stabilityai/stable-diffusion-2-base DDIM config is not compatible with current DDIM implementation

[Zeqiang-Lai]

Describe the bug

Assume 30 inference steps.

There are two 925, and the number of inference steps is 31. tensor([958, 925, 925, 892, 859, 826, 793, 760, 727, 694, 661, 628, 595, 562, 529, 496, 463, 430, 397, 364, 331, 298, 265, 232, 199, 166, 133, 100, 67, 34, 1], device='cuda:0')

{
  "_class_name": "DDIMScheduler",
  "_diffusers_version": "0.8.0",
  "beta_end": 0.012,
  "beta_schedule": "scaled_linear",
  "beta_start": 0.00085,
  "clip_sample": false,
  "num_train_timesteps": 1000,
  "set_alpha_to_one": false,
  "skip_prk_steps": true,
  "steps_offset": 1,
  "trained_betas": null
}

The correct one should be. tensor([958, 925, 892, 859, 826, 793, 760, 727, 694, 661, 628, 595, 562, 529, 496, 463, 430, 397, 364, 331, 298, 265, 232, 199, 166, 133, 100, 67, 34, 1], device='cuda:0')

{
  "_class_name": "DDIMScheduler",
  "_diffusers_version": "0.25.0",
  "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,
  "num_train_timesteps": 1000,
  "prediction_type": "epsilon",
  "rescale_betas_zero_snr": false,
  "sample_max_value": 1.0,
  "set_alpha_to_one": false,
  "steps_offset": 1,
  "thresholding": false,
  "timestep_spacing": "leading",
  "trained_betas": null
}

Reproduction

See above

Logs

No response

System Info

Version: 0.26.3

Who can help?

No response

[sayakpaul]

Cc: @yiyixuxu

[yiyixuxu]

can you share a reproducible script?

[github-actions[bot]]

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.