Skip to content

edit gemm_a4w8_asm api #672

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
valarLip merged 1 commit into from
Jul 17, 2025
Merged

edit gemm_a4w8_asm api #672

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
2 changes: 1 addition & 1 deletion aiter/ops/gemm_op_a4w4.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,7 +69,7 @@ def gemm_a4w4_asm(
A_scale: Tensor, # A_scale:[M, K/32] e8m0 paded
B_scale: Tensor, # B_scale:[N, K/32] e8m0 paded
out: Tensor, # Out:[M, N] bf16
bias: Tensor, # bias:[1, N] f32
bias: Optional[Tensor] = None, # bias:[1, N] f32
alpha: Optional[float] = 1.0,
beta: Optional[float] = 0.0,
bpreshuffle: Optional[bool] = True,
Expand Down
8 changes: 4 additions & 4 deletions csrc/include/asm_gemm_a4w4.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ torch::Tensor gemm_a4w4_asm(torch::Tensor& A, // A:[M, K/2] f4x2
torch::Tensor& A_scale, // A_scale:[M, K/32] e8m0 paded
torch::Tensor& B_scale, // B_scale:[N, K/32] e8m0 paded
torch::Tensor& out, // Out:[M, N] bf16
torch::Tensor& bias, // bias:[M, N] f32
std::optional<float> alpha = 1.0,
std::optional<float> beta = 0.0,
std::optional<bool> bpreshuffle = true);
std::optional<torch::Tensor> bias = std::nullopt, // bias:[M, N] f32
std::optional<float> alpha = 1.0,
std::optional<float> beta = 0.0,
std::optional<bool> bpreshuffle = true);
26 changes: 13 additions & 13 deletions csrc/include/rocm_ops.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -302,18 +302,18 @@
py::arg("pad_c") = 0, \
py::arg("splitK") = 0);

#define GEMM_A4W4_ASM_PYBIND \
m.def("gemm_a4w4_asm", \
&gemm_a4w4_asm, \
"Asm gemm a4w4", \
py::arg("A"), \
py::arg("B"), \
py::arg("A_scale"), \
py::arg("B_scale"), \
py::arg("out"), \
py::arg("bias"), \
py::arg("alpha") = 1.0, \
py::arg("beta") = 0.0, \
#define GEMM_A4W4_ASM_PYBIND \
m.def("gemm_a4w4_asm", \
&gemm_a4w4_asm, \
"Asm gemm a4w4", \
py::arg("A"), \
py::arg("B"), \
py::arg("A_scale"), \
py::arg("B_scale"), \
py::arg("out"), \
py::arg("bias") = std::nullopt, \
py::arg("alpha") = 1.0, \
py::arg("beta") = 0.0, \
py::arg("bpreshuffle") = true);

#define GEMM_A4W4_BLOCKSCALE_PYBIND \
Expand Down Expand Up @@ -760,7 +760,7 @@
py::arg("unit_size"), \
py::arg("local_expert_mask") = std::nullopt, \
py::arg("num_local_tokens") = std::nullopt, \
py::arg("dispatch_policy") = 0);
py::arg("dispatch_policy") = 0);

#define NORM_PYBIND \
m.def("layernorm2d_fwd", \
Expand Down
10 changes: 5 additions & 5 deletions csrc/py_itfs_cu/asm_gemm_a4w4.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,10 +64,10 @@ torch::Tensor gemm_a4w4_asm(torch::Tensor& A, // A:[M, K/2] f4x2
torch::Tensor& A_scale, // A_scale:[M, K/32] e8m0 paded
torch::Tensor& B_scale, // B_scale:[N, K/32] e8m0 paded
torch::Tensor& out, // Out:[M, N] bf16
torch::Tensor& bias, // bias:[M, N] f32
std::optional<float> alpha = 1.0,
std::optional<float> beta = 0.0,
std::optional<bool> bpreshuffle = true)
std::optional<torch::Tensor> bias = std::nullopt, // bias:[M, N] f32
std::optional<float> alpha = 1.0,
std::optional<float> beta = 0.0,
std::optional<bool> bpreshuffle = true)
{
TORCH_CHECK(
out.dtype() == torch::ScalarType::BFloat16, __func__, " only support BFloat16 output now!");
Expand All @@ -77,7 +77,7 @@ torch::Tensor gemm_a4w4_asm(torch::Tensor& A, // A:[M, K/2] f4x2
KernelArgs args;
size_t arg_size = sizeof(args);
args.ptr_D = (void*)out.data_ptr();
args.ptr_C = (void*)bias.data_ptr();
args.ptr_C = bias.has_value() ? (void*)bias.value().data_ptr() : nullptr;
args.ptr_A = (void*)A.data_ptr();
args.ptr_B = (void*)B.data_ptr();

Expand Down
2 changes: 1 addition & 1 deletion op_tests/test_gemm_a4w4.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,7 +81,7 @@ def test_gemm(dtype, M, N, K):
out1 = torch.empty(M, N, dtype=dtype)
out2 = torch.empty((M + 255) // 256 * 256, N, dtype=dtype)
out3 = torch.empty((M + 255) // 256 * 256, N, dtype=dtype)
bias_f32 = torch.zeros(M, N, dtype=dtype)
bias_f32 = None
x_scales = x_scales.view(torch.uint8)
w_scales = w_scales.view(torch.uint8)

Expand Down