Skip to content

ggml : implement set_rows with i32 index #16159

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 9 commits into from
Sep 22, 2025
Merged

ggml : implement set_rows with i32 index #16159

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
a7403b1
implement set_rows with i32 index
CISC Sep 21, 2025
9a7bff3
template fix
CISC Sep 21, 2025
7657ec3
test quantized path
CISC Sep 22, 2025
480b404
Apply suggestions from code review
CISC Sep 22, 2025
9899182
forgotten name change
CISC Sep 22, 2025
aa76826
deduplicate cuda/sycl and test-fix [no-ci]
CISC Sep 22, 2025
bd47cba
indent++ [no-ci]
CISC Sep 22, 2025
d8abd2d
vulkan: support set_rows with i32 index type (#16162)
jeffbolznv Sep 22, 2025
2510204
disable i32 index for webgpu for now [no ci]
CISC Sep 22, 2025
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
12 changes: 10 additions & 2 deletions ggml/src/ggml-cpu/ops.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4739,6 +4739,7 @@ void ggml_compute_forward_get_rows(
//}
}

template<typename idx_t>
static void ggml_compute_forward_set_rows_f32(
const ggml_compute_params * params,
ggml_tensor * dst) {
Expand Down Expand Up @@ -4777,7 +4778,7 @@ static void ggml_compute_forward_set_rows_f32(
const int64_t i11 = i02%ne11;
const int64_t i10 = i;

const int64_t i1 = *(int64_t *) ((char *) src1->data + i10*nb10 + i11*nb11 + i12*nb12);
const int64_t i1 = *(idx_t *) ((char *) src1->data + i10*nb10 + i11*nb11 + i12*nb12);

GGML_ASSERT(i1 >= 0 && i1 < ne1);

Expand All @@ -4794,11 +4795,18 @@ void ggml_compute_forward_set_rows(
ggml_tensor * dst) {

const ggml_tensor * src0 = dst->src[0];
const ggml_tensor * src1 = dst->src[1];

switch (src0->type) {
case GGML_TYPE_F32:
{
ggml_compute_forward_set_rows_f32(params, dst);
if (src1->type == GGML_TYPE_I64) {
ggml_compute_forward_set_rows_f32<int64_t>(params, dst);
} else if (src1->type == GGML_TYPE_I32) {
ggml_compute_forward_set_rows_f32<int32_t>(params, dst);
} else {
GGML_ABORT("src1->type = %d (%s) not supported", src1->type, ggml_type_name(src1->type));
}
} break;
default:
{
Expand Down
2 changes: 1 addition & 1 deletion ggml/src/ggml-cuda/ggml-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3427,7 +3427,7 @@ static bool ggml_backend_cuda_device_supports_op(ggml_backend_dev_t dev, const g
op->type == GGML_TYPE_Q4_0 || op->type == GGML_TYPE_Q4_1 || op->type == GGML_TYPE_Q5_0 ||
op->type == GGML_TYPE_Q5_1 || op->type == GGML_TYPE_Q8_0 || op->type == GGML_TYPE_IQ4_NL) &&
op->src[0]->type == GGML_TYPE_F32 &&
op->src[1]->type == GGML_TYPE_I64;
(op->src[1]->type == GGML_TYPE_I64 || op->src[1]->type == GGML_TYPE_I32);
} break;
case GGML_OP_CPY:
{
Expand Down
Loading
Loading