[Auto Sync] Update scheduler_metrics_mixin.py, collector.py (20251104) (#12647)

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: SangBin Cho <rkooo567@gmail.com>

python/sglang/srt/managers/scheduler_metrics_mixin.py

@@ -164,6 +164,8 @@ class SchedulerMetricsMixin:
             self.stats.token_usage = token_usage
             if self.is_hybrid:
                 self.stats.swa_token_usage = swa_token_usage
+            if self.is_hybrid_gdn:
+                self.stats.mamba_usage = mamba_usage
             self.stats.num_queue_reqs = len(self.waiting_queue)
             self.stats.num_grammar_queue_reqs = len(self.grammar_queue)
             self.stats.cache_hit_rate = cache_hit_rate
@@ -306,6 +308,8 @@ class SchedulerMetricsMixin:
             self.stats.token_usage = token_usage
             if self.is_hybrid:
                 self.stats.swa_token_usage = swa_token_usage
+            if self.is_hybrid_gdn:
+                self.stats.mamba_usage = mamba_usage
             self.stats.gen_throughput = self.last_gen_throughput
             self.stats.num_queue_reqs = len(self.waiting_queue)
             self.stats.num_grammar_queue_reqs = len(self.grammar_queue)

python/sglang/srt/metrics/collector.py

@@ -150,6 +150,7 @@ class SchedulerStats:
     token_usage: float = 0.0
     pending_prealloc_token_usage: float = 0.0
     swa_token_usage: float = 0.0
+    mamba_usage: float = 0.0
     gen_throughput: float = 0.0
     num_queue_reqs: int = 0
     num_grammar_queue_reqs: int = 0
@@ -225,6 +226,12 @@ class SchedulerMetricsCollector:
             labelnames=labels.keys(),
             multiprocess_mode="mostrecent",
         )
+        self.mamba_usage = Gauge(
+            name="sglang:mamba_usage",
+            documentation="The token usage for Mamba layers.",
+            labelnames=labels.keys(),
+            multiprocess_mode="mostrecent",
+        )
         self.gen_throughput = Gauge(
             name="sglang:gen_throughput",
             documentation="The generation throughput (token/s).",
@@ -581,6 +588,7 @@ class SchedulerMetricsCollector:
             self.pending_prealloc_token_usage, stats.pending_prealloc_token_usage
         )
         self._log_gauge(self.swa_token_usage, stats.swa_token_usage)
+        self._log_gauge(self.mamba_usage, stats.mamba_usage)
         self._log_gauge(self.gen_throughput, stats.gen_throughput)
         self._log_gauge(self.num_queue_reqs, stats.num_queue_reqs)
         self._log_gauge(self.num_grammar_queue_reqs, stats.num_grammar_queue_reqs)