AMD GPU not being utilized with automatic 1111:

[CaslerTheTesticle]

Is there an existing issue for this?

• [X] I have searched the existing issues and checked the recent builds/commits

What would your feature do ?

I would like the feature to allow automatic 1111 to use the AMD RYZEN GPU for quicker builds. If there is any way of using automatic 1111 with amd please comment

Proposed workflow

just like for intel but for amd

Additional information

No response

[FreeBlues]

It can, I am using a AMD GPU: RX 6600 8G on my MBP 2020 (x86 cpu)

you need use some parameters, just like below:

python webui.py --no-half --opt-split-attention-v1

[bsalberto77]

It can, I am using a AMD GPU: RX 6600 8G on my MBP 2020 (x86 cpu)

you need use some parameters, just like below:

python webui.py --no-half --opt-split-attention-v1

Linux o Windows? thanks

[FreeBlues]

It can, I am using a AMD GPU: RX 6600 8G on my MBP 2020 (x86 cpu) you need use some parameters, just like below:

python webui.py --no-half --opt-split-attention-v1

Linux o Windows? thanks

I am using macOS. It can work on Linux, need rocm, someone success, but it looks like it is hard to run on windows.

Here is my system:

[Jorginthesage]

I did not notice any speed increase when using —opt-split-attention-v1

[misterteetee]

It can, I am using a AMD GPU: RX 6600 8G on my MBP 2020 (x86 cpu) you need use some parameters, just like below:

python webui.py --no-half --opt-split-attention-v1

Linux o Windows? thanks

I am using macOS. It can work on Linux, need rocm, someone success, but it looks like it is hard to run on windows.

Here is my system:

Not seeing any improvements here either (dual 6900XTs i9 hackintosh). Consistently 1.3-1.4 it/sec