns ns-vllm
- name: Run chart-testing (install)
run: |
export AWS_ACCESS_KEY_ID=minioadmin
export AWS_SECRET_ACCESS_KEY=minioadmin
sleep 30 && kubectl -n ns-vllm logs -f "$(kubectl -n ns-vllm get pods | awk '/deployment/ {print $1;exit}')" &
helm install --wait --wait-for-jobs --timeout 5m0s --debug --create-namespace --namespace=ns-vllm test-vllm examples/online_serving/chart-helm -f examples/online_serving/chart-helm/values.yaml --set secrets.s3endpoint=http://minio:9000 --set secrets.s3bucketname=testbucket --set secrets.s3accesskeyid=$AWS_ACCESS_KEY_ID --set secrets.s3accesskey=$AWS_SECRET_ACCESS_KEY --set resources.requests.cpu=1 --set resources.requests.memory=4Gi --set resources.limits.cpu=2 --set resources.limits.memory=5Gi --set image.env[0].name=VLLM_CPU_KVCACHE_SPACE --set image.env[1].name=VLLM_LOGGING_LEVEL --set-string image.env[0].value="1" --set-string image.env[1].value="DEBUG" --set-string extraInit.s3modelpath="opt-125m/" --set-string 'resources.limits.nvidia\.com/gpu=0' --set-string 'resources.requests.nvidia\.com/gpu=0' --set-string image.repository="vllm-cpu-env"
- name: curl test
run: |
kubectl -n ns-vllm port-forward service/test-vllm-service 8001:80 &
sleep 10
CODE="$(curl -v -f --location http://localhost:8001/v1/completions \
--header "Content-Type: application/json" \
--data '{
"model": "opt-125m",
"prompt": "San Francisco is a",
"max_tokens": 7,
"temperature": 0
}'):$CODE"
echo "$CODE"

View File

@ -1,17 +0,0 @@
{
"problemMatcher": [
{
"owner": "markdownlint",
"pattern": [
{
"regexp": "^([^:]*):(\\d+):?(\\d+)?\\s([\\w-\\/]*)\\s(.*)$",
"file": 1,
"line": 2,
"column": 3,
"code": 4,
"message": 5
}
]
}
]
}

View File

@ -5,23 +5,15 @@ on:
push:
branches: [main]
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
permissions:
contents: read
jobs:
pre-commit:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
- uses: actions/setup-python@e797f83bcb11b83ae66e0230d6156d7c80228e7c # v6.0.0
- uses: actions/setup-python@42375524e23c412d93fb67b49958b491fce71c38 # v5.4.0
with:
python-version: "3.12"
- run: echo "::add-matcher::.github/workflows/matchers/actionlint.json"
- run: echo "::add-matcher::.github/workflows/matchers/markdownlint.json"
- run: echo "::add-matcher::.github/workflows/matchers/mypy.json"
- uses: pre-commit/action@2c7b3805fd2a0fd8c1884dcaebf91fc102a13ecd # v3.0.1
with:

111
.github/workflows/publish.yml vendored Normal file
View File

@ -0,0 +1,111 @@
# This workflow will upload a Python Package to Release asset
# For more information see: https://help.github.com/en/actions/language-and-framework-guides/using-python-with-github-actions
name: Create Release
on:
push:
tags:
- v*
# Needed to create release and upload assets
permissions:
contents: write
jobs:
release:
# Retrieve tag and create release
name: Create Release
runs-on: ubuntu-latest
outputs:
upload_url: ${{ steps.create_release.outputs.upload_url }}
steps:
- name: Checkout
uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
- name: Extract branch info
shell: bash
run: |
echo "release_tag=${GITHUB_REF#refs/*/}" >> "$GITHUB_ENV"
- name: Create Release
id: create_release
uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
env:
RELEASE_TAG: ${{ env.release_tag }}
with:
github-token: "${{ secrets.GITHUB_TOKEN }}"
script: |
const script = require('.github/workflows/scripts/create_release.js')
await script(github, context, core)
# NOTE(simon): No longer build wheel using GitHub Actions. See buildkite's release workflow.
# wheel:
# name: Build Wheel
# runs-on: ${{ matrix.os }}
# needs: release
# strategy:
# fail-fast: false
# matrix:
# os: ['ubuntu-20.04']
# python-version: ['3.9', '3.10', '3.11', '3.12']
# pytorch-version: ['2.4.0'] # Must be the most recent version that meets requirements/cuda.txt.
# cuda-version: ['11.8', '12.1']
# steps:
# - name: Checkout
# uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
# - name: Setup ccache
# uses: hendrikmuhs/ccache-action@ed74d11c0b343532753ecead8a951bb09bb34bc9 # v1.2.14
# with:
# create-symlink: true
# key: ${{ github.job }}-${{ matrix.python-version }}-${{ matrix.cuda-version }}
# - name: Set up Linux Env
# if: ${{ runner.os == 'Linux' }}
# run: |
# bash -x .github/workflows/scripts/env.sh
# - name: Set up Python
# uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
# with:
# python-version: ${{ matrix.python-version }}
# - name: Install CUDA ${{ matrix.cuda-version }}
# run: |
# bash -x .github/workflows/scripts/cuda-install.sh ${{ matrix.cuda-version }} ${{ matrix.os }}
# - name: Install PyTorch ${{ matrix.pytorch-version }} with CUDA ${{ matrix.cuda-version }}
# run: |
# bash -x .github/workflows/scripts/pytorch-install.sh ${{ matrix.python-version }} ${{ matrix.pytorch-version }} ${{ matrix.cuda-version }}
# - name: Build wheel
# shell: bash
# env:
# CMAKE_BUILD_TYPE: Release # do not compile with debug symbol to reduce wheel size
# run: |
# bash -x .github/workflows/scripts/build.sh ${{ matrix.python-version }} ${{ matrix.cuda-version }}
# wheel_name=$(find dist -name "*whl" -print0 | xargs -0 -n 1 basename)
# asset_name=${wheel_name//"linux"/"manylinux1"}
# echo "wheel_name=${wheel_name}" >> "$GITHUB_ENV"
# echo "asset_name=${asset_name}" >> "$GITHUB_ENV"
# - name: Upload Release Asset
# uses: actions/upload-release-asset@e8f9f06c4b078e705bd2ea027f0926603fc9b4d5 # v1.0.2
# env:
# GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
# with:
# upload_url: ${{ needs.release.outputs.upload_url }}
# asset_path: ./dist/${{ env.wheel_name }}
# asset_name: ${{ env.asset_name }}
# asset_content_type: application/*
# (Danielkinz): This last step will publish the .whl to pypi. Warning: untested
# - name: Publish package
# uses: pypa/gh-action-pypi-publish@release/v1.8
# with:
# repository-url: https://test.pypi.org/legacy/
# password: ${{ secrets.PYPI_API_TOKEN }}
# skip-existing: true

View File

@ -1,6 +1,4 @@
name: PR Reminder Comment Bot
permissions:
pull-requests: write
on:
pull_request_target:
types: [opened]
@ -12,43 +10,16 @@ jobs:
uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
with:
script: |
try {
// Get the PR author
const prAuthor = context.payload.pull_request.user.login;
// Check if this is the author's first PR in this repository
// Use GitHub's search API to find all PRs by this author
const { data: searchResults } = await github.rest.search.issuesAndPullRequests({
q: `repo:${context.repo.owner}/${context.repo.repo} type:pr author:${prAuthor}`,
per_page: 100
});
const authorPRCount = searchResults.total_count;
console.log(`Found ${authorPRCount} PRs by ${prAuthor}`);
// Only post comment if this is the first PR (only one PR by this author)
if (authorPRCount === 1) {
console.log(`Posting welcome comment for first-time contributor: ${prAuthor}`);
await github.rest.issues.createComment({
owner: context.repo.owner,
repo: context.repo.repo,
issue_number: context.issue.number,
body: '👋 Hi! Thank you for contributing to the vLLM project.\n\n' +
'💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.\n\n' +
'Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run `fastcheck` CI which starts running only a small and essential subset of CI tests to quickly catch errors. \n\n' +
'You ask your reviewers to trigger select CI tests on top of `fastcheck` CI. \n\n' +
'Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.\n\n' +
'To run CI, PR reviewers can either: Add `ready` label to the PR or enable auto-merge.\n\n' +
'If you have any questions, please reach out to us on Slack at https://slack.vllm.ai.\n\n' +
'🚀'
});
} else {
console.log(`Skipping comment for ${prAuthor} - not their first PR (${authorPRCount} PRs found)`);
}
} catch (error) {
console.error('Error checking PR history or posting comment:', error);
// Don't fail the workflow, just log the error
}
github.rest.issues.createComment({
owner: context.repo.owner,
repo: context.repo.repo,
issue_number: context.issue.number,
body: '👋 Hi! Thank you for contributing to the vLLM project.\n\n' +
'💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.\n\n' +
'Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run `fastcheck` CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your `fastcheck` build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping `simon-mo` or `khluu` to add you in our Buildkite org.\n\n' +
'Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.\n\n' +
'To run CI, PR reviewers can either: Add `ready` label to the PR or enable auto-merge.\n\n' +
'🚀'
})
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

View File

@ -15,6 +15,7 @@ $python_executable -m pip install -r requirements/build.txt -r requirements/cuda
export MAX_JOBS=1
# Make sure release wheels are built for the following architectures
export TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 8.9 9.0+PTX"
export VLLM_FA_CMAKE_GPU_ARCHES="80-real;90-real"
bash tools/check_repo.sh

15
.gitignore vendored
View File

@ -4,9 +4,6 @@
# vllm-flash-attn built from source
vllm/vllm_flash_attn/*
# triton jit
.triton
# Byte-compiled / optimized / DLL files
__pycache__/
*.py[cod]
@ -80,6 +77,10 @@ instance/
# Scrapy stuff:
.scrapy
# Sphinx documentation
docs/_build/
docs/source/getting_started/examples/
# PyBuilder
.pybuilder/
target/
@ -149,9 +150,6 @@ venv.bak/
# mkdocs documentation
/site
docs/argparse
docs/examples/*
!docs/examples/README.md
# mypy
.mypy_cache/
@ -205,8 +203,5 @@ benchmarks/**/*.json
actionlint
shellcheck*/
# Ignore moe/marlin_moe gen code
# Ingore moe/marlin_moe gen code
csrc/moe/marlin_moe_wna16/kernel_*
# Ignore ep_kernels_workspace folder
ep_kernels_workspace/

View File

@ -1,13 +0,0 @@
MD007:
indent: 4
MD013: false
MD024:
siblings_only: true
MD033: false
MD042: false
MD045: false
MD046: false
MD051: false
MD052: false
MD053: false
MD059: false

View File

@ -11,59 +11,51 @@ repos:
hooks:
- id: yapf
args: [--in-place, --verbose]
# Keep the same list from yapfignore here to avoid yapf failing without any inputs
exclude: '(.buildkite|benchmarks|build|examples)/.*'
- repo: https://github.com/astral-sh/ruff-pre-commit
rev: v0.11.7
rev: v0.9.3
hooks:
- id: ruff
args: [--output-format, github, --fix]
- id: ruff-format
files: ^(.buildkite|benchmarks|examples)/.*
- repo: https://github.com/crate-ci/typos
rev: v1.35.5
- repo: https://github.com/codespell-project/codespell
rev: v2.4.0
hooks:
- id: typos
- id: codespell
additional_dependencies: ['tomli']
args: ['--toml', 'pyproject.toml']
- repo: https://github.com/PyCQA/isort
rev: 6.0.1
rev: 0a0b7a830386ba6a31c2ec8316849ae4d1b8240d # 6.0.0
hooks:
- id: isort
- repo: https://github.com/pre-commit/mirrors-clang-format
rev: v20.1.3
rev: v19.1.7
hooks:
- id: clang-format
exclude: 'csrc/(moe/topk_softmax_kernels.cu|quantization/gguf/(ggml-common.h|dequantize.cuh|vecdotq.cuh|mmq.cuh|mmvq.cuh))|vllm/third_party/.*'
types_or: [c++, cuda]
args: [--style=file, --verbose]
- repo: https://github.com/igorshubovych/markdownlint-cli
rev: v0.45.0
- repo: https://github.com/jackdewinter/pymarkdown
rev: v0.9.27
hooks:
- id: markdownlint
exclude: '.*\.inc\.md'
stages: [manual] # Only run in CI
- id: pymarkdown
args: [fix]
- repo: https://github.com/rhysd/actionlint
rev: v1.7.7
hooks:
- id: actionlint
- repo: https://github.com/astral-sh/uv-pre-commit
rev: 0.6.17
rev: 0.6.2
hooks:
- id: pip-compile
args: [requirements/test.in, -o, requirements/test.txt, --index-strategy, unsafe-best-match, --torch-backend, cu128]
args: [requirements/test.in, -o, requirements/test.txt]
files: ^requirements/test\.(in|txt)$
- repo: local
hooks:
- id: format-torch-nightly-test
name: reformat nightly_torch_test.txt to be in sync with test.in
language: python
entry: python tools/generate_nightly_torch_test.py
files: ^requirements/test\.(in|txt)$
- id: mypy-local
name: Run mypy for local Python installation
entry: tools/mypy.sh 0 "local"
language: python
types: [python]
additional_dependencies: &mypy_deps [mypy==1.11.1, types-cachetools, types-setuptools, types-PyYAML, types-requests, pydantic]
additional_dependencies: &mypy_deps [mypy==1.11.1, types-cachetools, types-setuptools, types-PyYAML, types-requests]
stages: [pre-commit] # Don't run in CI
- id: mypy-3.9 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
name: Run mypy for Python 3.9
@ -109,8 +101,8 @@ repos:
args:
- -c
- |
if ! grep -q "^Signed-off-by: $(git config user.name) <$(git config user.email)>" "$(git rev-parse --git-path COMMIT_EDITMSG)"; then
printf "\nSigned-off-by: $(git config user.name) <$(git config user.email)>\n" >> "$(git rev-parse --git-path COMMIT_EDITMSG)"
if ! grep -q "^Signed-off-by: $(git config user.name) <$(git config user.email)>" .git/COMMIT_EDITMSG; then
printf "\nSigned-off-by: $(git config user.name) <$(git config user.email)>\n" >> .git/COMMIT_EDITMSG
fi
language: system
verbose: true
@ -120,11 +112,6 @@ repos:
entry: python tools/check_spdx_header.py
language: python
types: [python]
- id: check-root-lazy-imports
name: Check root lazy imports
entry: python tools/check_init_lazy_imports.py
language: python
types: [python]
- id: check-filenames
name: Check for spaces in all filenames
entry: bash
@ -138,39 +125,12 @@ repos:
name: Update Dockerfile dependency graph
entry: tools/update-dockerfile-graph.sh
language: script
- id: enforce-import-regex-instead-of-re
name: Enforce import regex as re
entry: python tools/enforce_regex_import.py
language: python
types: [python]
files: ^docker/Dockerfile$
pass_filenames: false
additional_dependencies: [regex]
# forbid directly import triton
- id: forbid-direct-triton-import
name: "Forbid direct 'import triton'"
entry: python tools/check_triton_import.py
language: python
types: [python]
pass_filenames: false
additional_dependencies: [regex]
- id: check-pickle-imports
name: Prevent new pickle/cloudpickle imports
entry: python tools/check_pickle_imports.py
language: python
types: [python]
pass_filenames: false
additional_dependencies: [pathspec, regex]
- id: validate-config
name: Validate configuration has default values and that each field has a docstring
entry: python tools/validate_config.py
language: python
types: [python]
pass_filenames: true
files: vllm/config.py|tests/test_config.py|vllm/entrypoints/openai/cli_args.py
# Keep `suggestion` last
- id: suggestion
name: Suggestion
entry: bash -c 'echo "To bypass all the pre-commit hooks, add --no-verify to git commit. To skip a specific hook, prefix the commit command with SKIP=<hook-id>."'
entry: bash -c 'echo "To bypass pre-commit hooks, add --no-verify to git commit."'
language: system
verbose: true
pass_filenames: false

View File

@ -7,12 +7,13 @@ build:
os: ubuntu-22.04
tools:
python: "3.12"
jobs:
post_checkout:
- git fetch --unshallow || true
mkdocs:
configuration: mkdocs.yaml
sphinx:
configuration: docs/source/conf.py
fail_on_warning: true
# If using Sphinx, optionally build your docs in additional formats such as PDF
formats: []
# Optionally declare the Python requirements required to build your docs
python:

View File

@ -15,6 +15,7 @@ project(vllm_extensions LANGUAGES CXX)
# CUDA by default, can be overridden by using -DVLLM_TARGET_DEVICE=... (used by setup.py)
set(VLLM_TARGET_DEVICE "cuda" CACHE STRING "Target device backend for vLLM")
message(STATUS "Build type: ${CMAKE_BUILD_TYPE}")
message(STATUS "Target device: ${VLLM_TARGET_DEVICE}")
@ -23,14 +24,14 @@ include(${CMAKE_CURRENT_LIST_DIR}/cmake/utils.cmake)
# Suppress potential warnings about unused manually-specified variables
set(ignoreMe "${VLLM_PYTHON_PATH}")
# Prevent installation of dependencies (cutlass) by default.
install(CODE "set(CMAKE_INSTALL_LOCAL_ONLY TRUE)" ALL_COMPONENTS)
#
# Supported python versions. These versions will be searched in order, the
# first match will be selected. These should be kept in sync with setup.py.
#
set(PYTHON_SUPPORTED_VERSIONS "3.9" "3.10" "3.11" "3.12" "3.13")
set(PYTHON_SUPPORTED_VERSIONS "3.9" "3.10" "3.11" "3.12")
# Supported NVIDIA architectures.
set(CUDA_SUPPORTED_ARCHS "7.0;7.2;7.5;8.0;8.6;8.7;8.9;9.0;10.0;10.1;12.0")
# Supported AMD GPU architectures.
set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1101;gfx1200;gfx1201")
@ -45,8 +46,8 @@ set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1
# requirements.txt files and should be kept consistent. The ROCm torch
# versions are derived from docker/Dockerfile.rocm
#
set(TORCH_SUPPORTED_VERSION_CUDA "2.8.0")
set(TORCH_SUPPORTED_VERSION_ROCM "2.8.0")
set(TORCH_SUPPORTED_VERSION_CUDA "2.6.0")
set(TORCH_SUPPORTED_VERSION_ROCM "2.6.0")
#
# Try to find python package with an executable that exactly matches
@ -79,15 +80,6 @@ endif()
#
find_package(Torch REQUIRED)
# Supported NVIDIA architectures.
# This check must happen after find_package(Torch) because that's when CMAKE_CUDA_COMPILER_VERSION gets defined
if(DEFINED CMAKE_CUDA_COMPILER_VERSION AND
CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 12.8)
set(CUDA_SUPPORTED_ARCHS "7.0;7.2;7.5;8.0;8.6;8.7;8.9;9.0;10.0;10.1;12.0")
else()
set(CUDA_SUPPORTED_ARCHS "7.0;7.2;7.5;8.0;8.6;8.7;8.9;9.0")
endif()
#
# Forward the non-CUDA device extensions to external CMake scripts.
#
@ -171,6 +163,7 @@ if(NVCC_THREADS AND VLLM_GPU_LANG STREQUAL "CUDA")
list(APPEND VLLM_GPU_FLAGS "--threads=${NVCC_THREADS}")
endif()
#
# Use FetchContent for C++ dependencies that are compiled as part of vLLM's build process.
# setup.py will override FETCHCONTENT_BASE_DIR to play nicely with sccache.
@ -181,6 +174,9 @@ include(FetchContent)
file(MAKE_DIRECTORY ${FETCHCONTENT_BASE_DIR}) # Ensure the directory exists
message(STATUS "FetchContent base directory: ${FETCHCONTENT_BASE_DIR}")
#
# Set rocm version dev int.
#
if(VLLM_GPU_LANG STREQUAL "HIP")
#
# Overriding the default -O set up by cmake, adding ggdb3 for the most verbose devug info
@ -188,6 +184,7 @@ if(VLLM_GPU_LANG STREQUAL "HIP")
set(CMAKE_${VLLM_GPU_LANG}_FLAGS_DEBUG "${CMAKE_${VLLM_GPU_LANG}_FLAGS_DEBUG} -O0 -ggdb3")
set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -O0 -ggdb3")
#
# Certain HIP functions are marked as [[nodiscard]], yet vllm ignores the result which generates
# a lot of warnings that always mask real issues. Suppressing until this is properly addressed.
@ -230,33 +227,31 @@ endif()
#
set(VLLM_EXT_SRC
"csrc/mamba/mamba_ssm/selective_scan_fwd.cu"
"csrc/cache_kernels.cu"
"csrc/attention/paged_attention_v1.cu"
"csrc/attention/paged_attention_v2.cu"
"csrc/attention/merge_attn_states.cu"
"csrc/attention/vertical_slash_index.cu"
"csrc/pos_encoding_kernels.cu"
"csrc/activation_kernels.cu"
"csrc/layernorm_kernels.cu"
"csrc/layernorm_quant_kernels.cu"
"csrc/sampler.cu"
"csrc/cuda_view.cu"
"csrc/quantization/gptq/q_gemm.cu"
"csrc/quantization/compressed_tensors/int8_quant_kernels.cu"
"csrc/quantization/fp8/common.cu"
"csrc/quantization/fused_kernels/fused_layernorm_dynamic_per_token_quant.cu"
"csrc/quantization/gguf/gguf_kernel.cu"
"csrc/quantization/activation_kernels.cu"
"csrc/cuda_utils_kernels.cu"
"csrc/prepare_inputs/advance_step.cu"
"csrc/custom_all_reduce.cu"
"csrc/torch_bindings.cpp")
if(VLLM_GPU_LANG STREQUAL "CUDA")
SET(CUTLASS_ENABLE_HEADERS_ONLY ON CACHE BOOL "Enable only the header library")
# Set CUTLASS_REVISION. Used for FetchContent. Also fixes some bogus messages when building.
set(CUTLASS_REVISION "v4.0.0" CACHE STRING "CUTLASS revision to use")
# Set CUTLASS_REVISION manually -- its revision detection doesn't work in this case.
# Please keep this in sync with FetchContent_Declare line below.
set(CUTLASS_REVISION "v3.9.0" CACHE STRING "CUTLASS revision to use")
# Use the specified CUTLASS source directory for compilation if VLLM_CUTLASS_SRC_DIR is provided
if (DEFINED ENV{VLLM_CUTLASS_SRC_DIR})
@ -274,7 +269,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
cutlass
GIT_REPOSITORY https://github.com/nvidia/cutlass.git
# Please keep this in sync with CUTLASS_REVISION line above.
GIT_TAG ${CUTLASS_REVISION}
GIT_TAG v3.9.0
GIT_PROGRESS TRUE
# Speed up CUTLASS download by retrieving only the specified GIT_TAG instead of the history.
@ -286,16 +281,17 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
FetchContent_MakeAvailable(cutlass)
list(APPEND VLLM_EXT_SRC
"csrc/mamba/mamba_ssm/selective_scan_fwd.cu"
"csrc/mamba/causal_conv1d/causal_conv1d.cu"
"csrc/quantization/aqlm/gemm_kernels.cu"
"csrc/quantization/awq/gemm_kernels.cu"
"csrc/permute_cols.cu"
"csrc/quantization/cutlass_w8a8/scaled_mm_entry.cu"
"csrc/quantization/fp4/nvfp4_quant_entry.cu"
"csrc/quantization/fp4/nvfp4_scaled_mm_entry.cu"
"csrc/quantization/fp4/nvfp4_blockwise_moe_kernel.cu"
"csrc/sparse/cutlass/sparse_scaled_mm_entry.cu"
"csrc/cutlass_extensions/common.cpp"
"csrc/attention/mla/cutlass_mla_entry.cu"
"csrc/quantization/fp8/per_token_group_quant.cu")
"csrc/attention/mla/cutlass_mla_entry.cu")
set_gencode_flags_for_srcs(
SRCS "${VLLM_EXT_SRC}"
@ -304,72 +300,20 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
# Only build Marlin kernels if we are building for at least some compatible archs.
# Keep building Marlin for 9.0 as there are some group sizes and shapes that
# are not supported by Machete yet.
# 9.0 for latest bf16 atomicAdd PTX
cuda_archs_loose_intersection(MARLIN_ARCHS "8.0;8.7;9.0+PTX" "${CUDA_ARCHS}")
cuda_archs_loose_intersection(MARLIN_ARCHS "8.0;8.6;8.7;8.9;9.0;10.0;10.1;12.0" "${CUDA_ARCHS}")
if (MARLIN_ARCHS)
#
# For the Marlin kernels we automatically generate sources for various
# preselected input type pairs and schedules.
# Generate sources:
set(MARLIN_GEN_SCRIPT
${CMAKE_CURRENT_SOURCE_DIR}/csrc/quantization/gptq_marlin/generate_kernels.py)
file(MD5 ${MARLIN_GEN_SCRIPT} MARLIN_GEN_SCRIPT_HASH)
message(STATUS "Marlin generation script hash: ${MARLIN_GEN_SCRIPT_HASH}")
message(STATUS "Last run Marlin generate script hash: $CACHE{MARLIN_GEN_SCRIPT_HASH}")
if (NOT DEFINED CACHE{MARLIN_GEN_SCRIPT_HASH}
OR NOT $CACHE{MARLIN_GEN_SCRIPT_HASH} STREQUAL ${MARLIN_GEN_SCRIPT_HASH})
execute_process(
COMMAND ${CMAKE_COMMAND} -E env
PYTHONPATH=$PYTHONPATH
${Python_EXECUTABLE} ${MARLIN_GEN_SCRIPT}
RESULT_VARIABLE marlin_generation_result
OUTPUT_VARIABLE marlin_generation_result
OUTPUT_FILE ${CMAKE_CURRENT_BINARY_DIR}/marlin_generation.log
ERROR_FILE ${CMAKE_CURRENT_BINARY_DIR}/marlin_generation.log
)
if (NOT marlin_generation_result EQUAL 0)
message(FATAL_ERROR "Marlin generation failed."
" Result: \"${marlin_generation_result}\""
"\nCheck the log for details: "
"${CMAKE_CURRENT_BINARY_DIR}/marlin_generation.log")
else()
set(MARLIN_GEN_SCRIPT_HASH ${MARLIN_GEN_SCRIPT_HASH}
CACHE STRING "Last run Marlin generate script hash" FORCE)
message(STATUS "Marlin generation completed successfully.")
endif()
else()
message(STATUS "Marlin generation script has not changed, skipping generation.")
endif()
file(GLOB MARLIN_TEMPLATE_KERNEL_SRC "csrc/quantization/gptq_marlin/kernel_*.cu")
set_gencode_flags_for_srcs(
SRCS "${MARLIN_TEMPLATE_KERNEL_SRC}"
CUDA_ARCHS "${MARLIN_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8)
set_source_files_properties(${MARLIN_TEMPLATE_KERNEL_SRC}
PROPERTIES COMPILE_FLAGS "-static-global-template-stub=false")
endif()
list(APPEND VLLM_EXT_SRC ${MARLIN_TEMPLATE_KERNEL_SRC})
set(MARLIN_SRCS
"csrc/quantization/fp8/fp8_marlin.cu"
"csrc/quantization/marlin/dense/marlin_cuda_kernel.cu"
"csrc/quantization/marlin/sparse/marlin_24_cuda_kernel.cu"
"csrc/quantization/marlin/qqq/marlin_qqq_gemm_kernel.cu"
"csrc/quantization/gptq_marlin/gptq_marlin.cu"
"csrc/quantization/gptq_marlin/gptq_marlin_repack.cu"
"csrc/quantization/gptq_marlin/awq_marlin_repack.cu")
set_gencode_flags_for_srcs(
SRCS "${MARLIN_SRCS}"
CUDA_ARCHS "${MARLIN_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8)
set_source_files_properties("csrc/quantization/gptq_marlin/gptq_marlin.cu"
PROPERTIES COMPILE_FLAGS "-static-global-template-stub=false")
endif()
list(APPEND VLLM_EXT_SRC "${MARLIN_SRCS}")
message(STATUS "Building Marlin kernels for archs: ${MARLIN_ARCHS}")
else()
message(STATUS "Not building Marlin kernels as no compatible archs found"
@ -397,7 +341,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
# The cutlass_scaled_mm kernels for Hopper (c3x, i.e. CUTLASS 3.x) require
# CUDA 12.0 or later
cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a;" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0 AND SCALED_MM_ARCHS)
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.0 AND SCALED_MM_ARCHS)
set(SRCS
"csrc/quantization/cutlass_w8a8/scaled_mm_c3x_sm90.cu"
"csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm90_fp8.cu"
@ -413,7 +357,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
list(APPEND SCALED_MM_3X_ARCHS "${SCALED_MM_ARCHS}")
message(STATUS "Building scaled_mm_c3x_sm90 for archs: ${SCALED_MM_ARCHS}")
else()
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0 AND SCALED_MM_ARCHS)
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.0 AND SCALED_MM_ARCHS)
message(STATUS "Not building scaled_mm_c3x_sm90 as CUDA Compiler version is "
"not >= 12.0, we recommend upgrading to CUDA 12.0 or "
"later if you intend on running FP8 quantized models on "
@ -424,45 +368,13 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
endif()
endif()
# The cutlass_scaled_mm kernels for Geforce Blackwell SM120 (c3x, i.e. CUTLASS 3.x) require
# The cutlass_scaled_mm kernels for Blackwell (c3x, i.e. CUTLASS 3.x) require
# CUDA 12.8 or later
cuda_archs_loose_intersection(SCALED_MM_ARCHS "12.0;12.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
set(SRCS
"csrc/quantization/cutlass_w8a8/scaled_mm_c3x_sm120.cu"
"csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm120_fp8.cu"
"csrc/quantization/cutlass_w8a8/c3x/scaled_mm_blockwise_sm120_fp8.cu"
)
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${SCALED_MM_ARCHS}")
list(APPEND VLLM_EXT_SRC "${SRCS}")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_SCALED_MM_SM120=1")
# Let scaled_mm_c2x know it doesn't need to build these arches
list(APPEND SCALED_MM_3X_ARCHS "${SCALED_MM_ARCHS}")
message(STATUS "Building scaled_mm_c3x_sm120 for archs: ${SCALED_MM_ARCHS}")
else()
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
message(STATUS "Not building scaled_mm_c3x_sm120 as CUDA Compiler version is "
"not >= 12.8, we recommend upgrading to CUDA 12.8 or "
"later if you intend on running FP8 quantized models on "
"Blackwell.")
else()
message(STATUS "Not building scaled_mm_c3x_120 as no compatible archs found "
"in CUDA target architectures")
endif()
endif()
# The cutlass_scaled_mm kernels for Blackwell SM100 (c3x, i.e. CUTLASS 3.x)
# require CUDA 12.8 or later
cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a;10.1a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a;10.1a;12.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND SCALED_MM_ARCHS)
set(SRCS
"csrc/quantization/cutlass_w8a8/scaled_mm_c3x_sm100.cu"
"csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm100_fp8.cu"
"csrc/quantization/cutlass_w8a8/c3x/scaled_mm_blockwise_sm100_fp8.cu"
)
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
@ -473,7 +385,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
list(APPEND SCALED_MM_3X_ARCHS "${SCALED_MM_ARCHS}")
message(STATUS "Building scaled_mm_c3x_sm100 for archs: ${SCALED_MM_ARCHS}")
else()
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND SCALED_MM_ARCHS)
message(STATUS "Not building scaled_mm_c3x_sm100 as CUDA Compiler version is "
"not >= 12.8, we recommend upgrading to CUDA 12.8 or "
"later if you intend on running FP8 quantized models on "
@ -487,9 +399,8 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
#
# For the cutlass_scaled_mm kernels we want to build the c2x (CUTLASS 2.x)
# kernels for the remaining archs that are not already built for 3x.
# (Build 8.9 for FP8)
cuda_archs_loose_intersection(SCALED_MM_2X_ARCHS
"7.5;8.0;8.7;8.9+PTX" "${CUDA_ARCHS}")
"7.5;8.0;8.6;8.7;8.9;9.0;10.0;10.1;12.0" "${CUDA_ARCHS}")
# subtract out the archs that are already built for 3x
list(REMOVE_ITEM SCALED_MM_2X_ARCHS ${SCALED_MM_3X_ARCHS})
if (SCALED_MM_2X_ARCHS)
@ -516,7 +427,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
# The 2:4 sparse kernels cutlass_scaled_sparse_mm and cutlass_compressor
# require CUDA 12.2 or later (and only work on Hopper).
cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a;" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.2 AND SCALED_MM_ARCHS)
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.2 AND SCALED_MM_ARCHS)
set(SRCS "csrc/sparse/cutlass/sparse_scaled_mm_c3x.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
@ -525,7 +436,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_SPARSE_SCALED_MM_C3X=1")
message(STATUS "Building sparse_scaled_mm_c3x for archs: ${SCALED_MM_ARCHS}")
else()
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.2 AND SCALED_MM_ARCHS)
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.2 AND SCALED_MM_ARCHS)
message(STATUS "Not building sparse_scaled_mm_c3x kernels as CUDA Compiler version is "
"not >= 12.2, we recommend upgrading to CUDA 12.2 or later "
"if you intend on running FP8 sparse quantized models on Hopper.")
@ -535,41 +446,17 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
endif()
endif()
# The nvfp4_scaled_mm_sm120 kernels for Geforce Blackwell SM120 require
# CUDA 12.8 or later
cuda_archs_loose_intersection(FP4_ARCHS "12.0;12.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND FP4_ARCHS)
set(SRCS
"csrc/quantization/fp4/nvfp4_quant_kernels.cu"
"csrc/quantization/fp4/activation_nvfp4_quant_fusion_kernels.cu"
"csrc/quantization/fp4/nvfp4_scaled_mm_sm120_kernels.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${FP4_ARCHS}")
list(APPEND VLLM_EXT_SRC "${SRCS}")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_NVFP4_SM120=1")
message(STATUS "Building NVFP4 for archs: ${FP4_ARCHS}")
else()
message(STATUS "Not building NVFP4 as no compatible archs were found.")
# clear FP4_ARCHS
set(FP4_ARCHS)
endif()
# FP4 Archs and flags
cuda_archs_loose_intersection(FP4_ARCHS "10.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND FP4_ARCHS)
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND FP4_ARCHS)
set(SRCS
"csrc/quantization/fp4/nvfp4_quant_kernels.cu"
"csrc/quantization/fp4/activation_nvfp4_quant_fusion_kernels.cu"
"csrc/quantization/fp4/nvfp4_experts_quant.cu"
"csrc/quantization/fp4/nvfp4_scaled_mm_kernels.cu"
"csrc/quantization/fp4/nvfp4_blockwise_moe_kernel.cu")
"csrc/quantization/fp4/nvfp4_scaled_mm_kernels.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${FP4_ARCHS}")
list(APPEND VLLM_EXT_SRC "${SRCS}")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_NVFP4_SM100=1")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_CUTLASS_MOE_SM100=1")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_NVFP4=1")
message(STATUS "Building NVFP4 for archs: ${FP4_ARCHS}")
else()
message(STATUS "Not building NVFP4 as no compatible archs were found.")
@ -579,10 +466,9 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
# CUTLASS MLA Archs and flags
cuda_archs_loose_intersection(MLA_ARCHS "10.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND MLA_ARCHS)
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND MLA_ARCHS)
set(SRCS
"csrc/attention/mla/cutlass_mla_kernels.cu"
"csrc/attention/mla/sm100_cutlass_mla_kernel.cu")
"csrc/attention/mla/cutlass_mla_kernels.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${MLA_ARCHS}")
@ -600,12 +486,13 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
# CUTLASS MoE kernels
# The MoE kernel cutlass_moe_mm requires CUDA 12.3 or later (and ONLY works
# on Hopper). get_cutlass_(pplx_)moe_mm_data should only be compiled
# if it's possible to compile MoE kernels that use its output.
cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a" "${CUDA_ARCHS}")
# The MoE kernel cutlass_moe_mm requires CUDA 12.3 or later (and only works
# on Hopper). get_cutlass_moe_mm_data should only be compiled if it's possible
# to compile MoE kernels that use its output.
cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a;" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND SCALED_MM_ARCHS)
set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x_sm90.cu")
set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cu"
"csrc/quantization/cutlass_w8a8/moe/moe_data.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${SCALED_MM_ARCHS}")
@ -619,66 +506,6 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
"if you intend on running FP8 quantized MoE models on Hopper.")
else()
message(STATUS "Not building grouped_mm_c3x as no compatible archs found "
"in CUDA target architectures.")
endif()
endif()
cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x_sm100.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${SCALED_MM_ARCHS}")
list(APPEND VLLM_EXT_SRC "${SRCS}")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_CUTLASS_MOE_SM100=1")
message(STATUS "Building grouped_mm_c3x for archs: ${SCALED_MM_ARCHS}")
else()
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
message(STATUS "Not building grouped_mm_c3x kernels as CUDA Compiler version is "
"not >= 12.8, we recommend upgrading to CUDA 12.8 or later "
"if you intend on running FP8 quantized MoE models on Blackwell.")
else()
message(STATUS "Not building grouped_mm_c3x as no compatible archs found "
"in CUDA target architectures.")
endif()
endif()
# moe_data.cu is used by all CUTLASS MoE kernels.
cuda_archs_loose_intersection(CUTLASS_MOE_DATA_ARCHS "9.0a;10.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND CUTLASS_MOE_DATA_ARCHS)
set(SRCS "csrc/quantization/cutlass_w8a8/moe/moe_data.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${CUTLASS_MOE_DATA_ARCHS}")
list(APPEND VLLM_EXT_SRC "${SRCS}")
message(STATUS "Building moe_data for archs: ${CUTLASS_MOE_DATA_ARCHS}")
else()
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND CUTLASS_MOE_DATA_ARCHS)
message(STATUS "Not building moe_data as CUDA Compiler version is "
"not >= 12.3, we recommend upgrading to CUDA 12.3 or later "
"if you intend on running FP8 quantized MoE models on Hopper or Blackwell.")
else()
message(STATUS "Not building moe_data as no compatible archs found "
"in CUDA target architectures.")
endif()
endif()
cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
set(SRCS "csrc/quantization/cutlass_w8a8/moe/blockwise_scaled_group_mm_sm100.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${SCALED_MM_ARCHS}")
list(APPEND VLLM_EXT_SRC "${SRCS}")
list(APPEND VLLM_GPU_FLAGS "-DENABLE_CUTLASS_MOE_SM100=1")
message(STATUS "Building blockwise_scaled_group_mm_sm100 for archs: ${SCALED_MM_ARCHS}")
else()
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
message(STATUS "Not building blockwise_scaled_group_mm_sm100 kernels as CUDA Compiler version is "
"not >= 12.8, we recommend upgrading to CUDA 12.8 or later "
"if you intend on running FP8 quantized MoE models on Blackwell.")
else()
message(STATUS "Not building blockwise_scaled_group_mm_sm100 as no compatible archs found "
"in CUDA target architectures")
endif()
endif()
@ -689,7 +516,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
# The machete kernels only work on hopper and require CUDA 12.0 or later.
# Only build Machete kernels if we are building for something compatible with sm90a
cuda_archs_loose_intersection(MACHETE_ARCHS "9.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0 AND MACHETE_ARCHS)
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.0 AND MACHETE_ARCHS)
#
# For the Machete kernels we automatically generate sources for various
# preselected input type pairs and schedules.
@ -741,7 +568,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
message(STATUS "Building Machete kernels for archs: ${MACHETE_ARCHS}")
else()
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.0
AND MACHETE_ARCHS)
message(STATUS "Not building Machete kernels as CUDA Compiler version is "
"not >= 12.0, we recommend upgrading to CUDA 12.0 or "
@ -752,44 +579,9 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
"found in CUDA target architectures")
endif()
endif()
# Only build W4A8 kernels if we are building for something compatible with sm90a
cuda_archs_loose_intersection(W4A8_ARCHS "9.0a" "${CUDA_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0 AND W4A8_ARCHS)
set(SRCS
"csrc/quantization/cutlass_w4a8/w4a8_mm_entry.cu")
set_gencode_flags_for_srcs(
SRCS "${SRCS}"
CUDA_ARCHS "${W4A8_ARCHS}")
list(APPEND VLLM_EXT_SRC "${SRCS}")
message(STATUS "Building W4A8 kernels for archs: ${W4A8_ARCHS}")
else()
if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0
AND W4A8_ARCHS)
message(STATUS "Not building W4A8 kernels as CUDA Compiler version is "
"not >= 12.0, we recommend upgrading to CUDA 12.0 or "
"later if you intend on running w4a16 quantized models on "
"Hopper.")
else()
message(STATUS "Not building W4A8 kernels as no compatible archs "
"found in CUDA target architectures")
endif()
endif()
# if CUDA endif
endif()
if (VLLM_GPU_LANG STREQUAL "HIP")
# Add QuickReduce kernels
list(APPEND VLLM_EXT_SRC
"csrc/custom_quickreduce.cu"
)
# if ROCM endif
endif()
message(STATUS "Enabling C extension.")
define_gpu_extension_target(
_C
@ -819,17 +611,7 @@ set(VLLM_MOE_EXT_SRC
"csrc/moe/topk_softmax_kernels.cu")
if(VLLM_GPU_LANG STREQUAL "CUDA")
list(APPEND VLLM_MOE_EXT_SRC
"csrc/moe/moe_wna16.cu"
"csrc/moe/grouped_topk_kernels.cu")
endif()
if(VLLM_GPU_LANG STREQUAL "CUDA")
set(MOE_PERMUTE_SRC
"csrc/moe/permute_unpermute_kernels/moe_permute_unpermute_kernel.cu"
"csrc/moe/moe_permute_unpermute_op.cu")
list(APPEND VLLM_MOE_EXT_SRC "${MOE_PERMUTE_SRC}")
list(APPEND VLLM_MOE_EXT_SRC "csrc/moe/moe_wna16.cu")
endif()
set_gencode_flags_for_srcs(
@ -845,8 +627,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
CUDA_ARCHS "${CUDA_ARCHS}")
list(APPEND VLLM_MOE_EXT_SRC "${VLLM_MOE_WNA16_SRC}")
# 9.0 for latest bf16 atomicAdd PTX
cuda_archs_loose_intersection(MARLIN_MOE_ARCHS "8.0;8.7;9.0+PTX" "${CUDA_ARCHS}")
cuda_archs_loose_intersection(MARLIN_MOE_ARCHS "8.0;8.6;8.7;8.9;9.0;10.0;10.1;12.0" "${CUDA_ARCHS}")
if (MARLIN_MOE_ARCHS)
#
@ -864,7 +645,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
OR NOT $CACHE{MOE_MARLIN_GEN_SCRIPT_HASH} STREQUAL ${MOE_MARLIN_GEN_SCRIPT_HASH})
execute_process(
COMMAND ${CMAKE_COMMAND} -E env
PYTHONPATH=$PYTHONPATH
PYTHONPATH=${CMAKE_CURRENT_SOURCE_DIR}/csrc/cutlass_extensions/:${CUTLASS_DIR}/python/:${VLLM_PYTHON_PATH}:$PYTHONPATH
${Python_EXECUTABLE} ${MOE_MARLIN_GEN_SCRIPT}
RESULT_VARIABLE moe_marlin_generation_result
OUTPUT_VARIABLE moe_marlin_generation_output
@ -890,10 +671,6 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
set_gencode_flags_for_srcs(
SRCS "${MOE_WNAA16_MARLIN_SRC}"
CUDA_ARCHS "${MARLIN_MOE_ARCHS}")
if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8)
set_source_files_properties(${MOE_WNAA16_MARLIN_SRC}
PROPERTIES COMPILE_FLAGS "-static-global-template-stub=false")
endif()
list(APPEND VLLM_MOE_EXT_SRC ${MOE_WNAA16_MARLIN_SRC})
@ -912,8 +689,6 @@ define_gpu_extension_target(
SOURCES ${VLLM_MOE_EXT_SRC}
COMPILE_FLAGS ${VLLM_GPU_FLAGS}
ARCHITECTURES ${VLLM_GPU_ARCHES}
INCLUDE_DIRECTORIES ${CUTLASS_INCLUDE_DIR}
INCLUDE_DIRECTORIES ${CUTLASS_TOOLS_UTIL_INCLUDE_DIR}
USE_SABI 3
WITH_SOABI)
@ -940,7 +715,5 @@ endif()
# For CUDA we also build and ship some external projects.
if (VLLM_GPU_LANG STREQUAL "CUDA")
include(cmake/external_projects/flashmla.cmake)
# vllm-flash-attn should be last as it overwrites some CMake functions
include(cmake/external_projects/vllm_flash_attn.cmake)
endif ()

View File

@ -1,3 +1,3 @@
# Contributing to vLLM
You may find information about contributing to vLLM on [docs.vllm.ai](https://docs.vllm.ai/en/latest/contributing).
You may find information about contributing to vLLM on [docs.vllm.ai](https://docs.vllm.ai/en/latest/contributing/overview.html).

View File

@ -2,6 +2,7 @@ include LICENSE
include requirements/common.txt
include requirements/cuda.txt
include requirements/rocm.txt
include requirements/neuron.txt
include requirements/cpu.txt
include CMakeLists.txt

View File

@ -1,8 +1,7 @@
<!-- markdownlint-disable MD001 MD041 -->
<p align="center">
<picture>
<source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/vllm-project/vllm/main/docs/assets/logos/vllm-logo-text-dark.png">
<img alt="vLLM" src="https://raw.githubusercontent.com/vllm-project/vllm/main/docs/assets/logos/vllm-logo-text-light.png" width=55%>
<source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/vllm-project/vllm/main/docs/source/assets/logos/vllm-logo-text-dark.png">
<img alt="vLLM" src="https://raw.githubusercontent.com/vllm-project/vllm/main/docs/source/assets/logos/vllm-logo-text-light.png" width=55%>
</picture>
</p>
@ -17,26 +16,18 @@ Easy, fast, and cheap LLM serving for everyone
---
*Latest News* 🔥
- [2025/08] We hosted [vLLM Shenzhen Meetup](https://mp.weixin.qq.com/s/k8ZBO1u2_2odgiKWH_GVTQ) focusing on the ecosystem around vLLM! Please find the meetup slides [here](https://drive.google.com/drive/folders/1Ua2SVKVSu-wp5vou_6ElraDt2bnKhiEA).
- [2025/08] We hosted [vLLM Singapore Meetup](https://www.sginnovate.com/event/vllm-sg-meet). We shared V1 updates, disaggregated serving and MLLM speedups with speakers from Embedded LLM, AMD, WekaIO, and A*STAR. Please find the meetup slides [here](https://drive.google.com/drive/folders/1ncf3GyqLdqFaB6IeB834E5TZJPLAOiXZ?usp=sharing).
- [2025/08] We hosted [vLLM Shanghai Meetup](https://mp.weixin.qq.com/s/pDmAXHcN7Iqc8sUKgJgGtg) focusing on building, developing, and integrating with vLLM! Please find the meetup slides [here](https://drive.google.com/drive/folders/1OvLx39wnCGy_WKq8SiVKf7YcxxYI3WCH).
- [2025/05] vLLM is now a hosted project under PyTorch Foundation! Please find the announcement [here](https://pytorch.org/blog/pytorch-foundation-welcomes-vllm/).
- [2025/01] We are excited to announce the alpha release of vLLM V1: A major architectural upgrade with 1.7x speedup! Clean code, optimized execution loop, zero-overhead prefix caching, enhanced multimodal support, and more. Please check out our blog post [here](https://blog.vllm.ai/2025/01/27/v1-alpha-release.html).
<details>
<summary>Previous News</summary>
- [2025/08] We hosted [vLLM Korea Meetup](https://luma.com/cgcgprmh) with Red Hat and Rebellions! We shared the latest advancements in vLLM along with project spotlights from the vLLM Korea community. Please find the meetup slides [here](https://drive.google.com/file/d/1bcrrAE1rxUgx0mjIeOWT6hNe2RefC5Hm/view).
- [2025/08] We hosted [vLLM Beijing Meetup](https://mp.weixin.qq.com/s/dgkWg1WFpWGO2jCdTqQHxA) focusing on large-scale LLM deployment! Please find the meetup slides [here](https://drive.google.com/drive/folders/1Pid6NSFLU43DZRi0EaTcPgXsAzDvbBqF) and the recording [here](https://www.chaspark.com/#/live/1166916873711665152).
- [2025/05] We hosted [NYC vLLM Meetup](https://lu.ma/c1rqyf1f)! Please find the meetup slides [here](https://docs.google.com/presentation/d/1_q_aW_ioMJWUImf1s1YM-ZhjXz8cUeL0IJvaquOYBeA/edit?usp=sharing).
- [2025/04] We hosted [Asia Developer Day](https://www.sginnovate.com/event/limited-availability-morning-evening-slots-remaining-inaugural-vllm-asia-developer-day)! Please find the meetup slides from the vLLM team [here](https://docs.google.com/presentation/d/19cp6Qu8u48ihB91A064XfaXruNYiBOUKrBxAmDOllOo/edit?usp=sharing).
- [2025/03] We hosted [vLLM x Ollama Inference Night](https://lu.ma/vllm-ollama)! Please find the meetup slides from the vLLM team [here](https://docs.google.com/presentation/d/16T2PDD1YwRnZ4Tu8Q5r6n53c5Lr5c73UV9Vd2_eBo4U/edit?usp=sharing).
- [2025/03] We hosted [the first vLLM China Meetup](https://mp.weixin.qq.com/s/n77GibL2corAtQHtVEAzfg)! Please find the meetup slides from vLLM team [here](https://docs.google.com/presentation/d/1REHvfQMKGnvz6p3Fd23HhSO4c8j5WPGZV0bKYLwnHyQ/edit?usp=sharing).
- [2025/03] We hosted [the East Coast vLLM Meetup](https://lu.ma/7mu4k4xx)! Please find the meetup slides [here](https://docs.google.com/presentation/d/1NHiv8EUFF1NLd3fEYODm56nDmL26lEeXCaDgyDlTsRs/edit#slide=id.g31441846c39_0_0).
- [2025/02] We hosted [the ninth vLLM meetup](https://lu.ma/h7g3kuj9) with Meta! Please find the meetup slides from vLLM team [here](https://docs.google.com/presentation/d/1jzC_PZVXrVNSFVCW-V4cFXb6pn7zZ2CyP_Flwo05aqg/edit?usp=sharing) and AMD [here](https://drive.google.com/file/d/1Zk5qEJIkTmlQ2eQcXQZlljAx3m9s7nwn/view?usp=sharing). The slides from Meta will not be posted.
- [2025/01] We are excited to announce the alpha release of vLLM V1: A major architectural upgrade with 1.7x speedup! Clean code, optimized execution loop, zero-overhead prefix caching, enhanced multimodal support, and more. Please check out our blog post [here](https://blog.vllm.ai/2025/01/27/v1-alpha-release.html).
- [2025/01] We hosted [the eighth vLLM meetup](https://lu.ma/zep56hui) with Google Cloud! Please find the meetup slides from vLLM team [here](https://docs.google.com/presentation/d/1epVkt4Zu8Jz_S5OhEHPc798emsYh2BwYfRuDDVEF7u4/edit?usp=sharing), and Google Cloud team [here](https://drive.google.com/file/d/1h24pHewANyRL11xy5dXUbvRC9F9Kkjix/view?usp=sharing).
- [2024/12] vLLM joins [pytorch ecosystem](https://pytorch.org/blog/vllm-joins-pytorch)! Easy, Fast, and Cheap LLM Serving for Everyone!
<details>
<summary>Previous News</summary>
- [2024/11] We hosted [the seventh vLLM meetup](https://lu.ma/h0qvrajz) with Snowflake! Please find the meetup slides from vLLM team [here](https://docs.google.com/presentation/d/1e3CxQBV3JsfGp30SwyvS3eM_tW-ghOhJ9PAJGK6KR54/edit?usp=sharing), and Snowflake team [here](https://docs.google.com/presentation/d/1qF3RkDAbOULwz9WK5TOltt2fE9t6uIc_hVNLFAaQX6A/edit?usp=sharing).
- [2024/10] We have just created a developer slack ([slack.vllm.ai](https://slack.vllm.ai)) focusing on coordinating contributions and discussing features. Please feel free to join us there!
- [2024/10] Ray Summit 2024 held a special track for vLLM! Please find the opening talk slides from the vLLM team [here](https://docs.google.com/presentation/d/1B_KQxpHBTRa_mDF-tR6i8rWdOU5QoTZNcEg2MKZxEHM/edit?usp=sharing). Learn more from the [talks](https://www.youtube.com/playlist?list=PLzTswPQNepXl6AQwifuwUImLPFRVpksjR) from other vLLM contributors and users!
@ -53,7 +44,6 @@ Easy, fast, and cheap LLM serving for everyone
</details>
---
## About
vLLM is a fast and easy-to-use library for LLM inference and serving.
@ -66,27 +56,28 @@ vLLM is fast with:
- Efficient management of attention key and value memory with [**PagedAttention**](https://blog.vllm.ai/2023/06/20/vllm.html)
- Continuous batching of incoming requests
- Fast model execution with CUDA/HIP graph
- Quantizations: [GPTQ](https://arxiv.org/abs/2210.17323), [AWQ](https://arxiv.org/abs/2306.00978), [AutoRound](https://arxiv.org/abs/2309.05516), INT4, INT8, and FP8
- Optimized CUDA kernels, including integration with FlashAttention and FlashInfer
- Quantizations: [GPTQ](https://arxiv.org/abs/2210.17323), [AWQ](https://arxiv.org/abs/2306.00978), INT4, INT8, and FP8.
- Optimized CUDA kernels, including integration with FlashAttention and FlashInfer.
- Speculative decoding
- Chunked prefill
**Performance benchmark**: We include a performance benchmark at the end of [our blog post](https://blog.vllm.ai/2024/09/05/perf-update.html). It compares the performance of vLLM against other LLM serving engines ([TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM), [SGLang](https://github.com/sgl-project/sglang) and [LMDeploy](https://github.com/InternLM/lmdeploy)). The implementation is under [nightly-benchmarks folder](.buildkite/nightly-benchmarks/) and you can [reproduce](https://github.com/vllm-project/vllm/issues/8176) this benchmark using our one-click runnable script.
vLLM is flexible and easy to use with:
- Seamless integration with popular Hugging Face models
- High-throughput serving with various decoding algorithms, including *parallel sampling*, *beam search*, and more
- Tensor, pipeline, data and expert parallelism support for distributed inference
- Tensor parallelism and pipeline parallelism support for distributed inference
- Streaming outputs
- OpenAI-compatible API server
- Support NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs and GPUs, PowerPC CPUs, TPU, and AWS Neuron
- Support NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs and GPUs, PowerPC CPUs, TPU, and AWS Neuron.
- Prefix caching support
- Multi-LoRA support
- Multi-lora support
vLLM seamlessly supports most popular open-source models on HuggingFace, including:
- Transformer-like LLMs (e.g., Llama)
- Mixture-of-Expert LLMs (e.g., Mixtral, Deepseek-V2 and V3)
- Embedding Models (e.g., E5-Mistral)
- Embedding Models (e.g. E5-Mistral)
- Multi-modal LLMs (e.g., LLaVA)
Find the full list of supported models [here](https://docs.vllm.ai/en/latest/models/supported_models.html).
@ -100,7 +91,6 @@ pip install vllm
```
Visit our [documentation](https://docs.vllm.ai/en/latest/) to learn more.
- [Installation](https://docs.vllm.ai/en/latest/getting_started/installation.html)
- [Quickstart](https://docs.vllm.ai/en/latest/getting_started/quickstart.html)
- [List of Supported Models](https://docs.vllm.ai/en/latest/models/supported_models.html)
@ -108,16 +98,15 @@ Visit our [documentation](https://docs.vllm.ai/en/latest/) to learn more.
## Contributing
We welcome and value any contributions and collaborations.
Please check out [Contributing to vLLM](https://docs.vllm.ai/en/latest/contributing/index.html) for how to get involved.
Please check out [Contributing to vLLM](https://docs.vllm.ai/en/stable/contributing/overview.html) for how to get involved.
## Sponsors
vLLM is a community project. Our compute resources for development and testing are supported by the following organizations. Thank you for your support!
<!-- Note: Please sort them in alphabetical order. -->
<!-- Note: Please keep these consistent with docs/community/sponsors.md -->
<!-- Note: Please keep these consistent with docs/source/community/sponsors.md -->
Cash Donations:
- a16z
- Dropbox
- Sequoia Capital
@ -125,8 +114,6 @@ Cash Donations:
- ZhenFund
Compute Resources:
- Alibaba Cloud
- AMD
- Anyscale
- AWS
@ -165,14 +152,12 @@ If you use vLLM for your research, please cite our [paper](https://arxiv.org/abs
## Contact Us
<!-- --8<-- [start:contact-us] -->
- For technical questions and feature requests, please use GitHub [Issues](https://github.com/vllm-project/vllm/issues)
- For technical questions and feature requests, please use GitHub [Issues](https://github.com/vllm-project/vllm/issues) or [Discussions](https://github.com/vllm-project/vllm/discussions)
- For discussing with fellow users, please use the [vLLM Forum](https://discuss.vllm.ai)
- For coordinating contributions and development, please use [Slack](https://slack.vllm.ai)
- coordinating contributions and development, please use [Slack](https://slack.vllm.ai)
- For security disclosures, please use GitHub's [Security Advisories](https://github.com/vllm-project/vllm/security/advisories) feature
- For collaborations and partnerships, please contact us at [vllm-questions@lists.berkeley.edu](mailto:vllm-questions@lists.berkeley.edu)
<!-- --8<-- [end:contact-us] -->
## Media Kit
- If you wish to use vLLM's logo, please refer to [our media kit repo](https://github.com/vllm-project/media-kit)
- If you wish to use vLLM's logo, please refer to [our media kit repo](https://github.com/vllm-project/media-kit).

View File

@ -52,39 +52,3 @@ After branch cut, we approach finalizing the release branch with clear criteria
* Release branch specific changes (e.g. change version identifiers or CI fixes)
Please note: **No feature work allowed for cherry picks**. All PRs that are considered for cherry-picks need to be merged on trunk, the only exception are Release branch specific changes.
## Manual validations
### E2E Performance Validation
Before each release, we perform end-to-end performance validation to ensure no regressions are introduced. This validation uses the [vllm-benchmark workflow](https://github.com/pytorch/pytorch-integration-testing/actions/workflows/vllm-benchmark.yml) on PyTorch CI.
**Current Coverage:**
* Models: Llama3, Llama4, and Mixtral
* Hardware: NVIDIA H100 and AMD MI300x
* _Note: Coverage may change based on new model releases and hardware availability_
**Performance Validation Process:**
**Step 1: Get Access**
Request write access to the [pytorch/pytorch-integration-testing](https://github.com/pytorch/pytorch-integration-testing) repository to run the benchmark workflow.
**Step 2: Review Benchmark Setup**
Familiarize yourself with the benchmark configurations:
* [CUDA setup](https://github.com/pytorch/pytorch-integration-testing/tree/main/vllm-benchmarks/benchmarks/cuda)
* [ROCm setup](https://github.com/pytorch/pytorch-integration-testing/tree/main/vllm-benchmarks/benchmarks/rocm)
**Step 3: Run the Benchmark**
Navigate to the [vllm-benchmark workflow](https://github.com/pytorch/pytorch-integration-testing/actions/workflows/vllm-benchmark.yml) and configure:
* **vLLM branch**: Set to the release branch (e.g., `releases/v0.9.2`)
* **vLLM commit**: Set to the RC commit hash
**Step 4: Review Results**
Once the workflow completes, benchmark results will be available on the [vLLM benchmark dashboard](https://hud.pytorch.org/benchmark/llms?repoName=vllm-project%2Fvllm) under the corresponding branch and commit.
**Step 5: Performance Comparison**
Compare the current results against the previous release to verify no performance regressions have occurred. Here is an
example of [v0.9.1 vs v0.9.2](https://hud.pytorch.org/benchmark/llms?startTime=Thu%2C%2017%20Apr%202025%2021%3A43%3A50%20GMT&stopTime=Wed%2C%2016%20Jul%202025%2021%3A43%3A50%20GMT&granularity=week&lBranch=releases/v0.9.1&lCommit=b6553be1bc75f046b00046a4ad7576364d03c835&rBranch=releases/v0.9.2&rCommit=a5dd03c1ebc5e4f56f3c9d3dc0436e9c582c978f&repoName=vllm-project%2Fvllm&benchmarkName=&modelName=All%20Models&backendName=All%20Backends&modeName=All%20Modes&dtypeName=All%20DType&deviceName=All%20Devices&archName=All%20Platforms).

View File

@ -1,50 +1,11 @@
# Security Policy
## Reporting security issues
## Reporting a Vulnerability
Please report security issues privately using [the vulnerability submission form](https://github.com/vllm-project/vllm/security/advisories/new).
If you believe you have found a security vulnerability in vLLM, we encourage you to let us know right away. We will investigate all legitimate reports and do our best to quickly fix the problem.
## Issue triage
Please report security issues privately using [the vulnerability submission form](https://github.com/vllm-project/vllm/security/advisories/new). Reports will then be triaged by the [vulnerability management team](https://docs.vllm.ai/en/latest/contributing/vulnerability_management.html).
Reports will then be triaged by the [vulnerability management team](https://docs.vllm.ai/en/latest/contributing/vulnerability_management.html).
## Threat model
Please see the [Security Guide in the vLLM documentation](https://docs.vllm.ai/en/latest/usage/security.html) for more information on vLLM's security assumptions and recommendations.
---
Please see [PyTorch's Security Policy](https://github.com/pytorch/pytorch/blob/main/SECURITY.md) for more information and recommendations on how to securely interact with models.
## Issue severity
We will determine the risk of each issue, taking into account our experience dealing with past issues, versions affected, common defaults, and use cases. We use the following severity categories:
### CRITICAL Severity
Vulnerabilities that allow remote attackers to execute arbitrary code, take full control of the system, or significantly compromise confidentiality, integrity, or availability without any interaction or privileges needed, examples include remote code execution via network, deserialization issues that allow exploit chains. Generally those issues which are rated as CVSS ≥9.0.
### HIGH Severity
Serious security flaws that allow elevated impact—like RCE in specific, limited contexts or significant data loss—but require advanced conditions or some trust, examples include RCE in advanced deployment modes (e.g. multi-node), or high impact issues where some sort of privileged network access is required. These issues typically have CVSS scores between 7.0 and 8.9
### MODERATE Severity
Vulnerabilities that cause denial of service or partial disruption, but do not allow arbitrary code execution or data breach and have limited impact. These issues have a CVSS rating between 4.0 and 6.9
### LOW Severity
Minor issues such as informational disclosures, logging errors, non-exploitable flaws, or weaknesses that require local or high-privilege access and offer negligible impact. Examples include side channel attacks or hash collisions. These issues often have CVSS scores less than 4.0
## Prenotification policy
For certain security issues of CRITICAL, HIGH, or MODERATE severity level, we may prenotify certain organizations or vendors that ship vLLM. The purpose of this prenotification is to allow for a coordinated release of fixes for severe issues.
* This prenotification will be in the form of a private email notification. It may also include adding security contacts to the GitHub security advisory, typically a few days before release.
* If you wish to be added to the prenotification group, please send an email copying all the members of the [vulnerability management team](https://docs.vllm.ai/en/latest/contributing/vulnerability_management.html). Each vendor contact will be analyzed on a case-by-case basis.
* Organizations and vendors who either ship or use vLLM, are eligible to join the prenotification group if they meet at least one of the following qualifications
* Substantial internal deployment leveraging the upstream vLLM project.
* Established internal security teams and comprehensive compliance measures.
* Active and consistent contributions to the upstream vLLM project.
* We may withdraw organizations from receiving future prenotifications if they release fixes or any other information about issues before they are public. Group membership may also change based on policy refinements for who may be included.

View File

@ -22,25 +22,6 @@ become available.
<td style="text-align: center;"></td>
<td><code>wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json</code></td>
</tr>
<tr>
<td><strong>ShareGPT4V (Image)</strong></td>
<td style="text-align: center;"></td>
<td style="text-align: center;"></td>
<td>
<code>wget https://huggingface.co/datasets/Lin-Chen/ShareGPT4V/blob/main/sharegpt4v_instruct_gpt4-vision_cap100k.json</code>
<br>
<div>Note that the images need to be downloaded separately. For example, to download COCO's 2017 Train images:</div>
<code>wget http://images.cocodataset.org/zips/train2017.zip</code>
</td>
</tr>
<tr>
<td><strong>ShareGPT4Video (Video)</strong></td>
<td style="text-align: center;"></td>
<td style="text-align: center;"></td>
<td>
<code>git clone https://huggingface.co/datasets/ShareGPT4Video/ShareGPT4Video</code>
</td>
</tr>
<tr>
<td><strong>BurstGPT</strong></td>
<td style="text-align: center;"></td>
@ -48,7 +29,7 @@ become available.
<td><code>wget https://github.com/HPMLL/BurstGPT/releases/download/v1.1/BurstGPT_without_fails_2.csv</code></td>
</tr>
<tr>
<td><strong>Sonnet (deprecated)</strong></td>
<td><strong>Sonnet</strong></td>
<td style="text-align: center;"></td>
<td style="text-align: center;"></td>
<td>Local file: <code>benchmarks/sonnet.txt</code></td>
@ -59,18 +40,6 @@ become available.
<td style="text-align: center;"></td>
<td><code>synthetic</code></td>
</tr>
<tr>
<td><strong>RandomMultiModal (Image/Video)</strong></td>
<td style="text-align: center;">🟡</td>
<td style="text-align: center;">🚧</td>
<td><code>synthetic</code> </td>
</tr>
<tr>
<td><strong>Prefix Repetition</strong></td>
<td style="text-align: center;"></td>
<td style="text-align: center;"></td>
<td><code>synthetic</code></td>
</tr>
<tr>
<td><strong>HuggingFace-VisionArena</strong></td>
<td style="text-align: center;"></td>
@ -95,12 +64,6 @@ become available.
<td style="text-align: center;"></td>
<td><code>lmms-lab/LLaVA-OneVision-Data</code>, <code>Aeala/ShareGPT_Vicuna_unfiltered</code></td>
</tr>
<tr>
<td><strong>Custom</strong></td>
<td style="text-align: center;"></td>
<td style="text-align: center;"></td>
<td>Local file: <code>data.jsonl</code></td>
</tr>
</tbody>
</table>
@ -110,24 +73,15 @@ become available.
🚧: to be supported
**Note**: HuggingFace dataset's `dataset-name` should be set to `hf`.
For local `dataset-path`, please set `hf-name` to its Hugging Face ID like
**Note**: HuggingFace dataset's `dataset-name` should be set to `hf`
```bash
--dataset-path /datasets/VisionArena-Chat/ --hf-name lmarena-ai/VisionArena-Chat
```
## 🚀 Example - Online Benchmark
<details>
<summary>Show more</summary>
<br/>
---
## Example - Online Benchmark
First start serving your model
```bash
vllm serve NousResearch/Hermes-3-Llama-3.1-8B
vllm serve NousResearch/Hermes-3-Llama-3.1-8B --disable-log-requests
```
Then run the benchmarking script
@ -135,7 +89,7 @@ Then run the benchmarking script
```bash
# download dataset
# wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
vllm bench serve \
python3 vllm/benchmarks/benchmark_serving.py \
--backend vllm \
--model NousResearch/Hermes-3-Llama-3.1-8B \
--endpoint /v1/completions \
@ -146,74 +100,40 @@ vllm bench serve \
If successful, you will see the following output
```text
```
============ Serving Benchmark Result ============
Successful requests: 10
Benchmark duration (s): 5.78
Total input tokens: 1369
Total generated tokens: 2212
Request throughput (req/s): 1.73
Output token throughput (tok/s): 382.89
Total Token throughput (tok/s): 619.85
Successful requests: 10
Benchmark duration (s): 5.78
Total input tokens: 1369
Total generated tokens: 2212
Request throughput (req/s): 1.73
Output token throughput (tok/s): 382.89
Total Token throughput (tok/s): 619.85
---------------Time to First Token----------------
Mean TTFT (ms): 71.54
Median TTFT (ms): 73.88
P99 TTFT (ms): 79.49
Mean TTFT (ms): 71.54
Median TTFT (ms): 73.88
P99 TTFT (ms): 79.49
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms): 7.91
Median TPOT (ms): 7.96
P99 TPOT (ms): 8.03
Mean TPOT (ms): 7.91
Median TPOT (ms): 7.96
P99 TPOT (ms): 8.03
---------------Inter-token Latency----------------
Mean ITL (ms): 7.74
Median ITL (ms): 7.70
P99 ITL (ms): 8.39
Mean ITL (ms): 7.74
Median ITL (ms): 7.70
P99 ITL (ms): 8.39
==================================================
```
### Custom Dataset
If the dataset you want to benchmark is not supported yet in vLLM, even then you can benchmark on it using `CustomDataset`. Your data needs to be in `.jsonl` format and needs to have "prompt" field per entry, e.g., data.jsonl
```json
{"prompt": "What is the capital of India?"}
{"prompt": "What is the capital of Iran?"}
{"prompt": "What is the capital of China?"}
```
```bash
# start server
VLLM_USE_V1=1 vllm serve meta-llama/Llama-3.1-8B-Instruct
```
```bash
# run benchmarking script
vllm bench serve --port 9001 --save-result --save-detailed \
--backend vllm \
--model meta-llama/Llama-3.1-8B-Instruct \
--endpoint /v1/completions \
--dataset-name custom \
--dataset-path <path-to-your-data-jsonl> \
--custom-skip-chat-template \
--num-prompts 80 \
--max-concurrency 1 \
--temperature=0.3 \
--top-p=0.75 \
--result-dir "./log/"
```
You can skip applying chat template if your data already has it by using `--custom-skip-chat-template`.
### VisionArena Benchmark for Vision Language Models
```bash
# need a model with vision capability here
vllm serve Qwen/Qwen2-VL-7B-Instruct
vllm serve Qwen/Qwen2-VL-7B-Instruct --disable-log-requests
```
```bash
vllm bench serve \
python3 vllm/benchmarks/benchmark_serving.py \
--backend openai-chat \
--endpoint-type openai-chat \
--model Qwen/Qwen2-VL-7B-Instruct \
--endpoint /v1/chat/completions \
--dataset-name hf \
@ -226,13 +146,14 @@ vllm bench serve \
``` bash
VLLM_USE_V1=1 vllm serve meta-llama/Meta-Llama-3-8B-Instruct \
--speculative-config $'{"method": "ngram",
"num_speculative_tokens": 5, "prompt_lookup_max": 5,
"prompt_lookup_min": 2}'
--speculative-model "[ngram]" \
--ngram_prompt_lookup_min 2 \
--ngram-prompt-lookup-max 5 \
--num_speculative_tokens 5
```
``` bash
vllm bench serve \
python3 benchmarks/benchmark_serving.py \
--model meta-llama/Meta-Llama-3-8B-Instruct \
--dataset-name hf \
--dataset-path likaixin/InstructCoder \
@ -242,15 +163,14 @@ vllm bench serve \
### Other HuggingFaceDataset Examples
```bash
vllm serve Qwen/Qwen2-VL-7B-Instruct
vllm serve Qwen/Qwen2-VL-7B-Instruct --disable-log-requests
```
`lmms-lab/LLaVA-OneVision-Data`:
**`lmms-lab/LLaVA-OneVision-Data`**
```bash
vllm bench serve \
python3 vllm/benchmarks/benchmark_serving.py \
--backend openai-chat \
--endpoint-type openai-chat \
--model Qwen/Qwen2-VL-7B-Instruct \
--endpoint /v1/chat/completions \
--dataset-name hf \
@ -260,12 +180,11 @@ vllm bench serve \
--num-prompts 10
```
`Aeala/ShareGPT_Vicuna_unfiltered`:
**`Aeala/ShareGPT_Vicuna_unfiltered`**
```bash
vllm bench serve \
python3 vllm/benchmarks/benchmark_serving.py \
--backend openai-chat \
--endpoint-type openai-chat \
--model Qwen/Qwen2-VL-7B-Instruct \
--endpoint /v1/chat/completions \
--dataset-name hf \
@ -274,10 +193,10 @@ vllm bench serve \
--num-prompts 10
```
`AI-MO/aimo-validation-aime`:
**`AI-MO/aimo-validation-aime`**
``` bash
vllm bench serve \
python3 vllm/benchmarks/benchmark_serving.py \
--model Qwen/QwQ-32B \
--dataset-name hf \
--dataset-path AI-MO/aimo-validation-aime \
@ -285,23 +204,13 @@ vllm bench serve \
--seed 42
```
`philschmid/mt-bench`:
``` bash
vllm bench serve \
--model Qwen/QwQ-32B \
--dataset-name hf \
--dataset-path philschmid/mt-bench \
--num-prompts 80
```
### Running With Sampling Parameters
When using OpenAI-compatible backends such as `vllm`, optional sampling
parameters can be specified. Example client command:
```bash
vllm bench serve \
python3 vllm/benchmarks/benchmark_serving.py \
--backend vllm \
--model NousResearch/Hermes-3-Llama-3.1-8B \
--endpoint /v1/completions \
@ -313,34 +222,11 @@ vllm bench serve \
--num-prompts 10
```
### Running With Ramp-Up Request Rate
The benchmark tool also supports ramping up the request rate over the
duration of the benchmark run. This can be useful for stress testing the
server or finding the maximum throughput that it can handle, given some latency budget.
Two ramp-up strategies are supported:
- `linear`: Increases the request rate linearly from a start value to an end value.
- `exponential`: Increases the request rate exponentially.
The following arguments can be used to control the ramp-up:
- `--ramp-up-strategy`: The ramp-up strategy to use (`linear` or `exponential`).
- `--ramp-up-start-rps`: The request rate at the beginning of the benchmark.
- `--ramp-up-end-rps`: The request rate at the end of the benchmark.
</details>
## 📈 Example - Offline Throughput Benchmark
<details>
<summary>Show more</summary>
<br/>
---
## Example - Offline Throughput Benchmark
```bash
vllm bench throughput \
python3 vllm/benchmarks/benchmark_throughput.py \
--model NousResearch/Hermes-3-Llama-3.1-8B \
--dataset-name sonnet \
--dataset-path vllm/benchmarks/sonnet.txt \
@ -349,7 +235,7 @@ vllm bench throughput \
If successful, you will see the following output
```text
```
Throughput: 7.15 requests/s, 4656.00 total tokens/s, 1072.15 output tokens/s
Total num prompt tokens: 5014
Total num output tokens: 1500
@ -357,8 +243,8 @@ Total num output tokens: 1500
### VisionArena Benchmark for Vision Language Models
```bash
vllm bench throughput \
``` bash
python3 vllm/benchmarks/benchmark_throughput.py \
--model Qwen/Qwen2-VL-7B-Instruct \
--backend vllm-chat \
--dataset-name hf \
@ -369,7 +255,7 @@ vllm bench throughput \
The `num prompt tokens` now includes image token counts
```text
```
Throughput: 2.55 requests/s, 4036.92 total tokens/s, 326.90 output tokens/s
Total num prompt tokens: 14527
Total num output tokens: 1280
@ -380,7 +266,7 @@ Total num output tokens: 1280
``` bash
VLLM_WORKER_MULTIPROC_METHOD=spawn \
VLLM_USE_V1=1 \
vllm bench throughput \
python3 vllm/benchmarks/benchmark_throughput.py \
--dataset-name=hf \
--dataset-path=likaixin/InstructCoder \
--model=meta-llama/Meta-Llama-3-8B-Instruct \
@ -388,12 +274,13 @@ vllm bench throughput \
--output-len=100 \
--num-prompts=2048 \
--async-engine \
--speculative-config $'{"method": "ngram",
"num_speculative_tokens": 5, "prompt_lookup_max": 5,
"prompt_lookup_min": 2}'
--speculative-model="[ngram]" \
--ngram_prompt_lookup_min=2 \
--ngram-prompt-lookup-max=5 \
--num_speculative_tokens=5
```
```text
```
Throughput: 104.77 requests/s, 23836.22 total tokens/s, 10477.10 output tokens/s
Total num prompt tokens: 261136
Total num output tokens: 204800
@ -401,10 +288,10 @@ Total num output tokens: 204800
### Other HuggingFaceDataset Examples
`lmms-lab/LLaVA-OneVision-Data`:
**`lmms-lab/LLaVA-OneVision-Data`**
```bash
vllm bench throughput \
python3 vllm/benchmarks/benchmark_throughput.py \
--model Qwen/Qwen2-VL-7B-Instruct \
--backend vllm-chat \
--dataset-name hf \
@ -414,10 +301,10 @@ vllm bench throughput \
--num-prompts 10
```
`Aeala/ShareGPT_Vicuna_unfiltered`:
**`Aeala/ShareGPT_Vicuna_unfiltered`**
```bash
vllm bench throughput \
python3 vllm/benchmarks/benchmark_throughput.py \
--model Qwen/Qwen2-VL-7B-Instruct \
--backend vllm-chat \
--dataset-name hf \
@ -426,10 +313,10 @@ vllm bench throughput \
--num-prompts 10
```
`AI-MO/aimo-validation-aime`:
**`AI-MO/aimo-validation-aime`**
```bash
vllm bench throughput \
python3 benchmarks/benchmark_throughput.py \
--model Qwen/QwQ-32B \
--backend vllm \
--dataset-name hf \
@ -438,12 +325,12 @@ vllm bench throughput \
--num-prompts 10
```
Benchmark with LoRA adapters:
### Benchmark with LoRA Adapters
``` bash
# download dataset
# wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
vllm bench throughput \
python3 vllm/benchmarks/benchmark_throughput.py \
--model meta-llama/Llama-2-7b-hf \
--backend vllm \
--dataset_path <your data path>/ShareGPT_V3_unfiltered_cleaned_split.json \
@ -454,354 +341,3 @@ vllm bench throughput \
--enable-lora \
--lora-path yard1/llama-2-7b-sql-lora-test
```
</details>
## 🛠️ Example - Structured Output Benchmark
<details>
<summary>Show more</summary>
<br/>
Benchmark the performance of structured output generation (JSON, grammar, regex).
### Server Setup
```bash
vllm serve NousResearch/Hermes-3-Llama-3.1-8B
```
### JSON Schema Benchmark
```bash
python3 benchmarks/benchmark_serving_structured_output.py \
--backend vllm \
--model NousResearch/Hermes-3-Llama-3.1-8B \
--dataset json \
--structured-output-ratio 1.0 \
--request-rate 10 \
--num-prompts 1000
```
### Grammar-based Generation Benchmark
```bash
python3 benchmarks/benchmark_serving_structured_output.py \
--backend vllm \
--model NousResearch/Hermes-3-Llama-3.1-8B \
--dataset grammar \
--structure-type grammar \
--request-rate 10 \
--num-prompts 1000
```
### Regex-based Generation Benchmark
```bash
python3 benchmarks/benchmark_serving_structured_output.py \
--backend vllm \
--model NousResearch/Hermes-3-Llama-3.1-8B \
--dataset regex \
--request-rate 10 \
--num-prompts 1000
```
### Choice-based Generation Benchmark
```bash
python3 benchmarks/benchmark_serving_structured_output.py \
--backend vllm \
--model NousResearch/Hermes-3-Llama-3.1-8B \
--dataset choice \
--request-rate 10 \
--num-prompts 1000
```
### XGrammar Benchmark Dataset
```bash
python3 benchmarks/benchmark_serving_structured_output.py \
--backend vllm \
--model NousResearch/Hermes-3-Llama-3.1-8B \
--dataset xgrammar_bench \
--request-rate 10 \
--num-prompts 1000
```
</details>
## 📚 Example - Long Document QA Benchmark
<details>
<summary>Show more</summary>
<br/>
Benchmark the performance of long document question-answering with prefix caching.
### Basic Long Document QA Test
```bash
python3 benchmarks/benchmark_long_document_qa_throughput.py \
--model meta-llama/Llama-2-7b-chat-hf \
--enable-prefix-caching \
--num-documents 16 \
--document-length 2000 \
--output-len 50 \
--repeat-count 5
```
### Different Repeat Modes
```bash
# Random mode (default) - shuffle prompts randomly
python3 benchmarks/benchmark_long_document_qa_throughput.py \
--model meta-llama/Llama-2-7b-chat-hf \
--enable-prefix-caching \
--num-documents 8 \
--document-length 3000 \
--repeat-count 3 \
--repeat-mode random
# Tile mode - repeat entire prompt list in sequence
python3 benchmarks/benchmark_long_document_qa_throughput.py \
--model meta-llama/Llama-2-7b-chat-hf \
--enable-prefix-caching \
--num-documents 8 \
--document-length 3000 \
--repeat-count 3 \
--repeat-mode tile
# Interleave mode - repeat each prompt consecutively
python3 benchmarks/benchmark_long_document_qa_throughput.py \
--model meta-llama/Llama-2-7b-chat-hf \
--enable-prefix-caching \
--num-documents 8 \
--document-length 3000 \
--repeat-count 3 \
--repeat-mode interleave
```
</details>
## 🗂️ Example - Prefix Caching Benchmark
<details>
<summary>Show more</summary>
<br/>
Benchmark the efficiency of automatic prefix caching.
### Fixed Prompt with Prefix Caching
```bash
python3 benchmarks/benchmark_prefix_caching.py \
--model meta-llama/Llama-2-7b-chat-hf \
--enable-prefix-caching \
--num-prompts 1 \
--repeat-count 100 \
--input-length-range 128:256
```
### ShareGPT Dataset with Prefix Caching
```bash
# download dataset
# wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
python3 benchmarks/benchmark_prefix_caching.py \
--model meta-llama/Llama-2-7b-chat-hf \
--dataset-path /path/ShareGPT_V3_unfiltered_cleaned_split.json \
--enable-prefix-caching \
--num-prompts 20 \
--repeat-count 5 \
--input-length-range 128:256
```
### Prefix Repetition Dataset
```bash
vllm bench serve \
--backend openai \
--model meta-llama/Llama-2-7b-chat-hf \
--dataset-name prefix_repetition \
--num-prompts 100 \
--prefix-repetition-prefix-len 512 \
--prefix-repetition-suffix-len 128 \
--prefix-repetition-num-prefixes 5 \
--prefix-repetition-output-len 128
```
</details>
## ⚡ Example - Request Prioritization Benchmark
<details>
<summary>Show more</summary>
<br/>
Benchmark the performance of request prioritization in vLLM.
### Basic Prioritization Test
```bash
python3 benchmarks/benchmark_prioritization.py \
--model meta-llama/Llama-2-7b-chat-hf \
--input-len 128 \
--output-len 64 \
--num-prompts 100 \
--scheduling-policy priority
```
### Multiple Sequences per Prompt
```bash
python3 benchmarks/benchmark_prioritization.py \
--model meta-llama/Llama-2-7b-chat-hf \
--input-len 128 \
--output-len 64 \
--num-prompts 100 \
--scheduling-policy priority \
--n 2
```
</details>
## 👁️ Example - Multi-Modal Benchmark
<details>
<summary>Show more</summary>
<br/>
Benchmark the performance of multi-modal requests in vLLM.
### Images (ShareGPT4V)
Start vLLM:
```bash
python -m vllm.entrypoints.openai.api_server \
--model Qwen/Qwen2.5-VL-7B-Instruct \
--dtype bfloat16 \
--limit-mm-per-prompt '{"image": 1}' \
--allowed-local-media-path /path/to/sharegpt4v/images
```
Send requests with images:
```bash
python benchmarks/benchmark_serving.py \
--backend openai-chat \
--model Qwen/Qwen2.5-VL-7B-Instruct \
--dataset-name sharegpt \
--dataset-path /path/to/ShareGPT4V/sharegpt4v_instruct_gpt4-vision_cap100k.json \
--num-prompts 100 \
--save-result \
--result-dir ~/vllm_benchmark_results \
--save-detailed \
--endpoint /v1/chat/completion
```
### Videos (ShareGPT4Video)
Start vLLM:
```bash
python -m vllm.entrypoints.openai.api_server \
--model Qwen/Qwen2.5-VL-7B-Instruct \
--dtype bfloat16 \
--limit-mm-per-prompt '{"video": 1}' \
--allowed-local-media-path /path/to/sharegpt4video/videos
```
Send requests with videos:
```bash
python benchmarks/benchmark_serving.py \
--backend openai-chat \
--model Qwen/Qwen2.5-VL-7B-Instruct \
--dataset-name sharegpt \
--dataset-path /path/to/ShareGPT4Video/llava_v1_5_mix665k_with_video_chatgpt72k_share4video28k.json \
--num-prompts 100 \
--save-result \
--result-dir ~/vllm_benchmark_results \
--save-detailed \
--endpoint /v1/chat/completion
```
### Synthetic Random Images (random-mm)
Generate synthetic image inputs alongside random text prompts to stress-test vision models without external datasets.
Notes:
- Works only with online benchmark via the OpenAI backend (`--backend openai-chat`) and endpoint `/v1/chat/completions`.
- Video sampling is not yet implemented.
Start the server (example):
```bash
vllm serve Qwen/Qwen2.5-VL-3B-Instruct \
--dtype bfloat16 \
--max-model-len 16384 \
--limit-mm-per-prompt '{"image": 3, "video": 0}' \
--mm-processor-kwargs max_pixels=1003520
```
Benchmark. It is recommended to use the flag `--ignore-eos` to simulate real responses. You can set the size of the output via the arg `random-output-len`.
Ex.1: Fixed number of items and a single image resolution, enforcing generation of approx 40 tokens:
```bash
vllm bench serve \
--backend openai-chat \
--model Qwen/Qwen2.5-VL-3B-Instruct \
--endpoint /v1/chat/completions \
--dataset-name random-mm \
--num-prompts 100 \
--max-concurrency 10 \
--random-prefix-len 25 \
--random-input-len 300 \
--random-output-len 40 \
--random-range-ratio 0.2 \
--random-mm-base-items-per-request 2 \
--random-mm-limit-mm-per-prompt '{"image": 3, "video": 0}' \
--random-mm-bucket-config '{(224, 224, 1): 1.0}' \
--request-rate inf \
--ignore-eos \
--seed 42
```
The number of items per request can be controlled by passing multiple image buckets:
```bash
--random-mm-base-items-per-request 2 \
--random-mm-num-mm-items-range-ratio 0.5 \
--random-mm-limit-mm-per-prompt '{"image": 4, "video": 0}' \
--random-mm-bucket-config '{(256, 256, 1): 0.7, (720, 1280, 1): 0.3}' \
```
Flags specific to `random-mm`:
- `--random-mm-base-items-per-request`: base number of multimodal items per request.
- `--random-mm-num-mm-items-range-ratio`: vary item count uniformly in the closed integer range [floor(n·(1r)), ceil(n·(1+r))]. Set r=0 to keep it fixed; r=1 allows 0 items.
- `--random-mm-limit-mm-per-prompt`: per-modality hard caps, e.g. '{"image": 3, "video": 0}'.
- `--random-mm-bucket-config`: dict mapping (H, W, T) → probability. Entries with probability 0 are removed; remaining probabilities are renormalized to sum to 1. Use T=1 for images. Set any T>1 for videos (video sampling not yet supported).
Behavioral notes:
- If the requested base item count cannot be satisfied under the provided per-prompt limits, the tool raises an error rather than silently clamping.
How sampling works:
- Determine per-request item count k by sampling uniformly from the integer range defined by `--random-mm-base-items-per-request` and `--random-mm-num-mm-items-range-ratio`, then clamp k to at most the sum of per-modality limits.
- For each of the k items, sample a bucket (H, W, T) according to the normalized probabilities in `--random-mm-bucket-config`, while tracking how many items of each modality have been added.
- If a modality (e.g., image) reaches its limit from `--random-mm-limit-mm-per-prompt`, all buckets of that modality are excluded and the remaining bucket probabilities are renormalized before continuing.
This should be seen as an edge case, and if this behavior can be avoided by setting `--random-mm-limit-mm-per-prompt` to a large number. Note that this might result in errors due to engine config `--limit-mm-per-prompt`.
- The resulting request contains synthetic image data in `multi_modal_data` (OpenAI Chat format). When `random-mm` is used with the OpenAI Chat backend, prompts remain text and MM content is attached via `multi_modal_data`.
</details>

View File

@ -1,151 +0,0 @@
# Automated vLLM Server Parameter Tuning
This script automates the process of finding the optimal server parameter combination (`max-num-seqs` and `max-num-batched-tokens`) to maximize throughput for a vLLM server. It also supports additional constraints such as E2E latency and prefix cache hit rate.
## Table of Contents
- [Prerequisites](#prerequisites)
- [Configuration](#configuration)
- [How to Run](#how-to-run)
- [Example Use Cases](#example-use-cases)
- [Output](#output)
- [How It Works](#how-it-works)
## Prerequisites
Before running the script, please ensure the following steps are completed:
1. **Clone vLLM & Set Up Branch**: Clone the vLLM repository and check out to your desired branch.
```bash
git clone https://github.com/vllm-project/vllm.git
cd vllm
# git checkout <your-branch>
```
1. **Install Environment**: Install or update the correct running environment. For TPU usage, activate your `conda` environment and install the corresponding `torch` and `torch_xla` versions.
2. **Model Configuration**: If you are using a customized model, ensure its configuration files are correctly placed and accessible.
## Configuration
You must set the following variables at the top of the script before execution.
Note: You can also override the default values below via environment variables when running the script.
```bash
MODEL=meta-llama/Llama-3.3-70B-Instruct SYSTEM=TPU TP=8 DOWNLOAD_DIR='' INPUT_LEN=128 OUTPUT_LEN=2048 MAX_MODEL_LEN=2300 MIN_CACHE_HIT_PCT=0 MAX_LATENCY_ALLOWED_MS=100000000000 NUM_SEQS_LIST="128 256" NUM_BATCHED_TOKENS_LIST="1024 2048 4096" VLLM_LOGGING_LEVEL=DEBUG bash auto_tune.sh
```
| Variable | Description | Example Value |
| --- | --- | --- |
| `BASE` | **Required.** The absolute path to the parent directory of your vLLM repository directory. | `"$HOME"` |
| `MODEL` | **Required.** The Hugging Face model identifier to be served by vllm. | `"meta-llama/Llama-3.1-8B-Instruct"` |
| `SYSTEM`| **Required.** The hardware you are running on. Choices: `TPU` or `GPU`. (For other systems, it might not support saving profiles) | `"TPU"` |
| `TP` | **Required.** The tensor-parallelism size. | `1` |
| `DOWNLOAD_DIR` | **Required.** Directory to download and load model weights from. | `""` (default download path) |
| `INPUT_LEN` | **Required.** Request input length. | `4000` |
| `OUTPUT_LEN` | **Required.** Request output length. | `16` |
| `MAX_MODEL_LEN` | **Required.** Max model length. | `4096` |
| `MIN_CACHE_HIT_PCT` | Prefix cache hit rate in percentage (0-100). Set to `0` to disable. | `60` |
| `MAX_LATENCY_ALLOWED_MS` | The maximum allowed P99 end-to-end latency in milliseconds. Set to a very large number (e.g., `100000000000`) to effectively ignore the latency constraint. | `500` |
| `NUM_SEQS_LIST` | A space-separated string of `max-num-seqs` values to test. | `"128 256"` |
| `NUM_BATCHED_TOKENS_LIST` | A space-separated string of `max-num-batched-tokens` values to test. | `"1024 2048 4096"` |
**Note**: The default `NUM_SEQS_LIST` and `NUM_BATCHED_TOKENS_LIST` are set for medium-sized inputs/outputs. For very short contexts (e.g., 20 input, 20 output tokens), you may need to test larger values for `max-num-seqs`.
## How to Run
1. **Configure**: Edit the script and set the variables in the [Configuration](#configuration) section.
2. **Execute**: Run the script. Since the process can take a long time, it is highly recommended to use a terminal multiplexer like `tmux` or `screen` to prevent the script from stopping if your connection is lost.
```bash
cd <FOLDER_OF_THIS_SCRIPT>
bash auto_tune.sh
```
Please note that the `bash auto_tune.sh` command cannot contain full or partial path with keyword `vllm`, otherwise `pkill -f vllm` command will also kill this script itself.
## Example Use Cases
Here are a few examples of how to configure the script for different goals:
### 1. Maximize Throughput (No Latency Constraint)
- **Goal**: Find the best `max-num-seqs` and `max-num-batched-tokens` to get the highest possible throughput for 1800 input tokens and 20 output tokens.
- **Configuration**:
```bash
INPUT_LEN=1800
OUTPUT_LEN=20
MAX_MODEL_LEN=2048
MIN_CACHE_HIT_PCT=0
MAX_LATENCY_ALLOWED_MS=100000000000 # A very large number
```
#### 2. Maximize Throughput with a Latency Requirement
- **Goal**: Find the best server parameters when P99 end-to-end latency must be below 500ms.
- **Configuration**:
```bash
INPUT_LEN=1800
OUTPUT_LEN=20
MAX_MODEL_LEN=2048
MIN_CACHE_HIT_PCT=0
MAX_LATENCY_ALLOWED_MS=500
```
#### 3. Maximize Throughput with Prefix Caching and Latency Requirements
- **Goal**: Find the best server parameters assuming a 60% prefix cache hit rate and a latency requirement of 500ms.
- **Configuration**:
```bash
INPUT_LEN=1800
OUTPUT_LEN=20
MAX_MODEL_LEN=2048
MIN_CACHE_HIT_PCT=60
MAX_LATENCY_ALLOWED_MS=500
```
## Output
After the script finishes, you will find the results in a new, timestamped directory created inside `$BASE/auto-benchmark/`.
- **Log Files**: The directory (`$BASE/auto-benchmark/YYYY_MM_DD_HH_MM/`) contains detailed logs for each run:
- `vllm_log_...txt`: The log output from the vLLM server for each parameter combination.
- `bm_log_...txt`: The log output from the `vllm bench serve` command for each benchmark run.
- **Final Result Summary**: A file named `result.txt` is created in the log directory. It contains a summary of each tested combination and concludes with the overall best parameters found.
```text
# Example result.txt content
hash:a1b2c3d4...
max_num_seqs: 128, max_num_batched_tokens: 2048, request_rate: 10.0, e2el: 450.5, throughput: 9.8, goodput: 9.8
max_num_seqs: 128, max_num_batched_tokens: 4096 does not meet latency requirement 500
...
best_max_num_seqs: 256, best_num_batched_tokens: 2048, best_throughput: 12.5, profile saved in: /home/user/vllm/auto-benchmark/2024_08_01_10_30/profile
```
If it cannot find the best parameters, the final row will be `best_max_num_seqs: 0, best_num_batched_tokens: 0, best_throughput: 0`. This can be due to either the server not starting properly, or the latency requirement being too strict.
- **Profiler Trace**: A directory named `profile` is created inside the log directory. It contains the profiler trace file (e.g., `.xplane.pb` for TPU or a `.json` trace for GPU) from the single best-performing run.
## How It Works
The script follows a systematic process to find the optimal parameters:
1. **Find Max GPU Memory Utilization**: The script first determines the highest safe `gpu-memory-utilization` (starting from 0.98 and decreasing) that does not cause an Out-Of-Memory (OOM) error when launching the server. This ensures the benchmark runs use the maximum available memory without crashing.
2. **Iterate and Benchmark**: It then enters a nested loop, iterating through every combination of `max-num-seqs` and `max-num-batched-tokens` provided in the configuration lists.
3. **Latency-Aware Throughput Search**: For each parameter combination:
- The vLLM server is started.
- A benchmark is first run with an infinite request rate (`--request-rate inf`).
- If the resulting P99 E2E latency is within the `MAX_LATENCY_ALLOWED_MS` limit, this throughput is considered the maximum for this configuration.
- If the latency is too high, the script performs a search by iteratively decreasing the request rate until the latency constraint is met. This finds the highest sustainable throughput for the given parameters and latency requirement.
4. **Track Best Result**: Throughout the process, the script tracks the parameter combination that has yielded the highest valid throughput so far.
5. **Profile Collection**: For the best-performing run, the script saves the vLLM profiler output, which can be used for deep-dive performance analysis with tools like TensorBoard.

View File

@ -1,308 +0,0 @@
#!/bin/bash
# This script aims to tune the best server parameter combinations to maximize throughput for given requirement.
# See details in README (benchmarks/auto_tune/README.md).
TAG=$(date +"%Y_%m_%d_%H_%M")
SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
VLLM_LOGGING_LEVEL=${VLLM_LOGGING_LEVEL:-INFO}
BASE=${BASE:-"$SCRIPT_DIR/../../.."}
MODEL=${MODEL:-"meta-llama/Llama-3.1-8B-Instruct"}
SYSTEM=${SYSTEM:-"TPU"}
TP=${TP:-1}
DOWNLOAD_DIR=${DOWNLOAD_DIR:-""}
INPUT_LEN=${INPUT_LEN:-4000}
OUTPUT_LEN=${OUTPUT_LEN:-16}
MAX_MODEL_LEN=${MAX_MODEL_LEN:-4096}
MIN_CACHE_HIT_PCT=${MIN_CACHE_HIT_PCT:-0}
MAX_LATENCY_ALLOWED_MS=${MAX_LATENCY_ALLOWED_MS:-100000000000}
NUM_SEQS_LIST=${NUM_SEQS_LIST:-"128 256"}
NUM_BATCHED_TOKENS_LIST=${NUM_BATCHED_TOKENS_LIST:-"512 1024 2048 4096"}
LOG_FOLDER="$BASE/auto-benchmark/$TAG"
RESULT="$LOG_FOLDER/result.txt"
PROFILE_PATH="$LOG_FOLDER/profile"
echo "====================== AUTO TUNE PARAMETERS ===================="
echo "SCRIPT_DIR=$SCRIPT_DIR"
echo "BASE=$BASE"
echo "MODEL=$MODEL"
echo "SYSTEM=$SYSTEM"
echo "TP=$TP"
echo "DOWNLOAD_DIR=$DOWNLOAD_DIR"
echo "INPUT_LEN=$INPUT_LEN"
echo "OUTPUT_LEN=$OUTPUT_LEN"
echo "MAX_MODEL_LEN=$MAX_MODEL_LEN"
echo "MIN_CACHE_HIT_PCT=$MIN_CACHE_HIT_PCT"
echo "MAX_LATENCY_ALLOWED_MS=$MAX_LATENCY_ALLOWED_MS"
echo "NUM_SEQS_LIST=$NUM_SEQS_LIST"
echo "NUM_BATCHED_TOKENS_LIST=$NUM_BATCHED_TOKENS_LIST"
echo "VLLM_LOGGING_LEVEL=$VLLM_LOGGING_LEVEL"
echo "RESULT_FILE=$RESULT"
echo "====================== AUTO TUNEPARAMETERS ===================="
rm -rf $LOG_FOLDER
rm -rf $PROFILE_PATH
mkdir -p $LOG_FOLDER
mkdir -p $PROFILE_PATH
cd "$BASE/vllm"
pip install -q datasets
current_hash=$(git rev-parse HEAD)
echo "hash:$current_hash" >> "$RESULT"
echo "current_hash: $current_hash"
TOTAL_LEN=$((INPUT_LEN + OUTPUT_LEN))
RED='\033[0;31m'
if (( TOTAL_LEN > MAX_MODEL_LEN )); then
echo -e "${RED}FAILED: INPUT_LEN($INPUT_LEN) + OUTPUT_LEN($OUTPUT_LEN) = $TOTAL_LEN, which is > MAX_MODEL_LEN = $MAX_MODEL_LEN.\033[0m" >&2
exit 1
fi
best_throughput=0
best_max_num_seqs=0
best_num_batched_tokens=0
best_goodput=0
best_request_rate=0
start_server() {
local gpu_memory_utilization=$1
local max_num_seqs=$2
local max_num_batched_tokens=$3
local vllm_log=$4
local profile_dir=$5
pkill -if vllm
# Define the common arguments as a bash array.
# Each argument and its value are separate elements.
local common_args_array=(
"$MODEL"
"--disable-log-requests"
"--port" "8004"
"--gpu-memory-utilization" "$gpu_memory_utilization"
"--max-num-seqs" "$max_num_seqs"
"--max-num-batched-tokens" "$max_num_batched_tokens"
"--tensor-parallel-size" "$TP"
"--enable-prefix-caching"
"--load-format" "dummy"
"--download-dir" "$DOWNLOAD_DIR"
"--max-model-len" "$MAX_MODEL_LEN"
)
# Use the array expansion "${common_args_array[@]}"
# This correctly passes each element as a separate argument.
if [[ -n "$profile_dir" ]]; then
# Start server with profiling enabled
VLLM_USE_V1=1 VLLM_SERVER_DEV_MODE=1 VLLM_TORCH_PROFILER_DIR=$profile_dir \
vllm serve "${common_args_array[@]}" > "$vllm_log" 2>&1 &
else
# Start server without profiling
VLLM_USE_V1=1 VLLM_SERVER_DEV_MODE=1 \
vllm serve "${common_args_array[@]}" > "$vllm_log" 2>&1 &
fi
# wait for 10 minutes...
server_started=0
for i in {1..60}; do
RESPONSE=$(curl -s -X GET "http://0.0.0.0:8004/health" -w "%{http_code}" -o /dev/stdout)
STATUS_CODE=$(echo "$RESPONSE" | tail -n 1)
if [[ "$STATUS_CODE" -eq 200 ]]; then
server_started=1
break
else
sleep 10
fi
done
if (( ! server_started )); then
echo "server did not start within 10 minutes. Please check server log at $vllm_log".
return 1
else
return 0
fi
}
run_benchmark() {
local max_num_seqs=$1
local max_num_batched_tokens=$2
local gpu_memory_utilization=$3
echo "max_num_seq: $max_num_seqs, max_num_batched_tokens: $max_num_batched_tokens"
local vllm_log="$LOG_FOLDER/vllm_log_${max_num_seqs}_${max_num_batched_tokens}.txt"
echo "vllm_log: $vllm_log"
echo
rm -f $vllm_log
pkill -if vllm
echo "starting server..."
# Call start_server without a profile_dir to avoid profiling overhead
start_server $gpu_memory_utilization $max_num_seqs $max_num_batched_tokens $vllm_log ""
result=$?
if [[ "$result" -eq 1 ]]; then
echo "server failed to start. gpu_memory_utilization:$gpu_memory_utilization, max_num_seqs:$max_num_seqs, max_num_batched_tokens: $max_num_batched_tokens"
else
echo "server started."
fi
echo
echo "run benchmark test..."
meet_latency_requirement=0
# get a basic qps by using request-rate inf
bm_log="$LOG_FOLDER/bm_log_${max_num_seqs}_${max_num_batched_tokens}_requestrate_inf.txt"
prefix_len=$(( INPUT_LEN * MIN_CACHE_HIT_PCT / 100 ))
adjusted_input_len=$(( INPUT_LEN - prefix_len ))
# --profile flag is removed from this call
vllm bench serve \
--backend vllm \
--model $MODEL \
--dataset-name random \
--random-input-len $adjusted_input_len \
--random-output-len $OUTPUT_LEN \
--ignore-eos \
--disable-tqdm \
--request-rate inf \
--percentile-metrics ttft,tpot,itl,e2el \
--goodput e2el:$MAX_LATENCY_ALLOWED_MS \
--num-prompts 1000 \
--random-prefix-len $prefix_len \
--port 8004 &> "$bm_log"
throughput=$(grep "Request throughput (req/s):" "$bm_log" | sed 's/[^0-9.]//g')
e2el=$(grep "P99 E2EL (ms):" "$bm_log" | awk '{print $NF}')
goodput=$(grep "Request goodput (req/s):" "$bm_log" | sed 's/[^0-9.]//g')
if (( $(echo "$e2el <= $MAX_LATENCY_ALLOWED_MS" | bc -l) )); then
meet_latency_requirement=1
request_rate=inf
fi
if (( ! meet_latency_requirement )); then
# start from request-rate as int(throughput) + 1
request_rate=$((${throughput%.*} + 1))
while ((request_rate > 0)); do
# clear prefix cache
curl -X POST http://0.0.0.0:8004/reset_prefix_cache
sleep 5
bm_log="$LOG_FOLDER/bm_log_${max_num_seqs}_${max_num_batched_tokens}_requestrate_${request_rate}.txt"
vllm bench serve \
--backend vllm \
--model $MODEL \
--dataset-name random \
--random-input-len $adjusted_input_len \
--random-output-len $OUTPUT_LEN \
--ignore-eos \
--disable-tqdm \
--request-rate $request_rate \
--percentile-metrics ttft,tpot,itl,e2el \
--goodput e2el:$MAX_LATENCY_ALLOWED_MS \
--num-prompts 100 \
--random-prefix-len $prefix_len \
--port 8004 &> "$bm_log"
throughput=$(grep "Request throughput (req/s):" "$bm_log" | sed 's/[^0-9.]//g')
e2el=$(grep "P99 E2EL (ms):" "$bm_log" | awk '{print $NF}')
goodput=$(grep "Request goodput (req/s):" "$bm_log" | sed 's/[^0-9.]//g')
if (( $(echo "$e2el <= $MAX_LATENCY_ALLOWED_MS" | bc -l) )); then
meet_latency_requirement=1
break
fi
request_rate=$((request_rate-1))
done
fi
# write the results and update the best result.
if ((meet_latency_requirement)); then
echo "max_num_seqs: $max_num_seqs, max_num_batched_tokens: $max_num_batched_tokens, request_rate: $request_rate, e2el: $e2el, throughput: $throughput, goodput: $goodput"
echo "max_num_seqs: $max_num_seqs, max_num_batched_tokens: $max_num_batched_tokens, request_rate: $request_rate, e2el: $e2el, throughput: $throughput, goodput: $goodput" >> "$RESULT"
if (( $(echo "$throughput > $best_throughput" | bc -l) )); then
best_throughput=$throughput
best_max_num_seqs=$max_num_seqs
best_num_batched_tokens=$max_num_batched_tokens
best_goodput=$goodput
best_request_rate=$request_rate
fi
else
echo "max_num_seqs: $max_num_seqs, max_num_batched_tokens: $max_num_batched_tokens does not meet latency requirement ${MAX_LATENCY_ALLOWED_MS}"
echo "max_num_seqs: $max_num_seqs, max_num_batched_tokens: $max_num_batched_tokens does not meet latency requirement ${MAX_LATENCY_ALLOWED_MS}" >> "$RESULT"
fi
echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput"
pkill -if vllm
sleep 10
echo "===================="
return 0
}
read -r -a num_seqs_list <<< "$NUM_SEQS_LIST"
read -r -a num_batched_tokens_list <<< "$NUM_BATCHED_TOKENS_LIST"
# first find out the max gpu-memory-utilization without HBM OOM.
gpu_memory_utilization=0.98
find_gpu_memory_utilization=0
while (( $(echo "$gpu_memory_utilization >= 0.9" | bc -l) )); do
# Pass empty string for profile_dir argument
start_server $gpu_memory_utilization "${num_seqs_list[-1]}" "${num_batched_tokens_list[-1]}" "$LOG_FOLDER/vllm_log_gpu_memory_utilization_$gpu_memory_utilization.log" ""
result=$?
if [[ "$result" -eq 0 ]]; then
find_gpu_memory_utilization=1
break
else
gpu_memory_utilization=$(echo "$gpu_memory_utilization - 0.01" | bc)
fi
done
if [[ "$find_gpu_memory_utilization" -eq 1 ]]; then
echo "Using gpu_memory_utilization=$gpu_memory_utilization to serve model."
else
echo "Cannot find a proper gpu_memory_utilization over 0.9 to serve the model, please check logs in $LOG_FOLDER."
exit 1
fi
for num_seqs in "${num_seqs_list[@]}"; do
for num_batched_tokens in "${num_batched_tokens_list[@]}"; do
run_benchmark $num_seqs $num_batched_tokens $gpu_memory_utilization
done
done
echo "finish permutations"
# =================================================================================
# FINAL PROFILING RUN FOR THE BEST CONFIGURATION
# =================================================================================
if (( $(echo "$best_throughput > 0" | bc -l) )); then
echo
echo "Benchmark tuning finished. Now running profiling on the best configuration found..."
echo "Best config: max_num_seqs: $best_max_num_seqs, max_num_batched_tokens: $best_num_batched_tokens, throughput: $best_throughput"
echo
vllm_log="$LOG_FOLDER/vllm_log_BEST_PROFILE.txt"
bm_log="$LOG_FOLDER/bm_log_BEST_PROFILE.txt"
# Start server with the best params and profiling ENABLED
echo "Starting server for profiling..."
start_server $gpu_memory_utilization $best_max_num_seqs $best_num_batched_tokens "$vllm_log" "$PROFILE_PATH"
# Run benchmark with the best params and the --profile flag
echo "Running benchmark with profiling..."
prefix_len=$(( INPUT_LEN * MIN_CACHE_HIT_PCT / 100 ))
adjusted_input_len=$(( INPUT_LEN - prefix_len ))
vllm bench serve \
--backend vllm \
--model $MODEL \
--dataset-name random \
--random-input-len $adjusted_input_len \
--random-output-len $OUTPUT_LEN \
--ignore-eos \
--disable-tqdm \
--request-rate $best_request_rate \
--percentile-metrics ttft,tpot,itl,e2el \
--goodput e2el:$MAX_LATENCY_ALLOWED_MS \
--num-prompts 100 \
--random-prefix-len $prefix_len \
--port 8004 \
--profile &> "$bm_log"
else
echo "No configuration met the latency requirements. Skipping final profiling run."
fi
pkill -if vllm
echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH"
echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH" >> "$RESULT"

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import io
import json
@ -13,7 +12,8 @@ from typing import Optional, Union
import aiohttp
import huggingface_hub.constants
from tqdm.asyncio import tqdm
from transformers import AutoTokenizer, PreTrainedTokenizer, PreTrainedTokenizerFast
from transformers import (AutoTokenizer, PreTrainedTokenizer,
PreTrainedTokenizerFast)
# NOTE(simon): do not import vLLM here so the benchmark script
# can run without vLLM installed.
@ -31,10 +31,9 @@ class RequestFuncInput:
model_name: Optional[str] = None
logprobs: Optional[int] = None
extra_body: Optional[dict] = None
multi_modal_content: Optional[dict | list[dict]] = None
multi_modal_content: Optional[dict] = None
ignore_eos: bool = False
language: Optional[str] = None
request_id: Optional[str] = None
@dataclass
@ -44,7 +43,8 @@ class RequestFuncOutput:
latency: float = 0.0
output_tokens: int = 0
ttft: float = 0.0 # Time to first token
itl: list[float] = field(default_factory=list) # list of inter-token latencies
itl: list[float] = field(
default_factory=list) # list of inter-token latencies
tpot: float = 0.0 # avg next-token latencies
prompt_len: int = 0
error: str = ""
@ -57,9 +57,8 @@ async def async_request_tgi(
api_url = request_func_input.api_url
assert api_url.endswith("generate_stream")
async with aiohttp.ClientSession(
trust_env=True, timeout=AIOHTTP_TIMEOUT
) as session:
async with aiohttp.ClientSession(trust_env=True,
timeout=AIOHTTP_TIMEOUT) as session:
params = {
"max_new_tokens": request_func_input.output_len,
"do_sample": True,
@ -72,9 +71,6 @@ async def async_request_tgi(
"inputs": request_func_input.prompt,
"parameters": params,
}
headers = None
if request_func_input.request_id:
headers = {"x-request-id": request_func_input.request_id}
output = RequestFuncOutput()
output.prompt_len = request_func_input.prompt_len
if request_func_input.ignore_eos:
@ -86,9 +82,7 @@ async def async_request_tgi(
st = time.perf_counter()
most_recent_timestamp = st
try:
async with session.post(
url=api_url, json=payload, headers=headers
) as response:
async with session.post(url=api_url, json=payload) as response:
if response.status == 200:
async for chunk_bytes in response.content:
chunk_bytes = chunk_bytes.strip()
@ -111,7 +105,8 @@ async def async_request_tgi(
# Decoding phase
else:
output.itl.append(timestamp - most_recent_timestamp)
output.itl.append(timestamp -
most_recent_timestamp)
most_recent_timestamp = timestamp
@ -138,9 +133,8 @@ async def async_request_trt_llm(
api_url = request_func_input.api_url
assert api_url.endswith("generate_stream")
async with aiohttp.ClientSession(
trust_env=True, timeout=AIOHTTP_TIMEOUT
) as session:
async with aiohttp.ClientSession(trust_env=True,
timeout=AIOHTTP_TIMEOUT) as session:
payload = {
"accumulate_tokens": True,
"text_input": request_func_input.prompt,
@ -151,9 +145,6 @@ async def async_request_trt_llm(
}
if request_func_input.ignore_eos:
payload["min_length"] = request_func_input.output_len
headers = None
if request_func_input.request_id:
headers = {"x-request-id": request_func_input.request_id}
output = RequestFuncOutput()
output.prompt_len = request_func_input.prompt_len
@ -161,16 +152,15 @@ async def async_request_trt_llm(
st = time.perf_counter()
most_recent_timestamp = st
try:
async with session.post(
url=api_url, json=payload, headers=headers
) as response:
async with session.post(url=api_url, json=payload) as response:
if response.status == 200:
async for chunk_bytes in response.content:
chunk_bytes = chunk_bytes.strip()
if not chunk_bytes:
continue
chunk = chunk_bytes.decode("utf-8").removeprefix("data:")
chunk = chunk_bytes.decode("utf-8").removeprefix(
"data:")
data = json.loads(chunk)
output.generated_text += data["text_output"]
@ -182,7 +172,8 @@ async def async_request_trt_llm(
# Decoding phase
else:
output.itl.append(timestamp - most_recent_timestamp)
output.itl.append(timestamp -
most_recent_timestamp)
most_recent_timestamp = timestamp
@ -206,25 +197,15 @@ async def async_request_deepspeed_mii(
request_func_input: RequestFuncInput,
pbar: Optional[tqdm] = None,
) -> RequestFuncOutput:
api_url = request_func_input.api_url
assert api_url.endswith(("completions", "profile")), (
"OpenAI Completions API URL must end with 'completions' or 'profile'."
)
async with aiohttp.ClientSession(trust_env=True,
timeout=AIOHTTP_TIMEOUT) as session:
async with aiohttp.ClientSession(
trust_env=True, timeout=AIOHTTP_TIMEOUT
) as session:
payload = {
"model": request_func_input.model,
"prompt": request_func_input.prompt,
"max_tokens": request_func_input.output_len,
"temperature": 0.01, # deepspeed-mii does not accept 0.0 temp.
"top_p": 1.0,
}
headers = {"Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}"}
if request_func_input.request_id:
headers["x-request-id"] = request_func_input.request_id
output = RequestFuncOutput()
output.prompt_len = request_func_input.prompt_len
@ -235,21 +216,19 @@ async def async_request_deepspeed_mii(
st = time.perf_counter()
try:
async with session.post(
url=api_url, json=payload, headers=headers
) as response:
async with session.post(url=request_func_input.api_url,
json=payload) as response:
if response.status == 200:
parsed_resp = await response.json()
output.latency = time.perf_counter() - st
if "choices" in parsed_resp:
output.generated_text = parsed_resp["choices"][0]["text"]
output.generated_text = parsed_resp["choices"][0][
"text"]
elif "text" in parsed_resp:
output.generated_text = parsed_resp["text"][0]
else:
output.error = (
"Unexpected response format: "
"neither 'choices' nor 'text' found"
)
output.error = ("Unexpected response format: "
"neither 'choices' nor 'text' found")
output.success = False
output.success = True
else:
@ -270,20 +249,17 @@ async def async_request_openai_completions(
pbar: Optional[tqdm] = None,
) -> RequestFuncOutput:
api_url = request_func_input.api_url
assert api_url.endswith(("completions", "profile")), (
"OpenAI Completions API URL must end with 'completions' or 'profile'."
)
assert api_url.endswith(
("completions", "profile")
), "OpenAI Completions API URL must end with 'completions' or 'profile'."
async with aiohttp.ClientSession(
trust_env=True, timeout=AIOHTTP_TIMEOUT
) as session:
async with aiohttp.ClientSession(trust_env=True,
timeout=AIOHTTP_TIMEOUT) as session:
payload = {
"model": request_func_input.model_name
if request_func_input.model_name
else request_func_input.model,
"model": request_func_input.model_name \
if request_func_input.model_name else request_func_input.model,
"prompt": request_func_input.prompt,
"temperature": 0.0,
"repetition_penalty": 1.0,
"max_tokens": request_func_input.output_len,
"logprobs": request_func_input.logprobs,
"stream": True,
@ -295,9 +271,9 @@ async def async_request_openai_completions(
payload["ignore_eos"] = request_func_input.ignore_eos
if request_func_input.extra_body:
payload.update(request_func_input.extra_body)
headers = {"Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}"}
if request_func_input.request_id:
headers["x-request-id"] = request_func_input.request_id
headers = {
"Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}"
}
output = RequestFuncOutput()
output.prompt_len = request_func_input.prompt_len
@ -306,9 +282,8 @@ async def async_request_openai_completions(
st = time.perf_counter()
most_recent_timestamp = st
try:
async with session.post(
url=api_url, json=payload, headers=headers
) as response:
async with session.post(url=api_url, json=payload,
headers=headers) as response:
if response.status == 200:
first_chunk_received = False
async for chunk_bytes in response.content:
@ -316,7 +291,8 @@ async def async_request_openai_completions(
if not chunk_bytes:
continue
chunk = chunk_bytes.decode("utf-8").removeprefix("data: ")
chunk = chunk_bytes.decode("utf-8").removeprefix(
"data: ")
if chunk != "[DONE]":
data = json.loads(chunk)
@ -336,20 +312,21 @@ async def async_request_openai_completions(
# Decoding phase
else:
output.itl.append(timestamp - most_recent_timestamp)
output.itl.append(timestamp -
most_recent_timestamp)
most_recent_timestamp = timestamp
generated_text += text or ""
if usage := data.get("usage"):
output.output_tokens = usage.get("completion_tokens")
elif usage := data.get("usage"):
output.output_tokens = usage.get(
"completion_tokens")
if first_chunk_received:
output.success = True
else:
output.success = False
output.error = (
"Never received a valid chunk to calculate TTFT."
"This response will be marked as failed!"
)
"This response will be marked as failed!")
output.generated_text = generated_text
output.latency = most_recent_timestamp - st
else:
@ -370,30 +347,23 @@ async def async_request_openai_chat_completions(
pbar: Optional[tqdm] = None,
) -> RequestFuncOutput:
api_url = request_func_input.api_url
assert api_url.endswith(("chat/completions", "profile")), (
"OpenAI Chat Completions API URL must end with 'chat/completions'."
)
assert api_url.endswith(
("chat/completions", "profile")
), "OpenAI Chat Completions API URL must end with 'chat/completions'."
async with aiohttp.ClientSession(
trust_env=True, timeout=AIOHTTP_TIMEOUT
) as session:
async with aiohttp.ClientSession(trust_env=True,
timeout=AIOHTTP_TIMEOUT) as session:
content = [{"type": "text", "text": request_func_input.prompt}]
if request_func_input.multi_modal_content:
mm_content = request_func_input.multi_modal_content
if isinstance(mm_content, list):
content.extend(mm_content)
elif isinstance(mm_content, dict):
content.append(mm_content)
else:
raise TypeError(
"multi_modal_content must be a dict or list[dict] for openai-chat"
)
content.append(request_func_input.multi_modal_content)
payload = {
"model": request_func_input.model_name
if request_func_input.model_name
else request_func_input.model,
"model": request_func_input.model_name \
if request_func_input.model_name else request_func_input.model,
"messages": [
{"role": "user", "content": content},
{
"role": "user",
"content": content
},
],
"temperature": 0.0,
"max_completion_tokens": request_func_input.output_len,
@ -410,8 +380,6 @@ async def async_request_openai_chat_completions(
"Content-Type": "application/json",
"Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}",
}
if request_func_input.request_id:
headers["x-request-id"] = request_func_input.request_id
output = RequestFuncOutput()
output.prompt_len = request_func_input.prompt_len
@ -421,22 +389,16 @@ async def async_request_openai_chat_completions(
st = time.perf_counter()
most_recent_timestamp = st
try:
async with session.post(
url=api_url, json=payload, headers=headers
) as response:
async with session.post(url=api_url, json=payload,
headers=headers) as response:
if response.status == 200:
async for chunk_bytes in response.content:
chunk_bytes = chunk_bytes.strip()
if not chunk_bytes:
continue
chunk_bytes = chunk_bytes.decode("utf-8")
# NOTE: SSE comments (often used as pings) start with a colon.
# These are not JSON data payload and should be skipped.
if chunk_bytes.startswith(":"):
continue
chunk = chunk_bytes.removeprefix("data: ")
chunk = chunk_bytes.decode("utf-8").removeprefix(
"data: ")
if chunk != "[DONE]":
timestamp = time.perf_counter()
data = json.loads(chunk)
@ -450,11 +412,13 @@ async def async_request_openai_chat_completions(
# Decoding phase
else:
output.itl.append(timestamp - most_recent_timestamp)
output.itl.append(timestamp -
most_recent_timestamp)
generated_text += content or ""
elif usage := data.get("usage"):
output.output_tokens = usage.get("completion_tokens")
output.output_tokens = usage.get(
"completion_tokens")
most_recent_timestamp = timestamp
@ -480,36 +444,31 @@ async def async_request_openai_audio(
) -> RequestFuncOutput:
# Lazy import without PlaceholderModule to avoid vllm dep.
import soundfile
api_url = request_func_input.api_url
assert api_url.endswith(("transcriptions", "translations")), (
"OpenAI Chat Completions API URL must end with 'transcriptions' "
)
assert api_url.endswith(
("transcriptions", "translations"
)), "OpenAI Chat Completions API URL must end with 'transcriptions' "
"or `translations`."
async with aiohttp.ClientSession(
trust_env=True, timeout=AIOHTTP_TIMEOUT
) as session:
async with aiohttp.ClientSession(trust_env=True,
timeout=AIOHTTP_TIMEOUT) as session:
content = [{"type": "text", "text": request_func_input.prompt}]
payload = {
"model": request_func_input.model_name
if request_func_input.model_name
else request_func_input.model,
"model": request_func_input.model_name \
if request_func_input.model_name else request_func_input.model,
"temperature": 0.0,
"max_completion_tokens": request_func_input.output_len,
"stream": True,
"language": "en",
# Flattened due to multipart/form-data
"stream_include_usage": True,
"stream_continuous_usage_stats": True,
"stream_continuous_usage_stats": True
}
if request_func_input.extra_body:
payload.update(request_func_input.extra_body)
headers = {
"Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}",
}
if request_func_input.request_id:
headers["x-request-id"] = request_func_input.request_id
# Send audio file
def to_bytes(y, sr):
@ -518,12 +477,9 @@ async def async_request_openai_audio(
buffer.seek(0)
return buffer
mm_audio = request_func_input.multi_modal_content
if not isinstance(mm_audio, dict) or "audio" not in mm_audio:
raise TypeError("multi_modal_content must be a dict containing 'audio'")
with to_bytes(*mm_audio["audio"]) as f:
with to_bytes(*request_func_input.multi_modal_content['audio']) as f:
form = aiohttp.FormData()
form.add_field("file", f, content_type="audio/wav")
form.add_field('file', f, content_type='audio/wav')
for key, value in payload.items():
form.add_field(key, str(value))
@ -535,22 +491,24 @@ async def async_request_openai_audio(
st = time.perf_counter()
most_recent_timestamp = st
try:
async with session.post(
url=api_url, data=form, headers=headers
) as response:
async with session.post(url=api_url,
data=form,
headers=headers) as response:
if response.status == 200:
async for chunk_bytes in response.content:
chunk_bytes = chunk_bytes.strip()
if not chunk_bytes:
continue
chunk = chunk_bytes.decode("utf-8").removeprefix("data: ")
chunk = chunk_bytes.decode("utf-8").removeprefix(
"data: ")
if chunk != "[DONE]":
timestamp = time.perf_counter()
data = json.loads(chunk)
if choices := data.get("choices"):
content = choices[0]["delta"].get("content")
content = choices[0]["delta"].get(
"content")
# First token
if ttft == 0.0:
ttft = timestamp - st
@ -559,14 +517,12 @@ async def async_request_openai_audio(
# Decoding phase
else:
output.itl.append(
timestamp - most_recent_timestamp
)
timestamp - most_recent_timestamp)
generated_text += content or ""
elif usage := data.get("usage"):
output.output_tokens = usage.get(
"completion_tokens"
)
"completion_tokens")
most_recent_timestamp = timestamp
@ -587,7 +543,7 @@ async def async_request_openai_audio(
def get_model(pretrained_model_name_or_path: str) -> str:
if os.getenv("VLLM_USE_MODELSCOPE", "False").lower() == "true":
if os.getenv('VLLM_USE_MODELSCOPE', 'False').lower() == 'true':
from modelscope import snapshot_download
from vllm.model_executor.model_loader.weight_utils import get_lock
@ -598,8 +554,7 @@ def get_model(pretrained_model_name_or_path: str) -> str:
model_path = snapshot_download(
model_id=pretrained_model_name_or_path,
local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,
ignore_file_pattern=[".*.pt", ".*.safetensors", ".*.bin"],
)
ignore_file_pattern=[".*.pt", ".*.safetensors", ".*.bin"])
return model_path
return pretrained_model_name_or_path
@ -612,23 +567,23 @@ def get_tokenizer(
**kwargs,
) -> Union[PreTrainedTokenizer, PreTrainedTokenizerFast]:
if pretrained_model_name_or_path is not None and not os.path.exists(
pretrained_model_name_or_path
):
pretrained_model_name_or_path = get_model(pretrained_model_name_or_path)
pretrained_model_name_or_path):
pretrained_model_name_or_path = get_model(
pretrained_model_name_or_path)
if tokenizer_mode == "slow":
if kwargs.get("use_fast", False):
raise ValueError("Cannot use the fast tokenizer in slow tokenizer mode.")
raise ValueError(
"Cannot use the fast tokenizer in slow tokenizer mode.")
kwargs["use_fast"] = False
if tokenizer_mode == "mistral":
try:
from vllm.transformers_utils.tokenizer import MistralTokenizer
except ImportError as e:
raise ImportError(
"MistralTokenizer requires vllm package.\n"
"Please install it with `pip install vllm` "
"to use mistral tokenizer mode."
) from e
return MistralTokenizer.from_pretrained(str(pretrained_model_name_or_path))
raise ImportError("MistralTokenizer requires vllm package.\n"
"Please install it with `pip install vllm` "
"to use mistral tokenizer mode.") from e
return MistralTokenizer.from_pretrained(
str(pretrained_model_name_or_path))
else:
return AutoTokenizer.from_pretrained(
pretrained_model_name_or_path,
@ -648,11 +603,10 @@ ASYNC_REQUEST_FUNCS = {
"tensorrt-llm": async_request_trt_llm,
"scalellm": async_request_openai_completions,
"sglang": async_request_openai_completions,
"llama.cpp": async_request_openai_completions,
}
OPENAI_COMPATIBLE_BACKENDS = [
k
for k, v in ASYNC_REQUEST_FUNCS.items()
if v in (async_request_openai_completions, async_request_openai_chat_completions)
k for k, v in ASYNC_REQUEST_FUNCS.items()
if v in (async_request_openai_completions,
async_request_openai_chat_completions)
]

View File

@ -1,74 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import gc
from tabulate import tabulate
from benchmark_utils import TimeCollector
from vllm.utils import FlexibleArgumentParser
from vllm.v1.core.block_pool import BlockPool
def main(args):
rows = []
for allocate_block in args.allocate_blocks:
# Enforce a GC collect ahead to minimize the impact among runs
gc.collect()
block_pool = BlockPool(num_gpu_blocks=args.num_gpu_blocks, enable_caching=True)
get_blocks_times = TimeCollector(TimeCollector.US)
free_blocks_times = TimeCollector(TimeCollector.US)
for _ in range(args.num_iteration):
with get_blocks_times:
blocks = block_pool.get_new_blocks(allocate_block)
with free_blocks_times:
block_pool.free_blocks(blocks)
rows.append(
[get_blocks_times.cnt, args.num_gpu_blocks, allocate_block]
+ get_blocks_times.dump_avg_max()
+ free_blocks_times.dump_avg_max()
)
print(
tabulate(
rows,
headers=[
"Iterations",
"Total\nBlocks",
"Allocated\nBlocks",
"Get Blocks\nAvg (us)",
"Get Blocks\nMax (us)",
"Free Blocks\nAvg (us)",
"Free Blocks\nMax (us)",
],
tablefmt="grid",
floatfmt=".3f",
)
)
def invoke_main() -> None:
parser = FlexibleArgumentParser(
description="Benchmark the performance of BlockPool for KV Cache."
)
parser.add_argument("--num-gpu-blocks", type=int, default=100000)
parser.add_argument(
"--num-iteration",
type=int,
default=1000,
help="Number of iterations to run to stabilize final data readings",
)
parser.add_argument(
"--allocate-blocks",
type=int,
nargs="*",
default=[10, 50, 100, 500, 1000],
help="Number of blocks to allocate",
)
args = parser.parse_args()
main(args)
if __name__ == "__main__":
invoke_main() # pragma: no cover

File diff suppressed because it is too large Load Diff

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
"""Benchmark the latency of processing a single batch of requests."""
import argparse
@ -7,14 +6,14 @@ import dataclasses
import json
import os
import time
from pathlib import Path
from typing import Any, Optional
import numpy as np
from tqdm import tqdm
from typing_extensions import deprecated
import vllm.envs as envs
import torch
from benchmark_utils import convert_to_pytorch_benchmark_format, write_to_json
from tqdm import tqdm
from vllm import LLM, SamplingParams
from vllm.engine.arg_utils import EngineArgs
from vllm.inputs import PromptType
@ -22,23 +21,18 @@ from vllm.sampling_params import BeamSearchParams
from vllm.utils import FlexibleArgumentParser
def save_to_pytorch_benchmark_format(
args: argparse.Namespace, results: dict[str, Any]
) -> None:
def save_to_pytorch_benchmark_format(args: argparse.Namespace,
results: dict[str, Any]) -> None:
pt_records = convert_to_pytorch_benchmark_format(
args=args,
metrics={"latency": results["latencies"]},
extra_info={k: results[k] for k in ["avg_latency", "percentiles"]},
)
extra_info={k: results[k]
for k in ["avg_latency", "percentiles"]})
if pt_records:
pt_file = f"{os.path.splitext(args.output_json)[0]}.pytorch.json"
write_to_json(pt_file, pt_records)
@deprecated(
"benchmark_latency.py is deprecated and will be removed in a "
"future version. Please use 'vllm bench latency' instead.",
)
def main(args: argparse.Namespace):
print(args)
@ -48,11 +42,9 @@ def main(args: argparse.Namespace):
# the engine will automatically process the request in multiple batches.
llm = LLM(**dataclasses.asdict(engine_args))
assert llm.llm_engine.model_config.max_model_len >= (
args.input_len + args.output_len
), (
"Please ensure that max_model_len is greater than"
" the sum of input_len and output_len."
)
args.input_len +
args.output_len), ("Please ensure that max_model_len is greater than"
" the sum of input_len and output_len.")
sampling_params = SamplingParams(
n=args.n,
@ -63,16 +55,18 @@ def main(args: argparse.Namespace):
detokenize=not args.disable_detokenize,
)
print(sampling_params)
dummy_prompt_token_ids = np.random.randint(
10000, size=(args.batch_size, args.input_len)
)
dummy_prompts: list[PromptType] = [
{"prompt_token_ids": batch} for batch in dummy_prompt_token_ids.tolist()
]
dummy_prompt_token_ids = np.random.randint(10000,
size=(args.batch_size,
args.input_len))
dummy_prompts: list[PromptType] = [{
"prompt_token_ids": batch
} for batch in dummy_prompt_token_ids.tolist()]
def llm_generate():
if not args.use_beam_search:
llm.generate(dummy_prompts, sampling_params=sampling_params, use_tqdm=False)
llm.generate(dummy_prompts,
sampling_params=sampling_params,
use_tqdm=False)
else:
llm.beam_search(
dummy_prompts,
@ -85,9 +79,16 @@ def main(args: argparse.Namespace):
def run_to_completion(profile_dir: Optional[str] = None):
if profile_dir:
llm.start_profile()
llm_generate()
llm.stop_profile()
with torch.profiler.profile(
activities=[
torch.profiler.ProfilerActivity.CPU,
torch.profiler.ProfilerActivity.CUDA,
],
on_trace_ready=torch.profiler.tensorboard_trace_handler(
str(profile_dir)),
) as p:
llm_generate()
print(p.key_averages().table(sort_by="self_cuda_time_total"))
else:
start_time = time.perf_counter()
llm_generate()
@ -100,7 +101,10 @@ def main(args: argparse.Namespace):
run_to_completion(profile_dir=None)
if args.profile:
profile_dir = envs.VLLM_TORCH_PROFILER_DIR
profile_dir = args.profile_result_dir
if not profile_dir:
profile_dir = (Path(".") / "vllm_benchmark_result" /
f"latency_result_{time.time()}")
print(f"Profiling (results will be saved to '{profile_dir}')...")
run_to_completion(profile_dir=profile_dir)
return
@ -128,11 +132,10 @@ def main(args: argparse.Namespace):
save_to_pytorch_benchmark_format(args, results)
def create_argument_parser():
if __name__ == "__main__":
parser = FlexibleArgumentParser(
description="Benchmark the latency of processing a single batch of "
"requests till completion."
)
"requests till completion.")
parser.add_argument("--input-len", type=int, default=32)
parser.add_argument("--output-len", type=int, default=128)
parser.add_argument("--batch-size", type=int, default=8)
@ -149,14 +152,22 @@ def create_argument_parser():
default=10,
help="Number of iterations to run for warmup.",
)
parser.add_argument(
"--num-iters", type=int, default=30, help="Number of iterations to run."
)
parser.add_argument("--num-iters",
type=int,
default=30,
help="Number of iterations to run.")
parser.add_argument(
"--profile",
action="store_true",
help="profile the generation process of a single batch",
)
parser.add_argument(
"--profile-result-dir",
type=str,
default=None,
help=("path to save the pytorch profiler output. Can be visualized "
"with ui.perfetto.dev or Tensorboard."),
)
parser.add_argument(
"--output-json",
type=str,
@ -166,26 +177,10 @@ def create_argument_parser():
parser.add_argument(
"--disable-detokenize",
action="store_true",
help=(
"Do not detokenize responses (i.e. do not include "
"detokenization time in the latency measurement)"
),
help=("Do not detokenize responses (i.e. do not include "
"detokenization time in the latency measurement)"),
)
parser = EngineArgs.add_cli_args(parser)
# V1 enables prefix caching by default which skews the latency
# numbers. We need to disable prefix caching by default.
parser.set_defaults(enable_prefix_caching=False)
return parser
if __name__ == "__main__":
parser = create_argument_parser()
args = parser.parse_args()
if args.profile and not envs.VLLM_TORCH_PROFILER_DIR:
raise OSError(
"The environment variable 'VLLM_TORCH_PROFILER_DIR' is not set. "
"Please set it to a valid path to use torch profiler."
)
main(args)

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
"""
Offline benchmark to test the long document QA throughput.
@ -77,7 +76,7 @@ def repeat_prompts(prompts, repeat_count, mode: str):
- 'random': Shuffle the prompts randomly after repetition.
- 'tile': Repeat the entire prompt list in sequence.
Example: [1, 2, 3] -> [1, 2, 3, 1, 2, 3].
- 'interleave': Repeat each prompt consecutively before moving to
- 'interleave': Repeat each prompt consecutively before moving to
the next. Example: [1, 2, 3] -> [1, 1, 2, 2, 3, 3].
Returns:
@ -87,21 +86,20 @@ def repeat_prompts(prompts, repeat_count, mode: str):
ValueError: If an invalid mode is provided.
"""
print("Repeat mode: ", mode)
if mode == "random":
if mode == 'random':
repeated_prompts = prompts * repeat_count
random.shuffle(repeated_prompts)
return repeated_prompts
elif mode == "tile":
elif mode == 'tile':
return prompts * repeat_count
elif mode == "interleave":
elif mode == 'interleave':
repeated_prompts = []
for prompt in prompts:
repeated_prompts.extend([prompt] * repeat_count)
return repeated_prompts
else:
raise ValueError(
f"Invalid mode: {mode}, only support 'random', 'tile', 'interleave'"
)
raise ValueError(f"Invalid mode: {mode}, only support "
"'random', 'tile', 'interleave'")
def main(args):
@ -111,16 +109,16 @@ def main(args):
# we append the document id at the beginning to avoid any of the document
# being the prefix of other documents
prompts = [
str(i) + " ".join(["hi"] * args.document_length)
str(i) + ' '.join(['hi'] * args.document_length)
for i in range(args.num_documents)
]
prompts = repeat_prompts(prompts, args.repeat_count, mode=args.repeat_mode)
warmup_prompts = [
"This is warm up request " + str(i) + " ".join(["hi"] * args.document_length)
for i in range(args.num_documents)
]
"This is warm up request " + str(i) + \
' '.join(['hi'] * args.document_length)
for i in range(args.num_documents)]
# Create the LLM engine
engine_args = EngineArgs.from_cli_args(args)
@ -142,61 +140,45 @@ def main(args):
)
def create_argument_parser():
if __name__ == "__main__":
parser = FlexibleArgumentParser(
description="Benchmark the performance with or "
"without automatic prefix caching."
)
description=
'Benchmark the performance with or without automatic prefix caching.')
parser.add_argument(
"--document-length",
'--document-length',
type=int,
# Roughly the number of tokens for a system paper,
# excluding images
default=20000,
help="Range of input lengths for sampling prompts, "
'specified as "min:max" (e.g., "128:256").',
)
help='Range of input lengths for sampling prompts,'
'specified as "min:max" (e.g., "128:256").')
parser.add_argument(
"--num-documents",
type=int,
default=8,
help="Range of input lengths for sampling prompts, "
'specified as "min:max" (e.g., "128:256").',
)
parser.add_argument('--num-documents',
type=int,
default=8,
help='Range of input lengths for sampling prompts,'
'specified as "min:max" (e.g., "128:256").')
parser.add_argument("--output-len", type=int, default=10)
parser.add_argument('--output-len', type=int, default=10)
parser.add_argument(
"--repeat-count",
type=int,
default=2,
help="Number of times to repeat each prompt",
)
parser.add_argument('--repeat-count',
type=int,
default=2,
help='Number of times to repeat each prompt')
parser.add_argument(
"--repeat-mode",
type=str,
default="random",
help="The mode to repeat prompts. The supported "
'modes are "random", "tile", and "interleave". '
"See repeat_prompts() in the source code for details.",
)
parser.add_argument("--repeat-mode",
type=str,
default='random',
help='The mode to repeat prompts. The supported '
'modes are "random", "tile", and "interleave". '
'See repeat_prompts() in the source code for details.')
parser.add_argument(
"--shuffle-seed",
type=int,
default=0,
help='Random seed when the repeat mode is "random"',
)
parser.add_argument("--shuffle-seed",
type=int,
default=0,
help='Random seed when the repeat mode is "random"')
parser = EngineArgs.add_cli_args(parser)
return parser
if __name__ == "__main__":
parser = create_argument_parser()
args = parser.parse_args()
main(args)

View File

@ -1,112 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import gc
import numpy as np
from tabulate import tabulate
from benchmark_utils import TimeCollector
from vllm.config import ModelConfig, SpeculativeConfig, VllmConfig
from vllm.utils import FlexibleArgumentParser
from vllm.v1.spec_decode.ngram_proposer import NgramProposer
def main(args):
rows = []
for max_ngram in args.max_ngram:
collector = TimeCollector(TimeCollector.US)
model_config = ModelConfig(
model="facebook/opt-125m",
task="generate",
max_model_len=args.num_token + args.num_spec_token,
tokenizer="facebook/opt-125m",
tokenizer_mode="auto",
dtype="auto",
seed=None,
trust_remote_code=False,
)
proposer = NgramProposer(
vllm_config=VllmConfig(
model_config=model_config,
speculative_config=SpeculativeConfig(
prompt_lookup_min=args.min_ngram,
prompt_lookup_max=max_ngram,
num_speculative_tokens=args.num_spec_token,
method="ngram",
),
)
)
# Warm up
proposer.propose(np.random.randint(0, 20, (args.num_token,)))
gc.collect()
for _ in range(args.num_iteration):
tokens = np.random.randint(0, 20, (args.num_req, args.num_token))
with collector:
for i in range(args.num_req):
proposer.propose(tokens[i, :])
rows.append(
[args.num_req, args.num_token, args.min_ngram, max_ngram]
+ collector.dump_avg_max()
)
print(
tabulate(
rows,
headers=[
"# Request",
"# Token",
"Min Ngram",
"Max Ngram",
"Avg (us)",
"Max (us)",
],
tablefmt="grid",
floatfmt=".3f",
)
)
def invoke_main() -> None:
parser = FlexibleArgumentParser(
description="Benchmark the performance of N-gram speculative decode drafting"
)
parser.add_argument(
"--num-iteration",
type=int,
default=100,
help="Number of iterations to run to stabilize final data readings",
)
parser.add_argument(
"--num-req", type=int, default=128, help="Number of requests in the batch"
)
parser.add_argument(
"--num-token", type=int, default=1500, help="Number of tokens for each request"
)
parser.add_argument(
"--min-ngram",
type=int,
default=3,
help="Minimum n-gram to match",
)
parser.add_argument(
"--max-ngram",
type=int,
nargs="*",
default=[5, 7, 10, 15, 20],
help="Maximum n-gram to match",
)
parser.add_argument(
"--num-spec-token",
type=int,
default=3,
help="Number of speculative tokens to generate",
)
args = parser.parse_args()
main(args)
if __name__ == "__main__":
invoke_main() # pragma: no cover

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
"""
Benchmark the efficiency of prefix caching.
@ -64,7 +63,8 @@ class Request:
output_len: int
def sample_tokens(tokenizer: PreTrainedTokenizerBase, length: int) -> list[int]:
def sample_tokens(tokenizer: PreTrainedTokenizerBase,
length: int) -> list[int]:
vocab = tokenizer.get_vocab()
all_special_ids = set(tokenizer.all_special_ids)
@ -91,10 +91,8 @@ def sample_requests_from_dataset(
# Filter out the conversations with less than 2 turns.
dataset = [data for data in dataset if len(data["conversations"]) >= 2]
# Only keep the first two turns of each conversation.
dataset = [
(data["conversations"][0]["value"], data["conversations"][1]["value"])
for data in dataset
]
dataset = [(data["conversations"][0]["value"],
data["conversations"][1]["value"]) for data in dataset]
# Shuffle the dataset.
random.shuffle(dataset)
@ -115,9 +113,8 @@ def sample_requests_from_dataset(
completion = dataset[i][1]
completion_token_ids = tokenizer(completion).input_ids
prompt_len = len(prompt_token_ids)
output_len = (
len(completion_token_ids) if fixed_output_len is None else fixed_output_len
)
output_len = (len(completion_token_ids)
if fixed_output_len is None else fixed_output_len)
if min_len <= prompt_len <= max_len:
filtered_requests.append(Request(prompt, prompt_len, output_len))
@ -131,27 +128,27 @@ def sample_requests_from_random(
fixed_output_len: Optional[int],
prefix_len: int,
) -> list[Request]:
requests = []
prefix_token_ids = sample_tokens(tokenizer, prefix_len)
min_len, max_len = input_length_range
for i in range(num_requests):
unique_part_token_ids = sample_tokens(
tokenizer, random.randint(min_len - prefix_len, max_len - prefix_len)
)
tokenizer,
random.randint(min_len - prefix_len, max_len - prefix_len))
prompt_token_ids = prefix_token_ids + unique_part_token_ids
prompt = tokenizer.decode(prompt_token_ids)
prompt_len = len(prompt_token_ids)
assert min_len <= prompt_len <= max_len, (
f"prompt_len {prompt_len} out of range {min_len}:{max_len}"
)
assert (min_len <= prompt_len <= max_len
), f"prompt_len {prompt_len} out of range {min_len}:{max_len}"
requests.append(Request(prompt, prompt_len, fixed_output_len))
return requests
def repeat_and_sort_requests(
requests: list[Request], repeat_count: int, sort: bool = False
) -> list[str]:
def repeat_and_sort_requests(requests: list[Request],
repeat_count: int,
sort: bool = False) -> list[str]:
repeated_requests = requests * repeat_count
if sort:
repeated_requests.sort(key=lambda x: x[1])
@ -162,14 +159,14 @@ def repeat_and_sort_requests(
def main(args):
tokenizer = get_tokenizer(args.model, trust_remote_code=True)
input_length_range = tuple(map(int, args.input_length_range.split(":")))
input_length_range = tuple(map(int, args.input_length_range.split(':')))
random.seed(args.seed)
if args.dataset_path is not None:
if args.prefix_len > 0:
raise ValueError(
"prefix-len is not supported when dataset-path is provided."
)
print(f"Start to sample {args.num_prompts} prompts from {args.dataset_path}")
raise ValueError("prefix-len is not supported when "
"dataset-path is provided.")
print(f"Start to sample {args.num_prompts} prompts "
f"from {args.dataset_path}")
filtered_requests = sample_requests_from_dataset(
dataset_path=args.dataset_path,
num_requests=args.num_prompts,
@ -199,16 +196,14 @@ def main(args):
llm = LLM(**dataclasses.asdict(engine_args))
sampling_params = SamplingParams(
temperature=0,
max_tokens=args.output_len,
detokenize=not args.disable_detokenize,
)
sampling_params = SamplingParams(temperature=0,
max_tokens=args.output_len,
detokenize=not args.disable_detokenize)
print("Testing filtered requests")
prompts = repeat_and_sort_requests(
filtered_requests, repeat_count=args.repeat_count, sort=args.sort
)
prompts = repeat_and_sort_requests(filtered_requests,
repeat_count=args.repeat_count,
sort=args.sort)
print("------start generating------")
test_prefix(
@ -218,37 +213,31 @@ def main(args):
)
def create_argument_parser():
if __name__ == "__main__":
parser = FlexibleArgumentParser(
description="Benchmark the performance with or without "
"automatic prefix caching."
)
parser.add_argument(
"--dataset-path", type=str, default=None, help="Path to the dataset."
)
parser.add_argument("--output-len", type=int, default=10)
parser.add_argument(
"--num-prompts",
type=int,
required=True,
help="Number of the prompts sampled from dataset",
)
parser.add_argument(
"--repeat-count",
type=int,
default=1,
help="Number of times to repeat each prompt",
)
parser.add_argument(
"--sort", action="store_true", help="Sort prompts by input length"
)
parser.add_argument(
"--input-length-range",
type=str,
required=True,
help="Range of input lengths for sampling prompts,"
'specified as "min:max" (e.g., "128:256").',
)
description=
'Benchmark the performance with or without automatic prefix caching.')
parser.add_argument("--dataset-path",
type=str,
default=None,
help="Path to the dataset.")
parser.add_argument('--output-len', type=int, default=10)
parser.add_argument('--num-prompts',
type=int,
required=True,
help="Number of the prompts sampled from dataset")
parser.add_argument('--repeat-count',
type=int,
default=1,
help='Number of times to repeat each prompt')
parser.add_argument('--sort',
action='store_true',
help='Sort prompts by input length')
parser.add_argument('--input-length-range',
type=str,
required=True,
help='Range of input lengths for sampling prompts,'
'specified as "min:max" (e.g., "128:256").')
parser.add_argument(
"--prefix-len",
type=int,
@ -259,20 +248,12 @@ def create_argument_parser():
"when dataset-path is not provided.",
)
parser.add_argument(
"--disable-detokenize",
action="store_true",
help=(
"Do not detokenize responses (i.e. do not include "
"detokenization time in the latency measurement)"
),
'--disable-detokenize',
action='store_true',
help=("Do not detokenize responses (i.e. do not include "
"detokenization time in the latency measurement)"),
)
parser = EngineArgs.add_cli_args(parser)
return parser
if __name__ == "__main__":
parser = create_argument_parser()
args = parser.parse_args()
main(args)

View File

@ -1,7 +1,5 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
"""Benchmark offline prioritization."""
import argparse
import dataclasses
import json
@ -15,7 +13,7 @@ from vllm.engine.arg_utils import EngineArgs
from vllm.utils import FlexibleArgumentParser
# Select a equi-probable random priority
#Select a equi-probable random priority
def get_random_flag():
return 0 if random.random() < 0.5 else 1
@ -35,10 +33,8 @@ def sample_requests(
# Filter out the conversations with less than 2 turns.
dataset = [data for data in dataset if len(data["conversations"]) >= 2]
# Only keep the first two turns of each conversation.
dataset = [
(data["conversations"][0]["value"], data["conversations"][1]["value"])
for data in dataset
]
dataset = [(data["conversations"][0]["value"],
data["conversations"][1]["value"]) for data in dataset]
# Shuffle the dataset.
random.shuffle(dataset)
@ -55,9 +51,8 @@ def sample_requests(
completion = dataset[i][1]
completion_token_ids = tokenizer(completion).input_ids
prompt_len = len(prompt_token_ids)
output_len = (
len(completion_token_ids) if fixed_output_len is None else fixed_output_len
)
output_len = len(completion_token_ids
) if fixed_output_len is None else fixed_output_len
if prompt_len < 4 or output_len < 4:
# Prune too short sequences.
continue
@ -79,16 +74,13 @@ def run_vllm(
disable_detokenize: bool = False,
) -> float:
from vllm import LLM, SamplingParams
llm = LLM(**dataclasses.asdict(engine_args))
assert all(
llm.llm_engine.model_config.max_model_len >= (request[1] + request[2])
for request in requests
), (
"Please ensure that max_model_len is greater than the sum of"
" input_len and output_len for all requests."
)
for request in requests), (
"Please ensure that max_model_len is greater than the sum of"
" input_len and output_len for all requests.")
# Add the requests to the engine.
prompts = []
@ -105,8 +97,7 @@ def run_vllm(
ignore_eos=True,
max_tokens=output_len,
detokenize=not disable_detokenize,
)
)
))
start = time.perf_counter()
llm.generate(prompts, sampling_params, priority=priority, use_tqdm=True)
@ -120,33 +111,26 @@ def main(args: argparse.Namespace):
# Sample the requests.
tokenizer = AutoTokenizer.from_pretrained(
args.tokenizer, trust_remote_code=args.trust_remote_code
)
args.tokenizer, trust_remote_code=args.trust_remote_code)
if args.dataset is None:
# Synthesize a prompt with the given input length.
prompt = "hi" * (args.input_len - 1)
requests = [
(prompt, args.input_len, args.output_len, get_random_flag())
for _ in range(args.num_prompts)
]
requests = [(prompt, args.input_len, args.output_len,
get_random_flag()) for _ in range(args.num_prompts)]
else:
requests = sample_requests(
args.dataset, args.num_prompts, tokenizer, args.output_len
)
requests = sample_requests(args.dataset, args.num_prompts, tokenizer,
args.output_len)
if args.backend == "vllm":
elapsed_time = run_vllm(
requests, args.n, EngineArgs.from_cli_args(args), args.disable_detokenize
)
elapsed_time = run_vllm(requests, args.n,
EngineArgs.from_cli_args(args),
args.disable_detokenize)
else:
raise ValueError(f"Unknown backend: {args.backend}")
total_num_tokens = sum(
prompt_len + output_len for _, prompt_len, output_len, priority in requests
)
print(
f"Throughput: {len(requests) / elapsed_time:.2f} requests/s, "
f"{total_num_tokens / elapsed_time:.2f} tokens/s"
)
total_num_tokens = sum(prompt_len + output_len
for _, prompt_len, output_len, priority in requests)
print(f"Throughput: {len(requests) / elapsed_time:.2f} requests/s, "
f"{total_num_tokens / elapsed_time:.2f} tokens/s")
# Output JSON results if specified
if args.output_json:
@ -161,55 +145,46 @@ def main(args: argparse.Namespace):
json.dump(results, f, indent=4)
def create_argument_parser():
if __name__ == "__main__":
parser = FlexibleArgumentParser(description="Benchmark the throughput.")
parser.add_argument("--backend",
type=str,
choices=["vllm", "hf", "mii"],
default="vllm")
parser.add_argument("--dataset",
type=str,
default=None,
help="Path to the dataset.")
parser.add_argument("--input-len",
type=int,
default=None,
help="Input prompt length for each request")
parser.add_argument("--output-len",
type=int,
default=None,
help="Output length for each request. Overrides the "
"output length from the dataset.")
parser.add_argument("--n",
type=int,
default=1,
help="Number of generated sequences per prompt.")
parser.add_argument("--num-prompts",
type=int,
default=200,
help="Number of prompts to process.")
parser.add_argument(
"--backend", type=str, choices=["vllm", "hf", "mii"], default="vllm"
)
parser.add_argument(
"--dataset", type=str, default=None, help="Path to the dataset."
)
parser.add_argument(
"--input-len",
type=int,
default=None,
help="Input prompt length for each request",
)
parser.add_argument(
"--output-len",
type=int,
default=None,
help="Output length for each request. Overrides the "
"output length from the dataset.",
)
parser.add_argument(
"--n", type=int, default=1, help="Number of generated sequences per prompt."
)
parser.add_argument(
"--num-prompts", type=int, default=200, help="Number of prompts to process."
)
parser.add_argument(
"--output-json",
'--output-json',
type=str,
default=None,
help="Path to save the throughput results in JSON format.",
)
help='Path to save the throughput results in JSON format.')
parser.add_argument(
"--disable-detokenize",
action="store_true",
help=(
"Do not detokenize responses (i.e. do not include "
"detokenization time in the latency measurement)"
),
'--disable-detokenize',
action='store_true',
help=("Do not detokenize responses (i.e. do not include "
"detokenization time in the latency measurement)"),
)
parser = EngineArgs.add_cli_args(parser)
return parser
if __name__ == "__main__":
parser = create_argument_parser()
args = parser.parse_args()
if args.tokenizer is None:
args.tokenizer = args.model

File diff suppressed because it is too large Load Diff

View File

@ -1,10 +1,9 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
r"""Benchmark online serving throughput with structured outputs.
On the server side, run one of the following commands:
(vLLM OpenAI API server)
vllm serve <your_model>
vllm serve <your_model> --disable-log-requests
On the client side, run:
python benchmarks/benchmark_serving_structured_output.py \
@ -12,6 +11,7 @@ On the client side, run:
--model <your_model> \
--dataset json \
--structured-output-ratio 1.0 \
--structured-output-backend auto \
--request-rate 10 \
--num-prompts 1000
@ -19,7 +19,6 @@ On the client side, run:
--endpoint /generate_stream
to the end of the command above.
"""
import argparse
import asyncio
import copy
@ -37,15 +36,11 @@ from typing import Optional
import datasets
import numpy as np
import pandas as pd
from backend_request_func import (ASYNC_REQUEST_FUNCS, RequestFuncInput,
RequestFuncOutput)
from tqdm.asyncio import tqdm
from transformers import PreTrainedTokenizerBase
from backend_request_func import (
ASYNC_REQUEST_FUNCS,
RequestFuncInput,
RequestFuncOutput,
)
try:
from vllm.transformers_utils.tokenizer import get_tokenizer
except ImportError:
@ -57,8 +52,7 @@ except ImportError:
from argparse import ArgumentParser as FlexibleArgumentParser
from vllm.v1.structured_output.backend_xgrammar import (
has_xgrammar_unsupported_json_features,
)
has_xgrammar_unsupported_json_features)
MILLISECONDS_TO_SECONDS_CONVERSION = 1000
@ -104,7 +98,6 @@ class SampleRequest:
prompt_len: The length of the prompt in tokens.
expected_output_len: The expected length of the output in tokens.
"""
prompt: str
prompt_len: int
expected_output_len: int
@ -113,45 +106,45 @@ class SampleRequest:
completion: str = None
def sample_requests(
tokenizer: PreTrainedTokenizerBase, args: argparse.Namespace
) -> list[SampleRequest]:
if args.dataset == "json" or args.dataset == "json-unique":
def sample_requests(tokenizer: PreTrainedTokenizerBase,
args: argparse.Namespace) -> list[SampleRequest]:
if args.dataset == 'json' or args.dataset == 'json-unique':
if args.json_schema_path is None:
dir_path = os.path.dirname(os.path.realpath(__file__))
args.json_schema_path = os.path.join(
dir_path, "structured_schemas", "structured_schema_1.json"
)
args.json_schema_path = os.path.join(dir_path,
"structured_schemas",
"structured_schema_1.json")
json_schemas = []
with open(args.json_schema_path) as f:
schema = json.load(f)
if args.dataset == "json-unique":
json_schemas = [copy.deepcopy(schema) for _ in range(args.num_prompts)]
if args.dataset == 'json-unique':
json_schemas = [
copy.deepcopy(schema) for _ in range(args.num_prompts)
]
for i in range(len(json_schemas)):
if "properties" not in json_schemas[i]:
json_schemas[i]["properties"] = {}
json_schemas[i]["properties"][f"__optional_field_{uuid.uuid4()}"] = {
"type": "string",
"description": "An unique optional field to avoid cached schemas",
}
json_schemas[i]["properties"][
f"__optional_field_{uuid.uuid4()}"] = {
"type":
"string",
"description":
"An unique optional field to avoid cached schemas"
}
else:
json_schemas = [schema] * args.num_prompts
def gen_prompt(index: int):
return f"Generate an example of a brief user profile given the following schema: {json.dumps(get_schema(index))}" # noqa: E501
return f"Generate an example of a user profile given the following schema: {json.dumps(get_schema(index))}" # noqa: E501
def get_schema(index: int):
return json_schemas[index % len(json_schemas)]
requests = [
SampleRequest(
prompt=gen_prompt(i),
prompt_len=len(tokenizer(gen_prompt(i)).input_ids),
expected_output_len=args.output_len,
schema=get_schema(i),
structure_type=args.structure_type,
)
SampleRequest(prompt=gen_prompt(i),
prompt_len=len(tokenizer(gen_prompt(i)).input_ids),
expected_output_len=args.output_len,
schema=get_schema(i),
structure_type=args.structure_type)
for i in range(args.num_prompts)
]
@ -175,13 +168,11 @@ def sample_requests(
input_len = len(tokenizer(prompt).input_ids)
print(f"Input length of the prompt: {input_len} tokens")
requests = [
SampleRequest(
prompt=prompt,
prompt_len=input_len,
expected_output_len=args.output_len,
schema=schema,
structure_type=args.structure_type,
)
SampleRequest(prompt=prompt,
prompt_len=input_len,
expected_output_len=args.output_len,
schema=schema,
structure_type=args.structure_type)
for _ in range(args.num_prompts)
]
@ -195,13 +186,11 @@ def sample_requests(
input_len = len(tokenizer(prompt).input_ids)
print(f"Input length of the prompt: {input_len} tokens")
requests = [
SampleRequest(
prompt=prompt,
prompt_len=input_len,
expected_output_len=args.output_len,
schema=regex,
structure_type=args.structure_type,
)
SampleRequest(prompt=prompt,
prompt_len=input_len,
expected_output_len=args.output_len,
schema=regex,
structure_type=args.structure_type)
for _ in range(args.num_prompts)
]
@ -212,55 +201,47 @@ def sample_requests(
input_len = len(tokenizer(prompt).input_ids)
print(f"Input length of the prompt: {input_len} tokens")
requests = [
SampleRequest(
prompt=prompt,
prompt_len=input_len,
expected_output_len=args.output_len,
schema=choice,
structure_type=args.structure_type,
)
SampleRequest(prompt=prompt,
prompt_len=input_len,
expected_output_len=args.output_len,
schema=choice,
structure_type=args.structure_type)
for _ in range(args.num_prompts)
]
elif args.dataset == "xgrammar_bench":
requests: list[SampleRequest] = []
dataset = datasets.load_dataset("NousResearch/json-mode-eval", split="train")
dataset = datasets.load_dataset("NousResearch/json-mode-eval",
split="train")
full_dataset_len = len(dataset)
def _filter_func(item):
import json
schema = json.loads(item["schema"])
return not has_xgrammar_unsupported_json_features(schema)
dataset = dataset.filter(_filter_func)
num_filtered_out = full_dataset_len - len(dataset)
print(
f"dataset has {len(dataset)} entries after filtering "
f"out {num_filtered_out} entries with unsupported features"
)
print(f"dataset has {len(dataset)} entries after filtering "
f"out {num_filtered_out} entries with unsupported features")
len_dataset = len(dataset)
for data_point_idx in range(args.num_prompts):
idx = data_point_idx
while idx >= len_dataset:
idx -= len_dataset
schema = dataset["schema"][idx]
prompt = tokenizer.apply_chat_template(
dataset["prompt"][idx], tokenize=False, add_generation_prompt=True
)
prompt = tokenizer.apply_chat_template(dataset["prompt"][idx],
tokenize=False)
input_len = len(tokenizer(prompt).input_ids)
completion = dataset["completion"][idx]
requests.append(
SampleRequest(
prompt=prompt,
prompt_len=input_len,
expected_output_len=args.output_len,
schema=schema,
structure_type=args.structure_type,
completion=completion,
)
)
SampleRequest(prompt=prompt,
prompt_len=input_len,
expected_output_len=args.output_len,
schema=schema,
structure_type=args.structure_type,
completion=completion))
return requests
@ -292,8 +273,7 @@ async def get_request(
# Calculate scale parameter theta to maintain the desired request_rate.
assert burstiness > 0, (
f"A positive burstiness factor is expected, but given {burstiness}."
)
f"A positive burstiness factor is expected, but given {burstiness}.")
theta = 1.0 / (request_rate * burstiness)
for i, request in enumerate(input_requests):
@ -335,8 +315,8 @@ def calculate_metrics(
# multiple output tokens may be bundled together
# Note : this may inflate the output token count slightly
output_len = len(
tokenizer(outputs[i].generated_text, add_special_tokens=False).input_ids
)
tokenizer(outputs[i].generated_text,
add_special_tokens=False).input_ids)
actual_output_lens.append(output_len)
total_input += input_requests[i].prompt_len
tpot = 0
@ -360,19 +340,16 @@ def calculate_metrics(
if "ttft" in goodput_config_dict:
valid_metrics.append(ttfts)
slo_values.append(
goodput_config_dict["ttft"] / MILLISECONDS_TO_SECONDS_CONVERSION
)
slo_values.append(goodput_config_dict["ttft"] /
MILLISECONDS_TO_SECONDS_CONVERSION)
if "tpot" in goodput_config_dict:
valid_metrics.append(all_tpots)
slo_values.append(
goodput_config_dict["tpot"] / MILLISECONDS_TO_SECONDS_CONVERSION
)
slo_values.append(goodput_config_dict["tpot"] /
MILLISECONDS_TO_SECONDS_CONVERSION)
if "e2el" in goodput_config_dict:
valid_metrics.append(e2els)
slo_values.append(
goodput_config_dict["e2el"] / MILLISECONDS_TO_SECONDS_CONVERSION
)
slo_values.append(goodput_config_dict["e2el"] /
MILLISECONDS_TO_SECONDS_CONVERSION)
for req_metric in zip(*valid_metrics):
is_good_req = all([s >= r for s, r in zip(slo_values, req_metric)])
@ -383,8 +360,7 @@ def calculate_metrics(
warnings.warn(
"All requests failed. This is likely due to a misconfiguration "
"on the benchmark arguments.",
stacklevel=2,
)
stacklevel=2)
metrics = BenchmarkMetrics(
completed=completed,
total_input=total_input,
@ -393,31 +369,27 @@ def calculate_metrics(
request_goodput=good_completed / dur_s,
output_throughput=sum(actual_output_lens) / dur_s,
total_token_throughput=(total_input + sum(actual_output_lens)) / dur_s,
mean_ttft_ms=np.mean(ttfts or 0)
* 1000, # ttfts is empty if streaming is not supported by backend
mean_ttft_ms=np.mean(ttfts or 0) *
1000, # ttfts is empty if streaming is not supported by backend
std_ttft_ms=np.std(ttfts or 0) * 1000,
median_ttft_ms=np.median(ttfts or 0) * 1000,
percentiles_ttft_ms=[
(p, np.percentile(ttfts or 0, p) * 1000) for p in selected_percentiles
],
percentiles_ttft_ms=[(p, np.percentile(ttfts or 0, p) * 1000)
for p in selected_percentiles],
mean_tpot_ms=np.mean(tpots or 0) * 1000,
std_tpot_ms=np.std(tpots or 0) * 1000,
median_tpot_ms=np.median(tpots or 0) * 1000,
percentiles_tpot_ms=[
(p, np.percentile(tpots or 0, p) * 1000) for p in selected_percentiles
],
percentiles_tpot_ms=[(p, np.percentile(tpots or 0, p) * 1000)
for p in selected_percentiles],
mean_itl_ms=np.mean(itls or 0) * 1000,
std_itl_ms=np.std(itls or 0) * 1000,
median_itl_ms=np.median(itls or 0) * 1000,
percentiles_itl_ms=[
(p, np.percentile(itls or 0, p) * 1000) for p in selected_percentiles
],
percentiles_itl_ms=[(p, np.percentile(itls or 0, p) * 1000)
for p in selected_percentiles],
mean_e2el_ms=np.mean(e2els or 0) * 1000,
std_e2el_ms=np.std(e2els or 0) * 1000,
median_e2el_ms=np.median(e2els or 0) * 1000,
percentiles_e2el_ms=[
(p, np.percentile(e2els or 0, p) * 1000) for p in selected_percentiles
],
percentiles_e2el_ms=[(p, np.percentile(e2els or 0, p) * 1000)
for p in selected_percentiles],
)
return metrics, actual_output_lens
@ -439,6 +411,7 @@ async def benchmark(
ignore_eos: bool,
max_concurrency: Optional[int],
structured_output_ratio: float,
structured_output_backend: str,
goodput_config_dict: Optional[dict[str, float]] = None,
):
if backend in ASYNC_REQUEST_FUNCS:
@ -450,17 +423,18 @@ async def benchmark(
extra_body = {}
# Add the schema to the extra_body
extra_body[request.structure_type] = request.schema
# Add the specific structured_output_backend
extra_body["guided_decoding_backend"] = structured_output_backend
return extra_body
print("Starting initial single prompt test run...")
structured_output_req_idx = random.sample(
range(len(input_requests)), int(len(input_requests) * structured_output_ratio)
)
range(len(input_requests)),
int(len(input_requests) * structured_output_ratio))
test_request = input_requests[0]
test_req_extra_body = (
prepare_extra_body(test_request) if 0 in structured_output_req_idx else None
)
test_req_extra_body = (prepare_extra_body(test_request)
if 0 in structured_output_req_idx else None)
test_input = RequestFuncInput(
model=model_id,
prompt=test_request.prompt,
@ -474,8 +448,7 @@ async def benchmark(
if not test_output.success:
raise ValueError(
"Initial test run failed - Please make sure benchmark arguments "
f"are correctly specified. Error: {test_output.error}"
)
f"are correctly specified. Error: {test_output.error}")
else:
print("Initial test run completed. Starting main benchmark run...")
@ -494,7 +467,10 @@ async def benchmark(
if profile_output.success:
print("Profiler started")
distribution = "Poisson process" if burstiness == 1.0 else "Gamma distribution"
if burstiness == 1.0:
distribution = "Poisson process"
else:
distribution = "Gamma distribution"
print(f"Traffic request rate: {request_rate}")
print(f"Burstiness factor: {burstiness} ({distribution})")
@ -506,21 +482,24 @@ async def benchmark(
# and it will simplify the code in limited_request_func.
# semaphore = (asyncio.Semaphore(max_concurrency)
# if max_concurrency else contextlib.nullcontext())
semaphore = asyncio.Semaphore(max_concurrency) if max_concurrency else None
semaphore = (asyncio.Semaphore(max_concurrency)
if max_concurrency else None)
async def limited_request_func(request_func_input, pbar):
if semaphore is None:
return await request_func(request_func_input=request_func_input, pbar=pbar)
return await request_func(request_func_input=request_func_input,
pbar=pbar)
async with semaphore:
return await request_func(request_func_input=request_func_input, pbar=pbar)
return await request_func(request_func_input=request_func_input,
pbar=pbar)
benchmark_start_time = time.perf_counter()
tasks: list[asyncio.Task] = []
expected: list[str] = []
async for i, request in get_request(input_requests, request_rate, burstiness):
extra_body = (
prepare_extra_body(request) if i in structured_output_req_idx else None
)
async for i, request in get_request(input_requests, request_rate,
burstiness):
extra_body = prepare_extra_body(
request) if i in structured_output_req_idx else None
request_func_input = RequestFuncInput(
model=model_id,
prompt=request.prompt,
@ -533,11 +512,24 @@ async def benchmark(
expected.append(request.completion)
tasks.append(
asyncio.create_task(
limited_request_func(request_func_input=request_func_input, pbar=pbar)
)
)
limited_request_func(request_func_input=request_func_input,
pbar=pbar)))
outputs: list[RequestFuncOutput] = await asyncio.gather(*tasks)
if profile:
print("Stopping profiler...")
profile_input = RequestFuncInput(
model=model_id,
prompt=test_request.prompt,
api_url=base_url + "/stop_profile",
prompt_len=test_request.prompt_len,
output_len=test_request.expected_output_len,
extra_body={test_request.structure_type: test_request.schema},
)
profile_output = await request_func(request_func_input=profile_input)
if profile_output.success:
print("Profiler stopped")
if pbar is not None:
pbar.close()
@ -553,62 +545,54 @@ async def benchmark(
goodput_config_dict=goodput_config_dict,
)
print("{s:{c}^{n}}".format(s=" Serving Benchmark Result ", n=50, c="="))
print("{s:{c}^{n}}".format(s=' Serving Benchmark Result ', n=50, c='='))
print("{:<40} {:<10}".format("Successful requests:", metrics.completed))
if max_concurrency is not None:
print("{:<40} {:<10}".format("Maximum request concurrency:", max_concurrency))
if request_rate != float("inf"):
print("{:<40} {:<10.2f}".format("Request rate configured (RPS):", request_rate))
print("{:<40} {:<10.2f}".format("Benchmark duration (s):", benchmark_duration))
print("{:<40} {:<10.2f}".format("Benchmark duration (s):",
benchmark_duration))
print("{:<40} {:<10}".format("Total input tokens:", metrics.total_input))
print("{:<40} {:<10}".format("Total generated tokens:", metrics.total_output))
print(
"{:<40} {:<10.2f}".format(
"Request throughput (req/s):", metrics.request_throughput
)
)
print("{:<40} {:<10}".format("Total generated tokens:",
metrics.total_output))
print("{:<40} {:<10.2f}".format("Request throughput (req/s):",
metrics.request_throughput))
if goodput_config_dict:
print(
"{:<40} {:<10.2f}".format(
"Request goodput (req/s):", metrics.request_goodput
)
)
print(
"{:<40} {:<10.2f}".format(
"Output token throughput (tok/s):", metrics.output_throughput
)
)
print(
"{:<40} {:<10.2f}".format(
"Total Token throughput (tok/s):", metrics.total_token_throughput
)
)
print("{:<40} {:<10.2f}".format("Request goodput (req/s):",
metrics.request_goodput))
print("{:<40} {:<10.2f}".format("Output token throughput (tok/s):",
metrics.output_throughput))
print("{:<40} {:<10.2f}".format("Total Token throughput (tok/s):",
metrics.total_token_throughput))
result = {
"duration": benchmark_duration,
"completed": metrics.completed,
"total_input_tokens": metrics.total_input,
"total_output_tokens": metrics.total_output,
"request_throughput": metrics.request_throughput,
"output_throughput": metrics.output_throughput,
"total_token_throughput": metrics.total_token_throughput,
"ttft_description": pd.Series([output.ttft for output in outputs])
.describe()
.to_dict(),
"tpot_description": pd.Series([output.tpot for output in outputs])
.describe()
.to_dict(),
"duration":
benchmark_duration,
"completed":
metrics.completed,
"total_input_tokens":
metrics.total_input,
"total_output_tokens":
metrics.total_output,
"request_throughput":
metrics.request_throughput,
"output_throughput":
metrics.output_throughput,
"total_token_throughput":
metrics.total_token_throughput,
"ttft_description":
pd.Series([output.ttft for output in outputs]).describe().to_dict(),
"tpot_description":
pd.Series([output.tpot for output in outputs]).describe().to_dict(),
"input_lens": [output.prompt_len for output in outputs],
"output_lens": actual_output_lens,
"output_lens":
actual_output_lens,
"ttfts": [output.ttft for output in outputs],
"itls": [output.itl for output in outputs],
"errors": [output.error for output in outputs],
}
ret = [
{"generated": output.generated_text, "expected": gt}
for output, gt in zip(outputs, expected)
]
ret = [{
'generated': output.generated_text,
'expected': gt
} for output, gt in zip(outputs, expected)]
def process_one_metric(
# E.g., "ttft"
@ -622,65 +606,45 @@ async def benchmark(
# metric.
if metric_attribute_name not in selected_percentile_metrics:
return
print("{s:{c}^{n}}".format(s=metric_header, n=50, c="-"))
print(
"{:<40} {:<10.2f}".format(
f"Mean {metric_name} (ms):",
getattr(metrics, f"mean_{metric_attribute_name}_ms"),
)
)
print(
"{:<40} {:<10.2f}".format(
f"Median {metric_name} (ms):",
getattr(metrics, f"median_{metric_attribute_name}_ms"),
)
)
print("{s:{c}^{n}}".format(s=metric_header, n=50, c='-'))
print("{:<40} {:<10.2f}".format(
f"Mean {metric_name} (ms):",
getattr(metrics, f"mean_{metric_attribute_name}_ms")))
print("{:<40} {:<10.2f}".format(
f"Median {metric_name} (ms):",
getattr(metrics, f"median_{metric_attribute_name}_ms")))
result[f"mean_{metric_attribute_name}_ms"] = getattr(
metrics, f"mean_{metric_attribute_name}_ms"
)
metrics, f"mean_{metric_attribute_name}_ms")
result[f"median_{metric_attribute_name}_ms"] = getattr(
metrics, f"median_{metric_attribute_name}_ms"
)
metrics, f"median_{metric_attribute_name}_ms")
result[f"std_{metric_attribute_name}_ms"] = getattr(
metrics, f"std_{metric_attribute_name}_ms"
)
for p, value in getattr(metrics, f"percentiles_{metric_attribute_name}_ms"):
metrics, f"std_{metric_attribute_name}_ms")
for p, value in getattr(metrics,
f"percentiles_{metric_attribute_name}_ms"):
p_word = str(int(p)) if int(p) == p else str(p)
print("{:<40} {:<10.2f}".format(f"P{p_word} {metric_name} (ms):", value))
print("{:<40} {:<10.2f}".format(f"P{p_word} {metric_name} (ms):",
value))
result[f"p{p_word}_{metric_attribute_name}_ms"] = value
process_one_metric("ttft", "TTFT", "Time to First Token")
process_one_metric("tpot", "TPOT", "Time per Output Token (excl. 1st token)")
process_one_metric("tpot", "TPOT",
"Time per Output Token (excl. 1st token)")
process_one_metric("itl", "ITL", "Inter-token Latency")
process_one_metric("e2el", "E2EL", "End-to-end Latency")
print("=" * 50)
if profile:
print("Stopping profiler...")
profile_input = RequestFuncInput(
model=model_id,
prompt=test_request.prompt,
api_url=base_url + "/stop_profile",
prompt_len=test_request.prompt_len,
output_len=test_request.expected_output_len,
extra_body={test_request.structure_type: test_request.schema},
)
profile_output = await request_func(request_func_input=profile_input)
if profile_output.success:
print("Profiler stopped")
return result, ret
def evaluate(ret, args):
def _eval_correctness_json(expected, actual):
# extract json string from string using regex
import regex as re
actual = actual.replace("\n", "").replace(" ", "").strip()
import re
actual = actual.replace('\n', '').replace(' ', '').strip()
try:
actual = re.search(r"\{.*\}", actual).group()
actual = re.search(r'\{.*\}', actual).group()
actual = json.loads(actual)
except Exception:
return False
@ -691,33 +655,29 @@ def evaluate(ret, args):
return actual in args.choice
def _eval_correctness_regex(expected, actual):
import regex as re
import re
return re.match(args.regex, actual) is not None
def _eval_correctness(expected, actual):
if args.structure_type == "guided_json":
if args.structure_type == 'guided_json':
return _eval_correctness_json(expected, actual)
elif args.structure_type == "guided_regex":
elif args.structure_type == 'guided_regex':
return _eval_correctness_regex(expected, actual)
elif args.structure_type == "guided_choice":
elif args.structure_type == 'guided_choice':
return _eval_correctness_choice(expected, actual)
else:
return None
scores = []
for res in ret:
score = _eval_correctness(res["expected"], res["generated"])
res["correctness"] = score
score = _eval_correctness(res['expected'], res['generated'])
res['correctness'] = score
scores.append(score)
not_none_scores = [score for score in scores if score is not None]
return (
(sum(not_none_scores) / len(not_none_scores) * 100)
if len(not_none_scores) > 0
else None
)
return (sum(not_none_scores) / len(not_none_scores) *
100) if len(not_none_scores) > 0 else None
def parse_goodput(slo_pairs):
@ -729,10 +689,9 @@ def parse_goodput(slo_pairs):
except ValueError as err:
raise argparse.ArgumentTypeError(
"Invalid format found for service level objectives. "
'Specify service level objectives for goodput as "KEY:VALUE" '
"Specify service level objectives for goodput as \"KEY:VALUE\" "
"pairs, where the key is a metric name, and the value is a "
"number in milliseconds."
) from err
"number in milliseconds.") from err
return goodput_config_dict
@ -746,14 +705,12 @@ def check_goodput_args(args):
raise ValueError(
f"Invalid metric name found, {slo_name}: {slo_val}. "
"The service level objective name should be one of "
f"{str(VALID_NAMES)}. "
)
f"{str(VALID_NAMES)}. ")
if slo_val < 0:
raise ValueError(
f"Invalid value found, {slo_name}: {slo_val}. "
"The service level objective value should be "
"non-negative."
)
"non-negative.")
return goodput_config_dict
@ -779,19 +736,19 @@ def main(args: argparse.Namespace):
tokenizer_mode=args.tokenizer_mode,
)
if args.dataset == "grammar":
args.structure_type = "guided_grammar"
elif args.dataset == "regex":
args.structure_type = "guided_regex"
elif args.dataset == "choice":
args.structure_type = "guided_choice"
if args.dataset == 'grammar':
args.structure_type = 'guided_grammar'
elif args.dataset == 'regex':
args.structure_type = 'guided_regex'
elif args.dataset == 'choice':
args.structure_type = 'guided_choice'
else:
args.structure_type = "guided_json"
args.structure_type = 'guided_json'
if args.no_structured_output:
args.structured_output_ratio = 0
if args.save_results:
result_file_name = f"{args.structured_output_ratio}guided"
result_file_name = f'{args.structured_output_ratio}guided'
result_file_name += f"_{backend}"
result_file_name += f"_{args.request_rate}qps"
result_file_name += f"_{args.model.split('/')[-1]}"
@ -819,29 +776,37 @@ def main(args: argparse.Namespace):
disable_tqdm=args.disable_tqdm,
profile=args.profile,
selected_percentile_metrics=args.percentile_metrics.split(","),
selected_percentiles=[float(p) for p in args.metric_percentiles.split(",")],
selected_percentiles=[
float(p) for p in args.metric_percentiles.split(",")
],
ignore_eos=args.ignore_eos,
max_concurrency=args.max_concurrency,
structured_output_ratio=args.structured_output_ratio,
structured_output_backend=args.structured_output_backend,
goodput_config_dict=goodput_config_dict,
)
)
))
# Save config and results to json
score = evaluate(ret, args)
print("correct_rate(%)", score, "\n")
print("correct_rate(%)", score, '\n')
if args.save_results:
results = {
"backend": backend,
"model_id": model_id,
"tokenizer_id": tokenizer_id,
"num_prompts": args.num_prompts,
"request_rate": args.request_rate
if args.request_rate < float("inf")
else "inf",
"burstiness": args.burstiness,
"max_concurrency": args.max_concurrency,
"correct_rate(%)": score,
"backend":
backend,
"model_id":
model_id,
"tokenizer_id":
tokenizer_id,
"num_prompts":
args.num_prompts,
"request_rate":
args.request_rate if args.request_rate < float("inf") else "inf",
"burstiness":
args.burstiness,
"max_concurrency":
args.max_concurrency,
"correct_rate(%)":
score
}
results = {"outputs": ret, **results, **benchmark_result}
@ -850,14 +815,13 @@ def main(args: argparse.Namespace):
result_file_name = args.result_filename
if args.result_dir:
result_file_name = os.path.join(args.result_dir, result_file_name)
with open(result_file_name, "w", encoding="utf-8") as outfile:
with open(result_file_name, "w", encoding='utf-8') as outfile:
json.dump(results, outfile, indent=4)
def create_argument_parser():
if __name__ == "__main__":
parser = FlexibleArgumentParser(
description="Benchmark the online serving throughput."
)
description="Benchmark the online serving throughput.")
parser.add_argument(
"--backend",
type=str,
@ -879,14 +843,16 @@ def create_argument_parser():
default="/v1/completions",
help="API endpoint.",
)
parser.add_argument(
"--dataset",
default="json",
choices=["json", "json-unique", "grammar", "regex", "choice", "xgrammar_bench"],
)
parser.add_argument(
"--json-schema-path", type=str, default=None, help="Path to json schema."
)
parser.add_argument("--dataset",
default='json',
choices=[
'json', 'json-unique', 'grammar', 'regex',
'choice', 'xgrammar_bench'
])
parser.add_argument("--json_schema_path",
type=str,
default=None,
help="Path to json schema.")
parser.add_argument(
"--max-concurrency",
type=int,
@ -898,8 +864,7 @@ def create_argument_parser():
"initiated, this argument will control how many are actually allowed "
"to execute at a time. This means that when used in combination, the "
"actual request rate may be lower than specified with --request-rate, "
"if the server is not processing requests fast enough to keep up.",
)
"if the server is not processing requests fast enough to keep up.")
parser.add_argument(
"--model",
type=str,
@ -909,13 +874,15 @@ def create_argument_parser():
parser.add_argument(
"--tokenizer",
type=str,
help="Name or path of the tokenizer, if not using the default tokenizer.", # noqa: E501
help=
"Name or path of the tokenizer, if not using the default tokenizer.", # noqa: E501
)
parser.add_argument(
"--tokenizer-mode",
type=str,
default="auto",
help="Name or path of the tokenizer, if not using the default tokenizer.", # noqa: E501
help=
"Name or path of the tokenizer, if not using the default tokenizer.", # noqa: E501
)
parser.add_argument(
"--num-prompts",
@ -992,56 +959,52 @@ def create_argument_parser():
"--ignore-eos",
action="store_true",
help="Set ignore_eos flag when sending the benchmark request."
"Warning: ignore_eos is not supported in deepspeed_mii and tgi.",
)
"Warning: ignore_eos is not supported in deepspeed_mii and tgi.")
parser.add_argument(
"--percentile-metrics",
type=str,
default="ttft,tpot,itl",
help="Comma-separated list of selected metrics to report percentiles. "
help="Comma-separated list of selected metrics to report percentils. "
"This argument specifies the metrics to report percentiles. "
'Allowed metric names are "ttft", "tpot", "itl", "e2el". '
'Default value is "ttft,tpot,itl".',
)
"Allowed metric names are \"ttft\", \"tpot\", \"itl\", \"e2el\". "
"Default value is \"ttft,tpot,itl\".")
parser.add_argument(
"--metric-percentiles",
type=str,
default="99",
help="Comma-separated list of percentiles for selected metrics. "
'To report 25-th, 50-th, and 75-th percentiles, use "25,50,75". '
'Default value is "99". '
'Use "--percentile-metrics" to select metrics.',
"To report 25-th, 50-th, and 75-th percentiles, use \"25,50,75\". "
"Default value is \"99\". "
"Use \"--percentile-metrics\" to select metrics.",
)
parser.add_argument(
"--goodput",
nargs="+",
required=False,
help='Specify service level objectives for goodput as "KEY:VALUE" '
help="Specify service level objectives for goodput as \"KEY:VALUE\" "
"pairs, where the key is a metric name, and the value is in "
'milliseconds. Multiple "KEY:VALUE" pairs can be provided, '
"milliseconds. Multiple \"KEY:VALUE\" pairs can be provided, "
"separated by spaces. Allowed request level metric names are "
'"ttft", "tpot", "e2el". For more context on the definition of '
"\"ttft\", \"tpot\", \"e2el\". For more context on the definition of "
"goodput, refer to DistServe paper: https://arxiv.org/pdf/2401.09670 "
"and the blog: https://hao-ai-lab.github.io/blogs/distserve",
)
"and the blog: https://hao-ai-lab.github.io/blogs/distserve")
parser.add_argument(
"--no-structured-output",
action="store_true",
default=False,
help="Whether to disable JSON decoding or not.",
)
parser.add_argument(
"--structured-output-ratio",
type=float,
default=1.0,
help="Ratio of Structured Outputs requests",
)
parser.add_argument("--no-structured-output",
action='store_true',
default=False,
help="Whether to disable JSON decoding or not.")
parser.add_argument("--structured-output-ratio",
type=float,
default=1.0,
help="Ratio of Structured Outputs requests")
parser.add_argument("--structured-output-backend",
type=str,
choices=[
"outlines", "lm-format-enforcer", "xgrammar",
"guidance", "auto"
],
default="auto",
help="Backend to use for structured outputs")
return parser
if __name__ == "__main__":
parser = create_argument_parser()
args = parser.parse_args()
main(args)

View File

@ -1,7 +1,5 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
"""Benchmark offline inference throughput."""
import argparse
import dataclasses
import json
@ -13,26 +11,18 @@ from typing import Any, Optional, Union
import torch
import uvloop
from tqdm import tqdm
from transformers import AutoModelForCausalLM, AutoTokenizer, PreTrainedTokenizerBase
from typing_extensions import deprecated
from benchmark_dataset import (
AIMODataset,
BurstGPTDataset,
ConversationDataset,
InstructCoderDataset,
RandomDataset,
SampleRequest,
ShareGPTDataset,
SonnetDataset,
VisionArenaDataset,
)
from benchmark_dataset import (AIMODataset, BurstGPTDataset,
ConversationDataset, InstructCoderDataset,
RandomDataset, SampleRequest, ShareGPTDataset,
SonnetDataset, VisionArenaDataset)
from benchmark_utils import convert_to_pytorch_benchmark_format, write_to_json
from tqdm import tqdm
from transformers import (AutoModelForCausalLM, AutoTokenizer,
PreTrainedTokenizerBase)
from vllm.engine.arg_utils import AsyncEngineArgs, EngineArgs
from vllm.entrypoints.openai.api_server import (
build_async_engine_client_from_engine_args,
)
build_async_engine_client_from_engine_args)
from vllm.inputs import TextPrompt, TokensPrompt
from vllm.lora.request import LoRARequest
from vllm.outputs import RequestOutput
@ -47,30 +37,23 @@ def run_vllm(
disable_detokenize: bool = False,
) -> tuple[float, Optional[list[RequestOutput]]]:
from vllm import LLM, SamplingParams
llm = LLM(**dataclasses.asdict(engine_args))
assert all(
llm.llm_engine.model_config.max_model_len
>= (request.prompt_len + request.expected_output_len)
for request in requests
), (
"Please ensure that max_model_len is greater than the sum of"
" prompt_len and expected_output_len for all requests."
)
llm.llm_engine.model_config.max_model_len >= (
request.prompt_len + request.expected_output_len)
for request in requests), (
"Please ensure that max_model_len is greater than the sum of"
" prompt_len and expected_output_len for all requests.")
# Add the requests to the engine.
prompts: list[Union[TextPrompt, TokensPrompt]] = []
sampling_params: list[SamplingParams] = []
for request in requests:
prompts.append(
TokensPrompt(
prompt_token_ids=request.prompt["prompt_token_ids"],
multi_modal_data=request.multi_modal_data,
)
if "prompt_token_ids" in request.prompt
else TextPrompt(
prompt=request.prompt, multi_modal_data=request.multi_modal_data
)
)
TokensPrompt(prompt_token_ids=request.prompt["prompt_token_ids"],
multi_modal_data=request.multi_modal_data)
if "prompt_token_ids" in request.prompt else \
TextPrompt(prompt=request.prompt,
multi_modal_data=request.multi_modal_data))
sampling_params.append(
SamplingParams(
n=n,
@ -79,8 +62,7 @@ def run_vllm(
ignore_eos=True,
max_tokens=request.expected_output_len,
detokenize=not disable_detokenize,
)
)
))
lora_requests: Optional[list[LoRARequest]] = None
if engine_args.enable_lora:
lora_requests = [request.lora_request for request in requests]
@ -90,14 +72,16 @@ def run_vllm(
outputs = None
if not use_beam_search:
start = time.perf_counter()
outputs = llm.generate(
prompts, sampling_params, lora_request=lora_requests, use_tqdm=True
)
outputs = llm.generate(prompts,
sampling_params,
lora_request=lora_requests,
use_tqdm=True)
end = time.perf_counter()
else:
assert lora_requests is None, "BeamSearch API does not support LoRA"
prompts = [request.prompt for request in requests]
# output_len should be the same for all requests.
output_len = requests[0].expected_output_len
output_len = requests[0][2]
for request in requests:
assert request.expected_output_len == output_len
start = time.perf_counter()
@ -107,35 +91,30 @@ def run_vllm(
beam_width=n,
max_tokens=output_len,
ignore_eos=True,
),
)
))
end = time.perf_counter()
return end - start, outputs
def run_vllm_chat(
requests: list[SampleRequest],
n: int,
engine_args: EngineArgs,
disable_detokenize: bool = False,
) -> tuple[float, list[RequestOutput]]:
requests: list[SampleRequest],
n: int,
engine_args: EngineArgs,
disable_detokenize: bool = False) -> tuple[float, list[RequestOutput]]:
"""
Run vLLM chat benchmark. This function is recommended ONLY for benchmarking
multimodal models as it properly handles multimodal inputs and chat
formatting. For non-multimodal models, use run_vllm() instead.
"""
from vllm import LLM, SamplingParams
llm = LLM(**dataclasses.asdict(engine_args))
assert all(
llm.llm_engine.model_config.max_model_len
>= (request.prompt_len + request.expected_output_len)
for request in requests
), (
"Please ensure that max_model_len is greater than the sum of "
"prompt_len and expected_output_len for all requests."
)
llm.llm_engine.model_config.max_model_len >= (
request.prompt_len + request.expected_output_len)
for request in requests), (
"Please ensure that max_model_len is greater than the sum of "
"prompt_len and expected_output_len for all requests.")
prompts = []
sampling_params: list[SamplingParams] = []
@ -149,8 +128,7 @@ def run_vllm_chat(
ignore_eos=True,
max_tokens=request.expected_output_len,
detokenize=not disable_detokenize,
)
)
))
start = time.perf_counter()
outputs = llm.chat(prompts, sampling_params, use_tqdm=True)
end = time.perf_counter()
@ -167,18 +145,13 @@ async def run_vllm_async(
from vllm import SamplingParams
async with build_async_engine_client_from_engine_args(
engine_args,
disable_frontend_multiprocessing=disable_frontend_multiprocessing,
) as llm:
model_config = await llm.get_model_config()
engine_args, disable_frontend_multiprocessing) as llm:
assert all(
model_config.max_model_len
>= (request.prompt_len + request.expected_output_len)
for request in requests
), (
"Please ensure that max_model_len is greater than the sum of"
" prompt_len and expected_output_len for all requests."
)
llm.model_config.max_model_len >= (request.prompt_len +
request.expected_output_len)
for request in requests), (
"Please ensure that max_model_len is greater than the sum of"
" prompt_len and expected_output_len for all requests.")
# Add the requests to the engine.
prompts: list[Union[TextPrompt, TokensPrompt]] = []
@ -186,15 +159,11 @@ async def run_vllm_async(
lora_requests: list[Optional[LoRARequest]] = []
for request in requests:
prompts.append(
TokensPrompt(
prompt_token_ids=request.prompt["prompt_token_ids"],
multi_modal_data=request.multi_modal_data,
)
if "prompt_token_ids" in request.prompt
else TextPrompt(
prompt=request.prompt, multi_modal_data=request.multi_modal_data
)
)
TokensPrompt(prompt_token_ids=request.prompt["prompt_token_ids"],
multi_modal_data=request.multi_modal_data)
if "prompt_token_ids" in request.prompt else \
TextPrompt(prompt=request.prompt,
multi_modal_data=request.multi_modal_data))
sampling_params.append(
SamplingParams(
n=n,
@ -203,16 +172,17 @@ async def run_vllm_async(
ignore_eos=True,
max_tokens=request.expected_output_len,
detokenize=not disable_detokenize,
)
)
))
lora_requests.append(request.lora_request)
generators = []
start = time.perf_counter()
for i, (prompt, sp, lr) in enumerate(
zip(prompts, sampling_params, lora_requests)
):
generator = llm.generate(prompt, sp, lora_request=lr, request_id=f"test{i}")
for i, (prompt, sp,
lr) in enumerate(zip(prompts, sampling_params, lora_requests)):
generator = llm.generate(prompt,
sp,
lora_request=lr,
request_id=f"test{i}")
generators.append(generator)
all_gens = merge_async_iterators(*generators)
async for i, res in all_gens:
@ -231,8 +201,7 @@ def run_hf(
disable_detokenize: bool = False,
) -> float:
llm = AutoModelForCausalLM.from_pretrained(
model, torch_dtype=torch.float16, trust_remote_code=trust_remote_code
)
model, torch_dtype=torch.float16, trust_remote_code=trust_remote_code)
if llm.config.model_type == "llama":
# To enable padding in the HF backend.
tokenizer.pad_token = tokenizer.eos_token
@ -255,15 +224,14 @@ def run_hf(
# Check if we can add more requests to the batch.
next_prompt_len = requests[i + 1].prompt_len
next_output_len = requests[i + 1].expected_output_len
if (
max(max_prompt_len, next_prompt_len)
+ max(max_output_len, next_output_len)
) <= 2048:
if (max(max_prompt_len, next_prompt_len) +
max(max_output_len, next_output_len)) <= 2048:
# We can add more requests to the batch.
continue
# Generate the sequences.
input_ids = tokenizer(batch, return_tensors="pt", padding=True).input_ids
input_ids = tokenizer(batch, return_tensors="pt",
padding=True).input_ids
llm_outputs = llm.generate(
input_ids=input_ids.cuda(),
do_sample=True,
@ -293,7 +261,6 @@ def run_mii(
output_len: int,
) -> float:
from mii import client, serve
llm = serve(model, tensor_parallel=tensor_parallel_size)
prompts = [request.prompt for request in requests]
@ -305,9 +272,8 @@ def run_mii(
return end - start
def save_to_pytorch_benchmark_format(
args: argparse.Namespace, results: dict[str, Any]
) -> None:
def save_to_pytorch_benchmark_format(args: argparse.Namespace,
results: dict[str, Any]) -> None:
pt_records = convert_to_pytorch_benchmark_format(
args=args,
metrics={
@ -315,9 +281,9 @@ def save_to_pytorch_benchmark_format(
"tokens_per_second": [results["tokens_per_second"]],
},
extra_info={
k: results[k] for k in ["elapsed_time", "num_requests", "total_num_tokens"]
},
)
k: results[k]
for k in ["elapsed_time", "num_requests", "total_num_tokens"]
})
if pt_records:
# Don't use json suffix here as we don't want CI to pick it up
pt_file = f"{os.path.splitext(args.output_json)[0]}.pytorch.json"
@ -349,32 +315,30 @@ def get_requests(args, tokenizer):
sample_kwargs["enable_multimodal_chat"] = True
elif args.dataset_name == "sonnet":
assert tokenizer.chat_template or tokenizer.default_chat_template, (
"Tokenizer/model must have chat template for sonnet dataset."
)
"Tokenizer/model must have chat template for sonnet dataset.")
dataset_cls = SonnetDataset
sample_kwargs["prefix_len"] = args.prefix_len
sample_kwargs["return_prompt_formatted"] = True
elif args.dataset_name == "burstgpt":
dataset_cls = BurstGPTDataset
elif args.dataset_name == "hf":
common_kwargs["no_stream"] = args.no_stream
if args.dataset_path in VisionArenaDataset.SUPPORTED_DATASET_PATHS:
dataset_cls = VisionArenaDataset
common_kwargs["dataset_subset"] = None
common_kwargs["dataset_split"] = "train"
common_kwargs['dataset_subset'] = None
common_kwargs['dataset_split'] = "train"
sample_kwargs["enable_multimodal_chat"] = True
elif args.dataset_path in InstructCoderDataset.SUPPORTED_DATASET_PATHS:
dataset_cls = InstructCoderDataset
common_kwargs["dataset_split"] = "train"
common_kwargs['dataset_split'] = "train"
elif args.dataset_path in ConversationDataset.SUPPORTED_DATASET_PATHS:
dataset_cls = ConversationDataset
common_kwargs["dataset_subset"] = args.hf_subset
common_kwargs["dataset_split"] = args.hf_split
common_kwargs['dataset_subset'] = args.hf_subset
common_kwargs['dataset_split'] = args.hf_split
sample_kwargs["enable_multimodal_chat"] = True
elif args.dataset_path in AIMODataset.SUPPORTED_DATASET_PATHS:
dataset_cls = AIMODataset
common_kwargs["dataset_subset"] = None
common_kwargs["dataset_split"] = "train"
common_kwargs['dataset_subset'] = None
common_kwargs['dataset_split'] = "train"
else:
raise ValueError(f"Unknown dataset name: {args.dataset_name}")
# Remove None values
@ -382,10 +346,6 @@ def get_requests(args, tokenizer):
return dataset_cls(**common_kwargs).sample(**sample_kwargs)
@deprecated(
"benchmark_throughput.py is deprecated and will be removed in a "
"future version. Please use 'vllm bench throughput' instead.",
)
def main(args: argparse.Namespace):
if args.seed is None:
args.seed = 0
@ -393,10 +353,10 @@ def main(args: argparse.Namespace):
random.seed(args.seed)
# Sample the requests.
tokenizer = AutoTokenizer.from_pretrained(
args.tokenizer, trust_remote_code=args.trust_remote_code
)
args.tokenizer, trust_remote_code=args.trust_remote_code)
requests = get_requests(args, tokenizer)
is_multi_modal = any(request.multi_modal_data is not None for request in requests)
is_multi_modal = any(request.multi_modal_data is not None
for request in requests)
request_outputs: Optional[list[RequestOutput]] = None
if args.backend == "vllm":
if args.async_engine:
@ -407,34 +367,23 @@ def main(args: argparse.Namespace):
AsyncEngineArgs.from_cli_args(args),
args.disable_frontend_multiprocessing,
args.disable_detokenize,
)
)
))
else:
elapsed_time, request_outputs = run_vllm(
requests,
args.n,
EngineArgs.from_cli_args(args),
args.disable_detokenize,
)
requests, args.n, EngineArgs.from_cli_args(args),
args.disable_detokenize)
elif args.backend == "hf":
assert args.tensor_parallel_size == 1
elapsed_time = run_hf(
requests,
args.model,
tokenizer,
args.n,
args.hf_max_batch_size,
args.trust_remote_code,
args.disable_detokenize,
)
elapsed_time = run_hf(requests, args.model, tokenizer, args.n,
args.hf_max_batch_size, args.trust_remote_code,
args.disable_detokenize)
elif args.backend == "mii":
elapsed_time = run_mii(
requests, args.model, args.tensor_parallel_size, args.output_len
)
elapsed_time = run_mii(requests, args.model, args.tensor_parallel_size,
args.output_len)
elif args.backend == "vllm-chat":
elapsed_time, request_outputs = run_vllm_chat(
requests, args.n, EngineArgs.from_cli_args(args), args.disable_detokenize
)
requests, args.n, EngineArgs.from_cli_args(args),
args.disable_detokenize)
else:
raise ValueError(f"Unknown backend: {args.backend}")
@ -446,31 +395,28 @@ def main(args: argparse.Namespace):
for ro in request_outputs:
if not isinstance(ro, RequestOutput):
continue
total_prompt_tokens += (
len(ro.prompt_token_ids) if ro.prompt_token_ids else 0
)
total_output_tokens += sum(len(o.token_ids) for o in ro.outputs if o)
total_prompt_tokens += len(
ro.prompt_token_ids) if ro.prompt_token_ids else 0
total_output_tokens += sum(
len(o.token_ids) for o in ro.outputs if o)
total_num_tokens = total_prompt_tokens + total_output_tokens
else:
total_num_tokens = sum(r.prompt_len + r.expected_output_len for r in requests)
total_num_tokens = sum(r.prompt_len + r.expected_output_len
for r in requests)
total_output_tokens = sum(r.expected_output_len for r in requests)
total_prompt_tokens = total_num_tokens - total_output_tokens
if is_multi_modal and args.backend != "vllm-chat":
print(
"\033[91mWARNING\033[0m: Multi-modal request with "
f"{args.backend} backend detected. The "
"following metrics are not accurate because image tokens are not"
" counted. See vllm-project/vllm/issues/9778 for details."
)
print("\033[91mWARNING\033[0m: Multi-modal request with "
f"{args.backend} backend detected. The "
"following metrics are not accurate because image tokens are not"
" counted. See vllm-project/vllm/issues/9778 for details.")
# TODO(vllm-project/vllm/issues/9778): Count multi-modal token length.
# vllm-chat backend counts the image tokens now
print(
f"Throughput: {len(requests) / elapsed_time:.2f} requests/s, "
f"{total_num_tokens / elapsed_time:.2f} total tokens/s, "
f"{total_output_tokens / elapsed_time:.2f} output tokens/s"
)
print(f"Throughput: {len(requests) / elapsed_time:.2f} requests/s, "
f"{total_num_tokens / elapsed_time:.2f} total tokens/s, "
f"{total_output_tokens / elapsed_time:.2f} output tokens/s")
print(f"Total num prompt tokens: {total_prompt_tokens}")
print(f"Total num output tokens: {total_output_tokens}")
@ -498,8 +444,7 @@ def validate_args(args):
warnings.warn(
"The '--dataset' argument will be deprecated in the next release. "
"Please use '--dataset-name' and '--dataset-path' instead.",
stacklevel=2,
)
stacklevel=2)
args.dataset_path = args.dataset
if not getattr(args, "tokenizer", None):
@ -512,8 +457,9 @@ def validate_args(args):
# === Dataset Configuration ===
if not args.dataset and not args.dataset_path:
print("When dataset path is not set, it will default to random dataset")
args.dataset_name = "random"
print(
"When dataset path is not set, it will default to random dataset")
args.dataset_name = 'random'
if args.input_len is None:
raise ValueError("input_len must be provided for a random dataset")
@ -521,55 +467,41 @@ def validate_args(args):
# --hf-subset and --hf-split: only used
# when dataset_name is 'hf'
if args.dataset_name != "hf" and (
getattr(args, "hf_subset", None) is not None
or getattr(args, "hf_split", None) is not None
):
warnings.warn(
"--hf-subset and --hf-split will be ignored \
getattr(args, "hf_subset", None) is not None
or getattr(args, "hf_split", None) is not None):
warnings.warn("--hf-subset and --hf-split will be ignored \
since --dataset-name is not 'hf'.",
stacklevel=2,
)
stacklevel=2)
elif args.dataset_name == "hf":
if args.dataset_path in (
VisionArenaDataset.SUPPORTED_DATASET_PATHS.keys()
| ConversationDataset.SUPPORTED_DATASET_PATHS
):
assert args.backend == "vllm-chat", (
f"{args.dataset_path} needs to use vllm-chat as the backend."
) # noqa: E501
elif args.dataset_path in (
InstructCoderDataset.SUPPORTED_DATASET_PATHS
| AIMODataset.SUPPORTED_DATASET_PATHS
):
assert args.backend == "vllm", (
f"{args.dataset_path} needs to use vllm as the backend."
) # noqa: E501
VisionArenaDataset.SUPPORTED_DATASET_PATHS.keys()
| ConversationDataset.SUPPORTED_DATASET_PATHS):
assert args.backend == "vllm-chat", f"{args.dataset_path} needs to use vllm-chat as the backend." #noqa: E501
elif args.dataset_path in (InstructCoderDataset.SUPPORTED_DATASET_PATHS
| AIMODataset.SUPPORTED_DATASET_PATHS):
assert args.backend == "vllm", f"{args.dataset_path} needs to use vllm as the backend." #noqa: E501
else:
raise ValueError(f"{args.dataset_path} is not supported by hf dataset.")
raise ValueError(
f"{args.dataset_path} is not supported by hf dataset.")
# --random-range-ratio: only used when dataset_name is 'random'
if args.dataset_name != "random" and args.random_range_ratio is not None:
warnings.warn(
"--random-range-ratio will be ignored since \
if args.dataset_name != 'random' and args.random_range_ratio is not None:
warnings.warn("--random-range-ratio will be ignored since \
--dataset-name is not 'random'.",
stacklevel=2,
)
stacklevel=2)
# --prefix-len: only used when dataset_name is 'random', 'sonnet', or not
# set.
if (
args.dataset_name not in {"random", "sonnet", None}
and args.prefix_len is not None
):
warnings.warn(
"--prefix-len will be ignored since --dataset-name\
if args.dataset_name not in {"random", "sonnet", None
} and args.prefix_len is not None:
warnings.warn("--prefix-len will be ignored since --dataset-name\
is not 'random', 'sonnet', or not set.",
stacklevel=2,
)
stacklevel=2)
# === LoRA Settings ===
if getattr(args, "enable_lora", False) and args.backend != "vllm":
raise ValueError("LoRA benchmarking is only supported for vLLM backend")
raise ValueError(
"LoRA benchmarking is only supported for vLLM backend")
if getattr(args, "enable_lora", False) and args.lora_path is None:
raise ValueError("LoRA path must be provided when enable_lora is True")
@ -579,10 +511,8 @@ def validate_args(args):
if args.backend != "hf" and args.hf_max_batch_size is not None:
raise ValueError("HF max batch size is only for HF backend.")
if (
args.backend in {"hf", "mii"}
and getattr(args, "quantization", None) is not None
):
if args.backend in {"hf", "mii"} and getattr(args, "quantization",
None) is not None:
raise ValueError("Quantization is only for vLLM backend.")
if args.backend == "mii" and args.dtype != "auto":
@ -590,37 +520,29 @@ def validate_args(args):
if args.backend == "mii" and args.n != 1:
raise ValueError("n must be 1 for MII backend.")
if args.backend == "mii" and args.tokenizer != args.model:
raise ValueError("Tokenizer must be the same as the model for MII backend.")
raise ValueError(
"Tokenizer must be the same as the model for MII backend.")
# --data-parallel is not supported currently.
# https://github.com/vllm-project/vllm/issues/16222
if args.data_parallel_size > 1:
raise ValueError(
"Data parallel is not supported in offline benchmark, "
"please use benchmark serving instead"
)
"Data parallel is not supported in offline benchmark, \
please use benchmark serving instead")
def create_argument_parser():
if __name__ == "__main__":
parser = FlexibleArgumentParser(description="Benchmark the throughput.")
parser.add_argument(
"--backend",
type=str,
choices=["vllm", "hf", "mii", "vllm-chat"],
default="vllm",
)
parser.add_argument("--backend",
type=str,
choices=["vllm", "hf", "mii", "vllm-chat"],
default="vllm")
parser.add_argument(
"--dataset-name",
type=str,
choices=["sharegpt", "random", "sonnet", "burstgpt", "hf"],
help="Name of the dataset to benchmark on.",
default="sharegpt",
)
parser.add_argument(
"--no-stream",
action="store_true",
help="Do not load the dataset in streaming mode.",
)
default="sharegpt")
parser.add_argument(
"--dataset",
type=str,
@ -628,70 +550,57 @@ def create_argument_parser():
help="Path to the ShareGPT dataset, will be deprecated in\
the next release. The dataset is expected to "
"be a json in form of list[dict[..., conversations: "
"list[dict[..., value: <prompt_or_response>]]]]",
)
"list[dict[..., value: <prompt_or_response>]]]]")
parser.add_argument("--dataset-path",
type=str,
default=None,
help="Path to the dataset")
parser.add_argument("--input-len",
type=int,
default=None,
help="Input prompt length for each request")
parser.add_argument("--output-len",
type=int,
default=None,
help="Output length for each request. Overrides the "
"output length from the dataset.")
parser.add_argument("--n",
type=int,
default=1,
help="Number of generated sequences per prompt.")
parser.add_argument("--num-prompts",
type=int,
default=1000,
help="Number of prompts to process.")
parser.add_argument("--hf-max-batch-size",
type=int,
default=None,
help="Maximum batch size for HF backend.")
parser.add_argument(
"--dataset-path", type=str, default=None, help="Path to the dataset"
)
parser.add_argument(
"--input-len",
type=int,
default=None,
help="Input prompt length for each request",
)
parser.add_argument(
"--output-len",
type=int,
default=None,
help="Output length for each request. Overrides the "
"output length from the dataset.",
)
parser.add_argument(
"--n", type=int, default=1, help="Number of generated sequences per prompt."
)
parser.add_argument(
"--num-prompts", type=int, default=1000, help="Number of prompts to process."
)
parser.add_argument(
"--hf-max-batch-size",
type=int,
default=None,
help="Maximum batch size for HF backend.",
)
parser.add_argument(
"--output-json",
'--output-json',
type=str,
default=None,
help="Path to save the throughput results in JSON format.",
)
parser.add_argument(
"--async-engine",
action="store_true",
default=False,
help="Use vLLM async engine rather than LLM class.",
)
parser.add_argument(
"--disable-frontend-multiprocessing",
action="store_true",
default=False,
help="Disable decoupled async engine frontend.",
)
help='Path to save the throughput results in JSON format.')
parser.add_argument("--async-engine",
action='store_true',
default=False,
help="Use vLLM async engine rather than LLM class.")
parser.add_argument("--disable-frontend-multiprocessing",
action='store_true',
default=False,
help="Disable decoupled async engine frontend.")
parser.add_argument(
"--disable-detokenize",
action="store_true",
help=(
"Do not detokenize the response (i.e. do not include "
"detokenization time in the measurement)"
),
)
help=("Do not detokenize the response (i.e. do not include "
"detokenization time in the measurement)"))
# LoRA
parser.add_argument(
"--lora-path",
type=str,
default=None,
help="Path to the LoRA adapters to use. This can be an absolute path, "
"a relative path, or a Hugging Face model identifier.",
)
help="Path to the lora adapters to use. This can be an absolute path, "
"a relative path, or a Hugging Face model identifier.")
parser.add_argument(
"--prefix-len",
type=int,
@ -705,8 +614,7 @@ def create_argument_parser():
f"prefix_len (default: {SonnetDataset.DEFAULT_PREFIX_LEN}) "
"controls how much of the input is fixed lines versus "
"random lines, but the total input length remains approximately "
"input_len tokens.",
)
"input_len tokens.")
# random dataset
parser.add_argument(
"--random-range-ratio",
@ -719,21 +627,17 @@ def create_argument_parser():
"[length * (1 - range_ratio), length * (1 + range_ratio)].",
)
# hf dataset
parser.add_argument(
"--hf-subset", type=str, default=None, help="Subset of the HF dataset."
)
parser.add_argument(
"--hf-split", type=str, default=None, help="Split of the HF dataset."
)
# hf dtaset
parser.add_argument("--hf-subset",
type=str,
default=None,
help="Subset of the HF dataset.")
parser.add_argument("--hf-split",
type=str,
default=None,
help="Split of the HF dataset.")
parser = AsyncEngineArgs.add_cli_args(parser)
return parser
if __name__ == "__main__":
parser = create_argument_parser()
args = parser.parse_args()
if args.tokenizer is None:
args.tokenizer = args.model

View File

@ -1,17 +1,15 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
import json
import math
import os
import time
from types import TracebackType
from typing import Any, Optional, Union
from typing import Any
def convert_to_pytorch_benchmark_format(
args: argparse.Namespace, metrics: dict[str, list], extra_info: dict[str, Any]
) -> list:
def convert_to_pytorch_benchmark_format(args: argparse.Namespace,
metrics: dict[str, list],
extra_info: dict[str, Any]) -> list:
"""
Save the benchmark results in the format used by PyTorch OSS benchmark with
on metric per record
@ -39,12 +37,12 @@ def convert_to_pytorch_benchmark_format(
},
}
tp = record["benchmark"]["extra_info"]["args"].get("tensor_parallel_size")
tp = record["benchmark"]["extra_info"]["args"].get(
"tensor_parallel_size")
# Save tensor_parallel_size parameter if it's part of the metadata
if not tp and "tensor_parallel_size" in extra_info:
record["benchmark"]["extra_info"]["args"]["tensor_parallel_size"] = (
extra_info["tensor_parallel_size"]
)
record["benchmark"]["extra_info"]["args"][
"tensor_parallel_size"] = extra_info["tensor_parallel_size"]
records.append(record)
@ -52,6 +50,7 @@ def convert_to_pytorch_benchmark_format(
class InfEncoder(json.JSONEncoder):
def clear_inf(self, o: Any):
if isinstance(o, dict):
return {k: self.clear_inf(v) for k, v in o.items()}
@ -67,59 +66,4 @@ class InfEncoder(json.JSONEncoder):
def write_to_json(filename: str, records: list) -> None:
with open(filename, "w") as f:
json.dump(
records,
f,
cls=InfEncoder,
default=lambda o: f"<{type(o).__name__} object is not JSON serializable>",
)
# Collect time and generate time metrics
#
# Example Usage:
# collector = TimeCollector(TimeCollector.US)
# for _ in range(total_iteration):
# with collector:
# ...
# collector.dump_avg_max()
class TimeCollector:
NS: int = 1
US: int = NS * 1000
MS: int = US * 1000
S: int = MS * 1000
def __init__(self, scale: int) -> None:
self.cnt: int = 0
self._sum: int = 0
self._max: Optional[int] = None
self.scale = scale
self.start_time: int = time.monotonic_ns()
def collect(self, v: int) -> None:
self.cnt += 1
self._sum += v
if self._max is None:
self._max = v
else:
self._max = max(self._max, v)
def avg(self) -> Union[float, str]:
return self._sum * 1.0 / self.cnt / self.scale if self.cnt > 0 else "N/A"
def max(self) -> Union[float, str]:
return self._max / self.scale if self._max else "N/A"
def dump_avg_max(self) -> list[Union[float, str]]:
return [self.avg(), self.max()]
def __enter__(self) -> None:
self.start_time = time.monotonic_ns()
def __exit__(
self,
exc_type: Optional[type[BaseException]],
exc_value: Optional[BaseException],
exc_traceback: Optional[TracebackType],
) -> None:
self.collect(time.monotonic_ns() - self.start_time)
json.dump(records, f, cls=InfEncoder)

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import argparse
import copy
@ -24,9 +23,8 @@ DEFAULT_TP_SIZES = [1]
# bench
def bench_fn(
label: str, sub_label: str, description: str, fn: Callable, *args, **kwargs
) -> TMeasurement:
def bench_fn(label: str, sub_label: str, description: str, fn: Callable, *args,
**kwargs) -> TMeasurement:
min_run_time = 1
globals = {
@ -43,18 +41,16 @@ def bench_fn(
).blocked_autorange(min_run_time=min_run_time)
def bench_int8(
dtype: torch.dtype, m: int, k: int, n: int, label: str, sub_label: str
) -> Iterable[TMeasurement]:
def bench_int8(dtype: torch.dtype, m: int, k: int, n: int, label: str,
sub_label: str) -> Iterable[TMeasurement]:
assert dtype == torch.int8
b_compressed, e, a, b = make_rand_sparse_tensors(torch.int8, m, n, k)
scale_a = torch.tensor(1.0, device="cuda", dtype=torch.float32)
scale_b = torch.tensor(1.0, device="cuda", dtype=torch.float32)
bias = torch.zeros((n,), device="cuda", dtype=torch.bfloat16)
bias = torch.zeros((n, ), device="cuda", dtype=torch.bfloat16)
out = ops.cutlass_scaled_sparse_mm(
a, b_compressed, e, scale_a, scale_b, torch.bfloat16
)
out = ops.cutlass_scaled_sparse_mm(a, b_compressed, e, scale_a, scale_b,
torch.bfloat16)
out_ref = ops.cutlass_scaled_mm(a, b, scale_a, scale_b, torch.bfloat16)
if not torch.allclose(out, out_ref):
@ -67,107 +63,54 @@ def bench_int8(
timers = []
# pytorch impl - bfloat16
timers.append(
bench_fn(
label,
sub_label,
"pytorch_bf16_bf16_bf16_matmul-no-scales",
torch.mm,
a.to(dtype=torch.bfloat16),
b.to(dtype=torch.bfloat16),
)
)
bench_fn(label, sub_label, "pytorch_bf16_bf16_bf16_matmul-no-scales",
torch.mm, a.to(dtype=torch.bfloat16),
b.to(dtype=torch.bfloat16)))
# pytorch impl - float16
timers.append(
bench_fn(
label,
sub_label,
"pytorch_fp16_fp16_fp16_matmul-no-scales",
torch.mm,
a.to(dtype=torch.float16),
b.to(dtype=torch.float16),
)
)
bench_fn(label, sub_label,
"pytorch_fp16_fp16_fp16_matmul-no-scales", torch.mm,
a.to(dtype=torch.float16), b.to(dtype=torch.float16)))
# cutlass impl
timers.append(
bench_fn(
label,
sub_label,
"cutlass_i8_i8_bf16_scaled_mm",
ops.cutlass_scaled_mm,
a,
b,
scale_a,
scale_b,
torch.bfloat16,
)
)
bench_fn(label, sub_label, "cutlass_i8_i8_bf16_scaled_mm",
ops.cutlass_scaled_mm, a, b, scale_a, scale_b,
torch.bfloat16))
# cutlass with bias
timers.append(
bench_fn(
label,
sub_label,
"cutlass_i8_i8_bf16_scaled_mm_bias",
ops.cutlass_scaled_mm,
a,
b,
scale_a,
scale_b,
torch.bfloat16,
bias,
)
)
bench_fn(label, sub_label, "cutlass_i8_i8_bf16_scaled_mm_bias",
ops.cutlass_scaled_mm, a, b, scale_a, scale_b, torch.bfloat16,
bias))
# cutlass sparse impl
timers.append(
bench_fn(
label,
sub_label,
"cutlass_i8_i8_bf16_scaled_sparse_mm",
ops.cutlass_scaled_sparse_mm,
a,
b_compressed,
e,
scale_a,
scale_b,
torch.bfloat16,
)
)
bench_fn(label, sub_label, "cutlass_i8_i8_bf16_scaled_sparse_mm",
ops.cutlass_scaled_sparse_mm, a, b_compressed, e, scale_a,
scale_b, torch.bfloat16))
# cutlass sparse with bias
timers.append(
bench_fn(
label,
sub_label,
"cutlass_i8_i8_bf16_scaled_sparse_mm_bias",
ops.cutlass_scaled_sparse_mm,
a,
b_compressed,
e,
scale_a,
scale_b,
torch.bfloat16,
bias,
)
)
bench_fn(label, sub_label, "cutlass_i8_i8_bf16_scaled_sparse_mm_bias",
ops.cutlass_scaled_sparse_mm, a, b_compressed, e, scale_a,
scale_b, torch.bfloat16, bias))
return timers
def bench_fp8(
dtype: torch.dtype, m: int, k: int, n: int, label: str, sub_label: str
) -> Iterable[TMeasurement]:
def bench_fp8(dtype: torch.dtype, m: int, k: int, n: int, label: str,
sub_label: str) -> Iterable[TMeasurement]:
assert dtype == torch.float8_e4m3fn
b_compressed, e, a, b = make_rand_sparse_tensors(torch.float8_e4m3fn, m, n, k)
b_compressed, e, a, b = make_rand_sparse_tensors(torch.float8_e4m3fn, m, n,
k)
scale_a = torch.tensor(1.0, device="cuda", dtype=torch.float32)
scale_b = torch.tensor(1.0, device="cuda", dtype=torch.float32)
bias = torch.zeros((n,), device="cuda", dtype=torch.bfloat16)
bias = torch.zeros((n, ), device="cuda", dtype=torch.bfloat16)
out = ops.cutlass_scaled_sparse_mm(
a, b_compressed, e, scale_a, scale_b, torch.bfloat16
)
out = ops.cutlass_scaled_sparse_mm(a, b_compressed, e, scale_a, scale_b,
torch.bfloat16)
out_ref = ops.cutlass_scaled_mm(a, b, scale_a, scale_b, torch.bfloat16)
if not torch.allclose(out, out_ref):
@ -181,165 +124,97 @@ def bench_fp8(
# pytorch impl w. bf16
timers.append(
bench_fn(
label,
sub_label,
"pytorch_bf16_bf16_bf16_matmul-no-scales",
torch.mm,
a.to(dtype=torch.bfloat16, device="cuda"),
b.to(dtype=torch.bfloat16, device="cuda"),
)
)
bench_fn(label, sub_label, "pytorch_bf16_bf16_bf16_matmul-no-scales",
torch.mm, a.to(dtype=torch.bfloat16, device="cuda"),
b.to(dtype=torch.bfloat16, device="cuda")))
# pytorch impl: bf16 output, without fp8 fast accum
timers.append(
bench_fn(
label,
sub_label,
"pytorch_fp8_fp8_bf16_scaled_mm",
torch._scaled_mm,
a,
b,
scale_a=scale_a,
scale_b=scale_b,
out_dtype=torch.bfloat16,
)
)
bench_fn(label,
sub_label,
"pytorch_fp8_fp8_bf16_scaled_mm",
torch._scaled_mm,
a,
b,
scale_a=scale_a,
scale_b=scale_b,
out_dtype=torch.bfloat16))
# pytorch impl: bf16 output, with fp8 fast accum
timers.append(
bench_fn(
label,
sub_label,
"pytorch_fp8_fp8_bf16_scaled_mm_fast_accum",
torch._scaled_mm,
a,
b,
scale_a=scale_a,
scale_b=scale_b,
out_dtype=torch.bfloat16,
use_fast_accum=True,
)
)
bench_fn(label,
sub_label,
"pytorch_fp8_fp8_bf16_scaled_mm_fast_accum",
torch._scaled_mm,
a,
b,
scale_a=scale_a,
scale_b=scale_b,
out_dtype=torch.bfloat16,
use_fast_accum=True))
# pytorch impl: fp16 output, without fp8 fast accum
timers.append(
bench_fn(
label,
sub_label,
"pytorch_fp8_fp8_fp16_scaled_mm",
torch._scaled_mm,
a,
b,
scale_a=scale_a,
scale_b=scale_b,
out_dtype=torch.float16,
)
)
bench_fn(label,
sub_label,
"pytorch_fp8_fp8_fp16_scaled_mm",
torch._scaled_mm,
a,
b,
scale_a=scale_a,
scale_b=scale_b,
out_dtype=torch.float16))
# pytorch impl: fp16 output, with fp8 fast accum
timers.append(
bench_fn(
label,
sub_label,
"pytorch_fp8_fp8_fp16_scaled_mm_fast_accum",
torch._scaled_mm,
a,
b,
scale_a=scale_a,
scale_b=scale_b,
out_dtype=torch.float16,
use_fast_accum=True,
)
)
bench_fn(label,
sub_label,
"pytorch_fp8_fp8_fp16_scaled_mm_fast_accum",
torch._scaled_mm,
a,
b,
scale_a=scale_a,
scale_b=scale_b,
out_dtype=torch.float16,
use_fast_accum=True))
# cutlass impl: bf16 output
timers.append(
bench_fn(
label,
sub_label,
"cutlass_fp8_fp8_bf16_scaled_mm",
ops.cutlass_scaled_mm,
a,
b,
scale_a,
scale_b,
torch.bfloat16,
)
)
bench_fn(label, sub_label, "cutlass_fp8_fp8_bf16_scaled_mm",
ops.cutlass_scaled_mm, a, b, scale_a, scale_b,
torch.bfloat16))
# cutlass impl: bf16 output
timers.append(
bench_fn(
label,
sub_label,
"cutlass_fp8_fp8_bf16_scaled_sparse_mm",
ops.cutlass_scaled_sparse_mm,
a,
b_compressed,
e,
scale_a,
scale_b,
torch.bfloat16,
)
)
bench_fn(label, sub_label, "cutlass_fp8_fp8_bf16_scaled_sparse_mm",
ops.cutlass_scaled_sparse_mm, a, b_compressed, e, scale_a,
scale_b, torch.bfloat16))
# cutlass impl: fp16 output
timers.append(
bench_fn(
label,
sub_label,
"cutlass_fp8_fp8_fp16_scaled_sparse_mm",
ops.cutlass_scaled_sparse_mm,
a,
b_compressed,
e,
scale_a,
scale_b,
torch.float16,
)
)
bench_fn(label, sub_label, "cutlass_fp8_fp8_fp16_scaled_sparse_mm",
ops.cutlass_scaled_sparse_mm, a, b_compressed, e, scale_a,
scale_b, torch.float16))
# cutlass impl: bf16 output, with bias
timers.append(
bench_fn(
label,
sub_label,
"cutlass_fp8_fp8_bf16_scaled_sparse_mm_bias",
ops.cutlass_scaled_sparse_mm,
a,
b_compressed,
e,
scale_a,
scale_b,
torch.bfloat16,
bias,
)
)
bench_fn(label, sub_label,
"cutlass_fp8_fp8_bf16_scaled_sparse_mm_bias",
ops.cutlass_scaled_sparse_mm, a, b_compressed, e, scale_a,
scale_b, torch.bfloat16, bias))
# cutlass impl: fp16 output, with bias
timers.append(
bench_fn(
label,
sub_label,
"cutlass_fp8_fp8_fp16_scaled_sparse_mm_bias",
ops.cutlass_scaled_sparse_mm,
a,
b_compressed,
e,
scale_a,
scale_b,
torch.float16,
bias.to(dtype=torch.float16),
)
)
bench_fn(label, sub_label,
"cutlass_fp8_fp8_fp16_scaled_sparse_mm_bias",
ops.cutlass_scaled_sparse_mm, a, b_compressed, e, scale_a,
scale_b, torch.float16, bias.to(dtype=torch.float16)))
return timers
def bench(
dtype: torch.dtype, m: int, k: int, n: int, label: str, sub_label: str
) -> Iterable[TMeasurement]:
def bench(dtype: torch.dtype, m: int, k: int, n: int, label: str,
sub_label: str) -> Iterable[TMeasurement]:
if dtype == torch.int8:
return bench_int8(dtype, m, k, n, label, sub_label)
if dtype == torch.float8_e4m3fn:
@ -353,12 +228,12 @@ def print_timers(timers: Iterable[TMeasurement]):
compare.print()
def run(
dtype: torch.dtype, MKNs: Iterable[tuple[int, int, int]]
) -> Iterable[TMeasurement]:
def run(dtype: torch.dtype,
MKNs: Iterable[tuple[int, int, int]]) -> Iterable[TMeasurement]:
results = []
for m, k, n in MKNs:
timers = bench(dtype, m, k, n, f"scaled-{dtype}-gemm", f"MKN=({m}x{k}x{n})")
timers = bench(dtype, m, k, n, f"scaled-{dtype}-gemm",
f"MKN=({m}x{k}x{n})")
print_timers(timers)
results.extend(timers)
@ -366,12 +241,10 @@ def run(
# output makers
def make_output(
data: Iterable[TMeasurement],
MKNs: Iterable[tuple[int, int, int]],
base_description: str,
timestamp=None,
):
def make_output(data: Iterable[TMeasurement],
MKNs: Iterable[tuple[int, int, int]],
base_description: str,
timestamp=None):
print(f"== All Results {base_description} ====")
print_timers(data)
@ -385,7 +258,8 @@ def make_output(
def run_square_bench(args):
dim_sizes = list(range(args.dim_start, args.dim_end + 1, args.dim_increment))
dim_sizes = list(
range(args.dim_start, args.dim_end + 1, args.dim_increment))
MKNs = list(zip(dim_sizes, dim_sizes, dim_sizes))
data = run(args.dtype, MKNs)
@ -445,7 +319,7 @@ def run_model_bench(args):
pkl.dump(all_data, f)
if __name__ == "__main__":
if __name__ == '__main__':
def to_torch_dtype(dt):
if dt == "int8":
@ -470,15 +344,12 @@ Benchmark Cutlass GEMM.
Output:
- a .pkl file, that is a list of raw torch.benchmark.utils.Measurements for the pytorch and cutlass implementations for the various GEMMs.
""", # noqa: E501
formatter_class=argparse.RawTextHelpFormatter,
)
formatter_class=argparse.RawTextHelpFormatter)
parser.add_argument(
"--dtype",
type=to_torch_dtype,
required=True,
help="Available options are ['int8', 'fp8']",
)
parser.add_argument("--dtype",
type=to_torch_dtype,
required=True,
help="Available options are ['int8', 'fp8']")
subparsers = parser.add_subparsers(dest="cmd")
square_parser = subparsers.add_parser("square_bench")
@ -497,19 +368,19 @@ Benchmark Cutlass GEMM.
range_parser.set_defaults(func=run_range_bench)
model_parser = subparsers.add_parser("model_bench")
model_parser.add_argument(
"--models",
nargs="+",
type=str,
default=DEFAULT_MODELS,
choices=WEIGHT_SHAPES.keys(),
)
model_parser.add_argument(
"--tp-sizes", nargs="+", type=int, default=DEFAULT_TP_SIZES
)
model_parser.add_argument(
"--batch-sizes", nargs="+", type=int, default=DEFAULT_BATCH_SIZES
)
model_parser.add_argument("--models",
nargs="+",
type=str,
default=DEFAULT_MODELS,
choices=WEIGHT_SHAPES.keys())
model_parser.add_argument("--tp-sizes",
nargs="+",
type=int,
default=DEFAULT_TP_SIZES)
model_parser.add_argument("--batch-sizes",
nargs="+",
type=int,
default=DEFAULT_BATCH_SIZES)
model_parser.set_defaults(func=run_model_bench)
args = parser.parse_args()

View File

@ -1,5 +1,4 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
# Cutlass bench utils
from collections.abc import Iterable
@ -11,9 +10,8 @@ import vllm._custom_ops as ops
def to_fp8(tensor: torch.Tensor) -> torch.Tensor:
finfo = torch.finfo(torch.float8_e4m3fn)
return torch.round(tensor.clamp(min=finfo.min, max=finfo.max)).to(
dtype=torch.float8_e4m3fn
)
return torch.round(tensor.clamp(
min=finfo.min, max=finfo.max)).to(dtype=torch.float8_e4m3fn)
def to_int8(tensor: torch.Tensor) -> torch.Tensor:
@ -28,11 +26,10 @@ def to_fp16(tensor: torch.Tensor) -> torch.Tensor:
return tensor.to(dtype=torch.float16)
def make_rand_tensors(
dtype: torch.dtype, m: int, n: int, k: int
) -> tuple[torch.Tensor, torch.Tensor]:
a = torch.randn((m, k), device="cuda") * 5
b = torch.randn((n, k), device="cuda").t() * 5
def make_rand_tensors(dtype: torch.dtype, m: int, n: int,
k: int) -> tuple[torch.Tensor, torch.Tensor]:
a = torch.randn((m, k), device='cuda') * 5
b = torch.randn((n, k), device='cuda').t() * 5
if dtype == torch.int8:
return to_int8(a), to_int8(b)
@ -52,7 +49,9 @@ def prune_to_2_4(tensor):
# Create binary mask
mask = torch.zeros_like(reshaped)
mask.scatter_(dim=1, index=indices, src=torch.ones_like(indices, dtype=mask.dtype))
mask.scatter_(dim=1,
index=indices,
src=torch.ones_like(indices, dtype=mask.dtype))
# Apply mask and reshape back
pruned = reshaped * mask
@ -63,11 +62,10 @@ def prune_to_2_4(tensor):
return pruned.reshape(original_shape)
def make_rand_sparse_tensors(
dtype: torch.dtype, m: int, n: int, k: int
) -> tuple[torch.Tensor, torch.Tensor]:
a = torch.randn((m, k), device="cuda") * 5
b = torch.randn((n, k), device="cuda").t() * 5
def make_rand_sparse_tensors(dtype: torch.dtype, m: int, n: int,
k: int) -> tuple[torch.Tensor, torch.Tensor]:
a = torch.randn((m, k), device='cuda') * 5
b = torch.randn((n, k), device='cuda').t() * 5
b = prune_to_2_4(b.t()).t()
@ -88,9 +86,9 @@ def make_rand_sparse_tensors(
return b_compressed, e, a, b
def make_n_rand_sparse_tensors(
num_tensors: int, dtype: torch.dtype, m: int, n: int, k: int
) -> tuple[Iterable[torch.Tensor], Iterable[torch.Tensor]]:
def make_n_rand_sparse_tensors(num_tensors: int, dtype: torch.dtype,
m: int, n: int, k: int) -> \
tuple[Iterable[torch.Tensor], Iterable[torch.Tensor]]:
ABs = []
for _ in range(num_tensors):
b_comp, e, a, b = make_rand_sparse_tensors(dtype, m, n, k)

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