[Frontend][Paddle]Handle Exception in Op Conversion.

ngraph/frontend/paddlepaddle/src/frontend.cpp

@@ -49,8 +49,15 @@ NamedOutputs make_ng_node(const std::map<pdpd::TensorName, Output<Node>>& nodes,
             named_inputs[input_port.parameter()].push_back(node_it->second);
         }
     }
+    NamedOutputs outputs;
+    // In case the conversion function throws exception
+    try {
+        outputs = creator_it->second(NodeContext(DecoderPDPDProto(op_place), named_inputs));
+    } catch (std::exception& ex) {
+        FRONT_END_OP_CONVERSION_CHECK(false, "Fail to convert " + op_desc.type() + " Exception " + ex.what());
+    }
 
-    return creator_it->second(NodeContext(DecoderPDPDProto(op_place), named_inputs));
+    return outputs;
 }
 
 NamedOutputs make_framework_node(const std::map<pdpd::TensorName, Output<Node>>& nodes,

ngraph/frontend/paddlepaddle/src/model.cpp

@@ -289,7 +289,14 @@ InputModelPDPD::InputModelPDPDImpl::InputModelPDPDImpl(const std::basic_string<T
 
     FRONT_END_GENERAL_CHECK(pb_stream && pb_stream.is_open(), "Model file doesn't exist");
     FRONT_END_GENERAL_CHECK(m_fw_ptr->ParseFromIstream(&pb_stream), "Model can't be parsed");
-
+    // According to Paddle, the saved model has the framework version
+    // For example Paddle 2.1.0 is encoded as 2001000. 0 means the latest framework.
+    // https://github.com/PaddlePaddle/Paddle/blob/develop/cmake/version.cmake
+    // https://github.com/PaddlePaddle/Paddle/blob/2100816c5190693cc7dee181e96af72e9f0fbd1d/paddle/fluid/framework/program_desc.cc#L52
+    int64_t version = m_fw_ptr->version().version();
+    FRONT_END_GENERAL_CHECK(
+        version >= 2000000 || version == 0,
+        "[Frontend]Only Support Paddle greater than 2.0.0, current version " + std::to_string(version));
     loadPlaces();
     if (weights_stream && weights_stream.is_open()) {
         loadConsts(std::basic_string<T>{}, &weights_stream);
@@ -307,7 +314,10 @@ InputModelPDPD::InputModelPDPDImpl::InputModelPDPDImpl(const std::vector<std::is
                                 "Two streams are needed to load a model: model and weights streams");
     }
     FRONT_END_GENERAL_CHECK(m_fw_ptr->ParseFromIstream(streams[0]), "Model can't be parsed");
-
+    int64_t version = m_fw_ptr->version().version();
+    FRONT_END_GENERAL_CHECK(
+        version >= 2000000 || version == 0,
+        "[Frontend]Only Support Paddle greater than 2.0.0, current version " + std::to_string(version));
     loadPlaces();
     if (streams.size() > 1)
         loadConsts(std::string(), streams[1]);

ngraph/test/frontend/paddlepaddle/test_models/gen_scripts/generate_lower_version.py

@@ -0,0 +1,39 @@
+import paddle
+import numpy as np
+import os
+import sys
+from paddle.fluid.proto import framework_pb2
+
+paddle.enable_static()
+
+inp_blob = np.random.randn(1, 3, 4, 4).astype(np.float32)
+print(sys.path)
+main_program = paddle.static.Program()
+startup_program = paddle.static.Program()
+
+with paddle.static.program_guard(main_program, startup_program):
+    x = paddle.static.data(name='x', shape=[1, 3, 4, 4], dtype='float32')
+    test_layer = paddle.static.nn.conv2d(input=x, num_filters=5, filter_size=(1, 1), stride=(1, 1), padding=(1, 1),
+                                     dilation=(1, 1), groups=1, bias_attr=False)
+
+    cpu = paddle.static.cpu_places(1)
+    exe = paddle.static.Executor(cpu[0])
+    exe.run(startup_program)
+    inp_dict = {'x': inp_blob}
+    var = [test_layer]
+    res_paddle = exe.run(paddle.static.default_main_program(), fetch_list=var, feed=inp_dict)
+    paddle.static.save_inference_model(os.path.join(sys.argv[1], "lower_version/", "lower_version"), [x], [test_layer], exe, program=main_program)
+
+
+fw_model = framework_pb2.ProgramDesc()
+with open(os.path.join(sys.argv[1], "lower_version", "lower_version.pdmodel"), mode='rb') as file:
+    fw_model.ParseFromString(file.read())
+
+fw_model.version.version = 1800000
+print(fw_model.version.version)
+with open(os.path.join(sys.argv[1], "lower_version", "lower_version.pdmodel"), "wb") as f:
+    f.write(fw_model.SerializeToString())
+
+
+
+

ngraph/test/frontend/paddlepaddle/test_models/gen_scripts/generate_pad3d.py

@@ -57,6 +57,12 @@ def main():
     input_data = np.random.rand(*input_shape).astype(np.float32)
     pad3d("pad3d_test3", input_data, in_dtype, pad, data_format, mode)
 
+    input_shape = (2, 3, 4, 5, 6)
+    pad = [1, 2, 1, 1, 1, 2]
+    mode = "circular"
+    data_format = 'NDHWC'
+    input_data = np.random.rand(*input_shape).astype(np.float32)
+    pad3d("throw_in_conversion", input_data, in_dtype, pad, data_format, mode)
     # padding of type int feature only supported by PaddlePaddle 'develop' version(>=2.1.0)
 #    input_shape = (1, 2, 3, 4, 5)
 #    pad_int = 1

ngraph/test/frontend/paddlepaddle/throw_in_conversion.cpp

@@ -0,0 +1,39 @@
+// Copyright (C) 2018-2021 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include <frontend_manager/frontend_exceptions.hpp>
+#include <frontend_manager/frontend_manager.hpp>
+
+#include "common_test_utils/ngraph_test_utils.hpp"
+#include "paddle_utils.hpp"
+#include "utils.hpp"
+
+using namespace ngraph;
+using namespace ngraph::frontend;
+
+TEST(FrontEndConvertModelTest, throw_in_conversion) {
+    FrontEndManager fem;
+    FrontEnd::Ptr frontEnd;
+    InputModel::Ptr inputModel;
+    ASSERT_NO_THROW(frontEnd = fem.load_by_framework(PADDLE_FE));
+    ASSERT_NE(frontEnd, nullptr);
+    auto model_filename = FrontEndTestUtils::make_model_path(
+        std::string(TEST_PADDLE_MODELS_DIRNAME) + std::string("throw_in_conversion/throw_in_conversion.pdmodel"));
+    ASSERT_NO_THROW(inputModel = frontEnd->load(model_filename));
+    ASSERT_NE(inputModel, nullptr);
+    std::shared_ptr<ngraph::Function> function;
+    ASSERT_THROW(function = frontEnd->convert(inputModel), OpConversionFailure);
+}
+
+TEST(FrontEndConvertModelTest, unsupported_version) {
+    FrontEndManager fem;
+    FrontEnd::Ptr frontEnd;
+    InputModel::Ptr inputModel;
+    ASSERT_NO_THROW(frontEnd = fem.load_by_framework(PADDLE_FE));
+    ASSERT_NE(frontEnd, nullptr);
+    auto model_filename = FrontEndTestUtils::make_model_path(std::string(TEST_PADDLE_MODELS_DIRNAME) +
+                                                             std::string("lower_version/lower_version.pdmodel"));
+
+    ASSERT_THROW(inputModel = frontEnd->load(model_filename), GeneralFailure);
+}