Skip to content

Disable half2 for ExLlama when using HIP #2912

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jun 29, 2023
Merged

Conversation

ardfork
Copy link
Contributor

@ardfork ardfork commented Jun 28, 2023

Using kernels that rely on half2 produce gibberish output when using ROCm. ExLlama UI also disable half2 in the same way when using HIP.

Fix #2840

Using kernels that rely on half2 produce gibberish output when using
ROCm. ExLlama UI also disable half2 in the same way when using HIP.
@duncannah
Copy link

duncannah commented Jun 28, 2023

Thanks for your great work, I can confirm this works wonderfully.

One thing: exllama_hf.py is missing the torch version import.

@Koko2110
Copy link

Thank you for the extremely quick fix, I can also confirm it works as expected now.

@Dajinu
Copy link

Dajinu commented Jun 29, 2023

I've just tried this fix and it works. Man, you deserve all the love of the universe. Can't thank you enough!

@oobabooga oobabooga merged commit 3c076c3 into oobabooga:main Jun 29, 2023
jdehorty pushed a commit to jdehorty/text-generation-webui that referenced this pull request Jul 9, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

exllama gibberish output
5 participants