Set default worker process limit based on number of CPUs

Signed-off-by: Samuel Monson <[email protected]>

Author: sjmonson

src/guidellm/config.py

@@ -1,4 +1,5 @@
 import json
+import os
 from collections.abc import Sequence
 from enum import Enum
 from typing import Literal, Optional
@@ -18,6 +19,13 @@
 ]
 
 
+def get_avail_cpu() -> int:
+    """
+    Get the maximum number of worker processes based on the CPU count
+    """
+    return max(1, os.cpu_count() or 1) - 1
+
+
 class Environment(str, Enum):
     """
     Enum for the supported environments
@@ -131,7 +139,9 @@ class Settings(BaseSettings):
 
     # Scheduler settings
     max_concurrency: int = 512
-    max_worker_processes: int = 10
+    max_worker_processes: int = Field(
+        default_factory=lambda: max((os.cpu_count() or 1) - 1, 10)
+    )
     max_add_requests_per_loop: int = 20
     scheduler_start_delay: float = 5
 

src/guidellm/scheduler/strategy.py

@@ -1,5 +1,4 @@
 import math
-import os
 import random
 import time
 from collections.abc import Generator
@@ -72,9 +71,7 @@ def processes_limit(self) -> int:
 
         :return: The number of processes for the scheduling strategy.
         """
-        cpu_cores = os.cpu_count() or 1
-
-        return min(max(1, cpu_cores - 1), settings.max_worker_processes)
+        return settings.max_worker_processes
 
     @property
     def queued_requests_limit(self) -> Optional[int]:
@@ -231,9 +228,8 @@ def processes_limit(self) -> int:
         :return: {self.streams} for the concurrent scheduling strategy to limit
             the worker processes to the number of streams.
         """
-        cpu_cores = os.cpu_count() or 1
 
-        return min(max(1, cpu_cores - 1), self.streams)
+        return min(self.streams, settings.max_worker_processes)
 
     @property
     def queued_requests_limit(self) -> int: