Skip to content

add y_dtype op overload for expected_weights_desc #361

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
LuFinch merged 2 commits into from
Jul 23, 2026
+23 −4
Merged

add y_dtype op overload for expected_weights_desc #361

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
1fa0dc3
add y_dtype op overload for expected_weights_desc
agrawal-aka Jun 4, 2026
45ed6b6
Update include/ideep/operators/matmul.hpp
agrawal-aka Jul 23, 2026
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
27 changes: 23 additions & 4 deletions include/ideep/operators/matmul.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -659,9 +659,10 @@ struct matmul_forward : public dnnl::matmul,

static tensor::desc expected_weights_desc(
const dims& weights_dims,
const dims& src_dims = dims(),
data_type dtype = data_type::f32,
data_type x_dtype = data_type::f32,
const dims& src_dims,
data_type dtype,
data_type x_dtype,
data_type y_dtype,
const attr_t& attr = attr_t(),
const engine& aengine = engine::cpu_engine()) {
auto ndims = weights_dims.size();
Expand All @@ -670,7 +671,6 @@ struct matmul_forward : public dnnl::matmul,
x_dims[ndims-1] = weights_dims[ndims-2];
dims y_dims = (ndims == 3) ? dims({x_dims[0], x_dims[1], weights_dims[2]})
: dims({x_dims[0], weights_dims[1]});
auto y_dtype = (dtype != data_type::s8) ? dtype : data_type::s32;

IDEEP_ENFORCE(x_dims.size() == weights_dims.size(),
"Invalid dims for data and weights");
Expand All @@ -681,6 +681,25 @@ struct matmul_forward : public dnnl::matmul,
return pd.weights_desc();
}

static tensor::desc expected_weights_desc(
const dims& weights_dims,
const dims& src_dims = dims(),
data_type dtype = data_type::f32,
data_type x_dtype = data_type::f32,
const attr_t& attr = attr_t(),
const engine& aengine = engine::cpu_engine()) {
auto y_dtype = (dtype != data_type::s8) ? dtype : data_type::s32;

return expected_weights_desc(
weights_dims,
src_dims,
dtype,
x_dtype,
y_dtype,
attr,
aengine);
}

private:
// For 2-in-1 compute: prepare + compute
// Supports fp32, static int8 and dynamic int8
Expand Down