char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

Zijun Yuopenvino: OV 2026.2, context-shift, Q5_1 support, gemma4 dense/embedding, and -fa off (#24503)890f1a27e

char/aliceai-t5-moe
brancheschar/aliceai-t5-moeall refs →
693 B27 linesraw
1#include "../node_context.h"
2#include "../op_table.h"
3#include "../utils.h"
4
5#include <memory>
6#include <openvino/op/constant.hpp>
7#include <openvino/op/reduce_sum.hpp>
8
9namespace ov {
10namespace frontend {
11namespace ggml {
12namespace op {
13
14OutputVector translate_sum_rows(const NodeContext & context) {
15    num_inputs_check(context, 1, 1);
16
17    auto input = process_view_input_new(context, 0);
18    auto res = std::make_shared<ov::op::v1::ReduceSum>(
19        input, ov::op::v0::Constant::create(ov::element::i64, ov::Shape{1}, {-1}), true);
20
21    return rename_outputs_with_suffix({res}, context.get_name());
22}
23
24}  // namespace op
25}  // namespace ggml
26}  // namespace frontend
27}  // namespace ov