fix a bug in from_config

[yiyixuxu]

fix #7183

this script now runs expected

from diffusers import DiffusionPipeline, EulerAncestralDiscreteScheduler, LCMScheduler, AutoencoderKL
import torch

model_id = "stabilityai/stable-diffusion-xl-base-1.0"

pipe = DiffusionPipeline.from_pretrained(model_id, variant="fp16", torch_dtype=torch.float16)

#print correct value
print(f" test1: should print EulerAncestralDiscreteScheduler")
print(pipe.scheduler.config._class_name)
print(pipe.scheduler)
print(pipe.scheduler.config._class_name)

pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)

#gives Euler not LCM
print(f" test2: shuld print LCMScheduler")
print(pipe.scheduler.config._class_name)
print(pipe.scheduler)
print(pipe.scheduler.config._class_name)

pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)

#prints LCM not EulerA
print(f" test3: shuld print EulerAncestralDiscreteScheduler")
print(pipe.scheduler.config._class_name)
print(pipe.scheduler)
print(pipe.scheduler.config._class_name)

output

Loading pipeline components...: 100%|█████████████████████████████████████████████████████████████████████████████████████████████| 7/7 [00:02<00:00,  3.33it/s]
 test1: should print EulerAncestralDiscreteScheduler
EulerDiscreteScheduler
EulerDiscreteScheduler {
  "_class_name": "EulerDiscreteScheduler",
  "_diffusers_version": "0.27.0.dev0",
  "beta_end": 0.012,
  "beta_schedule": "scaled_linear",
  "beta_start": 0.00085,
  "clip_sample": false,
  "interpolation_type": "linear",
  "num_train_timesteps": 1000,
  "prediction_type": "epsilon",
  "rescale_betas_zero_snr": false,
  "sample_max_value": 1.0,
  "set_alpha_to_one": false,
  "sigma_max": null,
  "sigma_min": null,
  "skip_prk_steps": true,
  "steps_offset": 1,
  "timestep_spacing": "leading",
  "timestep_type": "discrete",
  "trained_betas": null,
  "use_karras_sigmas": false
}

EulerDiscreteScheduler
The config attributes {'skip_prk_steps': True} were passed to LCMScheduler, but are not expected and will be ignored. Please verify your scheduler_config.json configuration file.
 hidden_dict: {'interpolation_type': 'linear', 'use_karras_sigmas': False, '_class_name': 'LCMScheduler', '_diffusers_version': '0.27.0.dev0', 'skip_prk_steps': True}
 test2: shuld print LCMScheduler
LCMScheduler
LCMScheduler {
  "_class_name": "LCMScheduler",
  "_diffusers_version": "0.27.0.dev0",
  "beta_end": 0.012,
  "beta_schedule": "scaled_linear",
  "beta_start": 0.00085,
  "clip_sample": false,
  "clip_sample_range": 1.0,
  "dynamic_thresholding_ratio": 0.995,
  "interpolation_type": "linear",
  "num_train_timesteps": 1000,
  "original_inference_steps": 50,
  "prediction_type": "epsilon",
  "rescale_betas_zero_snr": false,
  "sample_max_value": 1.0,
  "set_alpha_to_one": false,
  "skip_prk_steps": true,
  "steps_offset": 1,
  "thresholding": false,
  "timestep_scaling": 10.0,
  "timestep_spacing": "leading",
  "trained_betas": null,
  "use_karras_sigmas": false
}

LCMScheduler
 hidden_dict: {'clip_sample': False, 'set_alpha_to_one': False, 'sample_max_value': 1.0, 'interpolation_type': 'linear', 'use_karras_sigmas': False, '_class_name': 'EulerAncestralDiscreteScheduler', '_diffusers_version': '0.27.0.dev0', 'skip_prk_steps': True}
 test3: shuld print EulerAncestralDiscreteScheduler
EulerAncestralDiscreteScheduler
EulerAncestralDiscreteScheduler {
  "_class_name": "EulerAncestralDiscreteScheduler",
  "_diffusers_version": "0.27.0.dev0",
  "beta_end": 0.012,
  "beta_schedule": "scaled_linear",
  "beta_start": 0.00085,
  "clip_sample": false,
  "interpolation_type": "linear",
  "num_train_timesteps": 1000,
  "prediction_type": "epsilon",
  "rescale_betas_zero_snr": false,
  "sample_max_value": 1.0,
  "set_alpha_to_one": false,
  "skip_prk_steps": true,
  "steps_offset": 1,
  "timestep_spacing": "leading",
  "trained_betas": null,
  "use_karras_sigmas": false
}

EulerAncestralDiscreteScheduler

[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]

LGTM 👍🏽