[Misc] Heterogeneous GPU Optimizer Logging Clean Up (#514)

logging cleaned
vllm-project · Dec 10, 2024 · 98b5ca4 · 98b5ca4
1 parent cc526f7
commit 98b5ca4
Show file tree

Hide file tree

Showing 4 changed files with 14 additions and 8 deletions.
diff --git a/python/aibrix/aibrix/gpu_optimizer/app.py b/python/aibrix/aibrix/gpu_optimizer/app.py
@@ -98,7 +98,7 @@ def start_serving_thread(watch_ver, deployment, watch_event: bool) -> bool:
             watch_ver, deployment_name, namespace, lambda: new_deployment(deployment)
         )
         logger.info(
-            f'Deployment "{deployment_name}" added to the model monitor for "{model_name}"'
+            f'Deployment "{deployment_name}" found in watch version {watch_ver}, added to the model monitor for "{model_name}"'
         )
         return False
 
@@ -353,7 +353,7 @@ def main(signal, timeout):
         logging.ERROR
     )  # Suppress kubenetes logs
     logging.getLogger("pulp.apis.core").setLevel(logging.INFO)  # Suppress pulp logs
-    logger = logging.getLogger("aibrix.gpuoptimizer")
+    logger = logging.getLogger("aibrix.gpu_optimizer")
 
     timeout = 600
     try:
@@ -374,10 +374,14 @@ def main(signal, timeout):
         app,
         host="0.0.0.0",
         port=8080,
+        log_level="info",
         log_config={
             "version": 1,
             "disable_existing_loggers": False,
-            "loggers": {"uvicorn.access": {"filters": ["dash_update_filter"]}},
+            "loggers": {
+                "uvicorn.access": {"filters": ["dash_update_filter"]},
+                "uvicorn.error": {"level": "INFO", "propagate": False},
+            },
             "filters": {
                 "dash_update_filter": {
                     "()": ExcludePathsFilter,

diff --git a/python/aibrix/aibrix/gpu_optimizer/load_monitor/monitor.py b/python/aibrix/aibrix/gpu_optimizer/load_monitor/monitor.py
@@ -32,7 +32,7 @@
 
 Empty_Array: Iterable = []
 
-logger = logging.getLogger("aibrix.gpuoptimizer.loadmonitor")
+logger = logging.getLogger("aibrix.gpu_optimizer.load_monitor")
 
 debug_gpu_profile = GPUProfile(
     gpu="default", cost=1.0, tputs=[[100]], indexes=[[10], [10]]
@@ -60,7 +60,9 @@ def __init__(self, name: str, replicas: int = 1, min_replicas: int = 0):
 
     @property
     def cost(self):
-        return 0.0 if self.profile is None else self.profile.cost * self.replicas
+        return (
+            float("inf") if self.profile is None else self.profile.cost * self.replicas
+        )
 
     @property
     def replicas(self):
@@ -502,7 +504,7 @@ def _minimize(self):
         self._cost = cost
         self._lock.release()
         logger.info(
-            f"{self.model_name} scaled to minimum, cost ${self._cost}: {list(self.deployments.values())}"
+            f"{self.model_name} scaled to minimum, total cost ${self._cost}. Detailed Configuration:{list(self.deployments.values())}"
         )
 
     @property

diff --git a/python/aibrix/aibrix/gpu_optimizer/load_monitor/profile_reader.py b/python/aibrix/aibrix/gpu_optimizer/load_monitor/profile_reader.py
@@ -17,7 +17,7 @@
 
 from aibrix.gpu_optimizer.optimizer import GPUProfile
 
-logger = logging.getLogger("aibrix.gpuoptimizer.profile_reader")
+logger = logging.getLogger("aibrix.gpu_optimizer.profile_reader")
 
 
 class ProfileReader(Protocol):

diff --git a/python/aibrix/aibrix/gpu_optimizer/load_monitor/visualizer.py b/python/aibrix/aibrix/gpu_optimizer/load_monitor/visualizer.py
@@ -83,7 +83,7 @@ def default_datasource(_: str) -> Optional[ModelMonitor]:
     "lock": threading.Lock(),
 }
 
-logger = logging.getLogger("aibrix.gpuoptimizer.loadmonitor.visualizer")
+logger = logging.getLogger("aibrix.gpu_optimizer.load_monitor.visualizer")
 
 
 def get_debug_model_montior(