Remove logger warnings for attention backends and hard error during runtime instead

[a-r-r-o-w]

• Fixes Passing enable_gqa flag in attention dispatcher should be version guarded #11959
• Instead of raising warning with logger and erroring out with a NoneType exception, we now check the imports during runtime and error out instead

import torch
from diffusers import FluxPipeline
from diffusers.utils import load_image

pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16
).to("cuda")

image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/flux_ip_adapter_input.jpg").resize((1024, 1024))

pipe.load_ip_adapter(
    "XLabs-AI/flux-ip-adapter",
    weight_name="ip_adapter.safetensors",
    image_encoder_pretrained_model_name_or_path="openai/clip-vit-large-patch14"
)
pipe.set_ip_adapter_scale(1.0)
pipe.transformer.set_attention_backend("_flash_3") # errors out if you don't have FA3
pipe.transformer.set_attention_backend("sage") # errors out if you don't have sage 

image = pipe(
    width=1024,
    height=1024,
    prompt="wearing sunglasses",
    negative_prompt="",
    true_cfg_scale=4.0,
    generator=torch.Generator().manual_seed(4444),
    ip_adapter_image=image,
).images[0]

image.save('output.png')

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[DN6]

Thank you 🙏🏽

[a-r-r-o-w]

Failing test is unrelated