Merge branch 'master' into CVS-164224

Author: CuriousPanCake

docs/articles_en/documentation/openvino-ir-format/operation-sets/operation-specs/generation/random-uniform-8.rst

@@ -42,7 +42,7 @@ The result of Philox is calculated by applying a fixed number of *key* and *coun
 This implementation uses 4x32_10 version of Philox algorithm, where number of rounds = 10.
 
 Suppose we have *n* which determines *n*-th 4 elements of random sequence.
-In each round *key*, *counter* and *n* are splitted to pairs of uint32 values:
+In each round *key*, *counter* and *n* are split to pairs of uint32 values:
 
 .. math::
 
@@ -220,7 +220,8 @@ Example 3. *RandomUniform* output with ``global_seed`` = 80, ``op_seed`` = 100,
 
 Mersenne-Twister Algorithm Explanation:
 
-Link to the original paper Mersenne Twister: Mersenne twister: a 623-dimensionally equidistributed uniform pseudo-random number generator <https://dl.acm.org/doi/10.1145/272991.272995>__.
+| Link to the original paper Mersenne Twister:
+| `Mersenne twister: a 623-dimensionally equidistributed uniform pseudo-random number generator <https://dl.acm.org/doi/10.1145/272991.272995>`__.
 
 The Mersenne-Twister algorithm generates random numbers by initializing a state array with a seed and then iterating through a series of transformations.
 Suppose we have n which determines the n-th element of the random sequence.
@@ -250,11 +251,11 @@ Whenever all state values are 'used', a new state array is generated recursively
 
 .. math::
 
-   current_state = state[i]
-   next_state    = state[i+1] if i+1 <= 623 else state[0]
-   next_m_state  = state[i+m] if i+m <= 623 else state[i+m-623]
+   current\_state = state[i]
+   next\_state    = state[i+1] if i+1 <= 623 else state[0]
+   next\_m_state  = state[i+m] if i+m <= 623 else state[i+m-623]
 
-   twisted_state = (((current_state & 0x80000000) | (next_state & 0x7fffffff)) >> 1) ^ (next_state & 1 ? 0x9908b0df : 0)
+   twisted\_state = (((current_state & 0x80000000) | (next_state & 0x7fffffff)) >> 1) ^ (next_state & 1 ? 0x9908b0df : 0)
    state[i] = next_m_state ^ twisted_state
 
 where m is a constant.

docs/articles_en/get-started/install-openvino/install-openvino-apt.rst

@@ -56,21 +56,21 @@ Step 1: Set Up the OpenVINO Toolkit APT Repository
 
          .. code-block:: sh
 
-            echo "deb https://apt.repos.intel.com/openvino/2025 ubuntu24 main" | sudo tee /etc/apt/sources.list.d/intel-openvino-2025.list
+            echo "deb https://apt.repos.intel.com/openvino ubuntu24 main" | sudo tee /etc/apt/sources.list.d/intel-openvino.list
 
       .. tab-item:: Ubuntu 22
          :sync: ubuntu-22
 
          .. code-block:: sh
 
-            echo "deb https://apt.repos.intel.com/openvino/2025 ubuntu22 main" | sudo tee /etc/apt/sources.list.d/intel-openvino-2025.list
+            echo "deb https://apt.repos.intel.com/openvino ubuntu22 main" | sudo tee /etc/apt/sources.list.d/intel-openvino.list
 
       .. tab-item:: Ubuntu 20
          :sync: ubuntu-20
 
          .. code-block:: sh
 
-            echo "deb https://apt.repos.intel.com/openvino/2025 ubuntu20 main" | sudo tee /etc/apt/sources.list.d/intel-openvino-2025.list
+            echo "deb https://apt.repos.intel.com/openvino ubuntu20 main" | sudo tee /etc/apt/sources.list.d/intel-openvino.list
 
 
 3. Update the list of packages via the update command:
@@ -126,7 +126,7 @@ Step 2: Install OpenVINO Runtime Using the APT Package Manager
          .. code-block:: sh
 
 
-            sudo apt install openvino-2025.0.0
+            sudo apt install openvino-2025.1.0
 
 .. note::
 
@@ -199,7 +199,7 @@ To uninstall OpenVINO Runtime via APT, run the following command based on your n
 
       .. code-block:: sh
 
-         sudo apt autoremove openvino-2025.0.0
+         sudo apt autoremove openvino-2025.1.0
 
 
 What's Next?

docs/articles_en/get-started/install-openvino/install-openvino-yum.rst

@@ -25,25 +25,25 @@ Step 1: Set Up the Repository
 +++++++++++++++++++++++++++++
 
 
-1. Create a YUM repository file (``openvino-2025.repo``) in the ``/tmp`` directory as a normal user:
+1. Create a YUM repository file (``openvino.repo``) in the ``/tmp`` directory as a normal user:
 
    .. code-block:: sh
 
-      tee > /tmp/openvino-2025.repo << EOF
+      tee > /tmp/openvino.repo << EOF
       [OpenVINO]
-      name=Intel(R) Distribution of OpenVINO 2025
-      baseurl=https://yum.repos.intel.com/openvino/2025
+      name=Intel(R) Distribution of OpenVINO
+      baseurl=https://yum.repos.intel.com/openvino
       enabled=1
       gpgcheck=1
       repo_gpgcheck=1
       gpgkey=https://yum.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
       EOF
 
-2. Move the new ``openvino-2025.repo`` file to the YUM configuration directory, i.e. ``/etc/yum.repos.d``:
+2. Move the new ``openvino.repo`` file to the YUM configuration directory, i.e. ``/etc/yum.repos.d``:
 
    .. code-block:: sh
 
-      sudo mv /tmp/openvino-2025.repo /etc/yum.repos.d
+      sudo mv /tmp/openvino.repo /etc/yum.repos.d
 
 3. Verify that the new repository is set up properly.
 
@@ -93,7 +93,7 @@ Install OpenVINO Runtime
       .. code-block:: sh
 
 
-         sudo yum install openvino-2025.0.0
+         sudo yum install openvino-2025.1.0
 
 
 
@@ -164,7 +164,7 @@ To uninstall OpenVINO Runtime via YUM, run the following command based on your n
 
       .. code-block:: sh
 
-         sudo yum autoremove openvino-2025.0.0
+         sudo yum autoremove openvino-2025.1.0
 
 
 

src/bindings/python/src/pyopenvino/core/tensor.cpp

@@ -216,7 +216,8 @@ void regclass_Tensor(py::module m) {
 
     cls.def(py::init([](py::object& image) {
                 if (!py::isinstance(image, py::module::import("PIL.Image").attr("Image"))) {
-                    throw py::type_error("Input must be a PIL.Image.Image object");
+                    throw py::type_error(
+                        "Input argument must be a PIL.Image.Image/numpy.array/List[int, float, str] object");
                 }
                 auto numpy = py::module::import("numpy");
                 py::array np_array = numpy.attr("array")(image);
@@ -225,19 +226,19 @@ void regclass_Tensor(py::module m) {
             }),
             py::arg("image"),
             R"(
-                Constructs Tensor from a Pillow Image.
+            Constructs Tensor from a Pillow Image.
 
-                :param image: Pillow Image to create the tensor from.
-                :type image: PIL.Image.Image
-                :Example:
-                .. code-block:: python
+            :param image: Pillow Image to create the tensor from.
+            :type image: PIL.Image.Image
+            :Example:
+            .. code-block:: python
 
-                    from PIL import Image
-                    import openvino as ov
+                from PIL import Image
+                import openvino as ov
 
-                    img = Image.open("example.jpg")
-                    tensor = ov.Tensor(img)
-            )");
+                img = Image.open("example.jpg")
+                tensor = ov.Tensor(img)
+        )");
 
     cls.def("get_element_type",
             &ov::Tensor::get_element_type,

src/bindings/python/src/pyopenvino/graph/op.cpp

@@ -53,7 +53,17 @@ bool PyOp::evaluate(ov::TensorVector& output_values, const ov::TensorVector& inp
 }
 
 bool PyOp::has_evaluate() const {
-    PYBIND11_OVERRIDE(bool, ov::op::Op, has_evaluate);
+    py::gil_scoped_acquire gil;  // Acquire the GIL while in this scope.
+    // Try to look up the overridden method on the Python side.
+    py::function overrided_py_method = pybind11::get_override(this, "has_evaluate");
+    if (overrided_py_method) {                                 // method is found
+        return static_cast<py::bool_>(overrided_py_method());  // Call the Python function.
+    }
+    py::function overrided_evaluate_method = pybind11::get_override(this, "evaluate");
+    if (overrided_evaluate_method) {
+        return true;
+    }
+    return false;
 }
 
 void PyOp::update_type_info() {

src/bindings/python/tests/test_graph/test_custom_op.py

@@ -106,6 +106,10 @@ def visit_attributes(self, visitor):
         visitor.on_attributes(self._attrs)
         return True
 
+    def evaluate(self, outputs, inputs):
+        inputs[0].copy_to(outputs[0])
+        return True
+
 
 # request - https://docs.pytest.org/en/7.1.x/reference/reference.html#request
 @pytest.fixture
@@ -156,6 +160,14 @@ def test_visit_attributes_custom_op(prepared_paths, attributes, expectation, rai
     if e is not None:
         assert raise_msg in str(e.value)
 
+    input_data = np.ones([2, 1], dtype=np.float32)
+    expected_output = np.maximum(0.0, input_data)
+
+    compiled_model = compile_model(model_with_op_attr)
+    input_tensor = Tensor(input_data)
+    results = compiled_model({"data1": input_tensor})
+    assert np.allclose(results[list(results)[0]], expected_output, 1e-4, 1e-4)
+
 
 def test_custom_add_op():
     data1 = np.array([1, 2, 3])

src/core/include/openvino/runtime/tensor.hpp

@@ -270,9 +270,12 @@ using TensorVector = std::vector<Tensor>;
 ///              length of the file content and `offset`. Default value is [?].
 /// \param offset_in_bytes Read file starting from specified offset. Default is 0. The remining size of the file should
 /// be compatible with shape.
+/// \param mmap Use mmap that postpones real read from file until data is accessed. If mmap is used, the file
+///             should not be modified until returned tensor is destroyed.
 OPENVINO_API
 Tensor read_tensor_data(const std::filesystem::path& file_name,
                         const element::Type& element_type = element::u8,
                         const PartialShape& shape = PartialShape::dynamic(1),
-                        std::size_t offset_in_bytes = 0);
+                        std::size_t offset_in_bytes = 0,
+                        bool mmap = true);
 }  // namespace ov

src/core/src/op/constant.cpp

@@ -658,10 +658,19 @@ bool Constant::has_evaluate() const {
 }
 
 bool Constant::evaluate_lower(TensorVector& outputs) const {
-    return evaluate(outputs, {});
+    if (!outputs.empty() && outputs[0].get_element_type() != m_element_type)
+        return evaluate(outputs, {});  // for TypeRelaxed<Constant>
+    outputs.resize(1);
+    outputs[0] = get_tensor_view();
+    return get_data_ptr() != nullptr;
 }
+
 bool Constant::evaluate_upper(TensorVector& outputs) const {
-    return evaluate(outputs, {});
+    if (!outputs.empty() && outputs[0].get_element_type() != m_element_type)
+        return evaluate(outputs, {});  // for TypeRelaxed<Constant>
+    outputs.resize(1);
+    outputs[0] = get_tensor_view();
+    return get_data_ptr() != nullptr;
 }
 
 bool Constant::can_constant_fold(const OutputVector& input_values) const {

src/core/src/runtime/tensor.cpp

@@ -162,26 +162,44 @@ ov::Shape calc_static_shape_for_file(const std::filesystem::path& file_name,
     dynamic_dimension = Dimension(new_dimension_size);
     return partial_shape_copy.get_shape();
 }
+
+void read_tensor_data(const std::filesystem::path& file_name, Tensor& tensor, size_t offset) {
+    OPENVINO_ASSERT(tensor.get_element_type() != ov::element::string);
+    std::ifstream fin(file_name, std::ios::binary);
+    fin.seekg(offset);
+    auto bytes_to_read = tensor.get_byte_size();
+    fin.read(static_cast<char*>(tensor.data()), bytes_to_read);
+    OPENVINO_ASSERT(static_cast<size_t>(fin.gcount()) == bytes_to_read,
+                    "Cannot read ",
+                    bytes_to_read,
+                    "bytes from ",
+                    file_name);
+}
 }  // namespace
 
 Tensor read_tensor_data(const std::filesystem::path& file_name,
                         const ov::element::Type& element_type,
                         const ov::PartialShape& partial_shape,
-                        size_t offset_in_bytes) {
+                        size_t offset_in_bytes,
+                        bool mmap) {
     OPENVINO_ASSERT(element_type != ov::element::string);
     auto static_shape = calc_static_shape_for_file(file_name, element_type, partial_shape, offset_in_bytes);
-
-    auto mapped_memory = ov::load_mmap_object(file_name);
-    auto shared_buffer =
-        std::make_shared<ov::SharedBuffer<std::shared_ptr<ov::MappedMemory>>>(mapped_memory->data() + offset_in_bytes,
-                                                                              mapped_memory->size() - offset_in_bytes,
-                                                                              mapped_memory);
-
-    auto view_tensor = Tensor(element_type, static_shape, shared_buffer->get_ptr());
-    auto impl = get_tensor_impl(view_tensor);
-    impl._so = shared_buffer;
-    view_tensor = make_tensor(impl);
-
-    return view_tensor;
+    if (mmap) {
+        auto mapped_memory = ov::load_mmap_object(file_name);
+        auto shared_buffer = std::make_shared<ov::SharedBuffer<std::shared_ptr<ov::MappedMemory>>>(
+            mapped_memory->data() + offset_in_bytes,
+            mapped_memory->size() - offset_in_bytes,
+            mapped_memory);
+
+        auto view_tensor = Tensor(element_type, static_shape, shared_buffer->get_ptr());
+        auto impl = get_tensor_impl(view_tensor);
+        impl._so = shared_buffer;
+        view_tensor = make_tensor(impl);
+        return view_tensor;
+    } else {
+        ov::Tensor tensor(element_type, static_shape);
+        read_tensor_data(file_name, tensor, offset_in_bytes);
+        return tensor;
+    }
 }
 }  // namespace ov