whispercpp 1.3.6 → 1.3.8

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

@@ -6,7 +6,7 @@
 #include <openvino/op/constant.hpp>
 #include <openvino/runtime/tensor.hpp>
 
-void unpack_32_4(const uint8_t* data, uint8_t* dst);
+void unpack_32_4(const uint8_t * data, uint8_t * dst);
 
 void extract_q4_0_data(const ggml_tensor * tensor,
                        ov::Tensor & weights_arr,
@@ -19,12 +19,18 @@ void extract_q4_1_data(const ggml_tensor * tensor,
                        ov::Tensor & zp_arr,
                        bool use_bias = false);
 
+void extract_q5_1_data(const ggml_tensor * tensor,
+                       ov::Tensor & weights_arr,
+                       ov::Tensor & scales_arr,
+                       ov::Tensor & zp_arr,
+                       bool use_bias = false);
+
 void extract_q8_0_data(const ggml_tensor * tensor,
                        ov::Tensor & weights_arr,
                        ov::Tensor & scales_arr,
                        ov::Tensor & zp_arr);
 
-void unpack_256_4(const uint8_t* data, uint8_t* dst);
+void unpack_256_4(const uint8_t * data, uint8_t * dst);
 
 void extract_q4_k_data(const ggml_tensor * tensor,
                        ov::Tensor & weights_arr,
@@ -145,8 +151,8 @@ namespace ov {
 namespace op {
 namespace util {
 // From <openvino>/src/common/transformations/include/transformations/utils/utils.hpp
-bool get_single_value(const std::shared_ptr<ov::op::v0::Constant>& const_node,
-                      float& value,
+bool get_single_value(const std::shared_ptr<ov::op::v0::Constant> & const_node,
+                      float & value,
                       bool check_value_range = true);
 }  // namespace util
 }  // namespace op

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

@@ -3,6 +3,8 @@
 #include <cstdint>
 #include <map>
 #include <openvino/core/node.hpp>
+#include <openvino/core/partial_shape.hpp>
+#include <openvino/core/shape.hpp>
 #include <openvino/frontend/decoder.hpp>
 #include <string>
 
@@ -12,22 +14,50 @@ namespace ggml {
 
 class GgmlDecoder : public DecoderBase {
 public:
-    virtual ov::Any get_attribute(const std::string& name) const = 0;
+    virtual ov::Any get_attribute(const std::string & name) const = 0;
 
-    virtual PartialShape get_input_shape(int node_idx, const std::string& name) const = 0;
+    virtual PartialShape get_input_shape(int node_idx, const std::string & name) const = 0;
 
-    virtual std::vector<size_t> get_input_stride(int node_idx, const std::string& name) const = 0;
+    virtual std::vector<size_t> get_input_stride(int node_idx, const std::string & name) const = 0;
 
-    virtual element::Type get_input_type(int node_idx, const std::string& name) const = 0;
+    virtual size_t get_view_input_size(int node_idx, const std::string & name) const = 0;
+
+    virtual size_t get_view_input_offset(int node_idx, const std::string & name, size_t view_index) const = 0;
+
+    virtual size_t get_view_input_src_offset(int node_idx, const std::string & name, size_t view_index) const = 0;
+
+    virtual std::vector<size_t> get_view_input_stride(int node_idx,
+                                                      const std::string & name,
+                                                      size_t view_index) const = 0;
+
+    virtual std::vector<size_t> get_view_input_src_stride(int node_idx,
+                                                          const std::string & name,
+                                                          size_t view_index) const = 0;
+
+    virtual Shape get_view_input_ggml_shape(int node_idx, const std::string & name, size_t view_index) const = 0;
+
+    virtual Shape get_view_input_src_ggml_shape(int node_idx, const std::string & name, size_t view_index) const = 0;
+
+    virtual PartialShape get_view_input_ov_shape(int node_idx, const std::string & name, size_t view_index) const = 0;
+
+    virtual PartialShape get_view_input_src_ov_shape(int node_idx,
+                                                     const std::string & name,
+                                                     size_t view_index) const = 0;
+
+    virtual std::string get_view_input_name(int node_idx, const std::string & name, size_t view_index) const = 0;
+
+    virtual std::string get_view_input_src_name(int node_idx, const std::string & name, size_t view_index) const = 0;
+
+    virtual element::Type get_input_type(int node_idx, const std::string & name) const = 0;
 
     virtual size_t get_input_size() const = 0;
 
     virtual size_t get_input_size(int node_idx) const = 0;
 
     virtual void get_input_node(size_t input_port_idx,
-                                std::string& producer_name,
-                                std::string& producer_output_port_name,
-                                size_t& producer_output_port_index) const = 0;
+                                std::string & producer_name,
+                                std::string & producer_output_port_name,
+                                size_t & producer_output_port_index) const = 0;
 
     virtual std::vector<std::string> get_input_names(int node_idx) const = 0;
 
@@ -35,30 +65,36 @@ public:
 
     virtual element::Type get_output_type(const int node_idx) const = 0;
 
-    virtual int32_t* get_input_op_params(int node_idx, const std::string& name) const = 0;
+    virtual std::vector<size_t> get_output_stride(int node_idx) const = 0;
+
+    virtual int32_t * get_input_op_params(int node_idx, const std::string & name) const = 0;
 
     virtual int32_t * get_output_op_params(int node_idx) const = 0;
 
+    virtual size_t get_output_op_offset(int node_idx) const = 0;
+
     virtual std::vector<std::string> get_output_names(int node_idx) const = 0;
 
-    virtual const std::string& get_op_type() const = 0;
+    virtual const std::string & get_op_type() const = 0;
 
-    virtual const std::string& get_op_type(int node_idx) const = 0;
+    virtual const std::string & get_op_type(int node_idx) const = 0;
 
-    virtual const std::string& get_op_name() const = 0;
+    virtual const std::string & get_op_name() const = 0;
 
-    virtual const std::string& get_op_name(int node_idx) const = 0;
+    virtual const std::string & get_op_name(int node_idx) const = 0;
 
     virtual void visit_subgraph(std::function<void(std::shared_ptr<GgmlDecoder>, int node_idx)> node_visitor) const = 0;
 
     virtual int get_op_case(int node_idx) const = 0;
 
-    virtual const std::map<std::string, std::shared_ptr<ov::Node>>& get_model_inputs() const = 0;
-    virtual const std::map<std::string, std::shared_ptr<ov::Node>>& get_model_extra_inputs() const = 0;
-    virtual const std::map<std::string, std::shared_ptr<ov::Node>>& get_model_weights() const = 0;
+    virtual const std::map<std::string, std::shared_ptr<ov::Node>> & get_model_inputs() const = 0;
+    virtual const std::map<std::string, std::shared_ptr<ov::Node>> & get_model_extra_inputs() const = 0;
+    virtual const std::map<std::string, std::shared_ptr<ov::Node>> & get_model_weights() const = 0;
     virtual std::vector<std::string> get_model_output_names() const = 0;
 
-    virtual int32_t* get_rope_params() const = 0;
+    virtual int32_t * get_rope_params() const = 0;
+
+    virtual bool has_mixed_rope_params() const = 0;
 
     virtual std::map<std::string, std::string> get_kv_param_res_names() const = 0;
 
@@ -66,7 +102,11 @@ public:
 
     virtual bool is_stateful() const = 0;
 
+    virtual bool is_splited_model() const = 0;
+
     virtual int is_swa_layer(int layer) const = 0;
+
+    virtual int32_t get_op_dynamic_dim(int node_idx) const = 0;
 };
 
 }  // namespace ggml

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

@@ -15,7 +15,7 @@ public:
     using Ptr = std::shared_ptr<FrontEnd>;
     FrontEnd();
 
-    static std::shared_ptr<Model> convert(const InputModel::Ptr& model, bool naive = false);
+    static std::shared_ptr<Model> convert(const InputModel::Ptr & model, bool naive = false);
 };
 
 }  // namespace ggml

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

@@ -1,9 +1,9 @@
 #pragma once
 
-#include <openvino/frontend/input_model.hpp>
-
 #include "decoder.h"
 
+#include <openvino/frontend/input_model.hpp>
+
 namespace ov {
 namespace frontend {
 namespace ggml {
@@ -16,9 +16,9 @@ class InputModel : public ov::frontend::InputModel {
     friend class ::ov::frontend::ggml::FrontEnd;
 
 public:
-    explicit InputModel(const std::shared_ptr<GgmlDecoder>& gdecoder);
+    explicit InputModel(const std::shared_ptr<GgmlDecoder> & gdecoder);
 
-    const std::shared_ptr<GgmlDecoder>& get_model_decoder() const;
+    const std::shared_ptr<GgmlDecoder> & get_model_decoder() const;
 
 private:
     std::shared_ptr<GgmlDecoder> m_decoder;

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

@@ -1,11 +1,11 @@
 #pragma once
 
+#include "decoder.h"
+
 #include <cstdint>
 #include <openvino/frontend/node_context.hpp>
 #include <string>
 
-#include "decoder.h"
-
 namespace ov {
 namespace frontend {
 namespace ggml {
@@ -16,28 +16,24 @@ typedef std::map<std::string, Output<Node>> TensorMap;
 
 class NodeContext : public frontend::NodeContext {
 public:
-    NodeContext(const std::shared_ptr<GgmlDecoder>& decoder,
-                std::shared_ptr<TensorMap>& tensor_map,
+    NodeContext(const std::shared_ptr<GgmlDecoder> & decoder,
+                std::shared_ptr<TensorMap> & tensor_map,
                 int node_idx,
-                TranslateSession* translate_session = nullptr)
-        : ov::frontend::NodeContext(decoder->get_op_type(node_idx)),
-          m_decoder(decoder),
-          m_tensor_map(tensor_map),
-          m_node_idx(node_idx),
-          m_translate_session(translate_session) {
+                TranslateSession * translate_session = nullptr) :
+        ov::frontend::NodeContext(decoder->get_op_type(node_idx)),
+        m_decoder(decoder),
+        m_tensor_map(tensor_map),
+        m_node_idx(node_idx),
+        m_translate_session(translate_session) {
         m_input_names = decoder->get_input_names(m_node_idx);
         m_output_names = decoder->get_output_names(m_node_idx);
     }
 
-    TranslateSession* get_translate_session() const {
-        return m_translate_session;
-    }
+    TranslateSession * get_translate_session() const { return m_translate_session; }
 
-    const std::vector<std::string>& get_input_names() const { return m_input_names; }
+    const std::vector<std::string> & get_input_names() const { return m_input_names; }
 
-    size_t get_input_size() const override {
-        return m_decoder->get_input_size(m_node_idx);
-    }
+    size_t get_input_size() const override { return m_decoder->get_input_size(m_node_idx); }
 
     ov::element::Type get_input_type(size_t index) const {
         return m_decoder->get_input_type(m_node_idx, m_input_names[index]);
@@ -55,42 +51,103 @@ public:
 
     PartialShape get_output_shape() const { return m_decoder->get_output_shape(m_node_idx); }
 
-    int32_t* get_input_op_params(size_t index) const {
+    int32_t * get_input_op_params(size_t index) const {
         return m_decoder->get_input_op_params(m_node_idx, m_input_names[index]);
     }
 
-    int32_t * get_output_op_params() const { return m_decoder->get_output_op_params(m_node_idx); }
+    size_t get_view_input_size(size_t index) const {
+        return m_decoder->get_view_input_size(m_node_idx, m_input_names[index]);
+    }
+
+    size_t get_view_input_offset(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_offset(m_node_idx, m_input_names[index], view_index);
+    }
 
-    ov::element::Type get_output_type() const {
-        return m_decoder->get_output_type(m_node_idx);
+    size_t get_view_input_src_offset(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_src_offset(m_node_idx, m_input_names[index], view_index);
     }
 
+    std::vector<size_t> get_view_input_stride(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_stride(m_node_idx, m_input_names[index], view_index);
+    }
+
+    std::vector<size_t> get_view_input_src_stride(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_src_stride(m_node_idx, m_input_names[index], view_index);
+    }
+
+    ov::Shape get_view_input_ggml_shape(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_ggml_shape(m_node_idx, m_input_names[index], view_index);
+    }
+
+    ov::Shape get_view_input_src_ggml_shape(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_src_ggml_shape(m_node_idx, m_input_names[index], view_index);
+    }
+
+    ov::PartialShape get_view_input_ov_shape(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_ov_shape(m_node_idx, m_input_names[index], view_index);
+    }
+
+    ov::PartialShape get_view_input_src_ov_shape(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_src_ov_shape(m_node_idx, m_input_names[index], view_index);
+    }
+
+    std::string get_view_input_name(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_name(m_node_idx, m_input_names[index], view_index);
+    }
+
+    std::string get_view_input_src_name(size_t index, size_t view_index) const {
+        return m_decoder->get_view_input_src_name(m_node_idx, m_input_names[index], view_index);
+    }
+
+    int32_t get_op_dynamic_dim() const { return m_decoder->get_op_dynamic_dim(m_node_idx); }
+
+    int32_t * get_output_op_params() const { return m_decoder->get_output_op_params(m_node_idx); }
+
+    size_t get_output_op_offset() const { return m_decoder->get_output_op_offset(m_node_idx); }
+
+    ov::element::Type get_output_type() const { return m_decoder->get_output_type(m_node_idx); }
+
+    std::vector<size_t> get_output_stride() const { return m_decoder->get_output_stride(m_node_idx); }
+
     Output<Node> get_input(int idx) const override {
+        // Check if this input is a VIEW
+        size_t view_input_size = m_decoder->get_view_input_size(m_node_idx, m_input_names[idx]);
+        if (view_input_size > 0) {
+            // This is a VIEW input, get the base tensor name (last element in the chain)
+            std::string base_name =
+                m_decoder->get_view_input_src_name(m_node_idx, m_input_names[idx], view_input_size - 1);
+            // Check if the VIEW has been resolved (translate_view produced a Slice)
+            auto view_it = m_tensor_map->find(m_input_names[idx]);
+            if (!base_name.empty() && view_it != m_tensor_map->end()) {
+                auto base_it = m_tensor_map->find(base_name);
+                if (base_it != m_tensor_map->end() &&
+                    view_it->second.get_node_shared_ptr() != base_it->second.get_node_shared_ptr()) {
+                    return view_it->second;
+                }
+                return base_it->second;
+            }
+            if (!base_name.empty()) {
+                return m_tensor_map->at(base_name);
+            }
+        }
+        // Not a VIEW or failed to get base name, use the original logic
         return m_tensor_map->at(m_input_names[idx]);
     }
 
-    Output<Node> get_input(const std::string& name) const override {
+    Output<Node> get_input(const std::string & name) const override {
         if (m_tensor_map->find(name) == m_tensor_map->end()) {
             throw std::runtime_error("'" + name + "' not found in tensor map.");
         }
         return m_tensor_map->at(name);
     }
 
-    bool has_input(const std::string& name) const {
-        return m_tensor_map->find(name) != m_tensor_map->end();
-    }
+    bool has_input(const std::string & name) const { return m_tensor_map->find(name) != m_tensor_map->end(); }
 
-    const std::string& get_name() const override {
-        return m_decoder->get_op_name(m_node_idx);
-    }
+    const std::string & get_name() const override { return m_decoder->get_op_name(m_node_idx); }
 
-    ov::Any get_attribute_as_any(const std::string& name) const override {
-        return m_decoder->get_attribute(name);
-    }
+    ov::Any get_attribute_as_any(const std::string & name) const override { return m_decoder->get_attribute(name); }
 
-    int get_op_case() const {
-        return m_decoder->get_op_case(m_node_idx);
-    }
+    int get_op_case() const { return m_decoder->get_op_case(m_node_idx); }
 
     bool is_static() const { return m_decoder->is_static(); }
 
@@ -98,14 +155,14 @@ public:
 
 private:
     std::shared_ptr<GgmlDecoder> m_decoder;
-    std::shared_ptr<TensorMap>& m_tensor_map;
+    std::shared_ptr<TensorMap> & m_tensor_map;
     int m_node_idx;
-    TranslateSession* m_translate_session;
+    TranslateSession * m_translate_session;
     std::vector<std::string> m_input_names;
     std::vector<std::string> m_output_names;
 };
 
-using CreatorFunction = std::function<ov::OutputVector(const ov::frontend::ggml::NodeContext&)>;
+using CreatorFunction = std::function<ov::OutputVector(const ov::frontend::ggml::NodeContext &)>;
 
 }  // namespace ggml
 }  // namespace frontend

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

@@ -0,0 +1,76 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <memory>
+#include <openvino/core/node.hpp>
+#include <openvino/core/node_output.hpp>
+#include <openvino/op/add.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/convert.hpp>
+#include <openvino/op/gather.hpp>
+#include <openvino/op/reshape.hpp>
+#include <openvino/op/shape_of.hpp>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+static ov::Output<ov::Node> reshape_add_id_input_to_2d(const ov::Output<ov::Node> & input,
+                                                       const ov::PartialShape & input_shape,
+                                                       const std::vector<int> & dims) {
+    const auto actual_shape = input.get_partial_shape();
+    if (actual_shape.rank().is_static() && actual_shape.rank().get_length() == 2) {
+        return input;
+    }
+
+    if (input_shape.rank().is_static() && input_shape.rank().get_length() == 2) {
+        return input;
+    }
+
+    auto shape = std::make_shared<ov::op::v3::ShapeOf>(input, ov::element::i64);
+    return std::make_shared<ov::op::v1::Reshape>(input, get_dimensions(shape, dims), false);
+}
+
+OutputVector translate_add_id(const NodeContext & context) {
+    num_inputs_check(context, 3, 3);
+
+    auto input = process_view_input_new(context, 0);
+    auto bias = process_view_input_new(context, 1);
+    auto ids = process_view_input_new(context, 2);
+
+    // OpenVINO uses reversed GGML dimensions:
+    //   input: [1, n_token, n_used, n_embd]
+    //   bias:  [1, 1, n_expert, n_embd]
+    //   ids:   [1, 1, n_token, n_used]
+    // Model bias constants may already be stored as [n_expert, n_embd].
+    bias = reshape_add_id_input_to_2d(bias, context.get_input_shape(1), {2, 3});
+    ids = reshape_add_id_input_to_2d(ids, context.get_input_shape(2), {2, 3});
+
+    if (ids.get_element_type() != ov::element::i32 && ids.get_element_type() != ov::element::i64) {
+        ids = std::make_shared<ov::op::v0::Convert>(ids, ov::element::i32);
+    }
+
+    auto gather_axis = ov::op::v0::Constant::create(ov::element::i32, ov::Shape{}, {0});
+    ov::Output<ov::Node> selected_bias = std::make_shared<ov::op::v8::Gather>(bias, ids, gather_axis);
+    selected_bias = std::make_shared<ov::op::v1::Reshape>(
+        selected_bias, std::make_shared<ov::op::v3::ShapeOf>(input, ov::element::i64), false);
+
+    if (selected_bias.get_element_type() != input.get_element_type()) {
+        selected_bias = std::make_shared<ov::op::v0::Convert>(selected_bias, input.get_element_type());
+    }
+
+    ov::Output<ov::Node> res = std::make_shared<ov::op::v1::Add>(input, selected_bias);
+    const auto output_type = context.get_output_type();
+    if (res.get_element_type() != output_type) {
+        res = std::make_shared<ov::op::v0::Convert>(res, output_type);
+    }
+
+    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/argsort.cpp

@@ -0,0 +1,47 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+#include "ggml.h"
+
+#include <openvino/frontend/exception.hpp>
+#include <openvino/op/constant.hpp>
+#include <openvino/op/squeeze.hpp>
+#include <openvino/op/topk.hpp>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_argsort(const NodeContext & context) {
+    num_inputs_check(context, 1, 1);
+
+    auto input = process_view_input_new(context, 0);
+
+    const int32_t order = context.get_output_op_params()[0];
+
+    ov::op::v11::TopK::Mode mode;
+    switch (order) {
+    case GGML_SORT_ORDER_ASC:
+        mode = ov::op::v11::TopK::Mode::MIN;
+        break;
+    case GGML_SORT_ORDER_DESC:
+        mode = ov::op::v11::TopK::Mode::MAX;
+        break;
+    default:
+        FRONT_END_OP_CONVERSION_CHECK(false, "Unsupported GGML_OP_ARGSORT order: ", order);
+    }
+
+    auto k = std::make_shared<ov::op::v0::Squeeze>(get_dimensions(input.get_node_shared_ptr(), {3}),
+                                                   ov::op::v0::Constant::create(ov::element::i64, {1}, {0}));
+
+    auto topk = std::make_shared<ov::op::v11::TopK>(input, k, 3, mode, ov::op::v11::TopK::SortType::SORT_VALUES,
+                                                    context.get_output_type(), false);
+
+    return rename_outputs_with_suffix({topk->output(1)}, context.get_name());
+}
+
+}  // namespace op
+}  // namespace ggml
+}  // namespace frontend
+}  // namespace ov

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

@@ -0,0 +1,33 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <cstring>
+#include <openvino/op/clamp.hpp>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_clamp(const NodeContext & context) {
+    num_inputs_check(context, 1, 1);
+
+    auto input = process_view_input_new(context, 0);
+
+    const int32_t * op_params = context.get_output_op_params();
+    FRONT_END_CHECK_IMPLEMENTED(op_params != nullptr, "CLAMP requires output op params");
+
+    float min;
+    float max;
+    std::memcpy(&min, reinterpret_cast<const float *>(op_params) + 0, sizeof(float));
+    std::memcpy(&max, reinterpret_cast<const float *>(op_params) + 1, sizeof(float));
+
+    auto res = std::make_shared<ov::op::v0::Clamp>(input, min, max);
+    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/concat.cpp

@@ -0,0 +1,48 @@
+#include "../node_context.h"
+#include "../op_table.h"
+#include "../utils.h"
+
+#include <memory>
+#include <openvino/frontend/exception.hpp>
+#include <openvino/op/concat.hpp>
+#include <openvino/op/convert.hpp>
+
+namespace ov {
+namespace frontend {
+namespace ggml {
+namespace op {
+
+OutputVector translate_concat(const NodeContext & context) {
+    num_inputs_check(context, 2, 2);
+
+    const int32_t * op_params = context.get_output_op_params();
+    FRONT_END_CHECK_IMPLEMENTED(op_params != nullptr, "CONCAT requires output op params");
+
+    const auto output_shape = context.get_output_shape();
+    FRONT_END_CHECK_IMPLEMENTED(output_shape.rank().is_static(), "CONCAT requires static output rank");
+
+    const auto rank = output_shape.rank().get_length();
+    const int32_t ggml_dim = op_params[0];
+    FRONT_END_CHECK_IMPLEMENTED(ggml_dim >= 0 && ggml_dim < rank, "CONCAT axis is out of range");
+
+    auto input_0 = process_view_input_new(context, 0);
+    auto input_1 = process_view_input_new(context, 1);
+    const auto output_type = context.get_output_type();
+
+    if (input_0.get_element_type() != output_type) {
+        input_0 = std::make_shared<ov::op::v0::Convert>(input_0, output_type);
+    }
+    if (input_1.get_element_type() != output_type) {
+        input_1 = std::make_shared<ov::op::v0::Convert>(input_1, output_type);
+    }
+
+    const auto axis = static_cast<int64_t>(rank - 1 - ggml_dim);
+    auto res = std::make_shared<ov::op::v0::Concat>(OutputVector{input_0, input_1}, axis);
+
+    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/cont.cpp

@@ -18,27 +18,19 @@ namespace op {
 OutputVector translate_cont(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, "Unsupported CONT case");
-
     auto src_shape = context.get_input_shape(0).to_shape();
     auto dst_shape = context.get_output_shape().to_shape();
-    ov::Output<Node> res;
 
-    if (op_case == 1) {
-        // The input comes from a PERMUTE
-        throw std::runtime_error("Code of this case might be outdated");
-        dst_shape[1] = -1;
-        res = std::make_shared<ov::op::v1::Reshape>(
-            context.get_input(0), ov::op::v0::Constant::create(ov::element::i64, {dst_shape.size()}, dst_shape), false);
-    } else if (op_case == 2) {
-        // The input comes from a TRANSPOSE
-        return {context.get_input(0)};
-    } else {
-        // The input comes from a VIEW
-        res = process_view_input(context, 0);
+    if (context.get_op_dynamic_dim() != -1) {
+        dst_shape[3 - context.get_op_dynamic_dim()] = -1;
     }
 
+    auto input = process_view_input_new(context, 0);
+
+    ov::Output<Node> res;
+    res = std::make_shared<ov::op::v1::Reshape>(
+        input, ov::op::v0::Constant::create(ov::element::i64, {dst_shape.size()}, dst_shape), false);
+
     return rename_outputs_with_suffix({res}, context.get_name());
 }
 

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

@@ -3,7 +3,9 @@
 #include "../utils.h"
 
 #include <memory>
+#include <openvino/op/constant.hpp>
 #include <openvino/op/convert.hpp>
+#include <openvino/op/reshape.hpp>
 
 namespace ov {
 namespace frontend {
@@ -11,7 +13,18 @@ namespace ggml {
 namespace op {
 
 OutputVector translate_cpy(const NodeContext & context) {
-    auto res = std::make_shared<ov::op::v0::Convert>(context.get_input(0), context.get_output_type());
+    auto input = process_view_input_new(context, 0);
+    auto input_shape = context.get_input_shape(0);
+    auto output_shape = context.get_output_shape();
+
+    // Non-cast CPY may need a reshape (e.g. [3,192,1,1] -> [576,1,1,1])
+    if (input_shape != output_shape) {
+        auto new_shape = ov::op::v0::Constant::create(
+            ov::element::i64, {static_cast<size_t>(output_shape.rank().get_length())}, output_shape.to_shape());
+        input = std::make_shared<ov::op::v1::Reshape>(input, new_shape, false);
+    }
+
+    auto res = std::make_shared<ov::op::v0::Convert>(input, context.get_output_type());
     return rename_outputs_with_suffix({res}, context.get_name());
 }