Cannot import name 'DIFFUSERS_SLOW_IMPORT' from 'diffusers.utils'

[Eneskoser]

Describe the bug

I got errors like below while I was using lpw_stable_diffusion

ImportError: cannot import name 'DIFFUSERS_SLOW_IMPORT' from 'diffusers.utils'
(/usr/local/lib/python3.10/dist-packages/diffusers/utils/init.py)

Reproduction

1 from typing import TYPE_CHECKING │
│ 2 │
│ ❱ 3 from ..utils import DIFFUSERS_SLOW_IMPORT, _LazyModule, deprecate │
│ 4 from ..utils.import_utils import is_torch_available, is_transformers_available

Logs

No response

System Info

diffusers==0.25.1

Who can help?

@sayakpaul @patrickvonplaten

[sayakpaul]

Cc: @DN6

[DN6]

@Eneskoser Could you share a reproducible code example please? I need to see exactly what you are running when you encounter the error.

[Eneskoser]

Hi
Its is like below

def get_pipeline(model_name_or_path,
                 vae_name_or_path=None,
                 text_encoder_name_or_path=None,
                 feature_extractor_name_or_path=None,
                 revision="fp16"):
    #scheduler = DPMSolverMultistepScheduler.from_pretrained(model_name_or_path, subfolder="scheduler")
    scheduler = DPMSolverMultistepScheduler(
        beta_start=0.00085,
        beta_end=0.012,
        beta_schedule="scaled_linear",
        num_train_timesteps=1000,
        trained_betas=None,
        prediction_type="epsilon",
        thresholding=False,
        algorithm_type="dpmsolver++",
        solver_type="midpoint",
        lower_order_final=True,
    )

    pipe = DiffusionPipeline.from_pretrained(
        model_name_or_path,
        custom_pipeline="lpw_stable_diffusion",
        safety_checker=None,
        revision=revision,
        scheduler=scheduler,
        vae=AutoencoderKL.from_pretrained(
            vae_name_or_path or model_name_or_path,
            subfolder=None if vae_name_or_path else "vae",
            revision=None if vae_name_or_path else revision,
            torch_dtype=torch.float16,
        ),
        feature_extractor=feature_extractor_name_or_path,
        torch_dtype=torch.float16
    ).to("cuda")

    #https://github.com/huggingface/diffusers/issues/1552
    #pipe.enable_attention_slicing()
    pipe.enable_xformers_memory_efficient_attention()
    return pipe
    
   model_name_or_path = 'runwayml/stable-diffusion-v1-5'
    vae_name_or_path = 'stabilityai/sd-vae-ft-mse'
    pipe = get_pipeline(MODEL_NAME_OR_PATH, vae_name_or_path=VAE_NAME_OR_PATH)

[patrickvonplaten]

Hey @Eneskoser,

We cannot execute this code - as it does run correctly. Please make sure to add a minimal, reproducible code snippet.

[hisushanta]

Hello @Eneskoser, I ran your code in Google Collab, and it works fine.

!pip3 install -U xformers --index-url https://download.pytorch.org/whl/cu121

from diffusers import DPMSolverMultistepScheduler , DiffusionPipeline, AutoencoderKL
import torch
def get_pipeline(model_name_or_path,
                 vae_name_or_path=None,
                 text_encoder_name_or_path=None,
                 feature_extractor_name_or_path=None,
                 revision="fp16"):
    #scheduler = DPMSolverMultistepScheduler.from_pretrained(model_name_or_path, subfolder="scheduler")
    scheduler = DPMSolverMultistepScheduler(
        beta_start=0.00085,
        beta_end=0.012,
        beta_schedule="scaled_linear",
        num_train_timesteps=1000,
        trained_betas=None,
        prediction_type="epsilon",
        thresholding=False,
        algorithm_type="dpmsolver++",
        solver_type="midpoint",
        lower_order_final=True,
    )

    pipe = DiffusionPipeline.from_pretrained(
        model_name_or_path,
        custom_pipeline="lpw_stable_diffusion",
        safety_checker=None,
        revision=revision,
        scheduler=scheduler,
        vae=AutoencoderKL.from_pretrained(
            vae_name_or_path or model_name_or_path,
            subfolder=None if vae_name_or_path else "vae",
            revision=None if vae_name_or_path else revision,
            torch_dtype=torch.float16,
        ),
        feature_extractor=feature_extractor_name_or_path,
        torch_dtype=torch.float16
    ).to("cuda")

    #https://github.com/huggingface/diffusers/issues/1552
    #pipe.enable_attention_slicing()
    pipe.enable_xformers_memory_efficient_attention()
    return pipe
    
model_name_or_path = 'runwayml/stable-diffusion-v1-5'
vae_name_or_path = 'stabilityai/sd-vae-ft-mse'
pipe = get_pipeline(model_name_or_path, vae_name_or_path=vae_name_or_path)

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[sayakpaul]

I guess this is not an issue anymore.