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

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
tests/v1/metrics/test_engine_logger_apis.py
View File

@@ -46,23 +46,22 @@ def log_stats_enabled_engine_args():
@pytest.mark.asyncio
async def test_async_llm_replace_default_loggers(
log_stats_enabled_engine_args):
async def test_async_llm_replace_default_loggers(log_stats_enabled_engine_args):
"""
RayPrometheusStatLogger should replace the default PrometheusStatLogger
"""
engine = AsyncLLM.from_engine_args(log_stats_enabled_engine_args,
stat_loggers=[RayPrometheusStatLogger])
assert isinstance(engine.logger_manager.prometheus_logger,
RayPrometheusStatLogger)
engine = AsyncLLM.from_engine_args(
log_stats_enabled_engine_args, stat_loggers=[RayPrometheusStatLogger]
)
assert isinstance(engine.logger_manager.prometheus_logger, RayPrometheusStatLogger)
engine.shutdown()
@pytest.mark.asyncio
async def test_async_llm_add_to_default_loggers(log_stats_enabled_engine_args):
"""
It's still possible to use custom stat loggers exclusively by passing
It's still possible to use custom stat loggers exclusively by passing
disable_log_stats=True in addition to a list of custom stat loggers.
"""
# Create engine_args with disable_log_stats=True for this test
@@ -70,12 +69,14 @@ async def test_async_llm_add_to_default_loggers(log_stats_enabled_engine_args):
disabled_log_engine_args.disable_log_stats = True
# Disable default loggers; pass custom stat logger to the constructor
engine = AsyncLLM.from_engine_args(disabled_log_engine_args,
stat_loggers=[DummyStatLogger])
engine = AsyncLLM.from_engine_args(
disabled_log_engine_args, stat_loggers=[DummyStatLogger]
)
assert len(engine.logger_manager.per_engine_logger_dict[0]) == 1
assert isinstance(engine.logger_manager.per_engine_logger_dict[0][0],
DummyStatLogger)
assert isinstance(
engine.logger_manager.per_engine_logger_dict[0][0], DummyStatLogger
)
# log_stats is still True, since custom stat loggers are used
assert engine.log_stats

44
tests/v1/metrics/test_metrics_reader.py
View File

@@ -4,8 +4,13 @@
import prometheus_client
import pytest
from vllm.v1.metrics.reader import (Counter, Gauge, Histogram, Vector,
get_metrics_snapshot)
from vllm.v1.metrics.reader import (
Counter,
Gauge,
Histogram,
Vector,
get_metrics_snapshot,
)
pytestmark = pytest.mark.cpu_test
@@ -20,10 +25,12 @@ def test_registry(monkeypatch):
@pytest.mark.parametrize("num_engines", [1, 4])
def test_gauge_metric(test_registry, num_engines):
g = prometheus_client.Gauge("vllm:test_gauge",
"Test gauge metric",
labelnames=["model", "engine_index"],
registry=test_registry)
g = prometheus_client.Gauge(
"vllm:test_gauge",
"Test gauge metric",
labelnames=["model", "engine_index"],
registry=test_registry,
)
for i in range(num_engines):
g.labels(model="foo", engine_index=str(i)).set(98.5)
@@ -41,10 +48,12 @@ def test_gauge_metric(test_registry, num_engines):
@pytest.mark.parametrize("num_engines", [1, 4])
def test_counter_metric(test_registry, num_engines):
c = prometheus_client.Counter("vllm:test_counter",
"Test counter metric",
labelnames=["model", "engine_index"],
registry=test_registry)
c = prometheus_client.Counter(
"vllm:test_counter",
"Test counter metric",
labelnames=["model", "engine_index"],
registry=test_registry,
)
for i in range(num_engines):
c.labels(model="bar", engine_index=str(i)).inc(19)
@@ -62,11 +71,13 @@ def test_counter_metric(test_registry, num_engines):
@pytest.mark.parametrize("num_engines", [1, 4])
def test_histogram_metric(test_registry, num_engines):
h = prometheus_client.Histogram("vllm:test_histogram",
"Test histogram metric",
labelnames=["model", "engine_index"],
buckets=[10, 20, 30, 40, 50],
registry=test_registry)
h = prometheus_client.Histogram(
"vllm:test_histogram",
"Test histogram metric",
labelnames=["model", "engine_index"],
buckets=[10, 20, 30, 40, 50],
registry=test_registry,
)
for i in range(num_engines):
hist = h.labels(model="blaa", engine_index=str(i))
hist.observe(42)
@@ -97,7 +108,8 @@ def test_vector_metric(test_registry, num_engines):
"vllm:spec_decode_num_accepted_tokens_per_pos",
"Vector-like counter metric",
labelnames=["position", "model", "engine_index"],
registry=test_registry)
registry=test_registry,
)
for i in range(num_engines):
c.labels(position="0", model="llama", engine_index=str(i)).inc(10)
c.labels(position="1", model="llama", engine_index=str(i)).inc(5)

48
tests/v1/metrics/test_ray_metrics.py
View File

@@ -8,8 +8,7 @@ import ray
from vllm.config import ModelDType
from vllm.sampling_params import SamplingParams
from vllm.v1.engine.async_llm import AsyncEngineArgs, AsyncLLM
from vllm.v1.metrics.ray_wrappers import (RayPrometheusMetric,
RayPrometheusStatLogger)
from vllm.v1.metrics.ray_wrappers import RayPrometheusMetric, RayPrometheusStatLogger
@pytest.fixture(scope="function", autouse=True)
@@ -17,7 +16,7 @@ def use_v1_only(monkeypatch):
"""
The change relies on V1 APIs, so set VLLM_USE_V1=1.
"""
monkeypatch.setenv('VLLM_USE_V1', '1')
monkeypatch.setenv("VLLM_USE_V1", "1")
MODELS = [
@@ -34,24 +33,23 @@ def test_engine_log_metrics_ray(
dtype: ModelDType,
max_tokens: int,
) -> None:
""" Simple smoke test, verifying this can be used without exceptions.
"""Simple smoke test, verifying this can be used without exceptions.
Need to start a Ray cluster in order to verify outputs."""
@ray.remote(num_gpus=1)
class EngineTestActor:
async def run(self):
# Set environment variable inside the Ray actor since environment
# variables from pytest fixtures don't propagate to Ray actors
os.environ['VLLM_USE_V1'] = '1'
os.environ["VLLM_USE_V1"] = "1"
engine_args = AsyncEngineArgs(model=model,
dtype=dtype,
disable_log_stats=False,
enforce_eager=True)
engine_args = AsyncEngineArgs(
model=model, dtype=dtype, disable_log_stats=False, enforce_eager=True
)
engine = AsyncLLM.from_engine_args(
engine_args, stat_loggers=[RayPrometheusStatLogger])
engine_args, stat_loggers=[RayPrometheusStatLogger]
)
for i, prompt in enumerate(example_prompts):
results = engine.generate(
@@ -73,32 +71,40 @@ def test_sanitized_opentelemetry_name():
# Only a-z, A-Z, 0-9, _, test valid characters are preserved
valid_name = "valid_metric_123_abcDEF"
assert RayPrometheusMetric._get_sanitized_opentelemetry_name(
valid_name) == valid_name
assert (
RayPrometheusMetric._get_sanitized_opentelemetry_name(valid_name) == valid_name
)
# Test dash, dot, are replaced
name_with_dash_dot = "metric-name.test"
expected = "metric_name_test"
assert RayPrometheusMetric._get_sanitized_opentelemetry_name(
name_with_dash_dot) == expected
assert (
RayPrometheusMetric._get_sanitized_opentelemetry_name(name_with_dash_dot)
== expected
)
# Test colon is replaced with underscore
name_with_colon = "metric:name"
expected = "metric_name"
assert RayPrometheusMetric._get_sanitized_opentelemetry_name(
name_with_colon) == expected
assert (
RayPrometheusMetric._get_sanitized_opentelemetry_name(name_with_colon)
== expected
)
# Test multiple invalid characters are replaced
name_with_invalid = "metric:name@with#special%chars"
expected = "metric_name_with_special_chars"
assert RayPrometheusMetric._get_sanitized_opentelemetry_name(
name_with_invalid) == expected
assert (
RayPrometheusMetric._get_sanitized_opentelemetry_name(name_with_invalid)
== expected
)
# Test mixed valid and invalid characters
complex_name = "vllm:engine_stats/time.latency_ms-99p"
expected = "vllm_engine_stats_time_latency_ms_99p"
assert RayPrometheusMetric._get_sanitized_opentelemetry_name(
complex_name) == expected
assert (
RayPrometheusMetric._get_sanitized_opentelemetry_name(complex_name) == expected
)
# Test empty string
assert RayPrometheusMetric._get_sanitized_opentelemetry_name("") == ""

26
tests/v1/metrics/test_stats.py
View File

@@ -6,16 +6,18 @@ from vllm.v1.metrics.stats import IterationStats
def test_iteration_stats_repr():
iteration_stats = IterationStats()
iteration_stats.iteration_timestamp = 0
expected_repr = ("IterationStats("
"iteration_timestamp=0, "
"num_generation_tokens=0, "
"num_prompt_tokens=0, "
"num_preempted_reqs=0, "
"finished_requests=[], "
"max_num_generation_tokens_iter=[], "
"n_params_iter=[], "
"time_to_first_tokens_iter=[], "
"inter_token_latencies_iter=[], "
"waiting_lora_adapters={}, "
"running_lora_adapters={})")
expected_repr = (
"IterationStats("
"iteration_timestamp=0, "
"num_generation_tokens=0, "
"num_prompt_tokens=0, "
"num_preempted_reqs=0, "
"finished_requests=[], "
"max_num_generation_tokens_iter=[], "
"n_params_iter=[], "
"time_to_first_tokens_iter=[], "
"inter_token_latencies_iter=[], "
"waiting_lora_adapters={}, "
"running_lora_adapters={})"
)
assert repr(iteration_stats) == expected_repr