chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Add realtime token counter metrics (#15198)

Browse Source
This commit is contained in:
fzyzcjy
2025-12-17 21:20:24 +08:00
committed by GitHub
parent b12b40dea4
commit 3e690cce53
2 changed files with 37 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/managers/scheduler_metrics_mixin.py
+10
View File
@@ -216,6 +216,11 @@ class SchedulerMetricsMixin:
self.disagg_decode_transfer_queue.queue
)
self.metrics_collector.increment_realtime_tokens(
prefill_compute_tokens=adder.log_input_tokens,
prefill_cache_tokens=adder.log_hit_tokens,
)
# Others
self.calculate_utilization()
self.metrics_collector.log_stats(self.stats)
@@ -227,6 +232,7 @@ class SchedulerMetricsMixin:
):
batch = running_batch or self.running_batch
last_num_generated_tokens = self.num_generated_tokens
gap_latency = time.perf_counter() - self.last_decode_stats_tic
self.last_decode_stats_tic = time.perf_counter()
self.last_gen_throughput = self.num_generated_tokens / gap_latency
@@ -367,6 +373,10 @@ class SchedulerMetricsMixin:
self.disagg_decode_transfer_queue.queue
)
self.metrics_collector.increment_realtime_tokens(
decode_tokens=last_num_generated_tokens
)
# Others
self.calculate_utilization()
self.metrics_collector.log_stats(self.stats)