The model loading has suddenly become slow.

[crapthings]

Describe the bug

The model loading has suddenly become slow.

Reproduction

text2imgPipe = StableDiffusionControlNetPipeline.from_single_file(
  './base.safetensors',
  controlnet = [
    depth_control,
    openpose_control,
    inpaint_control,
    canny_control,
  ],
  torch_dtype = torch.float16,
  enable_pag = True,
  pag_applied_layers = ['mid'],
)

Logs

No response

System Info

• 🤗 Diffusers version: 0.30.0.dev0
• Platform: Linux-6.5.0-41-generic-x86_64-with-glibc2.35
• Running on a notebook?: No
• Running on Google Colab?: No
• Python version: 3.10.12
• PyTorch version (GPU?): 2.1.1+cu118 (True)
• Flax version (CPU?/GPU?/TPU?): not installed (NA)
• Jax version: not installed
• JaxLib version: not installed
• Huggingface_hub version: 0.23.3
• Transformers version: 4.43.0.dev0
• Accelerate version: 0.32.0
• PEFT version: 0.10.0
• Bitsandbytes version: not installed
• Safetensors version: 0.4.3
• xFormers version: 0.0.23+cu118
• Accelerator: NVIDIA L40, 46068 MiB VRAM
• Using GPU in script?:
• Using distributed or parallel set-up in script?:

Who can help?

No response

[crapthings]

revert to works

pip install accelerate==0.31.0

[knoopx]

I just want to express how frustrating is for the users of the library to have stuff breaking every single time regardless it's a major, minor or patch release or a completely unrelated library (Transformers/Accelerate/PEFT). Please take semantic versioning seriously and don't rush releases. Like for example this release, a new model should not be part of a patch release: https://github.com/huggingface/diffusers/releases/tag/v0.28.1

[DN6]

@crapthings I believe a fix has been merged into accelerate and a patch release will be done soon
huggingface/accelerate#2914

[muellerzr]

The patch is out on pypi, apologies for the frustration it has caused (which is entirely understandable!)

[sayakpaul]

Closing this since this seems to have been solved. Apologies for causing the frustration.