ValueError - OnnxUI doesn't generate images with converted Waifu Diffusion model

[GitThirteen]

Good morning/afternoon/evening,

I have been trying to get OnnxUI to work with waifu diffusion v1.3, and even though I (somehow) managed to convert the initial .ckpt into an onnx, the console is shouting the following at me after I click on Generate in the GUI:

Traceback (most recent call last):
  File "C:\Users\<Name>\Desktop\stable_diff\virtualenv\lib\site-packages\gradio\routes.py", line 289, in run_predict
    output = await app.blocks.process_api(
  File "C:\Users\<Name>\Desktop\stable_diff\virtualenv\lib\site-packages\gradio\blocks.py", line 991, in process_api
    result = await self.call_function(fn_index, inputs, iterator)
  File "C:\Users\<Name>\Desktop\stable_diff\virtualenv\lib\site-packages\gradio\blocks.py", line 833, in call_function
    prediction = await anyio.to_thread.run_sync(
  File "C:\Users\<Name>\Desktop\stable_diff\virtualenv\lib\site-packages\anyio\to_thread.py", line 31, in run_sync
    return await get_asynclib().run_sync_in_worker_thread(
  File "C:\Users\<Name>\Desktop\stable_diff\virtualenv\lib\site-packages\anyio\_backends\_asyncio.py", line 937, in run_sync_in_worker_thread
    return await future
  File "C:\Users\<Name>\Desktop\stable_diff\virtualenv\lib\site-packages\anyio\_backends\_asyncio.py", line 867, in run
    result = context.run(func, *args)
  File "C:\Users\<Name>\Desktop\stable_diff\onnxUI.py", line 173, in generate_click
    pipe = OnnxStableDiffusionPipeline.from_pretrained(
  File "C:\Users\<Name>\Desktop\stable_diff\virtualenv\lib\site-packages\diffusers\pipeline_utils.py", line 556, in from_pretrained
    raise ValueError(
ValueError: Pipeline <class 'diffusers.pipelines.stable_diffusion.pipeline_onnx_stable_diffusion.OnnxStableDiffusionPipeline'> expected {'safety_checker', 'vae_decoder', 'scheduler', 'vae_encoder', 'text_encoder', 'feature_extractor', 'tokenizer', 'unet'}, but only {'safety_checker', 'vae_decoder', 'scheduler', 'text_encoder', 'feature_extractor', 'tokenizer', 'unet'} were passed.

Comparing the provided modules to the expected ones the 'vae_encoder' seems to be missing.
I did have to use diffusers 0.5.1 with an older version of the diffusers -> onnx script to convert without errors, so my theory is that it might have something to do with some change that occurred between diffusers v0.5.1 and v0.6.0 (especially as pulling the regular, pre-packed diffusion model and using it via the GUI works perfectly fine). I was hoping that it would at least work with 0.5.1, but the GUI doesn't like that version.

Working on Windows 11 with 32 GB RAM and an RX580 with 8 GB VRAM.

The waifu diffusion model: https://huggingface.co/hakurei/waifu-diffusion-v1-3/blob/main/wd-v1-3-float32.ckpt

Appreciate any ideas on how to make this work.

[GitThirteen]

Update: No idea what went wrong yesterday, but I just did everything again including the changes in the newest pull request by averad and it's working perfectly fine now. (Had to edit the setup.bat a bit so it wouldn't override the changes on -update) Amazing project!

[azuritecoin]

You're correct. There was a change in the conversion scripts between v0.5.1 and v0.6.0. You need to reconvert the models using diffusers v0.6.0. It seems like you got it to work in the end. BTW waifu diffusion v1.3 has a diffusers version as well as a .ckpt.