whispercpp 1.3.5 → 1.3.6

data/ext/sources/ggml/src/ggml-openvino/openvino/op/glu_geglu.cpp

@@ -0,0 +1,61 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <memory>
+#include <openvino/core/node_output.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/gelu.hpp>
+#include <openvino/op/multiply.hpp>
+#include <openvino/op/sigmoid.hpp>
+#include <openvino/op/slice.hpp>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_glu_geglu(const NodeContext & context) {
+    num_inputs_check(context, 1, 2);
+
+    ov::Output<ov::Node> src0;
+    ov::Output<ov::Node> src1;
+    if (context.get_input_size() == 2) {
+        src0 = context.get_input(0);
+        src1 = context.get_input(1);
+    } else {
+        // GGML splits along ne[0] (OV last axis) using floor division: nc = ne[0] / 2.
+        // Both halves are nc elements; if the dimension is odd, the last element is dropped.
+        // Use Slice instead of Split to handle odd dimensions correctly.
+        auto combined = context.get_input(0);
+        auto combined_shape = combined.get_partial_shape();
+        int64_t last_dim_val = combined_shape[combined_shape.rank().get_length() - 1].get_length();
+        int64_t nc = last_dim_val / 2;
+
+        auto axis   = ov::op::v0::Constant::create(ov::element::i64, {1}, {-1});
+        auto step   = ov::op::v0::Constant::create(ov::element::i64, {1}, {1});
+        auto start0 = ov::op::v0::Constant::create(ov::element::i64, {1}, {0});
+        auto stop0  = ov::op::v0::Constant::create(ov::element::i64, {1}, {nc});
+        auto start1 = ov::op::v0::Constant::create(ov::element::i64, {1}, {nc});
+        auto stop1  = ov::op::v0::Constant::create(ov::element::i64, {1}, {2 * nc});
+
+        src0 = std::make_shared<ov::op::v8::Slice>(combined, start0, stop0, step, axis);
+        src1 = std::make_shared<ov::op::v8::Slice>(combined, start1, stop1, step, axis);
+    }
+
+    int32_t * params = context.get_output_op_params();
+    const int32_t swapped = params[1];
+    if (swapped) {
+        std::swap(src0, src1);
+    }
+
+    auto gelu = std::make_shared<ov::op::v7::Gelu>(src0);
+    auto res = std::make_shared<ov::op::v1::Multiply>(gelu, src1);
+
+    return rename_outputs_with_suffix({res}, context.get_name());
+}
+
+}  // namespace op
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

data/ext/sources/ggml/src/ggml-openvino/openvino/op/glu_swiglu.cpp

@@ -0,0 +1,62 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <cstdint>
+#include <memory>
+#include <openvino/core/node_output.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/multiply.hpp>
+#include <openvino/op/sigmoid.hpp>
+#include <openvino/op/slice.hpp>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_glu_swiglu(const NodeContext & context) {
+    num_inputs_check(context, 1, 2);
+
+    ov::Output<ov::Node> src0;
+    ov::Output<ov::Node> src1;
+    if (context.get_input_size() == 2) {
+        src0 = context.get_input(0);
+        src1 = context.get_input(1);
+    } else {
+        // GGML splits along ne[0] (OV last axis) using floor division: nc = ne[0] / 2.
+        // Both halves are nc elements; if the dimension is odd, the last element is dropped.
+        // Use Slice instead of Split to handle odd dimensions correctly.
+        auto combined = context.get_input(0);
+        auto combined_shape = combined.get_partial_shape();
+        int64_t last_dim_val = combined_shape[combined_shape.rank().get_length() - 1].get_length();
+        int64_t nc = last_dim_val / 2;
+
+        auto axis   = ov::op::v0::Constant::create(ov::element::i64, {1}, {-1});
+        auto step   = ov::op::v0::Constant::create(ov::element::i64, {1}, {1});
+        auto start0 = ov::op::v0::Constant::create(ov::element::i64, {1}, {0});
+        auto stop0  = ov::op::v0::Constant::create(ov::element::i64, {1}, {nc});
+        auto start1 = ov::op::v0::Constant::create(ov::element::i64, {1}, {nc});
+        auto stop1  = ov::op::v0::Constant::create(ov::element::i64, {1}, {2 * nc});
+
+        src0 = std::make_shared<ov::op::v8::Slice>(combined, start0, stop0, step, axis);
+        src1 = std::make_shared<ov::op::v8::Slice>(combined, start1, stop1, step, axis);
+    }
+
+    int32_t * params = context.get_output_op_params();
+    const int32_t swapped = params[1];
+    if (swapped) {
+        std::swap(src0, src1);
+    }
+
+    auto sigmoid = std::make_shared<ov::op::v0::Sigmoid>(src0);
+    auto silu = std::make_shared<ov::op::v1::Multiply>(src0, sigmoid);
+    auto res = std::make_shared<ov::op::v1::Multiply>(silu, src1);
+
+    return rename_outputs_with_suffix({res}, context.get_name());
+}
+
+}  // namespace op
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

data/ext/sources/ggml/src/ggml-openvino/openvino/op/mulmat.cpp

@@ -0,0 +1,90 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <climits>
+#include <cstdint>
+#include <memory>
+#include <openvino/core/node.hpp>
+#include <openvino/core/node_output.hpp>
+#include <openvino/op/broadcast.hpp>
+#include <openvino/op/concat.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/convert.hpp>
+#include <openvino/op/matmul.hpp>
+#include <openvino/op/reshape.hpp>
+#include <openvino/op/slice.hpp>
+#include <openvino/op/transpose.hpp>
+#include <openvino/op/unsqueeze.hpp>
+#include <openvino/op/util/op_types.hpp>
+#include <vector>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_mulmat(const NodeContext & context) {
+    num_inputs_check(context, 2, 2);
+
+    int op_case = context.get_op_case();
+
+    ov::Output<Node> res;
+    ov::Output<ov::Node> B = context.get_input(0);
+    ov::Output<ov::Node> A = context.get_input(1);
+
+    bool transpose_b = true;
+    if (op_case == 2) {
+        B = B.get_node_shared_ptr()->input_value(0);
+        transpose_b = false;
+    } else if (op_case == 3) {
+        B = process_view_input(context, 0);
+        A = process_view_input(context, 1);
+    }
+    if (A.get_element_type() != B.get_element_type()) {
+        B = std::make_shared<ov::op::v0::Convert>(context.get_input(0), context.get_input_type(1));
+    }
+
+    auto B_shape = context.get_input_shape(0).to_shape();
+    auto A_shape = context.get_input_shape(1).to_shape();
+    int64_t A_batch = A_shape[1];
+    int64_t B_batch = B_shape[1];
+
+    auto A_batch_larger = A_batch > B_batch;
+    auto batch_large = A_batch_larger ? A_batch : B_batch;
+    auto batch_small = A_batch_larger ? B_batch : A_batch;
+
+    Output<Node> Z = A_batch_larger ? B : A;
+    int64_t factor = batch_large / batch_small;
+    if (factor > 1 && batch_small > 1) {
+        auto batch_large_node = ov::op::v0::Constant::create(ov::element::i64, {1}, std::vector<int64_t>{batch_large});
+        auto batch_small_node = ov::op::v0::Constant::create(ov::element::i64, {1}, std::vector<int64_t>{batch_small});
+        auto factor_node = ov::op::v0::Constant::create(ov::element::i64, {1}, std::vector<int64_t>{factor});
+
+        auto unsqueeze_axes = ov::op::v0::Constant::create(ov::element::i64, Shape{}, {2});
+        auto Z_unsqueezed = std::make_shared<ov::op::v0::Unsqueeze>(Z, unsqueeze_axes);
+
+        auto broadcast_shape = ov::op::v0::Constant::create(
+            ov::element::i64, {5}, {(int64_t) 1, (int64_t) 1, factor, (int64_t) 1, (int64_t) 1});
+        auto new_Z_shape = ov::op::v0::Constant::create(ov::element::i64, {4},
+                                                        {(int64_t) 0, batch_large, (int64_t) -1, (int64_t) A_shape[3]});
+
+        auto Z_broadcasted = std::make_shared<ov::op::v3::Broadcast>(Z_unsqueezed, broadcast_shape,
+                                                                     ov::op::BroadcastType::BIDIRECTIONAL);
+        Z = std::make_shared<ov::op::v1::Reshape>(Z_broadcasted, new_Z_shape, true);
+    }
+    if (A_batch_larger) {
+        B = Z;
+    } else {
+        A = Z;
+    }
+
+    res = std::make_shared<ov::op::v0::MatMul>(A, B, false, transpose_b);
+
+    return rename_outputs_with_suffix({res}, context.get_name());
+}
+
+}  // namespace op
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

data/ext/sources/ggml/src/ggml-openvino/openvino/op/permute.cpp

@@ -0,0 +1,102 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <climits>
+#include <cstdint>
+#include <memory>
+#include <openvino/core/node.hpp>
+#include <openvino/op/add.hpp>
+#include <openvino/op/concat.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/reshape.hpp>
+#include <openvino/op/slice.hpp>
+#include <openvino/op/transpose.hpp>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_permute(const NodeContext & context) {
+    num_inputs_check(context, 1, 1);
+
+    int op_case = context.get_op_case();
+    FRONT_END_CHECK_IMPLEMENTED(op_case == 1 || op_case == 2 || op_case == 3 || op_case == 4,
+                                "Unsupported PERMUTE case");
+
+    ov::Output<Node> res;
+    auto src = context.get_input(0);
+    auto perm = ov::op::v0::Constant::create(ov::element::i64, {4}, {0, 2, 1, 3});
+
+    if (op_case == 1 || context.is_stateful()) {
+        res = std::make_shared<ov::op::v1::Transpose>(src, perm);
+    } else if (op_case == 4) {
+        auto output_shape = context.get_output_shape().to_shape();
+        auto n_heads = ov::op::v0::Constant::create(ov::element::i64, {1}, {output_shape[1]});
+        auto head_size = ov::op::v0::Constant::create(ov::element::i64, {1}, {output_shape[3]});
+        auto n_seq_active = context.has_input("n_seq_active") ?
+                                context.get_input("n_seq_active") :
+                                ov::op::v0::Constant::create(ov::element::i64, {1}, {output_shape[0]});
+        auto neg_one = ov::op::v0::Constant::create(ov::element::i64, {1}, {-1});
+
+        auto new_shape =
+            std::make_shared<ov::op::v0::Concat>(ov::OutputVector{n_seq_active, neg_one, n_heads, head_size}, 0);
+
+        // // Alternative
+        // auto zero = ov::op::v0::Constant::create(ov::element::i64, {1}, {0});
+        // auto new_shape = std::make_shared<ov::op::v0::Concat>(ov::OutputVector{n_seq_active, neg_one, zero, zero}, 0);
+
+        auto reshaped = std::make_shared<ov::op::v1::Reshape>(src, new_shape, true);
+        res = std::make_shared<ov::op::v1::Transpose>(reshaped, perm);
+    } else {
+        auto cache_shape = src.get_partial_shape();
+        auto output_shape = context.get_output_shape().to_shape();
+        int64_t head_size = output_shape[3];
+        int64_t n_heads = output_shape[1];
+        int64_t ctx_per_seq = cache_shape[2].is_static() ? cache_shape[2].get_length() : -1;
+        int64_t n_seq = cache_shape[1].get_length();
+
+        Output<Node> attention_size;
+        if (!context.has_input("attention_size")) {
+            attention_size = ov::op::v0::Constant::create(ov::element::i64, {1}, {output_shape[2]});
+        } else if (op_case == 2) {
+            attention_size = context.get_input("attention_size");
+        } else {
+            attention_size = context.get_input("attention_size_swa");
+        }
+
+        Output<Node> seq_active_start;
+        Output<Node> seq_active_end;
+        if (context.has_input("seq_active_start")) {
+            seq_active_start = context.get_input("seq_active_start");
+            seq_active_end = context.get_input("seq_active_end");
+        } else {
+            int64_t n_seq_active = output_shape[0];
+            size_t offset = *((size_t *) context.get_input_op_params(0));
+            int64_t seq_active_start_val = offset / context.get_input_stride(0)[0];
+            int64_t seq_active_end_val = seq_active_start_val + n_seq_active;
+            seq_active_start = ov::op::v0::Constant::create(ov::element::i64, {1}, {seq_active_start_val});
+            seq_active_end = ov::op::v0::Constant::create(ov::element::i64, {1}, {seq_active_end_val});
+        }
+
+        // 1. reshape to [n_seq, ctx_per_seq, n_heads, head_size]
+        // 2. slice out the active sequences
+        // 3. slice out the attention part in each sequence
+        // 4. permute
+        auto zero = ov::op::v0::Constant::create(ov::element::i64, {1}, {0});
+        auto one = ov::op::v0::Constant::create(ov::element::i64, {1}, {1});
+
+        auto src_reshaped = std::make_shared<ov::op::v1::Reshape>(
+            src, ov::op::v0::Constant::create(ov::element::i64, {4}, {n_seq, ctx_per_seq, n_heads, head_size}), false);
+        auto slice1 = std::make_shared<ov::op::v8::Slice>(src_reshaped, seq_active_start, seq_active_end, one, zero);
+        auto slice2 = std::make_shared<ov::op::v8::Slice>(slice1, zero, attention_size, one, one);
+        res = std::make_shared<ov::op::v1::Transpose>(slice2, perm);
+    }
+    return rename_outputs_with_suffix({res}, context.get_name());
+}
+
+}  // namespace op
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

data/ext/sources/ggml/src/ggml-openvino/openvino/op/reshape.cpp

@@ -0,0 +1,83 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <cstdint>
+#include <memory>
+#include <openvino/core/node.hpp>
+#include <openvino/core/node_output.hpp>
+#include <openvino/frontend/exception.hpp>
+#include <openvino/op/concat.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/reshape.hpp>
+#include <stdexcept>
+#include <vector>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_reshape(const NodeContext & context) {
+    num_inputs_check(context, 1, 1);
+    if (context.get_input_shape(0) == context.get_output_shape()) {
+        return {context.get_input(0)};
+    }
+
+    int op_case = context.get_op_case();
+    FRONT_END_CHECK_IMPLEMENTED(
+        op_case == 1 || op_case == 2 || op_case == 3 || op_case == 4 || op_case == 5 || op_case == 6,
+        "Unsupported RESHAPE case");
+
+    auto output_shape = context.get_output_shape().to_shape();
+    std::shared_ptr<ov::Node> new_shape_node;
+    if (op_case == 1) {
+        if (context.is_stateful()) {
+            new_shape_node = ov::op::v0::Constant::create(
+                ov::element::i64, {3},
+                std::vector<int64_t>{-1, (int64_t) output_shape[2], (int64_t) output_shape[3]});
+        } else {
+            new_shape_node = ov::op::v0::Constant::create(
+                ov::element::i64, {4},
+                std::vector<int64_t>{(int64_t) output_shape[0], -1, (int64_t) output_shape[2], (int64_t) output_shape[3]});
+        }
+    } else if (op_case == 2) {
+        new_shape_node = ov::op::v0::Constant::create(
+            ov::element::i64, {4},
+            std::vector<int64_t>{(int64_t) output_shape[0], (int64_t) output_shape[1], -1, (int64_t) output_shape[3]});
+
+    } else if (op_case == 3) {
+        throw std::runtime_error("might be outdated RESHAPE case");
+        new_shape_node = ov::op::v0::Constant::create(
+            ov::element::i64, {4}, std::vector<int64_t>{(int64_t) output_shape[0], (int64_t) output_shape[1], -1, 1});
+
+    } else if (op_case == 4) {
+        return {context.get_input(0).get_node_shared_ptr()->input_value(0)};
+
+    } else if (op_case == 5) {
+        if (context.is_stateful()) {
+            std::vector<int64_t> shape_vec = {1, -1, (int64_t) context.get_output_shape().to_shape()[3]};
+            new_shape_node = ov::op::v0::Constant::create(ov::element::i64, {3}, shape_vec);
+        } else {
+            std::vector<int64_t> shape_vec = {1, 1, -1, (int64_t) context.get_output_shape().to_shape()[3]};
+            new_shape_node = ov::op::v0::Constant::create(ov::element::i64, {4}, shape_vec);
+        }
+
+        // // Alternative
+        // auto token_len = context.get_input("token_len");
+        // auto emb_size =
+        //     ov::op::v0::Constant::create(ov::element::i64, {1}, {(int64_t) context.get_output_shape().to_shape()[3]});
+        // auto one = ov::op::v0::Constant::create(ov::element::i64, {1}, {1});
+        // new_shape_node = std::make_shared<ov::op::v0::Concat>(ov::OutputVector{one, one, token_len, emb_size}, 0);
+
+    } else if (op_case == 6) {
+        new_shape_node = ov::op::v0::Constant::create(ov::element::i64, {4}, context.get_output_shape().to_shape());
+    }
+    auto res = std::make_shared<ov::op::v1::Reshape>(context.get_input(0), new_shape_node, false);
+    return rename_outputs_with_suffix({res}, context.get_name());
+}
+
+}  // namespace op
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

data/ext/sources/ggml/src/ggml-openvino/openvino/op/rms_norm.cpp

@@ -0,0 +1,46 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <memory>
+#include <openvino/op/add.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/divide.hpp>
+#include <openvino/op/multiply.hpp>
+#include <openvino/op/power.hpp>
+#include <openvino/op/reduce_mean.hpp>
+#include <openvino/op/sqrt.hpp>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_rms_norm(const NodeContext & context) {
+    num_inputs_check(context, 1, 1);
+
+    auto input_node = context.get_input(0);
+    auto square = std::make_shared<ov::op::v1::Power>(
+        input_node, ov::op::v0::Constant::create(ov::element::f32, ov::Shape{1}, {2.0f}));
+
+    auto mean = std::make_shared<ov::op::v1::ReduceMean>(
+        square, ov::op::v0::Constant::create(ov::element::i64, ov::Shape{1}, {-1}), true);
+
+    float eps;
+    memcpy(&eps, context.get_output_op_params(), sizeof(float));
+
+    auto rms = std::make_shared<ov::op::v0::Sqrt>(
+        std::make_shared<ov::op::v1::Add>(mean, ov::op::v0::Constant::create(ov::element::f32, ov::Shape{1}, {eps})));
+
+    auto reciprocal =
+        std::make_shared<ov::op::v1::Divide>(ov::op::v0::Constant::create(ov::element::f32, ov::Shape{1}, {1.0f}), rms);
+
+    auto res = std::make_shared<ov::op::v1::Multiply>(input_node, reciprocal);
+
+    return rename_outputs_with_suffix({res}, context.get_name());
+}
+
+}  // namespace op
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

data/ext/sources/ggml/src/ggml-openvino/openvino/op/rope.cpp

@@ -0,0 +1,123 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <cstdint>
+#include <memory>
+#include <openvino/core/node.hpp>
+#include <openvino/core/node_output.hpp>
+#include <openvino/op/add.hpp>
+#include <openvino/op/concat.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/multiply.hpp>
+#include <openvino/op/reshape.hpp>
+#include <openvino/op/shape_of.hpp>
+#include <openvino/op/slice.hpp>
+#include <openvino/op/split.hpp>
+#include <openvino/op/subtract.hpp>
+#include <openvino/op/unsqueeze.hpp>
+#include <vector>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_rope(const NodeContext & context) {
+    num_inputs_check(context, 2, 3);
+
+    int op_case = context.get_op_case();
+
+    ov::Output<Node> res;
+
+    auto data_node = context.get_input(0).get_node_shared_ptr();
+    auto output_shape = context.get_output_shape().to_shape();
+    int32_t * op_params = context.get_output_op_params();
+
+    Output<Node> cos_theta_node;
+    Output<Node> sin_theta_node;
+    if (context.has_input("rope_cos")) {
+        cos_theta_node = context.get_input("rope_cos");
+        sin_theta_node = context.get_input("rope_sin");
+    } else {
+        auto inp_pos = context.get_input(1).get_node_shared_ptr();
+        std::shared_ptr<ov::Node> rope_freqs_weight;
+        if (context.get_input_size() == 3) {
+            rope_freqs_weight = context.get_input(2).get_node_shared_ptr();
+        }
+        auto sin_cos = make_sin_cos(op_params, inp_pos, rope_freqs_weight);
+        sin_theta_node = sin_cos.first;
+        cos_theta_node = sin_cos.second;
+    }
+
+    if (op_case == 2) {
+        // The input comes from a VIEW
+        int slice_len = output_shape[2] * output_shape[3];
+        data_node = process_view_input(context, 0, slice_len).get_node_shared_ptr();
+        if (context.is_stateful()) {
+            auto data_shape = ov::op::v0::Constant::create(
+                ov::element::i64, {3}, std::vector<int64_t>{-1, (int64_t) output_shape[2], (int64_t) output_shape[3]});
+            data_node = std::make_shared<ov::op::v1::Reshape>(data_node, data_shape, false);
+        } else {
+            auto data_shape = ov::op::v0::Constant::create(
+                ov::element::i64, {4}, std::vector<int64_t>{1, -1, (int64_t) output_shape[2], (int64_t) output_shape[3]});
+            data_node = std::make_shared<ov::op::v1::Reshape>(data_node, data_shape, false);
+        }
+    }
+
+    const int mode = op_params[2];
+    constexpr int ROPE_TYPE_NORMAL = 0;
+    constexpr int ROPE_TYPE_NEOX = 2;
+
+    if (mode == ROPE_TYPE_NORMAL) {
+        auto neg_one = ov::op::v0::Constant::create(ov::element::i64, {1}, {-1});
+        auto zero = ov::op::v0::Constant::create(ov::element::i64, {1}, {0});
+        auto one = ov::op::v0::Constant::create(ov::element::i64, {1}, {1});
+        auto two = ov::op::v0::Constant::create(ov::element::i64, {1}, {2});
+        auto end = ov::op::v0::Constant::create(ov::element::i64, {1}, {output_shape[3]});
+        Output<Node> even_slice;
+        Output<Node> odd_slice;
+        int32_t unsqueeze_dim = context.is_stateful() ? 3 : 4;
+        even_slice = std::make_shared<ov::op::v8::Slice>(data_node, zero, end, two, neg_one);
+        odd_slice = std::make_shared<ov::op::v8::Slice>(data_node, one, end, two, neg_one);
+
+        Output<Node> first_half =
+            std::make_shared<ov::op::v1::Subtract>(std::make_shared<ov::op::v1::Multiply>(even_slice, cos_theta_node),
+                                                   std::make_shared<ov::op::v1::Multiply>(odd_slice, sin_theta_node));
+        Output<Node> second_half =
+            std::make_shared<ov::op::v1::Add>(std::make_shared<ov::op::v1::Multiply>(even_slice, sin_theta_node),
+                                              std::make_shared<ov::op::v1::Multiply>(odd_slice, cos_theta_node));
+
+        first_half = std::make_shared<ov::op::v0::Unsqueeze>(first_half,
+                                                             ov::op::v0::Constant::create(ov::element::i64, {1}, {unsqueeze_dim}));
+        second_half = std::make_shared<ov::op::v0::Unsqueeze>(second_half,
+                                                              ov::op::v0::Constant::create(ov::element::i64, {1}, {unsqueeze_dim}));
+        auto stack = std::make_shared<ov::op::v0::Concat>(OutputVector{first_half, second_half}, unsqueeze_dim);
+
+        auto data_shape = ov::op::v0::Constant::create(
+            ov::element::i64, {4}, std::vector<int64_t>{1, -1, (int64_t) output_shape[2], (int64_t) output_shape[3]});
+        res = std::make_shared<ov::op::v1::Reshape>(stack, data_shape, false);
+    } else if (mode == ROPE_TYPE_NEOX) {
+        auto data_split = std::make_shared<ov::op::v1::Split>(
+            data_node, ov::op::v0::Constant::create(ov::element::i64, ov::Shape{}, {-1}), 2);
+        Output<Node> slice_data_node_0 = data_split->outputs()[0];
+        Output<Node> slice_data_node_1 = data_split->outputs()[1];
+
+        auto first_half_node = std::make_shared<ov::op::v1::Subtract>(
+            std::make_shared<ov::op::v1::Multiply>(slice_data_node_0, cos_theta_node),
+            std::make_shared<ov::op::v1::Multiply>(slice_data_node_1, sin_theta_node));
+
+        auto second_half_node = std::make_shared<ov::op::v1::Add>(
+            std::make_shared<ov::op::v1::Multiply>(slice_data_node_0, sin_theta_node),
+            std::make_shared<ov::op::v1::Multiply>(slice_data_node_1, cos_theta_node));
+
+        res = std::make_shared<ov::op::v0::Concat>(ov::OutputVector{first_half_node, second_half_node}, -1);
+    }
+
+    return rename_outputs_with_suffix({res}, context.get_name());
+}
+
+}  // namespace op
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

data/ext/sources/ggml/src/ggml-openvino/openvino/op/scale.cpp

@@ -0,0 +1,41 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <openvino/op/add.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/multiply.hpp>
+#include <vector>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_scale(const NodeContext & context) {
+    num_inputs_check(context, 1, 1);
+
+    float scale;
+    float bias;
+    memcpy(&scale, (float *) context.get_output_op_params() + 0, sizeof(float));
+    memcpy(&bias, (float *) context.get_output_op_params() + 1, sizeof(float));
+
+    auto scale_node = std::make_shared<ov::op::v0::Constant>(ov::element::f32, ov::Shape{}, std::vector<float>{scale});
+    auto scaled = std::make_shared<ov::op::v1::Multiply>(context.get_input(0), scale_node);
+
+    std::shared_ptr<ov::Node> res;
+    if (bias != 0.0f) {
+        auto bias_node =
+            std::make_shared<ov::op::v0::Constant>(ov::element::f32, ov::Shape{}, std::vector<float>{bias});
+        res = std::make_shared<ov::op::v1::Add>(scaled, bias_node);
+    } else {
+        res = scaled;
+    }
+
+    return rename_outputs_with_suffix({res}, context.get_name());
+}
+
+}  // namespace op
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov