Apply GPU type verification on backend debug mode

backend-cli.py

@@ -448,7 +448,8 @@ def get_args():
     parser.add_argument("--precision", type=str, default="float32,float16,8bit,4bit", help="Precision to debug")
     parser.add_argument("--inference-framework", type=str, default="hf-chat", help="Inference framework to debug")
     parser.add_argument("--limit", type=int, default=None, help="Limit for the number of samples")
-    parser.add_argument("--gpu-type", type=str, default="NVIDIA-A100-PCIe-80GB", help="GPU type")
+    parser.add_argument("--gpu-type", type=str, default="NVIDIA-A100-PCIe-80GB", 
+                        help="GPU type. NVIDIA-A100-PCIe-80GB; NVIDIA-RTX-A5000-24G; NVIDIA-H100-PCIe-80G")
     return parser.parse_args()
 
 
@@ -480,6 +481,10 @@ if __name__ == "__main__":
                             inference_framework=args.inference_framework,  # Use inference framework from arguments
                             gpu_type=args.gpu_type
                         )
+                        curr_gpu_type = get_gpu_details()
+                        if eval_request.gpu_type != curr_gpu_type:
+                            print(f"GPU type mismatch: {eval_request.gpu_type} vs {curr_gpu_type}")
+                            raise Exception("GPU type mismatch")
                         results = process_evaluation(task, eval_request, limit=args.limit)
                     except Exception as e:
                         print(f"debug running error: {e}")