UI with errors

[vic-t]

Is there maybe more to getting the UI working than cloning it into the Stable-Diffusion-ONNX-FP16 project folder and running it as described in the guide?

I followed the guide to the letter and then installed the UI as instructed. The service gets started fine, I can navigate to the page and enter a prompt to run. I can see that the preparation looks similar to when generating one of the image samples but before it starts to actually generate the image, the process fails with the following error (log includes the last working instruction):

2023-05-08 10:37:27.9455751 [W:onnxruntime:, session_state.cc:1138 onnxruntime::VerifyEachNodeIsAssignedToAnEp] Rerunning with verbose output on a non-minimal build will show node assignments.
iteration 1/1
Traceback (most recent call last):
  File "D:\ai\amd310\lib\site-packages\gradio\routes.py", line 412, in run_predict
    output = await app.get_blocks().process_api(
  File "D:\ai\amd310\lib\site-packages\gradio\blocks.py", line 1299, in process_api
    result = await self.call_function(
  File "D:\ai\amd310\lib\site-packages\gradio\blocks.py", line 1021, in call_function
    prediction = await anyio.to_thread.run_sync(
  File "D:\ai\amd310\lib\site-packages\anyio\to_thread.py", line 31, in run_sync
    return await get_asynclib().run_sync_in_worker_thread(
  File "D:\ai\amd310\lib\site-packages\anyio\_backends\_asyncio.py", line 937, in run_sync_in_worker_thread
    return await future
  File "D:\ai\amd310\lib\site-packages\anyio\_backends\_asyncio.py", line 867, in run
    result = context.run(func, *args)
  File "D:\ai\amd310\Stable-Diffusion-ONNX-FP16\OnnxDiffusersUI\onnxUI.py", line 978, in generate_click
    images, status = run_diffusers(
  File "D:\ai\amd310\Stable-Diffusion-ONNX-FP16\OnnxDiffusersUI\onnxUI.py", line 110, in run_diffusers
    batch_images = pipe(
  File "D:\ai\amd310\lib\site-packages\diffusers\pipelines\stable_diffusion\pipeline_onnx_stable_diffusion.py", line 365, in __call__
    prompt_embeds = self._encode_prompt(
TypeError: _encode_prompt() got an unexpected keyword argument 'prompt_embeds'

Any advice? I'm on Win11 with Radeon RX Vega 64 8GB VRAM, I'm running all of this in a venv with Python 3.10.10.

[Amblyopius]

Hi, onnx diffusers UI seems to not be keeping up with diffusers anymore. The error is related to a change in newer versions of diffusers. I'll see if I can maintain a fixed fork and will look for a different UI that can be integrated or make a basic one myself.

[NeusZimmer]

I have to check on the new version, I'm still on the previous, and need to install and try by myself, but maybe you could test my UI version and see it it works for this version...
I will try to adapt to the new version during next days, will cross my fingers on it, hope they did not break something that make it work in my low profile 4gb card... (EDIT:I did a clean install, and it works with the torch nightly version, no need to adapt it)
https://github.com/NeusZimmer/ONNX-ModularUI

[NeusZimmer]

PD, looking at the error, could be related to the lpw_pipe.py substitution of the text encoder, I faced a similar problem when adding to my UI, and I had to add a fake keyword argument to pass over it.

[Amblyopius]

I changed the docs today and it now points to my fork of ONNXdiffusersUI which resolves the issue (it was indeed in lpw_pipe.py).