Stable Diffusion ONNX requires cpu `onnxruntime` even if CUDA version is installed

Describe the bug

The ONNX support doesn’t work with CUDAExecutionProvider I installed onnxruntime-gpu Running import onnxruntime as ort ort.get_device() results GPU and ort.get_available_providers() results ['CPUExecutionProvider', 'TensorrtExecutionProvider', 'CUDAExecutionProvider'] but diffusers complains onnxruntime not installed and wants me to install the cpu version(pip install onnxruntime).

Reproduction

Install pip install onnxruntime-gpu

and run

from diffusers import StableDiffusionOnnxPipeline

pipe = StableDiffusionOnnxPipeline.from_pretrained(
    "CompVis/stable-diffusion-v1-4",
    revision="onnx",
    provider="CUDAExecutionProvider",
    use_auth_token=true,
)

Logs

No response

System Info

• diffusers version: 0.3.0
• Platform: Linux-5.10.133±x86_64-with-Ubuntu-18.04-bionic
• Python version: 3.7.13
• PyTorch version (GPU?): 1.12.1+cu113 (True)
• Huggingface_hub version: 0.9.1
• Transformers version: 4.21.3
• Using GPU in script?: Yes
• Using distributed or parallel set-up in script?: No

[SkyTNT]

This is my PR trying to fix this issues: #440

[patrickvonplaten]

Thanks!