[PT FE] Add ModuleExtension (#23536)

### Details:
 - *Continuation of #22867*

### Tickets:
 - *CVS-133733*

---------

Co-authored-by: Sergey Lyalin <[email protected]>

src/bindings/python/src/openvino/frontend/pytorch/__init__.py

@@ -13,6 +13,7 @@
     from openvino.frontend.pytorch.py_pytorch_frontend import _Type as DecoderType
     from openvino.frontend.pytorch.py_pytorch_frontend import ConversionExtensionPytorch as ConversionExtension
     from openvino.frontend.pytorch.py_pytorch_frontend import OpExtensionPytorch as OpExtension
+    from openvino.frontend.pytorch.module_extension import ModuleExtension
 except ImportError as err:
     raise ImportError("OpenVINO PyTorch frontend is not available, please make sure the frontend is built."
                       "{}".format(err))

src/bindings/python/src/openvino/frontend/pytorch/gptq.py

@@ -1,4 +1,3 @@
-
 # Copyright (C) 2018-2024 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 

src/bindings/python/src/openvino/frontend/pytorch/module_extension.py

@@ -0,0 +1,39 @@
+# Copyright (C) 2018-2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+# flake8: noqa
+# mypy: ignore-errors
+
+class ModuleExtension:
+    def __init__(self, module, target_op, evaluate=None, convert=None):
+        """
+        Creates an extension that replaces entire PyTorch module by a single operation.
+        This functionality works with PyTorch models only. A module can be identified by
+        module type (e.g. torch.nn.Linear), module instance in the model or module name.
+
+        Args:
+            module (str, torch.nn.Module, type(torch.nn.Module)): PyTorch module to replace
+
+            target_op (str): a target operation that will be used as a replacer for the module,
+                            could be a name of the extension operation or existing PyTorch operation
+                            (with prim:: or aten:: prefix following TorchScript syntax).
+
+            evaluate (callable with args module, *args, **kwargs): a callable that will replace a target
+                            module in model execution it is responsible for producing valid output for
+                            the module to allow correct model tracing. By default it calls original module
+                            forward with the same arguments. The provided code will not be a part of the final
+                            traced model, it is used only to produce valid results in the tracing.
+
+            convert (callable with args target_op, *args, **kwargs): a callable that will be traced and become
+                            a part of the final model instead of the target module. It accepts target_op as
+                            the first parameter, target_op is callable that will appear as a single node in the
+                            graph, the type of the node is target_op provided as another argument above.
+        """
+        self.module = module
+        self.target_op = target_op
+        self.evaluate = evaluate
+        if self.evaluate is None:
+            self.evaluate = lambda module, *args, **kwargs: module(*args, **kwargs)
+        self.convert = convert
+        if self.convert is None:
+            self.convert = lambda module, target_op, *args, **kwargs: target_op(*args, **kwargs)

src/bindings/python/src/openvino/frontend/pytorch/patch_model.py

@@ -0,0 +1,78 @@
+# Copyright (C) 2018-2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+# flake8: noqa
+# mypy: ignore-errors
+
+import torch
+
+
+class no_jit_trace:
+    def __enter__(self):
+        self.state = torch._C._get_tracing_state()
+        torch._C._set_tracing_state(None)
+
+    def __exit__(self, *args):
+        torch._C._set_tracing_state(self.state)
+        self.state = None
+
+
+def patch_model(model, module_extensions, orig_forward_name):
+    for name, m in model.named_modules():
+        if hasattr(m, orig_forward_name):
+            # already patched, skipping with a warning because it is unexpected
+            print(f'[ WARNING ] Unexpectedly found already patched module {name} while applying ModuleExtension during PyTorch model conversion. '
+                  'Result of the conversion maybe broken. Depending on the exact issue it may lead to broken original model.')
+            continue
+        extension = None
+        if m in module_extensions:
+            extension = module_extensions[m]
+        elif m.__class__ in module_extensions:
+            extension = module_extensions[m.__class__]
+        elif name in module_extensions:
+            extension = module_extensions[name]
+
+        if extension:
+            # The Trampoline class is instantiated for every module replacement, so we can use class members individually for each module.
+            class Trampoline(torch.autograd.Function):
+                target_extension = extension
+                original_module = m
+                stashed_args = None
+                stashed_kwargs = None
+
+                @staticmethod
+                @torch.jit.ignore
+                def forward(*args, **kwargs):
+                    with no_jit_trace():
+                        # `module` is going to be passed to a user-defined function `evaluate`
+                        # `module` is patched: forward function was replaced, and we are actually in this patched function right in this code
+                        # if we pass `module` as-is to the user code below, and it happens to call forward it will lead to infinite recursion or fail
+                        # so we need to temporary patch the module back to the original forward and then return it back again
+                        # stash the current forward to be able to return it back
+                        patched_forward = m.forward
+                        # set original forward for the module
+                        m.forward = getattr(m, orig_forward_name)
+                        # call user code
+                        results = extension.evaluate(
+                            m, *Trampoline.stashed_args, **Trampoline.stashed_kwargs)
+                        m.forward = patched_forward  # return patched forward back
+                        return results
+
+            def new_forward(*args, **kwargs):
+                Trampoline.stashed_args = args
+                Trampoline.stashed_kwargs = kwargs
+                return extension.convert(m, Trampoline.apply, *args, **kwargs)
+            setattr(m, orig_forward_name, m.forward)
+            m.forward = new_forward
+
+
+def unpatch_model(model, orig_forward_name):
+    for _, m in model.named_modules():
+        if hasattr(m, orig_forward_name):
+            try:
+                m.forward = getattr(m, orig_forward_name)
+                delattr(m, orig_forward_name)
+            except Exception as error:
+                print('[ WARNING ] Exception raised during model unpatching. Depending on the exact issue it may lead to broken original model.')
+                print('Original exception details:')
+                print(error)

src/bindings/python/src/openvino/frontend/pytorch/ts_decoder.py

@@ -10,20 +10,32 @@
 from openvino.frontend.pytorch.utils import ivalue_to_constant, get_value_from_getattr, pt_to_ov_type_map, prepare_example_inputs_and_model, convert_quantized_tensor, graph_has_ops
 from openvino.runtime import opset11 as ops
 from openvino.frontend.pytorch import gptq
+from openvino.frontend.pytorch import patch_model
+from openvino.frontend.pytorch.module_extension import ModuleExtension
 
 import typing
 import torch
 
 
 class TorchScriptPythonDecoder (Decoder):
-    def __init__(self, pt_module, graph_element=None, example_input=None, alias_db=None, shared_memory=True, skip_freeze=False, constant_cache=None):
+    def __init__(
+            self,
+            pt_module,
+            graph_element=None,
+            example_input=None,
+            alias_db=None,
+            shared_memory=True,
+            skip_freeze=False,
+            constant_cache=None,
+            module_extensions=None):
         Decoder.__init__(self)
         # We store every decoder created by this decoder so that all them are not deleted until the first decoder is deleted
         self.m_decoders = []
         self._input_signature = None
         self._shared_memory = shared_memory
         self._input_is_list = False
         self.constant_cache = constant_cache if constant_cache is not None else dict()
+        self.module_extensions = module_extensions
         if graph_element is None:
             try:
                 pt_module = self._get_scripted_model(
@@ -89,14 +101,22 @@ def _get_scripted_model(self, pt_module, example_inputs=None, skip_freeze=False)
             input_params = inspect.signature(pt_module.forward if hasattr(
                 pt_module, "forward") else pt_module.__call__).parameters
             input_signature = list(input_params)
+
             if example_inputs is None:
+                if self.module_extensions:
+                    raise RuntimeError("ModuleExtension is not supported for scripting. Please provide valid example_input argument to run tracing.")
                 scripted = torch.jit.script(pt_module)
                 freeze_by_default = True
             else:
                 input_parameters, input_signature, pt_module, self._input_is_list = prepare_example_inputs_and_model(
                     example_inputs, input_params, pt_module)
-                gptq_patched = False
 
+                # name of attribute in a patched module where the original forward method is kept
+                orig_forward_name = '_openvino_module_extension_patch_orig_forward'
+                if self.module_extensions:
+                    patch_model.patch_model(pt_module, self.module_extensions, orig_forward_name)
+
+                gptq_patched = False
                 if gptq.detect_gptq_model(pt_module):
                     try:
                         gptq.patch_model(pt_module)
@@ -115,6 +135,8 @@ def _get_scripted_model(self, pt_module, example_inputs=None, skip_freeze=False)
                 finally:
                     if gptq_patched:
                         gptq.unpatch_model(pt_module)
+                    if self.module_extensions:
+                        patch_model.unpatch_model(pt_module, orig_forward_name)
 
             if not freeze_by_default and graph_has_ops(scripted.inlined_graph, ["prim::Uninitialized", "prim::unchecked_cast", "aten::append"]):
                 # freeze models with unsupported ops
@@ -232,7 +254,8 @@ def visit_subgraph(self, node_visitor) -> None:
                                                node,
                                                alias_db=self.alias_db,
                                                shared_memory=self._shared_memory,
-                                               constant_cache=self.constant_cache)
+                                               constant_cache=self.constant_cache,
+                                               module_extensions=self.module_extensions)
             self.m_decoders.append(decoder)
             node_visitor(decoder)
 
@@ -255,13 +278,28 @@ def get_subgraph_decoder(self, index: int):
         decoder = TorchScriptPythonDecoder(self.pt_module,
                                            self.get_subgraphs()[index],
                                            alias_db=self.alias_db,
-                                           shared_memory=self._shared_memory)
+                                           shared_memory=self._shared_memory,
+                                           module_extensions=self.module_extensions)
         self.m_decoders.append(decoder)
         return decoder
 
     def get_op_type(self) -> str:
         assert isinstance(
             self.graph_element, torch.Node), "Function can be called only when self.graph_element is of type torch.Node"
+        if self.graph_element.kind() == "prim::PythonOp":
+            if hasattr(self.graph_element, 'pyobj') and callable(self.graph_element.pyobj) and hasattr(self.graph_element.pyobj(), '__self__'):
+                trampoline = self.graph_element.pyobj().__self__
+                if hasattr(trampoline, 'target_extension') and isinstance(trampoline.target_extension, ModuleExtension):
+                    target_op = trampoline.target_extension.target_op
+                    if callable(target_op):
+                        target = target_op(trampoline.original_module)
+                    elif isinstance(target_op, str):
+                        target = target_op
+                    # TODO: Support target as a callable that will play a role of ConversionExtension for an entire module instead of a single op.
+                    # Without supporting target as a callable here, ConversionExtension functionality is still possible to implement
+                    # by combining two extensions: ModuleExtension that use temporary name as a target op and another extension of type ConversionExtension
+                    # that translates that particular temporary name to custom graph. But providing conversion code as a callable `target` is more convenient.
+                    return target
         return self.graph_element.kind()
 
     def get_schema(self) -> str:

src/bindings/python/src/openvino/frontend/pytorch/utils.py

@@ -1,4 +1,3 @@
-
 # Copyright (C) 2018-2024 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 

src/frontends/pytorch/src/frontend.cpp

@@ -290,7 +290,7 @@ ov::frontend::InputModel::Ptr FrontEnd::load_impl(const std::vector<ov::Any>& va
 }
 
 std::map<std::string, CreatorFunction> FrontEnd::get_supported_ops(const ov::frontend::InputModel::Ptr& model) const {
-    std::map<std::string, CreatorFunction> supported_ops = get_supported_ops_fx();
+    std::map<std::string, CreatorFunction> supported_ops;
     if (std::dynamic_pointer_cast<pytorch::InputModel>(model)->decoder_type_name() == "fx")
         supported_ops = get_supported_ops_fx();
     else

tests/layer_tests/py_frontend_tests/test_torch_frontend.py

@@ -291,17 +291,17 @@ def forward(self, x):
         "Parameter", "ReluCustom", "Result"]
 
 
-def test_op_extension():
-    from openvino.frontend.pytorch.ts_decoder import TorchScriptPythonDecoder
-    from openvino.frontend.pytorch import OpExtension
-
-    class CosModel(torch.nn.Module):
-        def __init__(self):
+class CosModel(torch.nn.Module):
+    def __init__(self):
             super(CosModel, self).__init__()
 
-        def forward(self, x):
+    def forward(self, x):
             return torch.cos(x.to(torch.float32))
 
+def test_op_extension():
+    from openvino.frontend.pytorch.ts_decoder import TorchScriptPythonDecoder
+    from openvino.frontend.pytorch import OpExtension
+
     model = CosModel()
     decoder = TorchScriptPythonDecoder(get_scripted_model(model))
 
@@ -327,13 +327,6 @@ def test_op_extension_generic():
     from openvino.frontend.pytorch.ts_decoder import TorchScriptPythonDecoder
     from openvino.frontend import OpExtension
 
-    class CosModel(torch.nn.Module):
-        def __init__(self):
-            super(CosModel, self).__init__()
-
-        def forward(self, x):
-            return torch.cos(x.to(torch.float32))
-
     model = CosModel()
     decoder = TorchScriptPythonDecoder(get_scripted_model(model))
 
@@ -355,6 +348,49 @@ def forward(self, x):
         "Parameter", "Convert", "Sin", "Result"]
 
 
+def test_module_extension():
+    from openvino.frontend.pytorch.ts_decoder import TorchScriptPythonDecoder
+    from openvino.frontend.pytorch import ModuleExtension
+    from openvino import convert_model
+
+    class ModelWithModule(torch.nn.Module):
+        def __init__(self):
+            super(ModelWithModule, self).__init__()
+            self.cos_module = CosModel()
+
+        def forward(self, x):
+            return self.cos_module(x)
+
+    model = ModelWithModule()
+    decoder = TorchScriptPythonDecoder(model)
+
+    fem = FrontEndManager()
+    fe = fem.load_by_framework(framework="pytorch")
+    assert fe
+
+    input_model = fe.load(decoder)
+    assert input_model
+    converted_model = fe.convert(input_model)
+    assert converted_model
+    assert [n.get_type_name() for n in converted_model.get_ordered_ops()] == [
+        "Parameter", "Convert", "Cos", "Result"]
+
+    converted_model = convert_model(model, example_input=(torch.randn(100),), extension=[ModuleExtension(CosModel, "aten::sin")])
+    assert converted_model
+    assert [n.get_type_name() for n in converted_model.get_ordered_ops()] == [
+        "Parameter", "Sin", "Result"]
+
+    converted_model = convert_model(model, example_input=(torch.randn(100),), extension=[ModuleExtension(model.cos_module, "aten::sin")])
+    assert converted_model
+    assert [n.get_type_name() for n in converted_model.get_ordered_ops()] == [
+        "Parameter", "Sin", "Result"]
+
+    converted_model = convert_model(model, example_input=(torch.randn(100),), extension=[ModuleExtension("cos_module", "aten::sin")])
+    assert converted_model
+    assert [n.get_type_name() for n in converted_model.get_ordered_ops()] == [
+        "Parameter", "Sin", "Result"]
+
+
 def test_pytorch_telemetry():
     from openvino.frontend import TelemetryExtension
     from openvino.frontend.pytorch.ts_decoder import TorchScriptPythonDecoder
@@ -547,7 +583,7 @@ def forward(self, x: float, y: torch.Tensor):
     assert PartialShape(pt_out_shape) == om.get_output_partial_shape(0)
 
 
-class TestModel1(torch.nn.Module):
+class ModelTest1(torch.nn.Module):
     def __init__(self, *args, **kwargs) -> None:
         super().__init__(*args, **kwargs)
         self.pool = torch.nn.AdaptiveAvgPool2d(1)
@@ -559,8 +595,7 @@ def forward(self, x):
 def test_output_dict_names():
     from openvino.frontend.pytorch.ts_decoder import TorchScriptPythonDecoder
 
-    input = torch.ones((1, 3, 224, 224))
-    model = TestModel1()
+    model = ModelTest1()
     decoder = TorchScriptPythonDecoder(
         model, example_input=(torch.randn(1, 3, 224, 224),))
     fe_manager = FrontEndManager()
@@ -570,7 +605,7 @@ def test_output_dict_names():
     assert om.outputs[0].any_name == "x1" and om.outputs[1].any_name == "x2", "Output dict names are not expected"
 
 
-class TestModel2(torch.nn.Module):
+class ModelTest2(torch.nn.Module):
     def __init__(self, *args, **kwargs) -> None:
         super().__init__(*args, **kwargs)
         self.pool = torch.nn.AdaptiveAvgPool2d(1)
@@ -582,8 +617,7 @@ def forward(self, x):
 def test_output_tuple_names():
     from openvino.frontend.pytorch.ts_decoder import TorchScriptPythonDecoder
 
-    input = torch.ones((1, 3, 224, 224))
-    model = TestModel2()
+    model = ModelTest2()
     decoder = TorchScriptPythonDecoder(
         model, example_input=(torch.randn(1, 3, 224, 224),))
     fe_manager = FrontEndManager()

tools/ovc/openvino/tools/ovc/convert_impl.py

@@ -38,6 +38,7 @@
 
 # pylint: disable=no-name-in-module,import-error
 from openvino.frontend import FrontEndManager, OpConversionFailure, TelemetryExtension
+from openvino.frontend.pytorch.module_extension import ModuleExtension
 from openvino.runtime import get_version as get_rt_version
 from openvino.runtime import Type, PartialShape
 
@@ -173,7 +174,8 @@ def prepare_ir(argv: argparse.Namespace):
         moc_front_end.add_extension(TelemetryExtension("ovc", t.send_event, t.send_error, t.send_stack_trace))
         if any_extensions_used(argv):
             for extension in argv.extension:
-                moc_front_end.add_extension(extension)
+                if not isinstance(extension, ModuleExtension):
+                    moc_front_end.add_extension(extension)
         ov_model = moc_pipeline(argv, moc_front_end)
         return ov_model