Fix fixture issues / other suggested changes

# Conflicts:
#	tests/fixtures/inference.py
#	utilities/constants.py

Author: Sandeep20013

.gitignore

@@ -178,3 +178,4 @@ QWEN.md
 
 # Must-Gather Artifacts
 must-gather-collected/
+oc.tar

tests/fixtures/guardrails.py

@@ -48,11 +48,11 @@ def guardrails_orchestrator(
             orchestrator_config = request.getfixturevalue(argname="orchestrator_config")
             gorch_kwargs["orchestrator_config"] = orchestrator_config.name
 
-        if request.param.get("orchestrator_config_gpu"):
+        elif request.param.get("orchestrator_config_gpu"):
             orchestrator_config = request.getfixturevalue(argname="orchestrator_config_gpu")
             gorch_kwargs["orchestrator_config"] = orchestrator_config.name
 
-        if request.param.get("orchestrator_config_builtin_gpu"):
+        elif request.param.get("orchestrator_config_builtin_gpu"):
             orchestrator_config = request.getfixturevalue(argname="orchestrator_config_builtin_gpu")
             gorch_kwargs["orchestrator_config"] = orchestrator_config.name
 

tests/model_explainability/guardrails/test_guardrails_gpu.py

@@ -66,6 +66,7 @@
     indirect=True,
 )
 @pytest.mark.smoke
+@pytest.mark.gpu
 @pytest.mark.rawdeployment
 @pytest.mark.usefixtures("patched_dsc_kserve_headed", "guardrails_gateway_config")
 class TestGuardrailsOrchestratorWithBuiltInDetectors:
@@ -206,8 +207,8 @@ def test_guardrails_builtin_detectors_negative_detection(
 class TestGuardrailsOrchestratorHuggingFaceGPU:
     """
     These tests verify that the GuardrailsOrchestrator works as expected when using HuggingFace detectors
-    Steps:/
-        - Deploy an LLM (Qwen2.5-0.5B-Instruct) using the vLLM SR.
+    Steps:
+        - Deploy an LLM (Qwen2.5-3B-Instruct) using the vLLM SR.
         - Deploy the GuardrailsOrchestrator.
         - Deploy a prompt injection detector using the HuggingFace SR.
         - Check that the detector works when we have an unsuitable input.

utilities/constants.py

@@ -497,7 +497,7 @@ class LLMdInferenceSimConfig:
     "service": {"hostname": f"{LLMdInferenceSimConfig.isvc_name}-predictor", "port": 80}
 }
 
-
+@dataclass
 class VLLMGPUConfig:
     name: str = "vllm-gpu"
     port: int = 80
@@ -509,11 +509,12 @@ class VLLMGPUConfig:
     def get_hostname(cls, namespace: str) -> str:
         return f"{cls.isvc_name}-predictor.{namespace}.svc.cluster.local"
 
-
 VLLM_CHAT_GENERATION_CONFIG: dict[str, Any] = {
-    "service": {"hostname": VLLMGPUConfig.get_hostname("test-guardrails-huggingface"), "port": VLLMGPUConfig.port}
+    "service": {
+        "hostname": VLLMGPUConfig.get_hostname("test-guardrails-huggingface"),
+        "port": VLLMGPUConfig.port
+    }
 }
 
-
 class PodNotFound(Exception):
     """Pod not found"""