Skip to content

[Metal] global scale for qmm #4458

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
nastya236 merged 6 commits into from
Sep 8, 2026
Merged

[Metal] global scale for qmm #4458

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
727c58e
added global scale for qmm
nastya236 Sep 3, 2026
f7c3404
fix metal::is_available
nastya236 Sep 3, 2026
d6f1a8e
Merge branch 'main' into gather-qmm-nvfp4-global-scale
nastya236 Sep 3, 2026
285ab89
fix shapes and errors
nastya236 Sep 4, 2026
a54b48a
Merge remote-tracking branch 'origin/main' into gather-qmm-nvfp4-glob…
nastya236 Sep 8, 2026
af85e41
address comments
nastya236 Sep 8, 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
7 changes: 6 additions & 1 deletion mlx/backend/cuda/quantized/quantized.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -156,11 +156,16 @@ void GatherQMM::eval_gpu(const std::vector<array>& inputs, array& out) {
auto& s = stream();
auto& encoder = cu::get_command_encoder(s);

if (mode_ != QuantizationMode::Affine && inputs.size() == 6) {
throw std::runtime_error(
"[GatherQMM] Global scale is only supported on the Metal backend.");
}

array x = ensure_row_contiguous(inputs[0], encoder, s);
const array& w = inputs[1];
const array& scales = inputs[2];
std::optional<array> biases;
if (inputs.size() == 6) {
if (mode_ == QuantizationMode::Affine) {
biases = inputs[3];
}
array lhs_indices =
Expand Down
84 changes: 58 additions & 26 deletions mlx/backend/metal/jit_kernels.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1054,28 +1054,44 @@ MTL::ComputePipelineState* get_gather_qmm_kernel(
int bk,
int wm,
int wn,
bool transpose) {
bool transpose,
bool has_global_scale) {
const auto& lib_name = kernel_name;
auto lib = d.get_library(lib_name, [&]() {
std::string kernel_source;
concatenate(
kernel_source, metal::utils(), metal::quantized_utils(), metal::gemm());
bool is_affine = mode == "affine";
// Only the fp kernel takes a global scale.
auto template_def = is_affine ? get_template_definition(
lib_name,
"affine_gather_qmm_rhs",
get_type_string(x.dtype()),
group_size,
bits,
bm,
bn,
bk,
wm,
wn,
transpose)
: get_template_definition(
lib_name,
"fp_gather_qmm_rhs",
get_type_string(x.dtype()),
group_size,
bits,
bm,
bn,
bk,
wm,
wn,
transpose,
has_global_scale);
concatenate(
kernel_source,
is_affine ? metal::quantized() : metal::fp_quantized(),
get_template_definition(
lib_name,
(is_affine ? "affine" : "fp") + std::string("_gather_qmm_rhs"),
get_type_string(x.dtype()),
group_size,
bits,
bm,
bn,
bk,
wm,
wn,
transpose));
template_def);
return kernel_source;
});
return d.get_kernel(kernel_name, lib, hash_name, func_consts);
Expand Down Expand Up @@ -1255,7 +1271,8 @@ MTL::ComputePipelineState* get_gather_qmm_nax_kernel(
int bk,
int wm,
int wn,
bool transpose) {
bool transpose,
bool has_global_scale) {
const auto& lib_name = kernel_name;
auto lib = d.get_library(lib_name, [&]() {
std::string kernel_source;
Expand All @@ -1265,21 +1282,36 @@ MTL::ComputePipelineState* get_gather_qmm_nax_kernel(
metal::gemm_nax(),
metal::quantized_utils());
bool is_affine = mode == "affine";
// Only the fp kernel takes a global scale.
auto template_def = is_affine ? get_template_definition(
lib_name,
"affine_gather_qmm_rhs_nax",
get_type_string(x.dtype()),
group_size,
bits,
bm,
bn,
bk,
wm,
wn,
transpose)
: get_template_definition(
lib_name,
"fp_gather_qmm_rhs_nax",
get_type_string(x.dtype()),
group_size,
bits,
bm,
bn,
bk,
wm,
wn,
transpose,
has_global_scale);
concatenate(
kernel_source,
is_affine ? metal::quantized_nax() : metal::fp_quantized_nax(),
get_template_definition(
lib_name,
(is_affine ? "affine" : "fp") + std::string("_gather_qmm_rhs_nax"),
get_type_string(x.dtype()),
group_size,
bits,
bm,
bn,
bk,
wm,
wn,
transpose));
template_def);
return kernel_source;
});
return d.get_kernel(kernel_name, lib, hash_name, func_consts);
Expand Down
6 changes: 4 additions & 2 deletions mlx/backend/metal/kernels.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -321,7 +321,8 @@ MTL::ComputePipelineState* get_gather_qmm_kernel(
int bk,
int wm,
int wn,
bool transpose);
bool transpose,
bool has_global_scale);

MTL::ComputePipelineState* get_steel_gemm_fused_nax_kernel(
metal::Device& d,
Expand Down Expand Up @@ -400,7 +401,8 @@ MTL::ComputePipelineState* get_gather_qmm_nax_kernel(
int bk,
int wm,
int wn,
bool transpose);
bool transpose,
bool has_global_scale);

MTL::ComputePipelineState* get_steel_attention_kernel(
metal::Device& d,
Expand Down
Loading
Loading