Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Running on AMD gpu on Windows without installing the full SDK #194

DGdev91 started this conversation in Show and tell
Discussion options

The ROCm version of KoboldCpp (https://github.com/YellowRoseCx/koboldcpp-rocm) copies it's runtime libs in a temporary folder called "_MEI" + some nubers, placed in C:\Users\username\AppData\Local\Temp. In those libs there are also hipblas.dll, rocblas.dll, and a "library" folder containing ROCm's tensile library files, and this allows it to run on amd gpus without the need of having the SDK installed.

I just checked, and this works just fine for stablediffussion.cpp too. just copy those libs from koboldcpp's temp folder and place it in the same place as sd.exe

Maybe i can also be a good idea shipping those libs in the rocm windows build

You must be logged in to vote

Replies: 1 comment

Comment options

Bless you! I was getting realy skill issued trying to get my ROCm build to link properly to the SDK, and this solved it for me, though things have changed a bit since you started that discussion.

I couldn't find rocblas.dll in the "_MEI" temp folder, only hipblas.dll, however in the more recent relseases of the ROCm version of KoboldCpp, there are zip files alongside the exe (koboldcpp_rocm_files.zip), which contains some of the libraries used by koboldcpp, including rocblas.dll.

You must be logged in to vote
0 replies
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
2 participants

AltStyle によって変換されたページ (->オリジナル) /