Skip to content

Add support for FP8 on compute capability >=8.0, <8.9 #2213

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.

Sign up for GitHub

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
Jul 11, 2024
Merged

Add support for FP8 on compute capability >=8.0, <8.9 #2213

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions server/marlin/marlin_kernels/__init__.pyi
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,3 +59,18 @@ def marlin_gemm(
Matrix multiplication using Marlin kernels.
"""
...

# fp8 marlin
def fp8_marlin_gemm(
a: torch.Tensor,
b_q_weight: torch.Tensor,
b_scales: torch.Tensor,
workspace: torch.Tensor,
num_bits: int,
size_m: int,
size_n: int,
size_k: int,
) -> torch.Tensor:
return torch.ops._C.fp8_marlin_gemm(
a, b_q_weight, b_scales, workspace, num_bits, size_m, size_n, size_k
)
2 changes: 2 additions & 0 deletions server/marlin/marlin_kernels/ext.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,4 +9,6 @@ PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
m.def("gptq_marlin_repack", &gptq_marlin_repack,
"Repack GPTQ parameters for Marlin");
m.def("marlin_gemm", &marlin_gemm, "Marlin gemm");
// fp8_marlin Optimized Quantized GEMM for FP8 weight-only.
m.def("fp8_marlin_gemm", &fp8_marlin_gemm);
}
5 changes: 5 additions & 0 deletions server/marlin/marlin_kernels/ext.hh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,4 +27,9 @@ torch::Tensor marlin_gemm(torch::Tensor &a, torch::Tensor &b_q_weight,
torch::Tensor &b_scales, torch::Tensor &workspace,
int64_t size_m, int64_t size_n, int64_t size_k);

torch::Tensor fp8_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
torch::Tensor& b_scales, torch::Tensor& workspace,
int64_t num_bits, int64_t size_m, int64_t size_n,
int64_t size_k);

#endif
Loading