Daniel BachranMar 7, 2024

Next attempt to get #stable_diffusion working with my AMD-based setup, using the following instructions:
https://community.amd.com/t5/ai/updated-how-to-running-optimized-automatic1111-stable-diffusion/ba-p/630252

Unfortunately, running into an error upon following the instructions when running webui.bat: "RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check"

Need to check where to set COMMANDLINE_ARGS, I guess... 😄

[UPDATED HOW-TO] Running Optimized Automatic1111 Stable Diffusion WebUI on AMD GPUs

[UPDATE]: The Automatic1111-directML branch now supports Microsoft Olive under the Automatic1111 WebUI interface, which allows for generating optimized models and running them all under the Automatic1111 WebUI, without a separate branch needed to optimize for AMD platforms. The original blog with ad...

AMD.com
Show threadDaniel BachranMar 7, 2024

Set COMMANDLINE_ARGS in the environement for an initial test, but also defined it in webui-user.bat, which calls webui.bat:

set COMMANDLINE_ARGS=--skip-torch-cuda-test

Show threadDaniel BachranMar 7, 2024

Next error message: "launch.py: error: unrecognized arguments: --onnx --backend directml"

Interestingly, just running webui-user.bat seems to progress further...?

> webui-user.bat --onnx --backend directml

Show threadDaniel Bachran

#stable_diffusion #automatic111 UI is opening fine, but without ONNX and Olive tabs... 🤔

Checked the settings and enabled Olive and ONNX there, but no change. 😕
(And yes, I have applied the settings, reloaded the UI, and even tried restarting the whole batch. No ONNX and Olive tabs in the UI.)

Mar 7, 2024 at 11:37amWeb
Show threadDaniel BachranMar 7, 2024

And... It doesn't work, just throwing the following error:

ONNX: Failed to convert model: model='v1-5-pruned-emaonly.safetensors', error=Exporting the operator 'aten::scaled_dot_product_attention' to ONNX opset version 14 is not supported.

Show threadDaniel BachranMar 7, 2024

This comment sounded promising, but to no avail: Still getting the same error. The only thing I didn't do, yet, was to switch the execution provider: Searched for "execution provider" in the #automatic111 settings and switched from "CPUExecutionProvider" to "DmlExecutionProvider".

Still getting the same error, though. 🤔

Show threadDaniel BachranMar 7, 2024

This sounds like the underlying issue so far: https://github.com/pytorch/pytorch/issues/120559

Will see when I have time to investigate further...

ONNX: Failed to convert model / ONNX opset version 14 is not supported · Issue #120559 · pytorch/pytorch

🐛 Describe the bug Hello, since a while i am trying to get Stable Diffusion running on my RX 7900 XTX. Finally it is working normal when generating with a normal model that is not opitmized. Now i ...

GitHub