[Metrics] Log multi-modal cache stats and fix reset (#26285)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-10-10 01:45:55 -07:00
committed by GitHub
parent 6f0f570c43
commit ad430a67ca
25 changed files with 586 additions and 235 deletions
+3 -4
View File
@@ -20,7 +20,6 @@ from vllm.v1.core.kv_cache_utils import (
BlockHash,
FreeKVCacheBlockQueue,
KVCacheBlock,
PrefixCachingMetrics,
estimate_max_model_len,
generate_block_hash_extra_keys,
generate_scheduler_kv_cache_config,
@@ -42,7 +41,7 @@ from vllm.v1.kv_cache_interface import (
SlidingWindowSpec,
UniformTypeKVCacheSpecs,
)
from vllm.v1.metrics.stats import PrefixCacheStats
from vllm.v1.metrics.stats import CachingMetrics, PrefixCacheStats
from vllm.v1.request import Request
pytestmark = pytest.mark.cpu_test
@@ -536,7 +535,7 @@ def test_metrics():
"""
Test the prefix caching metrics.
"""
metrics = PrefixCachingMetrics(max_recent_requests=5)
metrics = CachingMetrics(max_recent_requests=5)
assert metrics.hit_rate == 0.0
metrics.observe(_stats(1, 20, 9))
@@ -568,7 +567,7 @@ def test_metrics_empty_stats():
"""
Test the prefix caching metrics with empty stats.
"""
metrics = PrefixCachingMetrics(max_recent_requests=5)
metrics = CachingMetrics(max_recent_requests=5)
metrics.observe(_stats(0, 0, 0))
metrics.observe(_stats(1, 20, 9))
metrics.observe(_stats(0, 0, 0))
+2 -1
View File
@@ -17,7 +17,7 @@ from vllm.sampling_params import RequestOutputKind
from vllm.v1.engine.async_llm import AsyncLLM
from vllm.v1.engine.core_client import DPAsyncMPClient
from vllm.v1.metrics.loggers import StatLoggerBase
from vllm.v1.metrics.stats import IterationStats, SchedulerStats
from vllm.v1.metrics.stats import IterationStats, MultiModalCacheStats, SchedulerStats
DP_SIZE = int(os.getenv("DP_SIZE", 2))
@@ -93,6 +93,7 @@ async def test_load(
self,
scheduler_stats: Optional[SchedulerStats],
iteration_stats: Optional[IterationStats],
mm_cache_stats: Optional[MultiModalCacheStats] = None,
engine_idx: int = 0,
):
if iteration_stats: