biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Deprecation][2/N] Replace --task with --runner and --convert (#21470)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Cyrus Leung
2025-07-28 10:42:40 +08:00
committed by GitHub
parent 8f605ee309
commit 86ae693f20
94 changed files with 1117 additions and 1083 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
tests/distributed/test_expert_parallel.py
View File

@@ -6,7 +6,7 @@ from typing import Literal, NamedTuple, Optional
import pytest
from vllm.config import TaskOption
from vllm.config import RunnerOption
from vllm.logger import init_logger
from ..utils import compare_two_settings, create_new_process_for_each_test
@@ -31,14 +31,14 @@ class EPTestOptions(NamedTuple):
class EPTestSettings:
parallel_setups: list[ParallelSetup]
distributed_backends: list[str]
task: TaskOption
runner: RunnerOption
test_options: EPTestOptions
@staticmethod
def detailed(
*,
tp_base: int = 2,
task: TaskOption = "auto",
runner: RunnerOption = "auto",
trust_remote_code: bool = False,
tokenizer_mode: Optional[str] = None,
load_format: Optional[str] = None,
@@ -63,7 +63,7 @@ class EPTestSettings:
chunked_prefill=False),
],
distributed_backends=["mp", "ray"],
task=task,
runner=runner,
test_options=EPTestOptions(trust_remote_code=trust_remote_code,
tokenizer_mode=tokenizer_mode,
load_format=load_format,
@@ -74,7 +74,7 @@ class EPTestSettings:
def fast(
*,
tp_base: int = 2,
task: TaskOption = "auto",
runner: RunnerOption = "auto",
trust_remote_code: bool = False,
tokenizer_mode: Optional[str] = None,
load_format: Optional[str] = None,
@@ -87,7 +87,7 @@ class EPTestSettings:
chunked_prefill=False),
],
distributed_backends=["mp"],
task=task,
runner=runner,
test_options=EPTestOptions(trust_remote_code=trust_remote_code,
tokenizer_mode=tokenizer_mode,
load_format=load_format,
@@ -100,7 +100,7 @@ class EPTestSettings:
for parallel_setup in self.parallel_setups:
for distributed_backend in self.distributed_backends:
yield (model_name, parallel_setup, distributed_backend,
self.task, opts)
self.runner, opts)
# NOTE: You can adjust tp_base locally to fit the model in GPU
@@ -118,7 +118,7 @@ def _compare_tp(
model_name: str,
parallel_setup: ParallelSetup,
distributed_backend: str,
task: TaskOption,
runner: RunnerOption,
test_options: EPTestOptions,
num_gpus_available: int,
*,
@@ -154,8 +154,8 @@ def _compare_tp(
common_args.append("--enable-chunked-prefill")
if eager_mode:
common_args.append("--enforce-eager")
if task != "auto":
common_args.extend(["--task", task])
if runner != "auto":
common_args.extend(["--runner", runner])
if trust_remote_code:
common_args.append("--trust-remote-code")
if tokenizer_mode:
@@ -203,7 +203,7 @@ def _compare_tp(
@pytest.mark.parametrize(
("model_name", "parallel_setup", "distributed_backend", "task",
("model_name", "parallel_setup", "distributed_backend", "runner",
"test_options"),
[
params for model_name, settings in TEST_MODELS.items()
@@ -215,14 +215,14 @@ def test_ep(
model_name: str,
parallel_setup: ParallelSetup,
distributed_backend: str,
task: TaskOption,
runner: RunnerOption,
test_options: EPTestOptions,
num_gpus_available,
):
_compare_tp(model_name,
parallel_setup,
distributed_backend,
task,
runner,
test_options,
num_gpus_available,
method="generate")

44
tests/distributed/test_pipeline_parallel.py
View File

@@ -14,7 +14,7 @@ from typing import Literal, NamedTuple, Optional
import pytest
from vllm.config import _FLOAT16_NOT_SUPPORTED_MODELS, TaskOption
from vllm.config import _FLOAT16_NOT_SUPPORTED_MODELS, RunnerOption
from vllm.logger import init_logger
from vllm.transformers_utils.config import get_config
@@ -60,7 +60,7 @@ class PPTestSettings:
distributed_backends: list[str]
# vllm major version: "0" for V0, "1" for V1
vllm_major_versions: list[str]
task: TaskOption
runner: RunnerOption
test_options: PPTestOptions
def __post_init__(self):
@@ -76,7 +76,7 @@ class PPTestSettings:
tp_base: int = 1,
pp_base: int = 2,
multi_node_only: bool = False,
task: TaskOption = "auto",
runner: RunnerOption = "auto",
load_format: Optional[str] = None,
):
return PPTestSettings(
@@ -104,7 +104,7 @@ class PPTestSettings:
],
distributed_backends=["mp", "mp", "ray", "ray"],
vllm_major_versions=["0", "1", "0", "1"],
task=task,
runner=runner,
test_options=PPTestOptions(multi_node_only=multi_node_only,
load_format=load_format),
)
@@ -114,7 +114,7 @@ class PPTestSettings:
*,
tp_base: int = 1,
pp_base: int = 2,
task: TaskOption = "auto",
runner: RunnerOption = "auto",
multi_node_only: bool = False,
load_format: Optional[str] = None,
):
@@ -127,7 +127,7 @@ class PPTestSettings:
],
distributed_backends=["mp"],
vllm_major_versions=["0"],
task=task,
runner=runner,
test_options=PPTestOptions(multi_node_only=multi_node_only,
load_format=load_format),
)
@@ -139,7 +139,7 @@ class PPTestSettings:
for backend, vllm_major_version in zip(self.distributed_backends,
self.vllm_major_versions):
yield (model_id, parallel_setup, backend, vllm_major_version,
self.task, opts)
self.runner, opts)
# NOTE: You can adjust tp_base and/or pp_base locally to fit the model in GPU
@@ -211,10 +211,10 @@ TEXT_GENERATION_MODELS = {
EMBEDDING_MODELS = { # type: ignore[var-annotated]
# [Text-only]
"intfloat/e5-mistral-7b-instruct": PPTestSettings.fast(task="embed"),
"BAAI/bge-multilingual-gemma2": PPTestSettings.fast(task="embed"),
"intfloat/e5-mistral-7b-instruct": PPTestSettings.fast(runner="pooling"),
"BAAI/bge-multilingual-gemma2": PPTestSettings.fast(runner="pooling"),
"Qwen/Qwen2.5-Math-RM-72B": PPTestSettings.fast(
load_format="dummy", task="embed"
load_format="dummy", runner="pooling"
),
}
@@ -269,7 +269,7 @@ def _compare_tp(
parallel_setup: ParallelSetup,
distributed_backend: str,
vllm_major_version: str,
task: TaskOption,
runner: RunnerOption,
test_options: PPTestOptions,
num_gpus_available: int,
*,
@@ -335,8 +335,8 @@ def _compare_tp(
common_args.append("--enable-chunked-prefill")
if eager_mode:
common_args.append("--enforce-eager")
if task != "auto":
common_args.extend(["--task", task])
if runner != "auto":
common_args.extend(["--runner", runner])
if trust_remote_code:
common_args.append("--trust-remote-code")
if tokenizer_mode:
@@ -415,7 +415,7 @@ def _compare_tp(
@pytest.mark.parametrize(
("model_id", "parallel_setup", "distributed_backend", "vllm_major_version",
"task", "test_options"),
"runner", "test_options"),
[
params for model_id, settings in TEXT_GENERATION_MODELS.items()
for params in settings.iter_params(model_id) if model_id in TEST_MODELS
@@ -427,7 +427,7 @@ def test_tp_language_generation(
parallel_setup: ParallelSetup,
distributed_backend: str,
vllm_major_version: str,
task: TaskOption,
runner: RunnerOption,
test_options: PPTestOptions,
num_gpus_available,
):
@@ -435,7 +435,7 @@ def test_tp_language_generation(
parallel_setup,
distributed_backend,
vllm_major_version,
task,
runner,
test_options,
num_gpus_available,
method="generate",
@@ -444,7 +444,7 @@ def test_tp_language_generation(
@pytest.mark.parametrize(
("model_id", "parallel_setup", "distributed_backend", "vllm_major_version",
"task", "test_options"),
"runner", "test_options"),
[
params for model_id, settings in EMBEDDING_MODELS.items()
for params in settings.iter_params(model_id) if model_id in TEST_MODELS
@@ -456,7 +456,7 @@ def test_tp_language_embedding(
parallel_setup: ParallelSetup,
distributed_backend: str,
vllm_major_version: str,
task: TaskOption,
runner: RunnerOption,
test_options: PPTestOptions,
num_gpus_available,
):
@@ -464,7 +464,7 @@ def test_tp_language_embedding(
parallel_setup,
distributed_backend,
vllm_major_version,
task,
runner,
test_options,
num_gpus_available,
method="encode",
@@ -473,7 +473,7 @@ def test_tp_language_embedding(
@pytest.mark.parametrize(
("model_id", "parallel_setup", "distributed_backend", "vllm_major_version",
"task", "test_options"),
"runner", "test_options"),
[
params for model_id, settings in MULTIMODAL_MODELS.items()
for params in settings.iter_params(model_id) if model_id in TEST_MODELS
@@ -485,7 +485,7 @@ def test_tp_multimodal_generation(
parallel_setup: ParallelSetup,
distributed_backend: str,
vllm_major_version: str,
task: TaskOption,
runner: RunnerOption,
test_options: PPTestOptions,
num_gpus_available,
):
@@ -493,7 +493,7 @@ def test_tp_multimodal_generation(
parallel_setup,
distributed_backend,
vllm_major_version,
task,
runner,
test_options,
num_gpus_available,
method="generate",

30
tests/distributed/test_sequence_parallel.py
View File

@@ -14,7 +14,7 @@ from typing import Literal, NamedTuple, Optional
import pytest
from vllm.config import TaskOption
from vllm.config import RunnerOption
from vllm.logger import init_logger
from ..models.registry import HF_EXAMPLE_MODELS
@@ -48,7 +48,7 @@ class SPTestSettings:
distributed_backends: list[str]
# vllm major version: "0" for V0, "1" for V1
vllm_major_versions: list[str]
task: TaskOption
runner: RunnerOption
test_options: SPTestOptions
def __post_init__(self):
@@ -64,7 +64,7 @@ class SPTestSettings:
tp_base: int = 2,
pp_base: int = 1,
multi_node_only: bool = False,
task: TaskOption = "auto",
runner: RunnerOption = "auto",
load_format: Optional[str] = None,
):
parallel_setups = []
@@ -81,7 +81,7 @@ class SPTestSettings:
parallel_setups=parallel_setups,
distributed_backends=["mp", "ray"],
vllm_major_versions=["1", "1"],
task=task,
runner=runner,
test_options=SPTestOptions(multi_node_only=multi_node_only,
load_format=load_format),
)
@@ -91,7 +91,7 @@ class SPTestSettings:
*,
tp_base: int = 2,
pp_base: int = 1,
task: TaskOption = "auto",
runner: RunnerOption = "auto",
multi_node_only: bool = False,
load_format: Optional[str] = None,
):
@@ -109,7 +109,7 @@ class SPTestSettings:
parallel_setups=parallel_setups,
distributed_backends=["mp", "ray"],
vllm_major_versions=["1", "1"],
task=task,
runner=runner,
test_options=SPTestOptions(multi_node_only=multi_node_only,
load_format=load_format),
)
@@ -119,7 +119,7 @@ class SPTestSettings:
*,
tp_base: int = 2,
pp_base: int = 1,
task: TaskOption = "auto",
runner: RunnerOption = "auto",
multi_node_only: bool = False,
load_format: Optional[str] = None,
):
@@ -135,7 +135,7 @@ class SPTestSettings:
parallel_setups=parallel_setups,
distributed_backends=["mp", "ray"],
vllm_major_versions=["1", "1"],
task=task,
runner=runner,
test_options=SPTestOptions(multi_node_only=multi_node_only,
load_format=load_format),
)
@@ -147,7 +147,7 @@ class SPTestSettings:
for backend, vllm_major_version in zip(self.distributed_backends,
self.vllm_major_versions):
yield (model_id, parallel_setup, backend, vllm_major_version,
self.task, opts)
self.runner, opts)
def _compare_sp(
@@ -155,7 +155,7 @@ def _compare_sp(
parallel_setup: ParallelSetup,
distributed_backend: str,
vllm_major_version: str,
task: TaskOption,
runner: RunnerOption,
test_options: SPTestOptions,
num_gpus_available: int,
*,
@@ -217,8 +217,8 @@ def _compare_sp(
common_args.append("--enable-chunked-prefill")
if eager_mode:
common_args.append("--enforce-eager")
if task != "auto":
common_args.extend(["--task", task])
if runner != "auto":
common_args.extend(["--runner", runner])
if trust_remote_code:
common_args.append("--trust-remote-code")
if tokenizer_mode:
@@ -298,7 +298,7 @@ SP_TEST_MODELS = [
@pytest.mark.parametrize(
("model_id", "parallel_setup", "distributed_backend", "vllm_major_version",
"task", "test_options"),
"runner", "test_options"),
[
params for model_id, settings in SP_TEXT_GENERATION_MODELS.items()
for params in settings.iter_params(model_id)
@@ -311,7 +311,7 @@ def test_tp_sp_generation(
parallel_setup: ParallelSetup,
distributed_backend: str,
vllm_major_version: str,
task: TaskOption,
runner: RunnerOption,
test_options: SPTestOptions,
num_gpus_available,
):
@@ -319,7 +319,7 @@ def test_tp_sp_generation(
parallel_setup,
distributed_backend,
vllm_major_version,
task,
runner,
test_options,
num_gpus_available,
method="generate",