whispercpp 1.3.5 → 1.3.7

data/ext/sources/ggml/src/ggml-openvino/openvino/translate_session.cpp

@@ -0,0 +1,317 @@
+#include "translate_session.h"
+
+#include "ggml-openvino/openvino/node_context.h"
+#include "ggml-openvino/openvino/utils.h"
+#include "input_model.h"
+#include "pass/mark_decompression_convert_constant_folding.h"
+#include "pass/squeeze_matmul.h"
+#include "rt_info/weightless_caching_attributes.hpp"
+
+#include <cstdint>
+#include <cstdlib>
+#include <map>
+#include <memory>
+#include <openvino/core/node.hpp>
+#include <openvino/core/preprocess/pre_post_process.hpp>
+#include <openvino/op/add.hpp>
+#include <openvino/op/broadcast.hpp>
+#include <openvino/op/concat.hpp>
+#include <openvino/op/convert.hpp>
+#include <openvino/op/convert_like.hpp>
+#include <openvino/op/cos.hpp>
+#include <openvino/op/divide.hpp>
+#include <openvino/op/gather.hpp>
+#include <openvino/op/multiply.hpp>
+#include <openvino/op/parameter.hpp>
+#include <openvino/op/range.hpp>
+#include <openvino/op/reshape.hpp>
+#include <openvino/op/result.hpp>
+#include <openvino/op/sin.hpp>
+#include <openvino/op/slice.hpp>
+#include <openvino/op/squeeze.hpp>
+#include <openvino/op/strided_slice.hpp>
+#include <openvino/op/transpose.hpp>
+#include <openvino/op/unsqueeze.hpp>
+#include <openvino/pass/constant_folding.hpp>
+#include <openvino/pass/make_stateful.hpp>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+
+using namespace ov::op;
+
+namespace {
+
+ov::pass::MakeStateful::ParamResPairs get_kv_param_res_pairs(
+    const std::shared_ptr<ov::Model> & model,
+    const std::map<std::string, std::string> & kv_param_res_names) {
+    ov::pass::MakeStateful::ParamResPairs pairs;
+    const auto & params = model->get_parameters();
+    const auto & results = model->get_results();
+
+    for (const auto & param_res : kv_param_res_names) {
+        const auto & param_name = param_res.first;
+        const auto & res_name = param_res.second;
+
+        auto param_it = std::find_if(params.begin(), params.end(), [&](const std::shared_ptr<v0::Parameter> & node) {
+            return node->get_friendly_name() == param_name;
+        });
+
+        OPENVINO_ASSERT(param_it != params.end(), "The tensor name ", param_name,
+                        " is not associated with any of "
+                        "Parameters in the network.");
+
+        auto res_it = std::find_if(results.begin(), results.end(), [&](const std::shared_ptr<v0::Result> & node) {
+            return node->get_friendly_name() == res_name;
+        });
+
+        OPENVINO_ASSERT(res_it != results.end(), "The tensor name ", res_name,
+                        " is not associated with any of "
+                        "Results in the network.");
+
+        std::shared_ptr<ov::op::v0::Parameter> param = *param_it;
+        std::shared_ptr<ov::op::v0::Result> res = *res_it;
+        pairs.emplace_back(param, res);
+    }
+    return pairs;
+}
+
+void add_sliced_mask(TensorMap & tensor_map, GgmlDecoder & ggml_model_decoder) {
+
+    auto create_sliced_mask = [&](const std::string & mask_name, const std::string & sliced_name, bool is_static) {
+        if ((tensor_map.find(mask_name) != tensor_map.end()) &&
+            (tensor_map.find("token_len_per_seq") != tensor_map.end())) {
+            auto token_len_per_seq = tensor_map.at("token_len_per_seq").get_node_shared_ptr();
+            auto mask = tensor_map.at(mask_name).get_node_shared_ptr();
+            std::shared_ptr<ov::Node> mask_sliced;
+            if (is_static) {
+                mask_sliced = mask;
+            } else if (ggml_model_decoder.is_stateful()) {
+                auto zero_2d = ov::op::v0::Constant::create(ov::element::i64, {2}, {0,0});
+                auto one_2d = ov::op::v0::Constant::create(ov::element::i64, {2}, {1,1});
+                auto zero_1d = ov::op::v0::Constant::create(ov::element::i64, {1}, {0});
+                auto three_1d = ov::op::v0::Constant::create(ov::element::i64, {1}, {3});
+                auto neg_one_1d = ov::op::v0::Constant::create(ov::element::i64, {1}, {-1});
+                auto axes = ov::op::v0::Constant::create(ov::element::i64, {2}, {-2,-1});
+                auto inp_pos = tensor_map.at("inp_pos").get_node_shared_ptr();
+                auto gather_inp_pos = std::make_shared<ov::op::v8::Gather>(inp_pos, neg_one_1d, three_1d);
+                auto reshaped_inp_pos = std::make_shared<ov::op::v1::Reshape>(gather_inp_pos, ov::op::v0::Constant::create(ov::element::i64, {1}, {1}), false);
+                auto inp_pos_incremented = std::make_shared<ov::op::v1::Add>(reshaped_inp_pos, ov::op::v0::Constant::create(ov::element::i32, ov::Shape{1}, {1}));
+                auto stop = std::make_shared<ov::op::v0::Concat>(ov::OutputVector{token_len_per_seq, std::make_shared<v1::ConvertLike>(inp_pos_incremented, token_len_per_seq)}, 0);
+                mask_sliced =
+                    std::make_shared<ov::op::v8::Slice>(mask, zero_2d, stop, one_2d, axes);
+                mask_sliced = std::make_shared<ov::op::v0::Convert>(mask_sliced, ov::element::f16);
+                mask_sliced->set_friendly_name(sliced_name);
+            } else {
+                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});
+                mask_sliced = std::make_shared<ov::op::v8::Slice>(mask, zero, token_len_per_seq, one, two);
+                mask_sliced = std::make_shared<ov::op::v0::Convert>(mask_sliced, ov::element::f16);
+                mask_sliced->set_friendly_name(sliced_name);
+            }
+            tensor_map.insert({sliced_name, mask_sliced->output(0)});
+        }
+    };
+
+    create_sliced_mask("self_kq_mask", "KQ_mask_sliced", ggml_model_decoder.is_static());
+    create_sliced_mask("self_kq_mask_swa", "KQ_mask_swa_sliced", ggml_model_decoder.is_static());
+}
+
+void add_rope_sin_cos(TensorMap & tensor_map, GgmlDecoder & ggml_model_decoder) {
+    int32_t * rope_params = ggml_model_decoder.get_rope_params();
+    if (tensor_map.find("inp_pos") == tensor_map.end() || rope_params == nullptr) {
+        return;
+    }
+    auto inp_pos = tensor_map.at("inp_pos").get_node_shared_ptr();
+    std::shared_ptr<ov::Node> rope_freqs_weight;
+    if (tensor_map.find("rope_freqs.weight") != tensor_map.end()) {
+        rope_freqs_weight = tensor_map.at("rope_freqs.weight").get_node_shared_ptr();
+    }
+
+    auto sin_cos = make_sin_cos(rope_params, inp_pos, rope_freqs_weight);
+    auto sin_theta = sin_cos.first;
+    auto cos_theta = sin_cos.second;
+
+    cos_theta.get_node_shared_ptr()->set_friendly_name("rope_cos");
+    sin_theta.get_node_shared_ptr()->set_friendly_name("rope_sin");
+    tensor_map.insert({"rope_cos", cos_theta});
+    tensor_map.insert({"rope_sin", sin_theta});
+}
+
+// Create common patterns
+void preprocess(TensorMap & tensor_map, GgmlDecoder & ggml_model_decoder) {
+    add_sliced_mask(tensor_map, ggml_model_decoder);
+    add_rope_sin_cos(tensor_map, ggml_model_decoder);
+}
+
+}  // namespace
+
+TranslateSession::TranslateSession(const frontend::InputModel::Ptr & input_model,
+                                   const std::unordered_map<std::string, CreatorFunction> & translator_map,
+                                   bool naive) :
+    m_input_model(input_model),
+    m_translator_map(translator_map),
+    m_ov_model(nullptr),
+    m_naive(naive) {}
+
+std::shared_ptr<Model> TranslateSession::get_converted_model() {
+    if (m_ov_model) {
+        return m_ov_model;
+    }
+    m_ov_model = translate_graph(m_input_model);
+    return m_ov_model;
+}
+
+std::shared_ptr<Model> TranslateSession::translate_graph(const frontend::InputModel::Ptr & input_model) {
+    ov::ParameterVector params;
+    ov::ResultVector results;
+    auto tensor_map = std::make_shared<TensorMap>();
+    std::shared_ptr<Model> resulting_model;
+
+    const auto & ggml_model = std::dynamic_pointer_cast<InputModel>(input_model);
+    std::shared_ptr<GgmlDecoder> ggml_model_decoder = ggml_model->get_model_decoder();
+
+    for (const auto & it : ggml_model_decoder->get_model_inputs()) {
+        params.push_back(std::dynamic_pointer_cast<ov::op::v0::Parameter>(it.second));
+        (*tensor_map)[it.first] = it.second;
+    }
+
+    for (const auto & it : ggml_model_decoder->get_model_extra_inputs()) {
+        if (std::dynamic_pointer_cast<ov::op::v0::Parameter>(it.second)) {
+            params.push_back(std::dynamic_pointer_cast<ov::op::v0::Parameter>(it.second));
+        }
+        (*tensor_map)[it.first] = it.second;
+    }
+
+    for (const auto & it : ggml_model_decoder->get_model_weights()) {
+        (*tensor_map)[it.first] = it.second;
+    }
+
+    auto node_visitor = [&](std::shared_ptr<GgmlDecoder> decoder, int node_idx) {
+        auto operation_type = decoder->get_op_type(node_idx);
+        if (operation_type == "GGML_OP_NONE") {
+            return;
+        }
+
+        ov::OutputVector converted_outputs;
+        auto it = m_translator_map.find(operation_type);
+        FRONT_END_OP_CONVERSION_CHECK(it != m_translator_map.end(), "Translation for operation type ", operation_type,
+                                      " is not implemented.");
+        NodeContext node_context(decoder, tensor_map, node_idx, this);
+        converted_outputs = it->second(node_context);
+
+        const auto & node_output_names = decoder->get_output_names(node_idx);
+        FRONT_END_OP_CONVERSION_CHECK(node_output_names.size() == converted_outputs.size(), "Number of ",
+                                      operation_type, " outputs greater than number of converted outputs, which are ",
+                                      node_output_names.size(), " and ", converted_outputs.size(), " respectively.");
+
+        for (size_t i = 0; i < node_output_names.size(); ++i) {
+            auto output_name = node_output_names[i];
+            if (i < converted_outputs.size() && converted_outputs[i].get_node_shared_ptr() != nullptr) {
+                (*tensor_map)[output_name] = converted_outputs[i];
+            }
+        }
+    };
+
+    if (!m_naive) {
+        preprocess(*tensor_map, *ggml_model_decoder);
+    }
+    ggml_model_decoder->visit_subgraph(node_visitor);
+
+    for (const auto & name : ggml_model_decoder->get_model_output_names()) {
+        FRONT_END_GENERAL_CHECK(tensor_map->find(name) != tensor_map->end(),
+                                "Output name not found in tensor map: ", name);
+        auto result = std::make_shared<v0::Result>(tensor_map->at(name));
+        result->set_friendly_name(name);
+        results.push_back(result);
+    }
+
+    ov::ParameterVector used_params;
+    for (const auto & param : params) {
+        if (!param->output(0).get_target_inputs().empty()) {
+            used_params.push_back(param);
+        }
+    }
+    // if (auto diff = params.size() - used_params.size()) {
+    //     GGML_LOG_INFO("%zu parameters are not used in the model.", diff);
+    // }
+    resulting_model = std::make_shared<Model>(results, used_params);
+
+    apply_transformations(resulting_model);
+
+    // Set WeightlessCacheAttribute on large constants to avoid unnecessary memory copies
+    // in the NPUW plugin. Without this attribute, NPUW's LazyTensor constructor
+    // (lazy_tensor.cpp, op::Const::Const) will memcpy every constant "in case export
+    // occurs", doubling memory usage per compile_model call.
+    //
+    // The bin_offset field serves as a unique key (not a real file offset) — this is
+    // the same convention the GPU plugin uses for non-IR models (see
+    // Plugin::set_weightless_cache_attributes in intel_gpu/src/plugin/plugin.cpp).
+    // Each constant must have a distinct bin_offset, otherwise GPU's weightless cache
+    // import will map multiple constants to the same data.
+    //
+    // Small constants (< 16 elements) are excluded since they may be introduced by
+    // optimization patterns and the overhead is negligible.
+    size_t offset = 0;
+    for (auto & node : resulting_model->get_ordered_ops()) {
+        if (auto cnst = ov::as_type_ptr<ov::op::v0::Constant>(node);
+            cnst && cnst->get_byte_size() / cnst->get_element_type().size() >= 16) {
+            auto & rt_info = cnst->get_rt_info();
+            if (rt_info.find(ov::WeightlessCacheAttribute::get_type_info_static()) == rt_info.end()) {
+                rt_info[ov::WeightlessCacheAttribute::get_type_info_static()] =
+                    ov::WeightlessCacheAttribute(cnst->get_byte_size(), offset++, cnst->get_element_type());
+            }
+        }
+    }
+    return resulting_model;
+}
+
+std::shared_ptr<Model> TranslateSession::apply_transformations(std::shared_ptr<Model> model) {
+    auto ggml_model_decoder = std::dynamic_pointer_cast<InputModel>(m_input_model)->get_model_decoder();
+    {
+        ov::pass::Manager manager;
+        manager.set_per_pass_validation(true);
+        manager.register_pass<ov::pass::MarkCompressedFloatConstants>();
+
+        if (ggml_model_decoder->is_stateful()) {
+            const auto kv_param_res_names = ggml_model_decoder->get_kv_param_res_names();
+            const auto kv_param_res_pairs = get_kv_param_res_pairs(model, kv_param_res_names);
+            manager.register_pass<ov::pass::MakeStateful>(kv_param_res_pairs);
+        }
+
+        if (ggml_model_decoder->is_static()) {
+            manager.register_pass<pass::SqueezeMatmul>();
+        }
+        manager.run_passes(model);
+        if (ggml_model_decoder->is_stateful()) {
+            auto output_names = ggml_model_decoder->get_model_output_names();
+            std::map<std::string, int> model_output_indexes;
+            for (size_t i=0; i<output_names.size(); i++) {
+                model_output_indexes.insert(std::make_pair(output_names[i], i));
+            }
+            ov::preprocess::PrePostProcessor ppp(model);
+            for (size_t i=0; i<model->get_output_size(); i++) {
+                auto output_friendly_name = model->output(i).get_node_shared_ptr()->get_friendly_name();
+                auto output_id = model_output_indexes[output_friendly_name];
+                auto model_output_shape = model->output(i).get_partial_shape();
+                auto decoder_output_shape = ggml_model_decoder->get_output_shape(output_id);
+                if (model_output_shape.rank().is_static() && decoder_output_shape.rank().is_static()
+                    && model_output_shape.rank().get_length() + 1 == decoder_output_shape.rank().get_length()
+                    && decoder_output_shape[0].is_static() && decoder_output_shape[0].get_length() == 1) {
+                    ppp.output(i).postprocess().custom([](const ov::Output<ov::Node>& node) {
+                        auto axes = ov::op::v0::Constant::create(ov::element::i32, ov::Shape{1}, {0});
+                        return std::make_shared<ov::op::v0::Unsqueeze>(node, axes);
+                    });
+                }
+            }
+            model = ppp.build();
+        }
+    }
+    return model;
+}
+
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

data/ext/sources/ggml/src/ggml-openvino/openvino/translate_session.h

@@ -0,0 +1,28 @@
+#pragma once
+
+#include "input_model.h"
+#include "node_context.h"
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+
+class TranslateSession {
+public:
+    TranslateSession(const frontend::InputModel::Ptr& input_model,
+                     const std::unordered_map<std::string, CreatorFunction>& translator_map, bool naive = false);
+
+    std::shared_ptr<Model> get_converted_model();
+    std::shared_ptr<Model> translate_graph(const frontend::InputModel::Ptr& input_model);
+
+private:
+    std::shared_ptr<Model> apply_transformations(std::shared_ptr<Model> model);
+    const frontend::InputModel::Ptr m_input_model;
+    const std::unordered_map<std::string, CreatorFunction>& m_translator_map;
+    std::shared_ptr<Model> m_ov_model;
+    bool m_naive;
+};
+
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

data/ext/sources/ggml/src/ggml-openvino/openvino/utils.cpp

@@ -0,0 +1,257 @@
+#include "utils.h"
+
+#include "ggml-impl.h"
+
+#include <cmath>
+#include <cstddef>
+#include <ctime>
+#include <memory>
+#include <openvino/op/add.hpp>
+#include <openvino/op/clamp.hpp>
+#include <openvino/op/convert.hpp>
+#include <openvino/op/cos.hpp>
+#include <openvino/op/divide.hpp>
+#include <openvino/op/gather.hpp>
+#include <openvino/op/maximum.hpp>
+#include <openvino/op/multiply.hpp>
+#include <openvino/op/reshape.hpp>
+#include <openvino/op/shape_of.hpp>
+#include <openvino/op/sin.hpp>
+#include <openvino/op/squeeze.hpp>
+#include <openvino/op/subtract.hpp>
+#include <openvino/op/transpose.hpp>
+#include <string>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+
+std::string getCurrentTime() {
+    std::time_t now = std::time(nullptr);
+    char buf[100];
+    std::strftime(buf, sizeof(buf), "%Y-%m-%d %H:%M:%S", std::localtime(&now));
+    return buf;
+}
+
+void num_inputs_check(const NodeContext & context, size_t min_inputs, size_t max_inputs) {
+    auto input_size = context.get_input_size();
+    FRONT_END_OP_CONVERSION_CHECK(input_size >= min_inputs, "Got less inputs than expected");
+    FRONT_END_OP_CONVERSION_CHECK(input_size <= max_inputs, "Got more inputs than expected");
+}
+
+int non_cont_dim(std::vector<size_t> ne, std::vector<size_t> nb) {
+    int dim = nb.size() - 1;
+    size_t bytes = nb[dim];
+    for (int i = dim; i > 0; i--) {
+        bytes *= ne[i];
+        if (bytes != nb[i - 1]) {
+            return i;
+        }
+    }
+    return 0;
+}
+
+std::shared_ptr<ov::Node> get_dimensions(const std::shared_ptr<ov::op::v3::ShapeOf> & shape,
+                                         const std::vector<int> & dims) {
+    using namespace ov::op;
+    const auto zero = v0::Constant::create(ov::element::i32, ov::Shape{}, {0});
+    const auto dims_const = v0::Constant::create(ov::element::i32, ov::Shape{dims.size()}, dims);
+    return std::make_shared<v8::Gather>(shape, dims_const, zero);
+}
+
+std::shared_ptr<ov::Node> get_dimensions(const std::shared_ptr<ov::Node> & node, const std::vector<int> & dims) {
+    return get_dimensions(std::make_shared<ov::op::v3::ShapeOf>(node), dims);
+}
+
+OutputVector rename_outputs_with_suffix(const OutputVector & outputs, const std::string & suffix) {
+    for (const auto & output : outputs) {
+        auto node = output.get_node_shared_ptr();
+        std::string name = node->get_friendly_name();
+        name += "_";
+        name += suffix;
+        node->set_friendly_name(name);
+        // std::cout << name << "  " << output.get_partial_shape() << std::endl;
+    }
+    return outputs;
+}
+
+namespace {
+ov::Output<ov::Node> rope_yarn_ramp_mix(int n_dims, const float corr_dims[2], float ext_factor) {
+    int half_n_dims = n_dims / 2;
+    std::vector<float> dim_ids_vec(half_n_dims);
+    std::iota(dim_ids_vec.begin(), dim_ids_vec.end(), 0);
+    auto dim_ids = ov::op::v0::Constant::create(ov::element::f32, Shape{1, 1, 1, (size_t) half_n_dims}, dim_ids_vec);
+    auto corr_low = ov::op::v0::Constant::create(ov::element::f32, Shape{1, 1, 1, 1}, {corr_dims[0]});
+    auto corr_high = ov::op::v0::Constant::create(ov::element::f32, Shape{1, 1, 1, 1}, {corr_dims[1]});
+    auto denom = std::make_shared<ov::op::v1::Maximum>(
+        std::make_shared<ov::op::v1::Subtract>(corr_high, corr_low),
+        ov::op::v0::Constant::create(ov::element::f32, Shape{1, 1, 1, 1}, {0.001f}));
+    auto ramp_y =
+        std::make_shared<ov::op::v1::Divide>(std::make_shared<ov::op::v1::Subtract>(dim_ids, corr_low), denom);
+    auto ramp_clamped = std::make_shared<ov::op::v0::Clamp>(ramp_y, 0.0f, 1.0f);
+    // rope_yarn_ramp returns (1 - clamp(y)), so invert before scaling
+    auto one = ov::op::v0::Constant::create(ov::element::f32, Shape{1, 1, 1, 1}, {1.0f});
+    auto ramp_inverted = std::make_shared<ov::op::v1::Subtract>(one, ramp_clamped);
+    auto ext_factor_node = ov::op::v0::Constant::create(ov::element::f32, Shape{}, {ext_factor});
+    auto ramp_mix = std::make_shared<ov::op::v1::Multiply>(ramp_inverted, ext_factor_node);
+    return ramp_mix;
+}
+
+float ggml_rope_yarn_corr_dim(int n_dims, int n_ctx_orig, float n_rot, float base) {
+#ifndef M_PI
+#    define M_PI 3.14159265358979323846
+#endif
+    return n_dims * logf(n_ctx_orig / (n_rot * 2 * (float) M_PI)) / (2 * logf(base));
+}
+
+void ggml_rope_yarn_corr_dims(int n_dims,
+                              int n_ctx_orig,
+                              float freq_base,
+                              float beta_fast,
+                              float beta_slow,
+                              float dims[2]) {
+    float start = floorf(ggml_rope_yarn_corr_dim(n_dims, n_ctx_orig, beta_fast, freq_base));
+    float end = ceilf(ggml_rope_yarn_corr_dim(n_dims, n_ctx_orig, beta_slow, freq_base));
+    dims[0] = std::max(0.0f, start);
+    dims[1] = std::min(static_cast<float>(n_dims - 1), end);
+}
+}  // namespace
+
+std::pair<ov::Output<Node>, ov::Output<Node>> make_sin_cos(int32_t * rope_params,
+                                                           std::shared_ptr<ov::Node> inp_pos,
+                                                           std::shared_ptr<ov::Node> rope_freqs_weight,
+                                                           bool imrope,
+                                                           bool stateful) {
+    if (stateful) {
+        inp_pos = std::make_shared<ov::op::v0::Squeeze>(inp_pos, ov::op::v0::Constant::create(ov::element::i64, {1}, {0}));
+        inp_pos = std::make_shared<ov::op::v0::Convert>(inp_pos, ov::element::f32);
+        auto pos_perm =
+            std::make_shared<ov::op::v0::Constant>(ov::element::i64, ov::Shape{3}, std::vector<int64_t>{2, 1, 0});
+        inp_pos = std::make_shared<ov::op::v1::Transpose>(inp_pos, pos_perm);
+    } else if (imrope) {
+        inp_pos = std::make_shared<ov::op::v0::Convert>(inp_pos, ov::element::f32);
+        auto pos_shape = ov::op::v0::Constant::create(ov::element::i64, ov::Shape{5}, {0, 0, 0, 4, -1});
+        inp_pos = std::make_shared<ov::op::v1::Reshape>(inp_pos, pos_shape, true);
+        auto pos_transpose_shape =
+            std::make_shared<ov::op::v0::Constant>(ov::element::i64, ov::Shape{5}, std::vector<int64_t>{0, 1, 2, 4, 3});
+        inp_pos = std::make_shared<ov::op::v1::Transpose>(inp_pos, pos_transpose_shape);
+    } else {
+        inp_pos = std::make_shared<ov::op::v0::Convert>(inp_pos, ov::element::f32);
+        auto pos_perm =
+            std::make_shared<ov::op::v0::Constant>(ov::element::i64, ov::Shape{4}, std::vector<int64_t>{0, 3, 1, 2});
+        inp_pos = std::make_shared<ov::op::v1::Transpose>(inp_pos, pos_perm);
+    }
+
+    float freq_base;
+    float freq_scale;
+    float ext_factor;
+    float attn_factor;
+    float beta_fast;
+    float beta_slow;
+    const int n_dims = rope_params[1];
+    const size_t n_dims_half = n_dims >> 1;
+    const int n_ctx_orig = rope_params[4];
+    memcpy(&freq_base, rope_params + 5, sizeof(float));
+    memcpy(&freq_scale, rope_params + 6, sizeof(float));
+    memcpy(&ext_factor, rope_params + 7, sizeof(float));
+    memcpy(&attn_factor, rope_params + 8, sizeof(float));
+    memcpy(&beta_fast, rope_params + 9, sizeof(float));
+    memcpy(&beta_slow, rope_params + 10, sizeof(float));
+
+    const float theta_scale = powf(freq_base, -2.0f / n_dims);
+
+    std::vector<float> factor(n_dims_half);
+
+    Output<Node> freq_factors;
+
+    Output<Node> theta;
+    float mscale = attn_factor;
+    if (imrope) {
+        std::vector<int64_t> gather_indices(n_dims_half);
+        for (size_t j = 0; j < n_dims_half; j++) {
+            gather_indices[j] = j % 3;
+            factor[j] = std::pow(theta_scale, j);
+        }
+        auto gather_indices_const =
+            std::make_shared<ov::op::v0::Constant>(ov::element::i64, ov::Shape{n_dims_half}, gather_indices);
+        auto gather_axis = ov::op::v0::Constant::create(ov::element::i32, ov::Shape{}, {4});
+        inp_pos = std::make_shared<ov::op::v8::Gather>(inp_pos, gather_indices_const, gather_axis);
+        auto factor_const = std::make_shared<ov::op::v0::Constant>(ov::element::f32, ov::Shape{n_dims_half}, factor);
+        theta = std::make_shared<ov::op::v1::Multiply>(inp_pos, factor_const);
+    } else {
+        float corr_dims[2];
+        ggml_rope_yarn_corr_dims(n_dims, n_ctx_orig, freq_base, beta_fast, beta_slow, corr_dims);
+        factor[0] = 1.0f;
+        for (size_t i = 1; i < factor.size(); i++) {
+            factor[i] = theta_scale * factor[i - 1];
+        }
+        if (stateful) {
+            freq_factors =
+                std::make_shared<ov::op::v0::Constant>(ov::element::f32, ov::Shape{1, 1, factor.size()}, factor);
+        } else {
+            freq_factors =
+                std::make_shared<ov::op::v0::Constant>(ov::element::f32, ov::Shape{1, 1, 1, factor.size()}, factor);
+        }
+        if (rope_freqs_weight) {
+            freq_factors = std::make_shared<ov::op::v1::Divide>(freq_factors, rope_freqs_weight);
+        }
+
+        auto theta_extrap = std::make_shared<ov::op::v1::Multiply>(freq_factors, inp_pos);
+        auto theta_interp = std::make_shared<ov::op::v1::Multiply>(
+            theta_extrap, ov::op::v0::Constant::create(ov::element::f32, {1}, {freq_scale}));
+
+        if (ext_factor == 0.0f) {
+            theta = theta_interp;
+        } else {
+            auto ramp_mix = rope_yarn_ramp_mix(n_dims, corr_dims, ext_factor);
+            Output<Node> one;
+            if (stateful) {
+                one = ov::op::v0::Constant::create(ov::element::f32, Shape{1, 1, 1}, {1.0f});
+            } else {
+                one = ov::op::v0::Constant::create(ov::element::f32, Shape{1, 1, 1, 1}, {1.0f});
+            }
+            auto one_minus_ramp = std::make_shared<ov::op::v1::Subtract>(one, ramp_mix);
+
+            theta = std::make_shared<ov::op::v1::Add>(std::make_shared<ov::op::v1::Multiply>(theta_interp, one_minus_ramp),
+                                                      std::make_shared<ov::op::v1::Multiply>(theta_extrap, ramp_mix));
+            mscale *= (1.0f + 0.1f * std::log(1.0f / freq_scale));
+        }
+    }
+
+    Output<Node> cos_theta = std::make_shared<ov::op::v0::Cos>(theta);
+    Output<Node> sin_theta = std::make_shared<ov::op::v0::Sin>(theta);
+
+    if (!imrope) {
+        auto mscale_node = ov::op::v0::Constant::create(ov::element::f32, Shape{}, {mscale});
+
+        cos_theta = std::make_shared<ov::op::v1::Multiply>(cos_theta, mscale_node);
+        sin_theta = std::make_shared<ov::op::v1::Multiply>(sin_theta, mscale_node);
+    }
+
+    return std::make_pair(sin_theta, cos_theta);
+}
+
+ov::Output<ov::Node> process_view_input(const NodeContext & context, int input_index, int slice_len) {
+    // Only works for VIEW operations that slice at the lowest dimension
+    // If the VIEW also reshape the result, `slice_len` should be provided
+    auto input = context.get_input(input_index);
+    auto * op_params = (size_t *) context.get_input_op_params(input_index);
+    auto src1_stride = context.get_input_stride(input_index);
+
+    int64_t split_addr = op_params[0] / src1_stride[3];
+    if (slice_len == 0) {
+        slice_len = context.get_input_shape(input_index)[3].get_length();
+    }
+    int64_t slice_end = split_addr + slice_len;
+
+    auto begin = ov::op::v0::Constant::create(ov::element::i64, {1}, {split_addr});
+    auto end = ov::op::v0::Constant::create(ov::element::i64, {1}, {slice_end});
+    auto stride = ov::op::v0::Constant::create(ov::element::i64, {1}, {1});
+    auto axes = ov::op::v0::Constant::create(ov::element::i64, {1}, {context.is_stateful() ? 2 : 3});
+    auto sliced = std::make_shared<ov::op::v8::Slice>(input, begin, end, stride, axes);
+    return sliced;
+}
+
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov