Skip to content

Add Grouped GEMM for Mixed Dtype #457

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

Open
wants to merge 7 commits into
base: sycl-develop
Choose a base branch
from

Conversation

muhammad-tanvir-1211
Copy link

This PR adds Grouped GEMM support for mixed precision GEMM.

@muhammad-tanvir-1211 muhammad-tanvir-1211 requested a review from a team July 4, 2025 15:59
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you rename this example to list the types instead of "mixed_dtype"?

using ret_type = cute::conditional_t<sizeof_bits_v<ElementZero> >= 8, ElementZero, int8_t>;
ret_type a = [&]() {
if constexpr (sizeof_bits_v<QuantizedElement> >= 8) {
return (ret_type)(src_tensor(n, k, l));
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
return (ret_type)(src_tensor(n, k, l));
return static_cast<ret_type>(src_tensor(n, k, l));

And a few more examples of this below.

Tensor<EngineScales, LayoutScales>& tCrS_input,
Tensor<EngineZeros, LayoutZeros> tCrZ_input
) {
// TODO: add assert here because such cases not support for int4 now
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
// TODO: add assert here because such cases not support for int4 now
// TODO (Codeplay): add assert here because int4 is not currently supported

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.

2 participants