A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 5, 04:45 UTC
[Core][Metrics] Remove unused `SchedulerStats.encoder_cache_usage` (#39693)
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Mark McLoughlin committed
Apr 14, 2026 at 17:53 UTC
906a8c15d0beded2cca8b357877266c631340f74
2 files changed
-10
vllm/v1/core/sched/scheduler.py
-9
@@ -1961,7 +1961,6 @@ class Scheduler(SchedulerInterface):
1961
num_waiting_reqs=len(self.waiting),
1962
num_skipped_waiting_reqs=len(self.skipped_waiting),
1963
kv_cache_usage=self.kv_cache_manager.usage,
1964
- encoder_cache_usage=self._get_encoder_cache_usage(),
1964
prefix_cache_stats=prefix_cache_stats,
1965
connector_prefix_cache_stats=connector_prefix_cache_stats,
1966
kv_cache_eviction_events=eviction_events,
@@ -1971,14 +1970,6 @@ class Scheduler(SchedulerInterface):
1970
perf_stats=perf_stats,
1971
)
1972
1974
- def _get_encoder_cache_usage(self) -> float:
1975
- """Get encoder cache usage as a fraction (0.0 to 1.0)."""
1976
- ecm = self.encoder_cache_manager
1977
- if ecm.cache_size == 0:
1978
- return 0.0
1979
- used_slots = ecm.cache_size - ecm.num_free_slots
1980
- return used_slots / ecm.cache_size
1981
-
1973
def make_spec_decoding_stats(
1974
self,
1975
spec_decoding_stats: SpecDecodingStats | None,
vllm/v1/metrics/stats.py
-1
@@ -181,7 +181,6 @@ class SchedulerStats:
181
current_wave: int = 0
182
183
kv_cache_usage: float = 0.0
184
- encoder_cache_usage: float = 0.0
184
185
prefix_cache_stats: PrefixCacheStats = field(default_factory=PrefixCacheStats)
186
connector_prefix_cache_stats: PrefixCacheStats | None = None