Frequently Asked Questions¶
Common questions about GPUX.
General¶
What is GPUX?¶
GPUX is a Docker-like GPU runtime for ML inference that provides universal GPU compatibility.
Why use GPUX instead of serving frameworks like TorchServe or Triton?¶
GPUX focuses on simplicity and universal GPU compatibility, while TorchServe and Triton are more complex and NVIDIA-focused.
Is GPUX production-ready?¶
Yes! GPUX is built on mature technologies (ONNX Runtime) and is ready for production use.
Installation¶
Which Python versions are supported?¶
Python 3.11 and higher.
Do I need a GPU?¶
No! GPUX works on CPU-only machines with automatic fallback.
Performance¶
How fast is GPUX?¶
Performance depends on your hardware and model. Use gpux run --benchmark to measure.
Can I use TensorRT for optimization?¶
Yes! GPUX automatically uses TensorRT when available on NVIDIA GPUs.
More Questions?
Check our Help page or open a discussion.