refactor(amd): safe hsa detection

Signed-off-by: thxCode <[email protected]>

Author: thxCode

gpustack_runtime/detector/amd.py

@@ -108,7 +108,7 @@ def detect(self) -> Devices | None:
                     dev_uuid = f"GPU-{(asic_serial[2:]).lower()}"
                 else:
                     dev_uuid = f"GPU-{pyrocmsmi.rsmi_dev_unique_id_get(dev_idx)[2:]}"
-                dev_hsa_agent = hsa_agents.get(dev_uuid)
+                dev_hsa_agent = hsa_agents.get(dev_uuid, pyhsa.Agent())
 
                 dev_gpu_driver_info = pyamdsmi.amdsmi_get_gpu_driver_info(dev)
                 dev_driver_ver = dev_gpu_driver_info.get("driver_version")

gpustack_runtime/detector/hygon.py

@@ -108,7 +108,7 @@ def detect(self) -> Devices | None:
                 dev_index = dev_idx
 
                 dev_uuid = f"GPU-{pyrocmsmi.rsmi_dev_unique_id_get(dev_idx)[2:]}"
-                dev_hsa_agent = hsa_agents.get(dev_uuid)
+                dev_hsa_agent = hsa_agents.get(dev_uuid, pyhsa.Agent())
 
                 dev_name = dev_hsa_agent.name
                 if not dev_name:

gpustack_runtime/detector/pyhsa/__init__.py

@@ -446,13 +446,13 @@ def has_agent_get_asic_family_id(agent):
 
 @dataclass
 class Agent:
-    device_type: int
-    device_id: str
-    uuid: str
-    name: str
-    compute_capability: str
-    compute_units: int
-    asic_family_id: int | None
+    device_type: int | None = None
+    device_id: str | None = None
+    uuid: str | None = None
+    name: str | None = None
+    compute_capability: str | None = None
+    compute_units: int | None = None
+    asic_family_id: int | None = None
 
 
 def get_agents() -> list[Agent]: