Skip to content

x_ray_runtime_environment

inference_models.developer_tools.x_ray_runtime_environment cached 

x_ray_runtime_environment()

Inspect the runtime environment and available hardware/software.

Performs comprehensive introspection of the system to detect available GPUs, CUDA/TensorRT versions, installed ML frameworks, and execution providers. This is useful for debugging model loading issues and understanding what backends are available.

The function is cached, so subsequent calls return the same result without re-inspecting the environment.

Returns:

  • RuntimeXRayResult ( RuntimeXRayResult ) –

    Dataclass containing:

    • gpu_available (bool): Whether any GPU is available

    • gpu_devices (List[str]): Names of available GPU devices

    • gpu_devices_cc (List[Version]): Compute capability versions

    • driver_version (Optional[Version]): NVIDIA driver version

    • cuda_version (Optional[Version]): CUDA runtime version

    • trt_version (Optional[Version]): TensorRT version

    • jetson_type (Optional[str]): Jetson device type (if running on Jetson)

    • l4t_version (Optional[Version]): L4T version (Jetson only)

    • os_version (Optional[str]): Operating system version

    • torch_available (bool): Whether PyTorch is installed

    • torch_version (Optional[Version]): PyTorch version

    • torchvision_version (Optional[Version]): Torchvision version

    • onnxruntime_version (Optional[Version]): ONNX Runtime version

    • available_onnx_execution_providers (Optional[Set[str]]): Available ONNX EPs

    • hf_transformers_available (bool): Whether Hugging Face Transformers is installed

    • trt_python_package_available (bool): Whether TensorRT Python package is installed

    • mediapipe_available (bool): Whether MediaPipe is installed

Examples:

Inspect runtime environment:

>>> from inference_models.developer_tools import x_ray_runtime_environment
>>>
>>> env = x_ray_runtime_environment()
>>> print(f"GPU available: {env.gpu_available}")
>>> print(f"GPU devices: {env.gpu_devices}")
>>> print(f"CUDA version: {env.cuda_version}")
>>> print(f"PyTorch available: {env.torch_available}")
>>> print(f"ONNX providers: {env.available_onnx_execution_providers}")

Check for specific capabilities:

>>> env = x_ray_runtime_environment()
>>> if env.gpu_available and env.cuda_version:
...     print("CUDA is available, can use GPU models")
>>> if "CUDAExecutionProvider" in (env.available_onnx_execution_providers or []):
...     print("ONNX CUDA execution provider available")

Debug model loading issues:

>>> from inference_models import AutoModel
>>> from inference_models.developer_tools import x_ray_runtime_environment
>>>
>>> try:
...     model = AutoModel.from_pretrained("yolov8n-640", device="cuda")
... except Exception as e:
...     print(f"Model loading failed: {e}")
...     env = x_ray_runtime_environment()
...     print(f"Environment info: {env}")
See Also
  • AutoModel.describe_compute_environment(): Higher-level environment inspection