[Bug]: When working with Onnx/Olive -> TypeError: 'OnnxRawPipeline' object is not callable

Gonzalo1987 commented 2 months ago

Checklist

[ ] The issue exists after disabling all extensions
[X] The issue exists on a clean installation of webui
[ ] The issue is caused by an extension, but I believe it is caused by a bug in the webui
[X] The issue exists in the current version of the webui
[X] The issue has not been reported before recently
[ ] The issue has been reported before but has not been fixed yet

What happened?

I can't generate images using Olive since the February refactoring. I've been waiting for it to be fixed in a commit but it doesn't seem to happen.

I've some people with the same bug in: https://github.com/lshqqytiger/stable-diffusion-webui-directml/discussions/149 https://github.com/lshqqytiger/stable-diffusion-webui-directml/discussions/149#discussioncomment-8642882

I can't use Olive since the refactor of feb. I was generating images with Olive at ~8 it/s and I'm now without olive at ~1.15 it/s.

I can do any tests that are necessary to try to correct the problem.

Steps to reproduce the problem

git clone https://github.com/lshqqytiger/stable-diffusion-webui-directml 2 cd ./stable-diffusion-webui-directml
Start webui. [with --use-directml]
Go to ONNX Runtime tab in Settings.
Check Use ONNX Runtime instead of PyTorch implementation.
Change Execution Provider to "DmlExecutionProvider"
Check Enable Olive.
Check everything under Olive models to process. --
.\venv\Scripts\activate
pip uninstall torch torchvision torch-directml -y
pip install onnxruntime-directml
Start webui [with --use-directml]
Check DmlExecutionProvider
Go to txt2img tab and generate.

What should have happened?

It should generate the images using Olive.

What browsers do you use to access the UI ?

Mozilla Firefox

Sysinfo

sysinfo-2024-04-24-18-57.json

Console logs

https://pastebin.com/MNQzzD3Y  -- first run
https://pastebin.com/Y2c23Dfw  -- reload ui
https://pastebin.com/VysnbEFS  -- with extra instructions

Additional information

It worked with the Olive UI some months ago.

FauxPrada commented 2 months ago

I don't think you need --use-directml when using ONNX / OLive, here is what I use:

--skip-torch-cuda-test --use-cpu-torch --opt-sdp-attention --disable-nan-check

You might need to remove any caches as well, like the venv cache and ~/.cache

Gonzalo1987 commented 2 months ago

I don't think you need --use-directml when using ONNX / OLive, here is what I use:

--skip-torch-cuda-test --use-cpu-torch --opt-sdp-attention --disable-nan-check

You might need to remove any caches as well, like the venv cache and ~/.cache

sadly I got the same error (OnnxRawPipeline is not callable) :(

"--use-cpu-torch" is for use the cpu instead the gpu it isn't?

Thanks!

lshqqytiger commented 1 month ago

Please pull the latest commits. You should have the latest stable release of pytorch.

.\venv\Scripts\activate
pip uninstall torch torchvision torch-directml onnxruntime onnxruntime-directml -y
pip install torch torchvision onnxruntime
pip install onnxruntime-directml

Then, run webui with --use-cpu-torch.

Gonzalo1987 commented 1 month ago

It's working like a charm, thank you very much for your efforts!!

lshqqytiger / stable-diffusion-webui-amdgpu