Skip to content
This repository was archived by the owner on May 11, 2025. It is now read-only.

Conversation

IlyasMoutawwakil
Copy link
Collaborator

Using HIPified Exllama (1 & 2) kernels in casper-hansen/AutoAWQ_kernels#5

@IlyasMoutawwakil IlyasMoutawwakil mentioned this pull request Jan 22, 2024
30 tasks
@casper-hansen
Copy link
Owner

casper-hansen commented Jan 26, 2024

@IlyasMoutawwakil Assuming you have an AMD ROCm GPU available to you, could you test if the AutoAWQ_kernels build is working for you by running perplexity/inference examples?

You can download the artifact here: https://github.com/casper-hansen/AutoAWQ_kernels/actions/runs/7667011822#artifacts

image

@IlyasMoutawwakil
Copy link
Collaborator Author

I confirm that the ROCm wheels work fine on an AMD MI250

@casper-hansen
Copy link
Owner

casper-hansen commented Jan 31, 2024

Hi @IlyasMoutawwakil, I ended up getting this error. I am not sure if it's just a bad GPU that I got on RunPod or what could possibly be wrong. I just ran a normal pip install -e .. EDIT: will try again soon, looks like a bad GPU (I think)

@IlyasMoutawwakil
Copy link
Collaborator Author

yes that seems unrelated to AWQ

@casper-hansen
Copy link
Owner

Looks good to me! Tested it with quite a few configurations. Nice work @IlyasMoutawwakil

@casper-hansen casper-hansen merged commit f018d2b into main Feb 3, 2024
@casper-hansen casper-hansen deleted the amd-rocm-support branch February 12, 2024 14:21
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants