Fix CI test for test_worker.py

Author: OliverBryant

xinference/core/tests/test_worker.py

@@ -29,6 +29,13 @@ def __init__(
         cuda_devices: List[int],
     ):
         super().__init__(supervisor_address, main_pool, cuda_devices)
+        self._gpu_memory_info = {}
+        for gpu_idx in cuda_devices:
+            self._gpu_memory_info[gpu_idx] = {
+                "total": 24000,
+                "used": 0,
+                "available": 24000
+            }
 
     async def __post_create__(self):
         pass

xinference/core/worker.py

@@ -538,8 +538,13 @@ def allocate_devices(self, model_uid: str, n_gpu: int) -> List[int]:
             remaining_needed = n_gpu - len(selected_devices)
 
             # Get GPUs sorted by available memory (most available first)
+            # Exclude GPUs that are already allocated by user_specified models
             candidate_gpus = [
-                dev for dev in self._total_gpu_devices if dev not in selected_devices
+                dev
+                for dev in self._total_gpu_devices
+                if dev not in selected_devices
+                and dev not in self._gpu_to_model_uid
+                and dev not in user_specified_allocated_devices
             ]
 
             gpu_memory_list = []