[core] Avoid metrics log noise when idle - include speculative decodi…

…ng and prefix cache noise Signed-off-by: Charles Darke <[email protected]>
vllm-project · Dec 1, 2024 · 479b628 · 479b628
1 parent f877a7d
commit 479b628
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/vllm/engine/metrics.py b/vllm/engine/metrics.py
@@ -473,13 +473,13 @@ def log(self, stats: Stats) -> None:
             )
             if (stats.cpu_prefix_cache_hit_rate >= 0
                     or stats.gpu_prefix_cache_hit_rate >= 0):
-                logger.info(
+                log_fn(
                     "Prefix cache hit rate: GPU: %.2f%%, CPU: %.2f%%",
                     stats.gpu_prefix_cache_hit_rate * 100,
                     stats.cpu_prefix_cache_hit_rate * 100,
                 )
             if self.spec_decode_metrics is not None:
-                logger.info(
+                log_fn(
                     self._format_spec_decode_metrics_str(
                         self.spec_decode_metrics))